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
|
|---|---|---|---|---|---|
Schistosomiasis , one of the most prevalent neglected tropical diseases , is a life-threatening public health problem in Yemen especially in rural communities . This cross-sectional study aims to determine the prevalence and associated risk factors of schistosomiasis among children in rural Yemen . Urine and faecal samples were collected from 400 children . Urine samples were examined using filtration technique for the presence of Schistosoma haematobium eggs while faecal samples were examined using formalin-ether concentration and Kato Katz techniques for the presence of S . mansoni . Demographic , socioeconomic and environmental information were collected via a validated questionnaire . Overall , 31 . 8% of the participants were found to be positive for schistosomiasis; 23 . 8% were infected with S . haematobium and 9 . 3% were infected with S . mansoni . Moreover , 39 . 5% of the participants were anaemic whereas 9 . 5% had hepatosplenomegaly . The prevalence of schistosomiasis was significantly higher among children aged >10 years compared to those aged ≤10 years ( P<0 . 05 ) . Multivariate analysis confirmed that presence of other infected family member ( P<0 . 001 ) , low household monthly income ( P = 0 . 003 ) , using unsafe sources for drinking water ( P = 0 . 003 ) , living nearby stream/spring ( P = 0 . 006 ) and living nearby pool/pond ( P = 0 . 002 ) were the key factors significantly associated with schistosomiasis among these children . This study reveals that schistosomiasis is still highly prevalent in Yemen . These findings support an urgent need to start an integrated , targeted and effective schistosomiasis control programme with a mission to move towards the elimination phase . Besides periodic drug distribution , health education and community mobilisation , provision of clean and safe drinking water , introduction of proper sanitation are imperative among these communities in order to curtail the transmission and morbidity caused by schistosomiasis . Screening and treating other infected family members should also be adopted by the public health authorities in combating this infection in these communities .
Schistosomiasis or bilharzia , one of the most prevalent neglected tropical diseases ( NTDs ) , is still a public health problem in many developing countries in the tropics and subtropics with approximately 240 million infected people and about 700 million people worldwide are at risk of this infection [1] . Over 90% of the disease is currently found in sub-Saharan Africa , where more than 200 , 000 deaths are annually attributed to schistosomiasis , and Middle East and North Africa regions [2]–[4] . Despite intensive efforts to control the disease , schistosomiasis together with soil-transmitted helminthiasis continue to represent more than 40% of the disease burden caused by all tropical diseases , excluding malaria [5] . Schistosomiasis is mainly caused by three different species of blood-dwelling fluke worms of the genus Schistosoma namely Schistosoma haematobium ( causes urinary schistosomiasis ) , S . mansoni and S . japonicum ( both cause intestinal schistosomiasis ) . Clinical manifestations of schistosomiasis are associated with the species-specific oviposition sites and the burden of infection [6] . Urinary schistosomiasis is characterized by haematuria as a classical sign . It is associated with bladder and uretral fibrosis , sandy patches in the bladder mucosa and hydronephrosis that are commonly seen in chronic cases while bladder cancer is possible as late stage complication [7] . On the other hand , intestinal clinical manifestations include abdominal pain , diarrhea , and blood in the stool . In advanced cases , hepatosplenomegaly is common and is repeatedly associated with ascites and other signs of portal hypertension [8] , [9] . Among the Middle East countries , Yemen has the highest percentage of people living in poverty where more than 50% of the population of nearly 25 million people lives below the poverty line [10] . The country has been unstable for several years , suffering from civil wars , a deteriorating economy and severe depletion in water resources . With regards to NTDs , Yemen is endemic for at least 8 NTDs namely soil-transmitted helminthiasis , schistosomiasis , onchocerciasis , lymphatic filariasis , leishmaniasis , fascioliasis , trachoma and leprosy . Moreover , the country ranks first in trachoma; second in schistosomiasis , ascariasis , fascioliasis and leprosy; and fourth in trichuriasis and cutaneous leishmaniasis [4] . In 2008 , Yemen launched its first campaign to eliminate schistosomiasis as a national public health problem with the aim of eliminating schistosomiasis-related morbidity through annual treatment to school-age children with a financial support from the World Bank and World Health Organization ( WHO ) [11] . Despite of these support and efforts to control the disease in Yemen , the prevalence of schistosomiasis remains largely unchanged ( since 1970s ) with prominent morbidity [12]–[17] . Moreover , new foci of schistosomiasis transmission have been identified . Hence , the aims of the present study were to determine the prevalence and distribution of schistosomiasis and to identify the associated key factors of this disease among Yemeni children in rural areas which are undergoing active control and prevention surveillances . It is hoped that findings of this study will assist public health authorities to identify and implement integrated and effective control measures to reduce the prevalence and burden of schistosomiasis significantly in rural Yemen .
The study protocol was approved by the Medical Ethics Committee of the University of Malaya Medical Centre ( Ref . no: 968 . 4 ) . It was also approved by the Hodeidah University , Yemen and permission to start data collection was also given by the Yemen Schistosomiasis National Control Project . The head of households and children were informed about the study objectives and methods and the priority of the consent for inclusion of children . Moreover , they were informed that they could withdraw their children from the study without any consequences . Thus , written and signed or thumb-printed informed consents were obtained from all adult participants before starting the survey . Similarly , written and signed or thumb-printed informed consents were taken from parents or guardians , on behalf of their children . All the infected children were treated with a single dose of 40 mg/kg body weight praziquantel tablets . Each child swallows the tablets with some water , while being observed by the researcher and medical officer ( Direct Observed Therapy ) [18] . A cross-sectional community-based study was carried out among children aged ≤15 years in rural areas in Yemen . Data were collected in a period of seven months from January to July 2012 . In each province , two rural districts were selected randomly from the available district list and then two villages within the selected districts were considered in collaboration with the Schistosomiasis Control Project office in each province . The number of inhabitants per household was recorded and all of them were invited to participate in this study . Unique reference codes were assigned to each households and study participants . This study was carried out in five provinces in Yemen namely Taiz , Ibb , Dhamar , Sana'a and Hodiedah . These provinces are endemic for schistosomiasis and undergoing active surveillances by the schistosomiasis national control project . The highest prevalence of schistosomiasis was reported in Hajjah and Taiz provinces [15] , [17] . However , we could not collect samples from Hajjah during the sampling period due to civil war which occurred in 3 provinces including Hajjah . Sana'a and Dhamar represent the mountainous areas at an altitude of >2000 m above sea level with a total population of 4 million . Taiz , Hodiedah and Ibb represent the country's coastal plains and foothills at an altitude of <2000 m above sea level with a total population of 6 . 5 million . In Yemen , climate varies from hot and high humidity in the coastal areas to cold in the highlands . In the coastal areas , relative humidity ranges between 70% and 90% and mean annual rainfall is about 200 mm with two rainy seasons ( February–April and July–September ) . In the highlands , the relative humidity ranges between 20% and 50% , mean annual rainfall is about 800 mm , and the climate is moderate in summer and cold in winter . Ten districts were selected for this study namely Mosa and Almafer ( Taiz ) , Alsabrah and Alodien ( Ibb ) , Otmah and Gabal al sharq ( Dhamar ) , Alhemah and Manakhah ( Sana'a ) , and Gabal Ras and Bora ( Hodiedah ) ( Figure 1 ) . The inclusion criteria in selecting these study areas were rural areas and undergoing active control surveillance . Moreover , the selection process was done after discussion with the schistosomiasis national control project personnel . Rural areas are farmlands which depend on streams , underground wells and rain ( water tanks ) as the main source of water for domestic and irrigation purposes . Agriculture is the main occupation of the people in these areas and surface traditional irrigation system is still dominant and covers large agricultural areas creating favorable snail-breeding conditions . Snail populations of different genera were identified in different water sources at the study areas and heavily infested water sources were observed . Out of about 780 households , 250 households were selected randomly from the villages for this study . We attempted to enrol all available children ≤15 years of age from the selected households after acquiring consent from the head of the households . Although 632 children received stool and urine containers , only 430 ( 68 . 0% ) children delivered the containers for examinations . In this study , 202 ( 32 . 0% ) failed to submit samples and/or absent during questionnaire surveys and 30 ( 4 . 7% ) containers were returned empty . Hence , they were excluded from the study . Overall , 400 ( 63 . 3% ) children ( 59 . 5% males and 40 . 5% females ) who had delivered suitable samples for examination with complete questionnaire data were included in this study ( Figure 2 ) . Throughout many visits to the study areas , most of the children were observed to play outside without wearing shoes or slippers . Some of the children play and swim in the streams/pools after school and in their leisure time . Besides that , their personal hygienic practices were also poor . Poverty prevails at these areas with very poor housing and living conditions and was notably observed in areas from Hodeidah province . A pretested questionnaire was used to collect data about the demographic , socio-economic , environmental background , personal hygiene , clinical signs and symptoms of urinary or intestinal schistosomiasis and history of receiving anti-schistosomal treatment . Information was collected from the children's parents or adult guardians via face-to-face interview . In addition , participants underwent physical examination and direct observation in order to record more details about their weight , height , body temperature , hepatosplenomegaly and personal hygiene . During the interviews , direct observation was made by an assistant on the personal hygiene of the children and household cleanliness including the availability of functioning toilets , piped water , cutting nails , wearing shoes when outside the house and washing hands . Faecal and urine samples were collected from each subject , between 10 am and 2 pm when maximum eggs excretion occurs [19] , into 100 mL clean containers with wide mouth and screw-cap . The containers were placed into zipped plastic bags , kept in a protected ice box and transported for examination at the nearest health center laboratory within 4 hours of collection . The faecal samples were examined using Kato-Katz technique for the presence of S . mansoni eggs [18] . Negative faecal samples were re-examined by formalin ether sedimentation technique as described by Cheesbrough [20] before the negative results were confirmed . To determine the worm burden , egg counts were taken and recorded as eggs per gram of feces ( epg ) for each positive sample and the intensity of infections was graded as heavy , moderate or light according to the criteria proposed by the WHO [18] . On the other hand , urine samples were examined for the presence of S . haematobium eggs by filtration method using nucleopore membrane . Besides that , dipstick test was also used [21] . For quality control , urine and faecal samples examination were performed in duplicate for about 25% of the samples , selected randomly . A finger prick blood was obtained from each child and hemoglobin ( Hb ) level was assessed directly by using the HemoCue hemoglobinometer ( HemoCue , AB , Angelhom , Sweden ) . Children with Hb levels lower than 12 g/dl were considered to be anaemic [22] . Data was double-entered by two different researchers into Microsoft Office Excel 2007 spreadsheets . Then , research leader cross-checked the two data sets for accuracy and created a single data set . Data analysis was performed by using Statistical Package for Social Sciences for Windows ( SPSS ) version 18 . Presence of schistosomiasis , demographic , socioeconomic , environmental and behavioural characteristics were treated as categorical variables and presented as frequencies and percentages . For inferential statistics , the dependent variable was schistosomiasis whereas the independent variables were the demographic factors ( age and gender ) , socioeconomic factors ( fathers' educational levels , parents' employment status , household monthly income , family size ) and environmental factors ( sources of drinking water , sources of household water , presence of streams , dams , wells , ponds , tanks , pools , troughs or any other man-made water collection places ) . Chi-square test was used to examine the significance of the associations and differences in frequency distribution of variables . Odd ratios ( OR ) and 95% confidence intervals ( CI ) were computed . Multiple logistic regression analysis was used to identify the factors significantly associated with schistosomiasis; OR and its corresponding 95% CI were calculated based on the final model . All variables that showed significant difference with P≤0 . 25 in the univariate analyses were used to develop the multiple logistic regression “STEPWISE” models as suggested by Bendel and Afifi [23] . All tests were considered significant at P<0 . 05 .
Four hundred children aged ≤15 years with a mean age of 10 years ( 95% CI = 9 . 7 , 10 . 2 ) participated voluntarily in this study . The general characteristics of the participants and their families are shown in Table 1 . Overall , almost half of the fathers had no formal education and about half ( 48 . 8% ) of them were farmers whereas 42 . 7% were government employee and/or professionals . On the other hand , almost all the mothers had no formal education and were not working ( i . e . , housewives ) . Moreover , more than half ( 59 . 3% ) of the families had low household monthly income ( <YER20 , 000; US$1 = YER214 ) . Most of the houses were made of stones and mud whereas few houses are made of burned bricks; only one fifth and a quarter of the houses had piped water supply and electricity , respectively . During the visits to the villages , we observed that toddlers and young children were playing and swimming in the water streams and pools . Many other people were observed using this water for different purposes such as washing clothes and utensils , washing cars and motorcycles and watering animals . Urine and faecal samples were collected from 400 children and examined for the presence of Schistosoma species and other parasites . A total of 339 ( 84 . 8% ) children were found to be infected with at least one parasite species . Besides S . mansoni , Ascaris lumbricoides , Trichuris trichiura , Ancylostoma duodenale , Hymenolepis nana , Enterobius vermicularis , Taenia sp . , Fasciola sp . , Entamoeba histolytica/dispar , Giardia duodenalis and Blastocystis sp . were also detected in the faecal samples examined . The prevalence and intensity of schistosomiasis according to Schistosoma species are shown in Table 2 . Of the 400 participants , 127 ( 31 . 8% ) were found to be infected by either S . mansoni or S . haematobium . Out of these infected children , 3 . 9% had mixed infections ( i . e . , both Schistosoma species ) . Overall , the prevalence of S . haematobium infection was higher than S . mansoni ( 23 . 8% vs 9 . 3% ) . The prevalence of schistosomiasis was significantly higher among children aged >10 years compared to those aged ≤10 years ( 37 . 6% vs 27 . 0%; χ2 = 5 . 135; P = 0 . 023 ) . Similarly , male children had higher prevalence of schistosomiasis than females ( 33 . 6% vs 29 . 0% ) . However , the difference was not statistically significant ( χ2 = 0 . 942; P = 0 . 332 ) . With regards to the intensity of infections , 22 . 1% and 8 . 1% of S . haematobium and S . mansoni infections respectively were of heavy intensities ( Table 2 ) . Children who participated in this study underwent physical examination and haemoglobin level was measured . Hepatosplenomegaly and anaemia were reported in 9 . 5% ( 38/400 ) and 39 . 5% ( 158/400 ) of the children , respectively . Moreover , 15 . 8% ( 63/400 ) had fever whilst 24 . 1% ( 96/400 ) had diarrhea . Of these studied children , 26 . 0% ( 104/400 ) and/or 15 . 0% ( 60/400 ) claimed to have haematuria and bloody stool , respectively . The association between schistosomiasis and the presence of hepatosplenomegaly and anaemia was examined . Children with S . mansoni infection had a significantly higher rate of hepatosplenomegaly ( 18 . 9%; 95% CI = 9 . 5 , 34 . 2 ) when compared with those without S . mansoni infection ( 8 . 3%; 95% CI = 5 . 8 , 11 . 4 ) whereas no significant difference in the case of S . haematobium infection . A significant association between the intensity of S . mansoni infection and hepatosplenomegaly was also reported ( P = 0 . 033 ) . Moreover , the presence of hepatosplenomegaly was significantly higher among children with mixed infection ( both Schistosoma species ) compared to those with single infection ( P>0 . 05 ) . On the other hand , the association between schistosomiasis and anaemia among these children was not significant ( P>0 . 05 ) . Results of univariate and multivariate analyses for the association of schistosomiasis with demographic , socioeconomic , environmental and behavioural factors are shown in Tables 3 and 4 . Table 3 shows that children aged >10 years ( 37 . 6%; 95% CI = 30 . 8 , 44 . 5 ) had significantly higher prevalence of schistosomiasis when compared with those aged ≤10 years ( 27 . 0%; 95% CI = 21 . 6 , 33 . 2 ) . Similarly , the prevalence of schistosomiasis was significantly higher among children of non educated fathers ( 38 . 2%; 95% CI = 31 . 6 , 45 . 3 ) and those from families with low household monthly income ( 38 . 7%; 95% CI = 32 . 9 , 44 . 9 ) when compared with the children of fathers with at least 6 years of formal education ( 25 . 8%; 95% CI = 20 . 4 , 32 . 2 ) and those from families with household monthly income of ≥YER20 , 000 ( 19 . 7%; 95% CI = 14 . 1 , 26 . 9 ) . Moreover , it was found that the presence of other family members infected with schistosomiasis showed significant association with higher prevalence of schistosomiasis ( P<0 . 001 ) . Moreover , children who lived in houses without toilets ( 39 . 4%; 95% CI = 32 . 6 , 46 . 7 ) , those who use unsafe sources for drinking water ( 36 . 2%; 95% CI = 75 . 0 , 85 . 1 ) , those who lived in houses where water used for household purposes was fetched from unsafe sources ( e . g . , stream , rain , well , water collection tank , trough , etc ) ( 36 . 2%; 95% CI = 30 . 8 , 42 . 0 ) had higher prevalence of schistosomiasis when compared to those having toilets in their houses ( 25 . 5%; 95% CI = 20 . 1 , 31 . 6 ) , those who use piped water ( 21 . 5%; 95% CI = 57 . 2 , 69 . 1 ) and those living in houses with safe sources of household water ( 21 . 5%; 95% CI = 15 . 1 , 29 . 7 ) . Furthermore , the results showed that the prevalence of infection was significantly higher among children who lived nearby stream and/or spring ( 41 . 6%; 95% CI = 33 . 3 , 50 . 4 ) and nearby pool and/or pond ( 51 . 3%; 95% CI = 40 . 5 , 61 . 9 ) when compared to their counterparts . Interestingly , there was strong negative associations between the presence of water pump and visits by foreigners to the area as the children who lived in the presence of nearby water pump ( 23 . 1%; 95% CI = 16 . 8 , 31 . 0 ) and those from villages where foreigners were seen playing/swimming in the open water sources ( 23 . 2%; 95% CI = 17 . 0 , 30 . 9 ) had significantly lower prevalence of infection compared to those who lived in houses not close to water pump ( 36 . 1%; 95% CI = 30 . 6 , 42 . 0 ) and those from villages where no foreigners were seen playing/swimming in the open water sources ( 36 . 4%; 95% CI = 30 . 8 , 42 . 5 ) . Five factors associated significantly with schistosomiasis were retained by multiple logistic regression model analysis ( Table 4 ) . The presence of other family member infected with schistosomiasis increased the children's odds for the disease by 4 . 1 times ( 95% CI = 2 . 40 , 6 . 85 ) . Similarly , children who used unsafe sources for drinking water had significantly higher odds of having schistosomiasis when compared to those living in houses supplied with piped water ( OR = 2 . 5; 95% CI = 1 . 36 , 4 . 41 ) . Moreover , children from families with low household monthly income ( <YER20 , 000 ) had significantly higher odds of schistosomiasis when compared with those from families with higher household monthly income ( OR = 2 . 3; 95% CI = 1 . 33 , 3 . 83 ) . Furthermore , significantly higher odds of having schistosomiasis were identified among children who lived nearby stream/spring ( OR = 2 . 2; 95% CI = 1 . 24 , 3 . 63 ) and those who lived nearby pool/pond ( OR = 2 . 5; 95% CI = 1 . 39 , 4 . 43 ) when compared to their counterparts .
This study reveals an alarmingly high prevalence of schistosomiasis among rural children in Yemen and this supports an urgent need to re-evaluate the current control measures and implement an integrated , targeted and effective schistosomiasis control measures . Regional control programmes are essential to prevent the dissemination of the infection to new areas at neighbouring countries . Screening of other family members and treating the infected individuals should be adopted by the public health authorities in combating this infection in these communities . Besides periodic drug distribution , health education regarding good personal hygiene and good sanitary practices , provision of clean and safe drinking water , introduction of proper sanitation are imperative among these communities in order to curtail the transmission and morbidity caused by schistosomiasis .
|
Schistosomiasis remains one of the most serious and prevalent diseases worldwide . Despite intensive control efforts by the government and international bodies , schistosomiasis is the second cause of death , after malaria , in Yemen , with an estimated 3 million cases . We screened 400 children in rural areas of five provinces in Yemen for the presence of schistosomiasis . Overall , 31 . 8% of the children were found to be positive for schistosomiasis; 23 . 8% were infected by Schistosoma haematobium , and 9 . 3% were infected by S . mansoni . The study identified the presence of other family members infected with schistosomiasis , low household monthly income , using unsafe water supply as a source for drinking water , living nearby stream/spring and/or pool/pond as the key factors significantly associated with schistosomiasis in these communities . Innovative and integrated control measures to control this infection should be implemented among this population . Periodic school-based and community-based drug distribution , health education , provision of clean and safe drinking water , introduction of proper sanitation will help to reduce the prevalence and morbidity of schistosomiasis among these communities .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"infectious",
"diseases",
"schistosomiasis",
"neglected",
"tropical",
"diseases",
"parasitic",
"diseases"
] |
2013
|
Prevalence and Associated Factors of Schistosomiasis among Children in Yemen: Implications for an Effective Control Programme
|
Despite comparable levels of virus replication , simian immunodeficiency viruses ( SIV ) infection is non-pathogenic in natural hosts , such as sooty mangabeys ( SM ) , whereas it is pathogenic in non-natural hosts , such as rhesus macaques ( RM ) . Comparative studies of pathogenic and non-pathogenic SIV infection can thus shed light on the role of specific factors in SIV pathogenesis . Here , we determine the impact of target-cell limitation , CD8+ T cells , and Natural Killer ( NK ) cells on virus replication in the early SIV infection . To this end , we fit previously published data of experimental SIV infections in SMs and RMs with mathematical models incorporating these factors and assess to what extent the inclusion of individual factors determines the quality of the fits . We find that for both rhesus macaques and sooty mangabeys , target-cell limitation alone cannot explain the control of early virus replication , whereas including CD8+ T cells into the models significantly improves the fits . By contrast , including NK cells does only significantly improve the fits in SMs . These findings have important implications for our understanding of SIV pathogenesis as they suggest that the level of early CD8+ T cell responses is not the key difference between pathogenic and non-pathogenic SIV infection .
The simian immunodeficiency virus ( SIV ) occurs as a natural infection in several Old-world monkey species , such as sooty mangabeys ( SM ) or African green monkeys [1] , [2] . In striking contrast to HIV infection of humans , SIV infection does not cause disease in natural hosts . The levels of virus replication , however , are similarly high in natural hosts and non-natural hosts such as rhesus macaques ( RM ) , in which SIV causes AIDS-like symptoms . Comparative studies of SIV infection in natural and non-natural hosts provide the opportunity to investigate the interaction between the virus and the host immune system in pathogenic and non-pathogenic infection . Such a comparison might shed light on the mechanisms of disease progression in pathogenic SIV and by extrapolation on HIV . Although natural and non-natural hosts allow similar levels of virus replication , there are interesting immunological differences: SMs do not exhibit the increased CD4+ T cell turnover and the generalized immune activation that is characteristic for the SIV infection of RMs or HIV-infection in humans [3] , [4] . Thus , virus load alone cannot be the key to understanding pathogenesis . Silvestri and Feinberg [5] interpreted these findings in favor of the hypothesis that HIV disease progression is a result of generalized immune activation rather than of the destruction of CD4+ T cells by the virus alone . This view of HIV pathogenesis is a derivative of the immuno-pathological hypothesis [6] . Because primary HIV infection is a period critical for the future immune responses' capability of controlling the infection [7] , [8] , the potential differences between pathogenic and non-pathogenic SIV infection are likely to manifest themselves early in infection . In both RMs and SMs , the early SIV infection is divided into three phases . The first phase is characterized by a sharp increase of virus load soon after infection . The second phase describes the decline of virus load that follows the initial peak viremia . The third phase finally describes the largely stable equilibrium virus load that eventually establishes after the decline . This stable virus load is also referred to as the viral set point . The characteristic pattern of virus load in primary SIV infection can be explained either through the delayed action of cellular immunity [9] , [10] or through target cell limitation [11] or both . Note that in this context the term target-cell limitation refers to the hypothesis that the level of target cells on its own can explain the early virus-load dynamics [11] . Regoes et al . [9] investigated these hypotheses by fitting mathematical models to viral loads of SIVmac239-infected RMs that exhibited either normal or experimentally impaired cellular immunity as a result of co-stimulatory blockade . This analysis showed that target-cell limitation can explain the virus-load dynamics in the animals with impaired cellular immunity but not in those with a normal immune response . In the latter case , the models could explain the virus-loads only if cellular immunity is also taken into account . These results imply that target-cell limitation alone cannot explain the level of virus replication during primary SIVmac239 infection of RMs and thus suggest a role for cellular immunity in determining the post-peak decline of viremia . In this article , we use the method of Regoes et al . [9] to analyze the early virus dynamics in non-pathogenic SIV infection of sooty mangabeys ( SM ) . In particular , we sought to determine the roles that target-cell limitation , CD8+ T cell responses and NK cells play in primary infection of SMs , and to compare the impact of these factors with that in SIV-infected RMs . To this end we fit the measurements of virus load with population-dynamic models that differ as to whether they take factors such as cellular immunity or NK cells into account . Comparing the goodness of fit of these models , we can then evaluate the role of these factors in the primary infection of pathogenic and non-pathogenic SIV .
The analysis of the RM data reconfirms the results of Regoes et al . [9] in an extended dataset . In particular , we find that target-cell limitation alone cannot explain the virus dynamics . For all animals except one ( animal RPB8 ) , the best fit of the target-cell model predicts a steadily increasing virus load ( black lines in Figure 3 ) , i . e . the fit fails to explain the characteristic peak and the subsequent post-peak decline exhibited by the data . Moreover , the quality of the fit is poor even for the animal for which the target-cell model can predict a viral load decrease . Adding specific cellular immunity to the target-cell model does significantly improve the fit for RMs ( F-test , p = 2 . 8×10−18 ) . Importantly , the CD8+ T cell model can explain the characteristic post-peak decline of the viral load ( green lines in Figure 3 ) . The results of our analysis of the data from SIV infection of SMs are strikingly similar to those obtained for the rhesus macaques: The target-cell model fails to explain the virus dynamics for all eight animals ( Figure 3 ) , whereas the CD8+ T cell model provides a significantly better fit ( F-test , p = 1 . 3×10−11 ) , which can reproduce the qualitative patterns of the virus dynamics . The only exception is the animal FSS , for which both the target-cell and the CD8+ T cell model produce poor fits . The poor quality of these fits might be due to the fact that this animal exhibits a comparatively early increase of target-cell number and a comparatively late increase of CD8+ T-cell number ( see Figure 1 ) . The similarity of the results in SMs and RMs suggests that the relative importance of specific cellular immunity and target-cell limitation during early infection is comparable in pathogenic and non-pathogenic SIV hosts . In both cases , the temporal dependence of the viral load can only be explained if CD8+ T cells are taken into account . Table 1 shows the best-fit estimates and the confidence intervals for the parameters of the CD8+ T cell model . The parameters r and k quantify the per-cell impact of target-cells and CD8+ T-cells on the viral replication rate ( see equation 2 ) . Both parameters are on average higher for sooty mangabeys: r roughly by a factor 6 and k by a factor 3 . Furthermore , the intrinsic death rates of infected cells , a , were estimated to be 0 for most animals . This suggests that , for both SMs and RMs , most deaths of infected cells are caused by cellular immunity ( see [9] ) . The NK cell model and the CD8+ T cell & NK model are obtained from the target-cell and the CD8+ T cell model by adding NK cell number as an explanatory variable . We consider the fits of these extended models for two reasons: First , to test whether the above results are robust against adding NK cells to the model and , second , to investigate the role of an important effector mechanism of the innate immune system during primary SIV infection . In total , four types of statistical comparisons were performed ( see Figure 2 ) : Comparison i ) between the target-cell model and the CD8+ T cell model is the one discussed above . Comparison ii ) between the target-cell model and the NK model evaluates whether adding NK cells to target-cell limitation improves significantly the quality of fit . Comparison iii ) between the NK model and the CD8+ T cell-NK model evaluates whether taking cellular immunity into account improves the fit of the NK model . Finally , comparison iv ) assesses whether NK-cell number does significantly improve the fit of the CD8+ T cell model . NK cell counts were available for 8 SMs ( FWo , FYl , FWn , FFS , FRS , FSS , FUV , FWV ) and 4 RMs ( RPB8 , RSO8 , RYE8 , RZS8 ) . If the number of all NK cells is used as a proxy of NK cell activity , extending the target cell based model by NK cells ( comparison ii ) does improve the model fits significantly only for SM but not for RM ( F-test , p = 0 . 016 and p = 0 . 24 for SM and RM , respectively ) . Extending the CD8+ T cell model by NK cells failed for both species to improve the model fits significantly ( F-test , p = 0 . 98 and p = 0 . 33 for SM and RM , respectively ) . In contrast , extending the NK model by CD8+ T cells improves the fit significantly ( F-test , p = 2 . 4×10−5 and p = 5 . 7×10−4 for RM and SM , respectively ) . If the number of proliferating NK cells is used as a proxy of NK cell activity , including NK cells again significantly improves the target-cell based model only for SM ( F-test , p = 1 . 3×10−5 and p = 0 . 33 for SM and RM , respectively ) . In addition , including NK cell activity via this proxy also improves the CD8+ T cell model for SM ( F-test , p = 0 . 00013 and p = 0 . 97 for SM and RM , respectively ) . These results suggest that NK cells play a role in the early infection of SM but not of RM .
The role of cellular immunity in early SIV/HIV infection has been a debated topic since the suggestion of Phillips [11] that early virus replication might be controlled by target-cell limitation . Several lines of evidence suggest however that cellular immunity is an important force for the control of early SIV replication . First , the post-peak decline of virus load coincides temporally with the rise of CTLs [12] ( although this is also consistent with the alternative explanation of [11] ) . Second , [10] have shown that the post-peak decline of virus-load is significantly weakened if CD8+ T-cells are depleted . Third , the ubiquitous selection for mutants that escape CTL response [13] also suggests an important role of cellular immunity . Fourth , it has been shown that the patients' ability to control HIV depends strongly on the alleles at the HLA and KIR loci [14] , which control the action of CD8 T cells and NK cells , respectively . More recently , some of the authors of this paper [9] have shown that mathematical models can explain the early virus dynamics if they take both target-cells and CD8+ T-cells into account , but not if they take only target cells into account . Our study extends this previous work by considering the impact of NK cells , important effectors of innate immunity . In addition to the extended analysis of the early viral dynamics in pathogenic SIV infection , we here compare our results to non-pathogenic SIV infection in sooty mangabeys ( SMs ) . This comparison has important implications for our understanding of pathogenesis . Our analysis confirms the earlier finding of [9] that target-cell limitation alone cannot explain the virus dynamics in RMs . We find that , in SIV-infected sooty mangabeys , target-cell limitation is equally unable to explain the viral load dynamics during early infection . In both species , our model can only explain the virus dynamics if it takes cellular immunity into account . This suggests that specific cellular immunity plays an important role in determining the dynamics of virus replication during early infection in both species . We , however , also found that a model , which assumes a constant viral replication rate , independent of target cells , was unable to fit the virus-load data of all animals consistently ( results not shown ) . This implies that , although target cells alone cannot explain the virus-load dynamics , in particular the peak and the post-peak decline , temporal variation of target cells is nevertheless important . Overall , our results indicate that the relative impact of target-cell limitation and specific cellular immunity is similar in RMs and SMs . These results give rise to testable predictions . If , for example , one would selectively deplete NK cells during primary infection , the pattern of virus load should be affected in SM , but not in RMs . In contrast , selective depletion of CD8+ T cells is predicted to lead to a loss of control of virus replication in both species . Of note , all depletion experiments performed using an anti-CD8 antibody depleted CD8+ T cells as well as NK cells because both cell types express CD8 [10] . In RMs , treatment with a costimulatory inhibitor , which prevented the development of SIV-specific cellular and humoral immunity and reduced target cell levels , gave rise to target cell limited virus replication [9] . The similarity between the factors governing virus replication predicts that an analogous treatment of SMs would also lead to target cell limitation . Our conclusions about the role of cellular immunity and target-cell limitations are based on several assumptions . First , the virus loads and the immune-cell densities were measured in the blood , which is not the main compartment of SIV replication and lymphocytes . Our analysis , therefore , relies on the assumption that the measurements in the blood reflect the situation in the whole body . In this context , it has been suggested that target-cell depletion in the gut might play an important role in the early SIV infection [15] , [16] , [17] , [18] , [19] , [20] , [21] , [22] . However , a recent study has shown that in SIV infections of both SM and RM , the target-cell depletion in the gut occurs too early to explain the peak in virus-load [23] . Second , our models consider only the primary phase of SIV infection . Therefore , our conclusion that cellular immunity does not differ in pathogenic and non-pathogenic SIV , does only apply to this phase . It might thus be that cellular immunity at later phases plays a very different role in RMs and SMs , as suggested by numerous comparative studies [2] , [3] , [24] , [25] . As discussed in Regoes et al . [9] , it is difficult to extend the approach used here to later phases of infection , because immune-escape and antibody responses would require considerably more complicated models . Last , we cannot exclude that different cell compartments or cell types play the role of target cells in the SIV infections of sooty mangabeys and rhesus macaques . Indeed , our model fits result in larger replication rate constants , r , for SM than for RM , which either suggests a better target cell utilization in SM , or is an indication that Ki67+ CD4+ T cells do not play the same roles in SM and RM . Such an effect could systematically bias our analysis if our proxy ( i . e . proliferating CD4+ cells ) would be representative for target cells in one species but not in the other . Finally , the p values of the model comparisons rely on the assumptions of normality and independence , which might be violated in our data . Especially , autocorrelation in the virus-load and cell-numbers , might potentially lead to an overestimate of the degrees of freedom and thereby to an underestimate of those p-values . However , it should be noted that independently of the statistical evaluation , the least-squares approach is a simple and intuitive method to fit dynamical models to data , and these fits clearly ( Figure 3 ) show that for all animals except one RM ( RPB8 ) , the best fit of the target-cell-limitation model fails to predict a post-peak decrease in virus-load . This suggests that our results regarding the CTLs are robust against these ( in principle valid ) statistical concerns . By contrast , adding NK cells to the model leads to smaller improvements of the fits and therefore these findings may be more vulnerable to potential autocorrelations . One important caveat mentioned in the previous section is the uncertainty as to whether the measured cell populations ( e . g . Ki67+ CD4+ T-Cells , Ki67+ CD8+ T-Cells , NK cells ) can be identified with populations performing a specific function ( target cells , cytotoxic T cells , cytotoxic NK cells ) . This potential problem is substantially alleviated by the way these measurements are integrated into our model . Specifically , the quality of fit as measured by the residual sum of squares , is invariant with respect to a linear transformation of the variables . I . e . if we measure the cell population x but the active population is x′ = a x-b we will obtain the same quality of fit regardless of whether we incorporate x or x′ into our model . Therefore it does not matter whether only a fraction of the measured cells is active or whether a constant number of the measured cells is inactive . For practical reasons , however , it is important that the fraction of the active cells is not too small relative to the inactive cells , because then the noise in the latter is likely to overwhelm the signal in the former . This reasoning implies that the comparison of the quality of fit of the different models ( Figure 2 ) is much more robust than the parameter estimates ( Table 1 ) : In principle , the first type of analysis ( model comparison ) still works , even if the linear transformation ( relating measured cell populations to the cell-populations performing a specific function ) is different for each animal . By contrast , the second type of analysis ( parameter estimation ) requires that this transformation is similar in the animals compared . For these reasons , we conclude that not too much weight should be given to the parameter estimates , as they rely much stronger on a good match between measured cell populations and the populations actually performing a certain function , while we can assert that the model comparison is robust . The fundamental robustness of the method also explains why [9] found qualitatively similar results with Ki67+ CD8+ T-cells and tetramer positive T-cells as markers for SIV-specific cellular immunity . As SIV infection is pathogenic in rhesus macaques but non-pathogenic in sooty mangabeys , our results can be interpreted in the context of current theories of SIV pathogenesis , in particular with respect to reasons underlying the absence of disease progression in SIV-infected SMs . While an initial study suggested that acute SIV infection of SMs is characterized by limited to absent T cell activation [25] , a number of more recent studies that included a more comprehensive sample collection have shown very clearly that SMs exhibit substantial T cell activation during acute SIV infection [24] , [26] , [27] , [28] . However , in marked contrast with SIV-infected RMs , sooty mangabeys are able to rapidly and dramatically reduce the level of T cell activation during the early chronic infection ( i . e . , starting at day 30 post inoculation ) [24] , [26] , [27] , [28] . Although our model comparison did not directly test differences in the antigenicity of SIV between SM and RM , our results are more consistent with the latter observations and suggest that the divergent outcome of SIV infection in RMs and SMs is not caused by differences in CD8+ T-cell response during the early stages of infection .
All the experiments on non-human primates from which these data are sampled have been approved by the Institutional Animal Care and Use Committee ( IACUC ) . All these experiments have been described in previous publications . The data analyzed in this article were generated in experimental infections of SMs infected with the viral strain SIVsmm and of rhesus macaques infected with the strains SIVmac ( animals rbm , rvy , roz , ryt ) or SIVsmm ( animals RPB8 , RSO8 , RYE8 , RZS8 , RFT8 ) . A detailed description of the experiments can be found in Garber et al . [29] , Gordon et al . [30] , and Mandl et al . [4] . For the sake of comparability , we consider the same time-window as Regoes et al . , i . e . a window ranging from day 0 ( start of infection ) to day 30 . In one of the rhesus macaques ( animal RFT8 ) no SIV infection could be established . This animal was therefore excluded from further analysis . In total , we consider 8 SMs ( all infected with SIVsmm ) and 8 RMs ( 4 infected with SIVmac239 and 4 infected with SIVsmm ) . Figure 1 shows the measurements relevant for this study: the virus-load , the density of proliferating CD4+ T-cells , the density of proliferating CD8+ T-cells , and the density of NK cells . The fraction of proliferating CD4+ and CD8+ T cells was assessed by staining for the nuclear antigen Ki67 , which is expressed by cycling cells . We consider the density of proliferating CD4+ T-cells as representative for the size of the target cell population and the density of proliferating CD8+ T-cells as a surrogate measure for the SIV-specific cellular immunity . We will therefore refer to the density of proliferating CD4+ T-cells and of proliferating CD8+ T-cells also as “target cells” and “cellular immunity” , according to the functional role we assume these populations to play . Data on the density of NK cells were only available for all sooty mangabeys and for 4 out of 8 rhesus macaques ( RPB8 , RSO8 , RYE8 , RZS8 ) . The data were analyzed by using population dynamic models , which describe the virus dynamics as a function of target cells , CD8+ T-cells , and NK cells . The models are fitted to the virus load . Hereby , the measurements of target cells , CD8+ T-cells , and NK cells were used as explanatory variables . Importantly , the model does not aim to explain the measurements of these cell populations , but considers them only as factors that might explain viral replication . A detailed account of this approach can be found in [9] . In order to assess the role of target-cell limitation and cellular immunity in early SIV infection , we compared the fits of two nested models , which describe the virus dynamics by taking into account either target cells only or target cells and specific cellular immunity . These models are referred to as the target-cell model and the CD8+ T cell model , respectively . Mathematically , these models read ( 1 ) ( 2 ) where v is the virus load and T ( t ) and E ( t ) denote the number of proliferating CD4+ T-cells and of proliferating CD8+ T-cells , respectively . The parameters r , a , and k are chosen for each animal such that T ( t ) and E ( t ) give the best possible prediction of v ( see below ) . In order to test the impact of the non-adaptive immune system on our results , we extended the above models by adding NK cell number as an explaining factor . We incorporate the impact of NK cells by using two different proxies: either the total density of NK cells ( characterized as CD3− CD20− CD16+ cells ) or only the density of activated NK cells ( i . e . Ki67+ NK cells ) . The second approach is identical to the one used of CD8+ and CD4+ T-cells . The first approach can be justified by the fact that , in contrast to CD8+ T-cells , NK cells do not recognize specific antigens . Thus , every NK cell can potentially inhibit virus replication by either killing infected cells or by IFN-gamma production [31] , and their effect is most likely proportional to their level . We would like to emphasize that we do not assume that every NK cell is cytotoxic , or that every NK cell has anti-viral activity . We only assume that the impact of NK cells is proportional to their abundance ( see discussion ) . The extensions of the target-cell model and the CD8+ T cell model are referred to as the NK-model and the CD8+ T cell & NK model . Mathematically these models read ( 3 ) ( 4 ) where N ( t ) denotes the number of NK cells and the parameter n is chosen according to the best fit criterion . We illustrate the fitting-procedure for the CD8+ T cell & NK model: First the differential equation of the model ( 4 ) can be integrated to ( 5 ) If t0…tk , denote the time points for which measurements of v are available then the parameters r , k , a and n are chosen such that the residual sum of squares ( 6 ) is minimized . The integrals in the sum are computed from the measurements of the cell numbers T , E , and N by first interpolating these measurements by a piecewise linear function , resulting in the functions T ( t ) , E ( t ) , and N ( t ) , and then integrating these interpolating functions . As expression ( 5 ) is linear in the parameters r , k , a and n , the best fit can be found using a standard linear-model solver such as the lm ( ) routine of the R language [32] . Biologically , the parameters r , k , a and n must be larger than or equal to 0 . If the best fit of ( 5 ) does not fulfill these conditions , one or several of the parameters r , k , a and n is set to 0 and the fitting procedure is repeated with these reduced functions . From all the “reduced fits” , that one is chosen , which yields the minimal sum of squares while fulfilling the biological conditions . The fits for the target-cell , the CD8+ T cell , and the NK model are obtained in a similar way as for the CTL-NK model . In formula ( 5 ) the parameters that do not occur in the differential equation of the model ( i . e . equation 1 , 2 , or 3 for the target-cell , CD8+ T cell , and NK model respectively ) are set to 0 and the remaining parameters are chosen such that the corresponding sum of squares ( SSQtarget-cell , SSQCD8+ T cell , and SSQNK ) is minimized . We can statistically compare two of the above models , for instance model 1 and model 2 , if they are nested , i . e . if model 1 results from model 2 by setting one of the parameters to 0 . In such cases , model 2 will always provide a better fit than model 1 , because model 1 is included as a special case in model 2 . Whether this improvement in the quality of fit is significant can then be assessed by performing an F-test . The corresponding test statistic is Here SSQi denotes the residual sum of squares of the model i , and dfi refers to the corresponding degrees of freedom . The p value that corresponds to the value of F is then calculated from the Fisher Distribution with degrees of freedom df1-df2 and df2 , i . e F ( df1-df2 , df2 ) . This comparison between models can be made either for each animal individually , or , as we mostly do in this article , for all animals of a species taken together . In the latter case , the residual sum of squares obtained by fitting the models to each animal and their corresponding degrees of freedom have to be summed to perform the F-test . Figure 2 illustrates the statistical comparisons that are made in this article . The most important of these comparisons is the one between the target-cell model and the CD8+ T cell model ( comparison i in Figure 2 ) , which assesses the relative importance of target cells and specific cellular immunity for explaining the virus-load dynamics . If NK-cell counts are available , one can ask in addition whether taking NK cells into account improves the fit of the target-cell model ( comparison ii ) , whether taking specific cellular immunity into account improves the fit of the NK model ( comparison iii ) , and whether taking NK cells into account improves the fit of the CD8+ T cell model ( comparison iv ) .
|
Simian immunodeficiency viruses ( SIV ) are typically non-pathogenic in their natural hosts . However , if the same virus infects a non-natural host it often leads to AIDS-like symptoms . Therefore , comparing SIV infections in these two types of host might help explain the pathogenesis of SIV in non-natural hosts and thereby also that of HIV . We combined mathematical modeling with data on the levels of virus and immune cells early in infection , and compared both non-pathogenic SIV infections of sooty mangabeys and pathogenic SIV infection of rhesus macaques with respect to how the virus grows in them and to what extent it is controlled by the immune system . We found that the impact of the immune system on early virus replication is remarkably similar in both species . In particular , for both species virus replication can only be explained by the effect of CD8+ T cells . These findings have important implications for our understanding of SIV pathogenesis as they suggest that the impact of the early immune responses is not the key difference between pathogenic and non-pathogenic SIV infection .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"infectious",
"diseases/hiv",
"infection",
"and",
"aids",
"immunology/immune",
"response",
"infectious",
"diseases/viral",
"infections"
] |
2010
|
Similar Impact of CD8+ T Cell Responses on Early Virus Dynamics during SIV Infections of Rhesus Macaques and Sooty Mangabeys
|
Epstein-Barr virus ( EBV ) is a human herpesvirus that causes acute infectious mononucleosis and is associated with cancer and autoimmune disease . While many studies have been performed examining acute disease in adults following primary infection , little is known about the virological and immunological events during EBV’s lengthy 6 week incubation period owing to the challenge of collecting samples from this stage of infection . We conducted a prospective study in college students with special emphasis on frequent screening to capture blood and oral wash samples during the incubation period . Here we describe the viral dissemination and immune response in the 6 weeks prior to onset of acute infectious mononucleosis symptoms . While virus is presumed to be present in the oral cavity from time of transmission , we did not detect viral genomes in the oral wash until one week before symptom onset , at which time viral genomes were present in high copy numbers , suggesting loss of initial viral replication control . In contrast , using a sensitive nested PCR method , we detected viral genomes at low levels in blood about 3 weeks before symptoms . However , high levels of EBV in the blood were only observed close to symptom onset–coincident with or just after increased viral detection in the oral cavity . These data imply that B cells are the major reservoir of virus in the oral cavity prior to infectious mononucleosis . The early presence of viral genomes in the blood , even at low levels , correlated with a striking decrease in the number of circulating plasmacytoid dendritic cells well before symptom onset , which remained depressed throughout convalescence . On the other hand , natural killer cells expanded only after symptom onset . Likewise , CD4+ Foxp3+ regulatory T cells decreased two fold , but only after symptom onset . We observed no substantial virus specific CD8 T cell expansion during the incubation period , although polyclonal CD8 activation was detected in concert with viral genomes increasing in the blood and oral cavity , possibly due to a systemic type I interferon response . This study provides the first description of events during the incubation period of natural EBV infection in humans and definitive data upon which to formulate theories of viral control and disease pathogenesis .
Epstein-Barr virus ( EBV ) is a ubiquitous human herpesvirus . As with all herpesviruses , EBV causes lifelong infection in its host . Infection is associated with autoimmune diseases [1 , 2] and is known to cause several types of cancer , representing 1% of the worldwide cancer burden [3] . Primary infection in children is either asymptomatic or causes mild symptoms not readily associated with EBV . In adolescents and young adults , however , EBV is most commonly recognized as the etiologic agent of acute infectious mononucleosis ( AIM ) [4] . The virus is chiefly transmitted from person to person in oral secretions , although it can be acquired from blood transfusions or from receipt of allogeneic donor cells or tissue . There is currently no vaccine or effective treatment for AIM or other EBV related diseases . Because EBV infection is limited to primates , there are no small animal models of infection except humanized mice [5] . Neither humanized mice nor mice infected with the related gamma herpesvirus MHV68 exhibit true AIM . Therefore a detailed and accurate understanding of primary infection in humans is critical for developing therapeutic tools to treat EBV related diseases . Abundant data are available on AIM in humans , especially the most severe cases , as subjects are typically seen in clinic . Indeed , the antibody and cell mediated adaptive immune response to the virus , and how it wanes and changes after AIM presents is well established [6] . However a particular knowledge gap exists regarding the events that occur between transmission and symptom onset—the incubation period—which is unusually long , about six weeks [7 , 8] . In comparison , most other acute viral infections have incubation periods ranging from less than a day to a week [9] . Thus in particular , we lack knowledge about initial infection events and the innate immune response to EBV in humans; although these are presumed to be critical as EBV has a multitude of innate immune evasion mechanisms [10 , 11] . From in vitro studies , we know that EBV efficiently infects B cells through binding of viral gp350 and gp42 proteins to the B cell surface molecules CD21 and HLA Class II , respectively [12] . However , EBV can also infect oral epithelial cells , albeit much less efficiently [13] . It is unknown which cells are initially infected in the oral cavity during natural infection . One possibility is that the virus infects and replicates in oral epithelial cells early during primary infection; but AIM does not occur until B cells are later infected in the tonsils , and virus disseminates to the blood . The virally encoded LMP1 and LMP2 proteins are known to drive infected B cells to differentiate by acting as functional homologues of CD40 and the BCR respectively , and this also triggers migration from tonsils to the blood [14] . Another model proposes that both cell types are infected early in the oral cavity , and that cycles of infection and reactivation must occur during the incubation period to ultimately produce high levels of infected B cells in circulation , which drives AIM [15] . Alternatively , a third model is that B cells may be initially infected in tonsils where they vertically transmit virus at low levels [16] . Infection would be limited to B cells in the nasopharyngeal secondary lymphoid tissue until some stochastic event caused reactivation of the virus , spreading virus to epithelial cells and resulting in an acute increase in viral load , occurring weeks after initial infection . The dynamics of tissue tropism of EBV are interesting in this context , as the virus produced by epithelial cells is particularly efficient at infecting B cells , and vice versa [13] . Each of these three models make distinct predictions about the relative levels of virus in the oral cavity versus blood during the incubation period , and would suggest different strategies to combat infection therapeutically , as antiviral compounds only target actively replicating virus . Although it is critical to understand the viral and immunological events that occur during the incubation period , it is challenging to obtain both samples and comprehensive clinical data during this period . To address this need , we enrolled undergraduate volunteers who were naïve to EBV and monitored them routinely for natural infection , capturing timepoints within the incubation period by chance . Through frequent sampling , we were able to obtain 48 incubational samples from 40 young adult study participants . We sought to detect and quantify virus in the blood and oral cavity , with particular attention to when virus disseminates from nasopharyngeal tissue to the periphery . In addition , innate and adaptive immune responses during this period were examined , with particular emphasis on natural killer ( NK ) cells , plasmacytoid dendrtitic cells ( pDC ) , CD8 T cells , and Foxp3+ T regulatory ( Treg ) cells . Surprisingly , during the incubation period , viral genomes were detected at low levels in peripheral blood prior to detection in the oral cavity . A dramatic reduction in blood pDC numbers and a type I interferon response were observed roughly coincident with viral increases in the blood . In contrast , the dramatic increase in CD8 T cell numbers characteristic of AIM , the distinctive type II interferon/cell cycle gene expression signature of AIM , and changes in NK cell phenotype and Treg cell numbers previously reported to occur during AIM were not observed until symptom onset . We discuss possible mechanisms to explain these changes and implications for the treatment of EBV related diseases .
We previously described a prospective study of primary EBV infection in 66 undergraduates at the University of Minnesota [17] , 59 of whom were symptomatic . As an extension of this study , we enrolled a new prospective cohort with the specific aim of more frequent sampling in order to serendipitously capture more samples from the incubation period of primary EBV infection . From both cohorts combined , a total of 48 blood and oral wash samples from 40 subjects were obtained during the historically defined incubation period ( 42 days ) . These samples were the focus of this study . Since the exact date of infection with EBV is undefinable in a study of natural infection , we designated the date of symptom onset as day “zero . ” EBV is transmitted through salivary exchange in young adults and infection is established in the oral cavity , both in squamous epithelial tissue and lymphocytes of the Waldeyer’s ring [18] . Very little is known , however , about when and how EBV egresses from tonsillar tissue into the peripheral blood . In order to examine dissemination during the 6 week incubation period more closely we tested for the presence of viral genomes by both quantitative PCR ( qPCR ) and by a highly sensitive nested PCR . The nested PCR assay gave a much more sensitive but non-quantitative readout of viral presence , whereas the limit of detection for the quantitative PCR assay was 200 copies of EBV per milliliter of whole blood or 40 copies of viral genome per millilter of oral wash . Surprisingly , viral genomes were not detected in the oral cavity by either method until approximately a week prior to symptom onset , at which point large amounts of viral genomes were detected ( Fig 1A ) . These data are expressed at a per subject level as “time to first response” in Fig 1C . Viral genomes were detected in both cells from the oral wash as well as supernatant , suggesting that virus persists in the oral cavity at very low levels for the first 4–5 weeks after transmission , and then exhibits an explosive pattern of replication . In contrast , viral genomes were detected in peripheral blood as early as 22 days prior to symptom onset , but was only detected via the more sensitive nested PCR assay at this early stage ( Fig 1B and 1D ) . In fact , 10 subjects showed viral genome detection in the blood before the oral cavity when consecutive timepoints were evaluated ( Fig 1E ) . It should be noted that we were able to obtain substantially more DNA from blood cells than from oral wash cells , which could explain why low levels of viral genomes were not detected early in the oral cavity . Nonetheless , we detected dramatically higher viral loads in the oral wash at the time of symptom onset ( Fig 1A ) , suggesting efficient viral detection in oral samples . Virus detection in the blood using the less sensitive qPCR assay was delayed by at least a week ( Fig 1F ) , and viral genomes >200 copies/ml were not found in any individuals until on or after timepoints where high viral loads were detected in the oral cavity ( Fig 1E ) . These data provide the first description of EBV viral dynamics during the incubation period of natural infection in humans , and suggest a scenario where viral replication is self-limiting in the oral cavity for many weeks . Dissemination to the blood occurs during this “quiet period” . Closer to the time of symptom onset , virus replicates rapidly in the oral cavity , and subsequently high viral loads are detected in the blood . Previous work from our group revealed that distinct gene expression signatures were present in peripheral blood mononuclear cells early versus late after primary EBV infection [19] . We sought to expand this dataset by examining the additional incubational samples obtained from our most recent cohort . Samples were evaluated by PCR SuperArray consisting of 43 genes representative of gene changes initially observed by microarray [19] . Heirarchical clustering revealed three distinct patterns from incubational samples ( Fig 2 ) . Subjects exhibited either no change , a type I interferon ( IFN ) signature , or a type II IFN/cell cycle signature . These signatures clustered temporally , segregating into these approximate time frames: ( i ) no change seen from -42 to -7 days prior symptom onset , ( ii ) a type 1 IFN signature from -15 to -3 days , and ( iii ) the distinctive type II IFN/cell cycle signature associated with AIM within days of symptom onset . Notably , the type I IFN signature was present when viral genomes were detected only by nested PCR ( low viral loads ) in 3 out of 4 subjects , although 7 subjects showed no interferon response despite the presence of viral genomes in blood by nested PCR . By ELISA , we detected no substantial ( >25 pg/ml ) IFNa protein at any timepoints . Plasmacytoid dendritic cells ( pDC ) are major producers of type I IFN . Although a robust type I IFN response was observed during the incubation period in some study participants , the gene expression signature was relatively transient . Khanna’s group recently found pDC numbers to be reduced in acute IM patients [20] . Thus , we thus sought to examine pDC numbers during the incubation period . pDC were identified as BDCA-2+ CD123+ cells amongst non-lymphoid cells ( CD3 , CD20 , CD56 and CD14 negative ) and were HLA-DR+ and CD11c- . As an example , flow plots are shown for 6 timepoints from subject 5524 ( Fig 3A and 3B ) . Analysis of pDC from all subjects during the incubation period revealed a remarkable decline during the ten-day period leading up to symptom onset ( Fig 3C and 3E ) . A slight decline was observable before that , but was not statistically significant . In contrast , conventional myeloid derived DC ( cDC ) , identified as HLA-DR+ CD11c+ amongst non-lymphoid cells , were not significantly increased or decreased during either the incubation period or the early phase of AIM ( Fig 3D ) . The loss of pDC in circulation was strongly correlated with the presence of viral genomes in peripheral blood ( Fig 3F ) . Interestingly , the reduction in pDC was as profound at timepoints where only low levels of viral genomes were detected as when high levels of viral genomes were detected ( Fig 3F ) , and there was no significant correlation between the number of viral genomes present and the extent of pDC reduction . The importance of NK cells in EBV infection has become increasingly apparent in recent years [21] . Changes in NK cell phenotype during AIM were reported previously , and included a loss of CD56bright “immature” NK cells and a corresponding increase in CD56dim “mature” NK cells [22] . We also observed a reduction in the percentage and number of CD56bright NK cells in this cohort; however , unlike pDC changes , NK cell changes were not apparent until symptom onset ( Fig 4A ) . We further examined a specific CD56dim NKG2A+ KIR- cell subset reported to be expanded during AIM as a consequence of virus-induced proliferation [23] . We observed a similar expansion in our cohort , but it likewise was not detected until symptom onset , and remained elevated for at least 50 days ( Fig 4B ) . CD8 T cells provide vital immune control of EBV [6] , and although their expansion during AIM has been well documented , it is not known when they first become activated during primary infection . Using peptide:MHC I tetramers , we detected no EBV specific CD8 T cell expansion ( tetramer binding cells above . 05% of CD8 T cells ) until timepoints near the onset of symptoms ( Figs 5A and S1 ) . Similarly , upregulation of CD11a and downregulation of CD45RA on tetramer binding T cells , indicating antigen experience , were not seen until symptom onset . The expansion of EBV specific T cells was tightly concordant with total CD8 T cell expansion , as reflected by an increased CD8:CD4 ratio ( Fig 5A and 5C ) . Interestingly we detected upregulation of CD38 and granzyme B on total polyclonal CD8 T cells earlier , during the incubation period ( Figs 5A , 5B and S1 ) . These features of polyclonal activation correspond kinetically to when a type I IFN response was most strongly represented ( Fig 2 ) . Foxp3+ CD25+ Treg cells , are important for the maintenance of self-tolerance and dampening chronic inflammation . A reduction in the number of circulating CD25hi CD4+ T cells was previously reported in AIM patients [24] , but it was unknown when these changes began to manifest and how long they persisted through convalescence . Analysis of individual subjects over time in our study corroborated a decrease in Treg cells at the onset of AIM ( Fig 6A and 6B , shown for a representative subject ) . In all subjects , Treg cells were significantly decreased only during the first ten days of AIM ( Fig 6B ) . Numbers were depressed as well as frequency . The overall number of CD4+ T cells , in contrast , was unchanged ( Fig 6C ) . Although the fate of blood Treg cells during AIM is unknown ( e . g . whether they trafficked to tissues or died ) , previously reported histology of AIM tonsils would argue against local infiltration into tonsils , although the sample size of this study was very small [24] .
Our findings have important implications regarding how EBV infection progresses through natural routes in its native host . Despite an oral transmission mode , viral genomes were not detected in the oral cavity in appreciable quantities until subjects had presumably been infected five to six weeks . The lack of detectable viral genomes in oral wash argues against substantial lytic replication within squamous epithelial cells early during infection . Rather , it is consistent with the idea that B cells are a major cell type initially infected in the nasopharyngeum ( S2 Fig ) . EBV efficiently infects B cells , particularly when virus is derived from epithelial sources , which it likely would be during transmission , since virus produced by epithelial cells packages more gp42 into virions than virus produced by B cells [12] . Infected B cells are known to divide and differentiate , replicating the viral genome as an episome along with cellular division [18] . This “vertical” replication would be expected to expand viral load relatively slowly , compared to active viral replication in lytically infected cells . Starting approximately 1 week before symptom onset , viral genomes became detectable at high copy number in the oral wash . It is unclear what event precipitates this sharp increase , but it was not gradual like the decline in viral loads in the oral cavity during latency . It has been postulated that undefined signals may trigger viral reactivation in latently infected B cells [25] , which could then lead to high local production of virus and large-scale infection of epithelial cells . Interestingly , our data would also suggest that infected B cells begin to disseminate into circulation prior to events that precipitate large scale viral production in the oral cavity ( S2 Fig ) . Notably , in 10 subjects we detected low levels of viral genomes in peripheral blood at timepoints prior to detection in the oral cavity . Furthermore , in 7 subjects with low viral genomes present in the blood , there was no type I IFN response detected . This is consistent with the idea that virus is disseminated into circulation via latently infected memory B cells [26] where it goes undetected by the innate immune system . Indeed , infected B cells were shown to transition into a “latency 0” stage that closely resembles resting memory B cells , with altered trafficking patterns [27] . Another point that emerges clearly from these data , is that systemic innate and adaptive immune responses do not occur until viral loads rise relatively late in the incubation period , either in the oral cavity or the blood . The earliest responses detected were a type I interferon response ( Fig 2 ) and upregulation of CD38 on total CD8 T cells ( Fig 5 ) , which occurred during the 10 days prior to symptom onset . These two observations may be related , as it was previously shown that type I IFN can upregulate granzyme B in CD8 T cells , independent of activation through the antigen receptor [28] . Indeed , the proportion of CD8 T cells that upregulated CD38 and Granzyme B ( >80% in some individuals ) at these early time points is too high to be explained by T cells recognizing virus through their antigen receptor , as clonal expansion had not been detected at these time points . Of note is the fact the type I IFN response was relatively transient and not associated with symptoms in any of the study subjects . An adaptive immune response followed these early events , with expansion of virus specific CD8 T cells , and increased CD8:CD4 ratios rising in all subjects in the first 10 days following symptom onset . IgM responses to EBV viral capsid antigen were also detected in this time frame . As previously reported , IgG responses to VCA developed subsequent to IgM , and IgG responses to EBNA-1 were not maximal until after 3 months . Foxp3+ T regulatory cells were reduced during symptomatic IM as reported previously [24] . A similar reduction is observed in various infections in mice [29 , 30] , where reduction of effector cell IL-2 in the face of inflammatory cytokines was suggested to be the mechanism [31] . Too little Treg activity can result in immunopathology [31] , and we did observe an inverse correlation between Treg percentages during acute infection and disease severity ( Spearman r = 0 . 4871 , p = 0 . 0251 ) although whether this is causative remains to be explored in EBV . Curiously , blood NK responses were observed only after symptom onset and not earlier , although NK cells are thought to function early in infections . This result may not be entirely unexpected , as NK cells respond preferentially to lytically rather than latently infected cells [32] , and our results suggest that latently infected cells are introduced into circulation prior to lytically infected cells . NK cells are thought to play a protective role in AIM as evidenced by NK cell depletion in humanized mice infected with EBV , which resulted in higher levels of viral DNA in blood [32] . It is possible that NK cells in the tonsil play a critical role in humans , limiting viral spread amongst epithelial cells . Furthermore , NK cells have been hypothesized to play a role in the age dependence of symptomatic primary EBV infection . For example , newborns were reported to have more than twice as many circulating CD56dim NKG2A+ KIR- NK cells than adolescents [23] , which could explain why children experience less EBV associated morbidity in comparison with adolescents and young adults . However , it was recently reported that children with asymptomic primary EBV infection have blood viral loads as high as adults with AIM [33] , which is not consistent with a model of better NK control of EBV infected PBMC in infants . By closely observing the viral and immune dynamics during natural infection , we offer a new hypothesis on AIM pathogenesis , which proposes that explosive viral replication in the oral cavity creates a situation of exaggerated CD8 T cell response . It may be that children experience less AIM than adults despite ultimately achieving equally high levels of virus in the oral cavity and in blood , because infection in the oral cavity was not initially held in check . This allowed the adaptive immune response to develop by the time blood levels of virus increased . Indeed , memory CD8+ T cells specific for the virus , were observed in asymptomatic children concurrent with high viral loads [33] . Ironically , it may be heightened oral innate immune surveillance in adolescents and adults , compared with children , that puts them at risk for AIM . A final point of interest in our study was that circulating pDC percentages and numbers were significantly diminished during the viral incubation period . The pDC decrease began during the same period as a type I IFN response and polyclonal CD8 T cell activation were observed—the 10 days prior to symptom onset . Unlike the type I IFN response , which was transient , the pDC reduction was sustained for up to 50 days . The reduction was also observed at all timepoints ( except one ) that showed the presence of viral genomes , even low levels of viral genomes , and even when a type I IFN response was not present . From this we conclude that viral infection or viral products were responsible for the pDC reduction , but it was unlikely to be mediated by the host’s type I IFN response . This reduction could be related to the dynamics of pDC activation . Evidence in the literature suggests that pDC can mature or leave the circulation into tissues or secondary lymphoid organs during infections [34] . Upon activation , pDC enter a maturation program that can result in progression of pDC into antigen presenting cells [35]; however , we did not see corresponding increase in the MHC class II molecule HLA-DR . Alternatively , reduction of circulating pDC numbers may be related to the BamHI-A rightward frame 1 ( BARF1 ) protein secreted by EBV during lytic replication [36] . BARF1 enhances viral replication and persistence in part by binding to and inhibiting the signaling of colony stimulating factor ( M-CSF ) [37 , 38] an important factor for the survival and maintenance of pDC [39] . It would be interesting to determine if primates infected with a BARF-1 deficient form of lymphocryptovirus show pDC reductions or not . The functional consequences of pDC loss from the blood during primary infection remain to be explored . An important caveat to this work is that we have been limited to sampling peripheral blood and washings from the oropharynx . It is unknown whether or not we would be able to detect viral DNA or cellular responses to EBV ( like a type I interferon response ) if we were able to evaluate tonsillar tissue . The possibility that these cells may be sequestered in local tissue cannot be ruled out and remains to be investigated in future studies . In summary , we report several novel findings about the viral and immune dynamics during the lengthy incubation period of primary EBV infection . These include relatively early dissemination of virus into circulation in a form that does not elicit immune responses . A sharp increase in viral load subsequently occurs in the oral cavity and blood within 10 days of symptom onset . An early type I IFN response during this period is associated with a marked drop in blood pDC numbers and polyclonal CD8 T cell activation , without notable symptoms . Symptom onset coincides with a developing adaptive immune response and a strong type II interferon signature . Severity of illness correlates most strongly with increased CD8 T cell numbers , confirming the notion of AIM as an immunopathologic disease . The sharply increased viral loads that are presumed to drive an exuberant T cell response are already underway prior to symptom onset , providing a potential explanation for the lack of a clear-cut benefit from antiviral drugs in AIM [4] . We also speculate that pre-existing adaptive immunity to EBV would change the dynamics of infection in the oral cavity and thereby prevent IM in adolescents and adults .
Samples analyzed here were obtained from two studies: one with a large number of subjects and less frequent sampling ( Mono 5 ) [17] and another with a smaller number of subjects with more frequent sampling ( Class of 2016 ) . For the Mono 5 study , healthy undergraduate volunteers from the University of Minnesota were recruited in 2006 and 2007 . We screened 546 participants for IgG antibodies against EBV viral capsid antigen ( EBV VCA IgG ) and EBV nuclear antigen-1 ( EBNA-1 ) . Of the 202 eligible EBV-naïve subjects , 143 ( 71% ) were enrolled in the prospective study . Blood and oral washings were collected approximately every 4–8 weeks from enrolled participants during the academic year . Symptoms between visits were reported via an electronic monitoring journal . Subjects with symptoms consistent with acute primary EBV infection were asked for an additional visit which included a physical exam , laboratory-confirmation of primary EBV infection via heterophile , EBV-specific serology , and viral titer in the oral cavity or blood . Primary EBV infection was defined as a positive EBV antibody test and the presence of EBV DNA in the blood and/or oral cavity of a subject who was previously negative for both EBV antibodies and EBV DNA . All participants were monitored with follow-up visits . The Class of 2016 study was similar to the above except blood and/or oral washings were collected approximately every 2 weeks . We screened 279 participants and 87 EBV-naïve subjects were enrolled in 2012 , 16 of whom experienced primary EBV infection during their 9 month freshman year . All participants gave written informed consent and the University of Minnesota Institutional Review Board approved all protocols used . Subjects gave oral wash samples by gargling with 22mL of normal saline . Suspended oral cells were separated from supernatant by centrifugation and frozen in two aliquots at -80°C . Four 1ml aliquots of supernatant were saved and frozen at -80°C . Peripheral blood was obtained via venipuncture into EDTA Vacutainer tubes ( Fisher Scientific ) . Blood Peripheral blood mononuclear cells ( PBMCs ) were isolated by Accuspin System-Histopaque-1077 ( Sigma-Aldrich ) density gradient centrifugation per manufacturer’s instructions . PBMC were divided into 1x10^7 cells/mL aliquots in a 90% FBS and 10% dimethylsulfoxide solution to prevent cell damage ( Sigma-Aldrich ) . Vials were placed inside Mr . Frosty freezing containers ( Thermo Scientific ) and frozen at -80°C per the manufacturer’s instructions , then transferred to liquid nitrogen for long term storage . Cells were rapidly thawed in a 37°C water bath , diluted to 10ml in RPNK media supplemented with 50U/ml benzonase ( Novagen ) ( RPNK media: RPMI 1640 ( Cellgro ) supplemented with 10% FBS ( Atlanta Biologicals ) , 2% Penicillin—Streptomycin ( 5000U/ml , 5000μg/ml respectively , GIBCO , Invitrogen ) and 1% L-glutamine ( 29 . 2mg/ml , GIBCO ) ) . Cells were then counted using a hemocytometer and divided into separate fractions for flow cytometry , RNA processing , and/or DNA processing . Multiple time points were chosen from subjects who gave a blood sample during the incubation period . 1 to 2 x10^6 PBMCs from each of these time points were used in each stain . The following antibodies were used to identify relevant surface and intracellular markers: CD3 ( UCHT1 ) , CD4 ( RPA-T4 ) , CD11a ( HI111 ) , CD20 ( 2H7 ) , CD123 ( 6H6 ) , PD-1 ( MIH4 ) , CD25 ( BC96 ) , Foxp3 ( PCH101 ) ( eBioscience ) ; CD45RA ( HI100 ) , CD38 ( HIT2 ) , CD16 ( 3G8 ) , CD57 ( HCD57 ) , CD14 ( M5E2 ) , CD19 ( HIB19 ) , BDCA-2 ( 201A ) ( BioLegend ) ; CD56 ( NCAM16 . 2 ) , HLA-DR ( G46-6 ) , CD11c ( B-ly6 ) ( Becton Dickinson ) ; NKG2A ( Z199 ) , KIR2DL1/2DS1 ( EB6B ) , KIR2DL2/2DL3/2DS2 ( GL183 ) ( Beckman Coulter ) ; CD8 ( 3B5 ) ( Invitrogen ) ; KIR3DL1 ( DX9 ) , KIR3DL2 ( 539304 ) ( R&D systems ) . Intracellular granzyme B staining was performed using the Cytofix/Cytoperm kit per the manufacturer’s instructions ( BD ) . Intranuclear Foxp3 staining was performed using the Foxp3 / Transcription Factor Staining Buffer Set per the manufacturer’s instructions ( eBioscience ) . All samples were acquired using an LSR II ( BD ) and analyzed with FlowJo software ( TreeStar ) . During analysis , lymphocyte frequencies were normalized to the absolute number of lymphocytes in each sample . For each sample , 1–2 x 106 PMBCs were used for each RNA extraction . Cells were first homogenized using QIAshredder columns ( Qiagen ) per the manufacturer’s instructions . RNA was then extracted using RNeasy kit ( Qiagen ) with on-column DNase step ( Qiagen ) per the manufacturer’s instructions . RNA was then quantified using a Nanodrop 2000/2000c spectrophotometer ( Thermo Scientific ) and kept frozen at -80°C . DNA extractions were performed with the Qiagen QIAmp Blood Mini kit per the manufacture’s instructions , using either 200 μL of whole blood or 5x10^6 PBMC . An EBV BMLF1259–267 ( GLCTLVAML ) -A*0201 tetramer reagent was purchased from ProImmun . Other biotinylated MHC-peptide monomers were obtained from the National Institutes of Health ( NIH ) tetramer facility: EBV BRLF1109–117 ( YVLDHLIVV ) -A*0201 , EBV BRLF1147–155 ( RVRAYTYSK ) -A*03 , EBV BZLF1190–197 ( RAKFKQLL ) -B*08 , EBV EBNA3A325–333 ( FLRGRAYGL ) -B*08 , EBNA3A379–387 ( RPPIFIRRL ) -B*07 , and EBNA3A603–611 ( RLRAEAQVK ) -A*03 . Before use , APC-streptavidin ( Invitrogen ) was added to monomers at a 4:1 molar ratio overnight in the dark at 4°C to generate fluorescent pMHCI tetramer complexes . All tetramers were stored in the dark at 4°C . cDNA was generated with 100ng of starting RNA using the SuperScript III Platinum Two-Step qRT-PCR Kit ( Invitrogen ) per the manufacturer’s instructions . Samples were stored at -20°C . Standard quantitative PCR was performed with FastStart Universal SYBR Green Master ( Rox ) ( Roche ) per the manufacture’s instructions . Additional data were generated with precoated “SuperArray” PCR plates . 43 genes were selected for analysis by PCR from a larger list of changed genes in IM subjects and other acute viral infections that comprised relevant functional groupings as assessed by Ingenuity Pathway Analysis as previously described [19] . 384-well SuperArray plates pre-coated with primers for the desired 43 genes plus 5 controls were obtained as a custom order from SABiosciences . 10 μL of cDNA per subject was used with RT2 Real-time SYBR green/Rox PCR master mix ( SABiosciences ) for qRT-PCR analysis . Products were detected using an ABI Prism 7900HT Sequence Detection System ( Applied Biosystems ) . The genes ACTB , B2M , and RPL13A were used as housekeeping genes during the calculation of fold changes . Fold changes were calculated as: 2^ ( Δ Acute Housekeeping Control–Baseline Housekeeping Control ) /2^ ( Δ Acute Gene of Interest–Baseline Gene of Interest ) . Fold change values obtained by quantitative PCR were imported into the open source program Cluster 3 . 0 , [40] which clustered the genes hierarchically using a Pearson non-averaged correlation and average linkage . Heatmaps were then visualized using the program Java Treeview . [41] PCR was performed with the HotStarTaq master mix kit ( Qiagen ) per the manufacture’s instructions . Primers specific for EBNA1 [42] were used ( outer-F 5’-GTA GAA GGC CAT TTT TCC AC-3’; outer-R 5’-CTC CAT CGT CAA AGC TGC A-3’; inner-F 5’-AGA TGA CCC AGG AGA AGG CCC AAG C-3’; inner-R 5’-CAA AGG GGA GAC GAC TCA ATG GTG T-5’ ) .
|
Chronic viral infections are ubiquitous in the human population . Many of these viruses persist in spite of the host immune response . Epstein-Barr virus ( EBV ) is a human herpesvirus and the primary causative agent of acute infectious mononucleosis . The virus is primarily transmitted through salivary exchange yet the kinetics of dissemination and initial immune response remain poorly understood , especially during EBV’s lengthy six-week incubation period . By doing prospective analysis of natural infection in human subjects , we were able to examine responses during the incubation period . We found that virus disseminates into the blood from the oral cavity much earlier than previously predicted and often before large-scale viral replication in oral cells . This correlated with a systemic innate immune response in the form of type I interferon signaling . A subsequent decrease in circulating plasmacytoid dendritic cells was observed simultaneously with polyclonal CD8 T cell activation . These data suggest that EBV replication is self-limiting in the oral cavity and that infection is established for several weeks before virally infected cells traffic to peripheral blood and initiate innate and adaptive immune response .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
The Incubation Period of Primary Epstein-Barr Virus Infection: Viral Dynamics and Immunologic Events
|
Nef is the viral gene product employed by the majority of primate lentiviruses to overcome restriction by tetherin ( BST-2 or CD317 ) , an interferon-inducible transmembrane protein that inhibits the detachment of enveloped viruses from infected cells . Although the mechanisms of tetherin antagonism by HIV-1 Vpu and HIV-2 Env have been investigated in detail , comparatively little is known about tetherin antagonism by SIV Nef . Here we demonstrate a direct physical interaction between SIV Nef and rhesus macaque tetherin , define the residues in Nef required for tetherin antagonism , and show that the anti-tetherin activity of Nef is dependent on clathrin-mediated endocytosis . SIV Nef co-immunoprecipitated with rhesus macaque tetherin and the Nef core domain bound directly to a peptide corresponding to the cytoplasmic domain of rhesus tetherin by surface plasmon resonance . An analysis of alanine-scanning substitutions identified residues throughout the N-terminal , globular core and flexible loop regions of Nef that were required for tetherin antagonism . Although there was significant overlap with sequences required for CD4 downregulation , tetherin antagonism was genetically separable from this activity , as well as from other Nef functions , including MHC class I-downregulation and infectivity enhancement . Consistent with a role for clathrin and dynamin 2 in the endocytosis of tetherin , dominant-negative mutants of AP180 and dynamin 2 impaired the ability of Nef to downmodulate tetherin and to counteract restriction . Taken together , these results reveal that the mechanism of tetherin antagonism by Nef depends on a physical interaction between Nef and tetherin , requires sequences throughout Nef , but is genetically separable from other Nef functions , and leads to the removal of tetherin from sites of virus release at the plasma membrane by clathrin-mediated endocytosis .
Mammalian cells express a number of proteins that inhibit specific steps of virus replication . One such factor , tetherin ( BST-2 or CD317 ) , impairs the release of enveloped viruses from infected cells [1] , [2] , [3] , [4] , [5] . Tetherin is a type II integral membrane protein with a topology that allows both ends of the molecule to be anchored in lipid membranes [6] . It has an N-terminal cytoplasmic domain followed by a single-pass transmembrane domain , an extracellular coiled-coil domain and a C-terminal glycosyl-phosphatidylinositol ( GPI ) anchor [6] . Under conditions of interferon-induction , tetherin is upregulated and becomes incorporated into virus particles as they attempt to bud from infected cells [7] , [8] , [9] . Captured virions are then internalized and routed to lysosomal compartments for degradation by a mechanism that involves interactions between the cytoplasmic domain of tetherin and the endocytosis machinery of the cell [7] , [10] . Tetherin has played an important role in shaping the course of lentiviral evolution in primates , having selected for at least three different viral gene products to overcome this restriction factor ( reviewed in [11] , [12] ) . Whereas HIV-1 Vpu and HIV-2 Env antagonize human tetherin [4] , [5] , [13] , the majority of SIVs use Nef to counteract the tetherin proteins of their non-human primate hosts [14] , [15] , [16] . Indeed , HIV-1 Vpu and HIV-2 Env appear to have acquired the ability to antagonize tetherin due to the absence of sequences in the cytoplasmic domain of human tetherin that confer susceptibility to Nef [11] , [12] . Alternative models have been proposed for the mechanism of tetherin antagonism by HIV-1 Vpu . Vpu physically associates with tetherin via membrane-spanning domain interactions [17] , [18] , [19] , [20] , recruits ßTrCP-2 , a component of the Skp1-Cullin1-F-box ubiquitin ligase complex , promoting the ubiquitination of non-lysine residues in the cytoplasmic domain of tetherin [21] , [22] , and uses the ESCRT-mediated trafficking of tetherin [23] for degradation in lysosomes [17] , [19] , [24] , [25] , [26] . There is also evidence that Vpu may antagonize tetherin in the absence of degradation by sequestering the protein in a perinuclear compartment , either by retaining newly synthesized tetherin , or by blocking the recycling of tetherin to the plasma membrane [24] , [27] , [28] , [29] . Vpu-mediated downmodulation of tetherin and enhancement of virus release were also recently shown to be dependent in part on clathrin-mediated endocytosis [30] . The mechanism of tetherin antagonism by HIV-2 Env depends on a physical interaction between Env and tetherin , and a conserved tyrosine-based endocytosis motif in the cytoplasmic tail of gp41 [13] , [31] , [32] . The sequences required for Env interactions with tetherin are poorly defined , but appear to reside in the extracellular domains of both proteins , as indicated by analyses of recombinant forms of Env and tetherin [13] , [33] , and the identification of substitutions in the ectodomains of each protein that disrupt tetherin antagonism [32] , [33] , [34] , [35] . HIV-2 Env does not promote the degradation of tetherin , but leads to the internalization and sequestration of tetherin by a clathrin-dependent mechanism , consistent with the trapping of tetherin in recycling endosomes [13] , [30] , [31] . Comparatively little is known about the mechanism of tetherin antagonism by Nef . The Nef proteins of phylogenetically diverse SIVs , including SIVsmm/mac , SIVagm and SIVcpz , antagonize the tetherin proteins of their non-human primate hosts , but not human tetherin [14] , [15] , [16] . This specificity maps to a five amino acid sequence that is present in the cytoplasmic tails of non-human primate tetherin proteins ( G/D14DIWK18 in rhesus macaques , sooty mangabeys and chimpanzees ) , but absent from the corresponding region of human tetherin [15] , [16] . We previously reported that SIV Nef downregulates rhesus tetherin from the surface of transfected and infected cells [15] , [36] . Zhang et al . further demonstrated that this activity is AP-2-dependent [37] . Here we demonstrate a direct physical interaction between SIV Nef and rhesus tetherin , define residues throughout Nef required for tetherin antagonism , and demonstrate that the anti-tetherin activity of Nef is dependent , at least in part , on clathrin-mediated endocytosis .
SIV Nef was tested for a physical interaction with tetherin by co-immunoprecipitation . Tetherin was immunoprecipitated from lysates of 293T cells co-transfected with constructs expressing Nef and either human or rhesus macaque tetherin . Immunoprecipitated proteins were separated by SDS-PAGE , and western blots were probed with monoclonal antibodies to Nef and to tetherin . In accordance with the selective activity of Nef in opposing restriction by tetherin [15] , [16] , SIV Nef strongly co-immunoprecipitated with rhesus tetherin , but not with human tetherin ( Figure 1A ) . To determine if this interaction is direct , SIV Nef was tested for binding to peptides corresponding to the N-terminal cytoplasmic domains of rhesus and human tetherin by surface plasmon resonance ( SPR ) . Tetherin peptides were biotinylated at conserved cysteine residues ( C25 in rBST-2 and C20 in hBST-2 ) and coupled to the surface of neutravidin-coated CM5-BIAcore chips to mimic the native orientation of the N-terminus of tetherin on the inner leaflet of the plasma membrane . Recombinant SIVmac239 Nef proteins containing residues 4–263 ( Nef4–263 ) and 96–237 ( Nef96–237 ) were flowed over the immobilized peptides to assess binding . SIV Nef96–237 bound to the N-terminal peptide of rhesus tetherin ( Figure 1B ) , but not to the corresponding peptide of human tetherin ( Figure 1C ) . The dissociation constant and the maximum response for SIV Nef96–237 binding to rhesus tetherin were determined by equilibrium analysis ( Kd . app 401+/−114 µM ) ( Figure 1D ) . The nearly full-length Nef protein , Nef4–263 , also bound to rhesus tetherin ( data not shown ) . However , the Kd of this interaction could not be determined due to artifacts at protein concentrations greater than 300 µM that may reflect Nef dimerization . These results reveal a direct physical interaction between SIV Nef and the cytoplasmic domain of rhesus macaque tetherin . To identify sequences in SIV Nef that contribute to tetherin antagonism , 103 pair-wise alanine-scanning substitutions were introduced throughout the N-terminal , globular core and the flexible loop regions of SIVmac239 Nef ( residues 3–210 ) , and these mutants were tested for their ability to counteract rhesus tetherin in virus release assays ( Figure 2 ) . Mutations in the C-terminal domain ( residues 211–263 ) were not tested , since these sequences can be deleted without affecting the anti-tetherin activity of Nef ( data not shown ) . Virus release for 43 of the Nef mutants was reduced to a similar or greater extent than a myristoylation site mutant ( G2A ) , which was previously shown to impair tetherin antagonism [15] . These results were corroborated by western blot analyses comparing p55 Gag expression in cell lysates to the accumulation of p27 capsid ( CA ) in the cell culture supernatant ( Figure S1 ) . This approach identified 9 substitutions in the N-terminal domain ( Figure 2B ) , 27 substitutions in the globular core domain ( Figure 2C ) , and 7 substitutions in the flexible loop region of Nef ( Figure 2D ) that disrupt tetherin antagonism . To define residues in SIV Nef that contribute to interactions with tetherin , Nef mutants lacking anti-tetherin activity were tested for binding to rhesus tetherin by co-immunoprecipitation ( Figure 3A ) . The ratios of the band intensities for Nef and tetherin in immunoprecipitates were calculated to quantify differences in binding to tetherin ( Table 1 ) . Substitutions at positions 2 , 5–6 , 66–67 , 68–69 and 70–71 in the N-terminal domain , positions 116–117 and 174–175 in the globular core domain , and positions 181–182 , 193–194 , 195–196 and 199–200 in the flexible loop region diminished the co-immunoprecipitation of Nef with tetherin ( Figure 3A , 3B and Table 1 ) . Some of the substitutions in the globular core , particularly at positions 178–179 and 180–181 , resulted in reduced levels of Nef protein in cell lysates ( Figure 3A and Table 1 ) . Hence , the loss of tetherin binding in these instances may reflect decreased Nef protein stability or expression rather than a disruption of tetherin contact residues . However , many of the core domain mutations that decreased steady-state levels of Nef did not result in a corresponding decrease in binding to tetherin , and a few paradoxically appear to have increased the stability of this interaction ( Figure 3A and Table 1 ) . Since many of these mutations did not completely abrogate binding to tetherin , combinations of alanine substitutions were also tested . The co-immunoprecipitation of Nef with tetherin was reduced to nearly undetectable levels by combining substitutions in the N-terminal domain , either alone ( residues 66–71 ) , or together with the substitutions in the globular core and flexible loop ( x6: residues 66–71 , 116–117 , 174–175 and 181–182 ) ( Figure 3B ) . To determine if residues in the flexible loop of SIV Nef are needed for direct binding to rhesus tetherin , purified Nef proteins with deletions in the flexible loop were tested for binding to the cytoplasmic domain of rhesus tetherin by SPR . Recombinant SIVmac239 Nef96–237 lacking residues 181–200 , 181–195 , 197–205 and 181–205 were flowed over BIAcore chips coated with a 26 amino acid peptide corresponding to the cytoplasmic domain of rhesus tetherin , as described for Figure 1B . All of the deletion mutants bound to the peptide within a similar range of apparent Kd values ( Figure 3C; representative raw SPR data is shown in Figure S2A and S2B ) . The somewhat lower Kd . app estimates for three of the deletion mutants ( Δ181–200 , Δ181–195 and Δ181–205 ) may be due to technical limitations with testing these mutants at high concentrations as a result of protein aggregation , rather than an actual increase in binding affinity . Some of these mutants also showed lower apparent Kd values for binding to a TCRζ chain peptide ( Figure S2C–S2E ) , an interaction that reflects direct binding of the SIV Nef core domain as corroborated by three-dimensional structural data [38] . Hence , these results demonstrate that the flexible loop of Nef is not required for direct binding to rhesus tetherin , implying that surfaces of the core domain are sufficient for the low affinity interaction with the N-terminus of rhesus tetherin observed by SPR . Cytoplasmic domain variants of tetherin were also tested for binding to Nef by co-immunoprecipitation and SPR . Deletion of the first 10 amino acids of rhesus tetherin ( rΔ10 ) significantly reduced , but did not eliminate , binding to Nef in both assays ( Figures 4A , 4B and S3B ) , indicating that although these residues are not essential for binding to Nef , they contribute to the stability of the interaction . Consistent with previous studies mapping the anti-tetherin activity of Nef to a five amino acid sequence ( G/D14DIWK18 ) that is missing from human tetherin [15] , [16] , alanine substitutions at positions 14–18 of rhesus tetherin ( rA14-A18 ) diminished Nef binding , whereas the introduction of these residues into human tetherin partially restored binding ( Figures 4A , 4B , S3C and S3D ) . Thus , although the specificity of tetherin antagonism by Nef is dependent on residues 14–18 , and these sequences contribute to a physical interaction with Nef , they are not the sole determinant of Nef binding . These tetherin variants were also tested in virus release assays to determine how Nef binding relates to susceptibility to antagonism . In accordance with partial binding of Nef to rΔ10 and hDDIWK , restriction of virus release by each of these mutants was partially counteracted by Nef ( Figure 4C ) . However , despite a physical interaction between rA14-A18 and Nef that was detectable by co-immunoprecipitation and SPR assays , this mutant was resistant to antagonism and restricted virus release to an extent comparable to human tetherin ( Figure 4C ) . Therefore , although a physical interaction may be necessary for tetherin antagonism by Nef , it is not sufficient . This raises the possibility that the anti-tetherin activity of Nef may require the recruitment of one or more additional cellular factors that participate in interactions with the G/D14DIWK18 sequence . To determine if substitutions that impair tetherin antagonism also disrupt other activities of Nef , the Nef mutants were tested for CD4-downregulation , MHC I-downregulation and infectivity enhancement; three functional activities of Nef that require distinct protein sequences and cellular pathways [39] , [40] . CD4− and MHC class I-downregulation assays were performed by electroporating Jurkat cells with bicistronic constructs that express wild-type Nef , or a mutant Nef protein , together with green fluorescent protein ( GFP ) , and comparing the mean fluorescence intensity ( MFI ) of CD4 and MHC class I staining on the surface of cells expressing Nef to cells transfected with an empty vector ( pCGCG ) ( Figures S4 and S5 ) . Infectivity enhancement was measured by infecting GHOST X4/R5 cells , which express GFP in response to HIV-1 or SIV infection , with SIVmac239 Δnef trans-complemented with wild-type Nef or each of the Nef mutants , and measuring the percentage of infected GFP+ cells by flow cytometry ( Figure S6 ) . Of the 43 Nef mutants with impaired anti-tetherin activity , only 5 retained the ability to downregulate CD4 within 3 standard deviations of wild-type Nef ( Figure 5A , black dotted line ) . In contrast , 16 of the mutants retained the ability to downregulate MHC I within 3 standard deviations of wild-type Nef ( Figure 5B , black dotted line ) . Whereas substitutions in the N-terminal domain and flexible loop region , with the exception of substitutions at positions 5–6 and 74–75 , had little or no effect on MHC I-downregulation , many of the substitutions in the globular core impaired this activity ( Figure 4B and S5 ) . In most cases , the loss of MHC I-downregulation corresponded with a partial decrease in Nef protein levels ( Table 1 ) , suggesting that the effects of these mutations were not necessarily specific to this function of Nef . Nevertheless , five Nef mutants with impaired anti-tetherin activity , and no significant effects on protein stability , retained the ability to downregulate both CD4 and MHC I molecules . These included Nef mutants with substitutions at positions 106–107 , 181–182 , 193–194 , 199–200 and 209–210 ( Table 1 ) . Therefore , tetherin antagonism is separable from CD4− and MHC class I-downregulation . The infectivity of SIV Δnef trans-complemented with each of the Nef mutants relative to SIV Δnef trans-complemented with wild-type Nef was also determined to assess the effects of the substitutions on Nef-mediated infectivity enhancement . To control for assay-to-assay variation in the susceptibility of the GHOST X4/R5 cells to infection , the percentage of infected cells obtained for each of the Nef mutants was normalized to the percentage of infected cells obtained for wild-type Nef . Nef mutants were considered to retain the ability to enhance virus infectivity if the relative infectivity was at least 5 standard deviations above the infectivity of SIV Δnef trans-complemented with an empty vector ( pCGCG ) ( Figure 5C ) . This analysis identified 12 Nef mutants that were impaired for infectivity enhancement ( Figure S6A–S6C ) . In accordance with previous observations , the G2A substitution in SIV Nef did not have a significant effect on virus infectivity [41] . Consistent with a study of this function of HIV-1 Nef [40] , all of these mutants also lost the ability to bind to dynamin 2 ( Dyn2 ) ( Figure S6D ) . Of these 12 Nef mutants , 10 also exhibited impaired anti-tetherin activity , suggesting that tetherin antagonism and infectivity enhancement may be linked , perhaps by a common dependence on a physical interaction with Dyn2 . However , two of the substitutions in the core domain at positions 94–95 and 98–99 that disrupted infectivity enhancement did not significantly affect anti-tetherin activity ( Figure 2C and S6B ) . Moreover , three of the substitutions that disrupted binding to Dyn2 ( 82–83 , 146–147 and 168–169 ) did not impair binding to rhesus tetherin ( Table 2 ) . Thus , Nef appears to use distinct surfaces to bind Dyn2 and tetherin . In addition , since all but 10 of the 43 Nef mutants lacking anti-tetherin activity retained the ability to enhance infectivity , including the 5 mutants that retained both CD4− and MHC class I-downregulation , infectivity enhancement is independent of tetherin antagonism ( Figure 5C and Table 1 ) . AP-2 binds to a pair of conserved motifs in the flexible loop of Nef that are necessary for tetherin antagonism; a di-leucine motif and a di-acidic motif , corresponding to residues E191XXXLM195 and D204D205 of SIVmac239 Nef , respectively ( Figure 6A ) [37] , [42] , [43] . Consistent with previous observations [37] , substitutions of residues within either of these motifs ( positions 193–194 , 195–196 , 203–204 , 205–206 ) impaired tetherin downregulation by Nef ( Figure 6B and 6C ) . In addition , substitutions at positions 181–182 , 199–200 and 209–210 , not previously identified as AP-2 binding sites or known to be involved in the anti-tetherin activity of Nef , also impaired tetherin downregulation ( Figure 6B and 6D ) . These results confirm the role of the di-leucine and di-acidic motifs and identify additional sequences in the flexible loop of SIV Nef required for tetherin downmodulation . Since AP-2 binds to both Nef and tetherin [42] , [43] , [44] , the flexible loop mutants were also tested for their ability to interact with the α-adaptin ( α2 ) and μ2 subunits of AP-2 by co-immunoprecipitation . Endogenous α2 and μ2 were immunoprecipitated in parallel from lysates of parental 293T cells ( Figure 6E ) , or 293T cells that constitutively express HA-tagged rhesus tetherin ( Figure 6F ) , following transfection with Nef expression constructs . Immunoprecipitates were separated by electrophoresis and western blots were probed with antibodies to Nef , α2 , μ2 and tetherin . Wild-type Nef co-immunoprecipitated with α2 and μ2 , both in the absence and in the presence of tetherin ( Figures 6E and 6F ) . As expected , substitutions in the di-leucine and di-acidic motifs of Nef ( 193–194 , 195–196 , 203–204 , 205–206 ) greatly diminished binding to both subunits ( Figures 6E and 6F ) . Substitutions at positions flanking these motifs ( 181–182 and 209–210 ) also disrupted binding to α2 and to μ2 ( Figures 6E and 6F ) . Whereas in the absence of tetherin , substitutions at positions 181–182 , 203–204 , 205–206 and 209–210 eliminated Nef binding to μ2 , and substitutions at positions 193–194 , 195–196 , and 199–200 reduced Nef binding to μ2 , ( Figure 6E ) , the binding of these Nef mutants to μ2 was partially restored in the presence of tetherin ( Figure 6F ) . These results suggest that the loss of anti-tetherin activity for each of the flexible loop mutants reflects a deficit in Nef binding to AP-2 , and raises the possibility that AP-2 may form a multimeric complex with both Nef and tetherin that stabilizes an otherwise low affinity direct interaction between these two proteins . Since the activities of HIV-1 Vpu and HIV-2 Env in downregulating tetherin and facilitating virus release were recently shown to be dependent on AP180 , a component of the clathrin assembly complex [30] , and dynamin 2 ( Dyn2 ) , an ubiquitously expressed GTPase required for the scission of vesicular membranes [45] , we asked whether tetherin antagonism by Nef also requires AP180 and Dyn2 . The effects of dominant-negative mutants of Dyn2 ( Dyn2K44A ) and AP180 ( AP180C ) on the surface expression of tetherin , and on virus release , were therefore tested in 293T cells expressing HA-tagged rhesus macaque tetherin . As a control , we also included the dominant-negative mutant of dynamin 1 ( Dyn1K44A ) , which is exclusively expressed in neurons [46] . Changes in tetherin expression on the cell surface were assessed by flow cytometry after co-transfection with constructs expressing Dyn2K44A , Dyn2 , Dyn1K44A or AP180C , with or without Nef ( Figure 7A ) . In the absence of Nef , Dyn1K44A , Dyn2K44A and AP180C all slightly increased surface levels of tetherin , whereas wild-type Dyn2 decreased surface levels of tetherin ( Figure 7A ) , which may reflect a role for dynamin and clathrin in the constitutive endocytosis of tetherin [44] . In the presence of Nef , the effects of the dominant-negative mutants were more pronounced . Whereas SIV Nef reduced the surface expression of rhesus tetherin by 2- to 3-fold , as previously reported [15] , this effect was almost completely reversed by AP180C and Dyn2K44A ( Figure 7A ) . Although Dyn1K44A also increased the overall levels of tetherin at the cell surface , as shown in transfections with the empty vector , Nef was still able to downregulate tetherin in the presence of Dyn1K44A ( Figure 7A ) . Likewise , in the presence of wild-type Dyn2 , Nef further decreased the surface levels of tetherin . These results demonstrate that the downregulation of tetherin by Nef is dependent , at least in part , on clathrin-mediated endocytosis . To further confirm that Dyn2 and clathrin are required for tetherin antagonism by Nef , virus release for wild-type SIV versus SIV Δnef was compared in the presence and absence of each of the dominant-negative mutants . 293T cells were co-transfected with proviral DNA for SIVmac239 or SIVmac239 Δnef , together with a construct expressing rhesus macaque tetherin and expression constructs for AP180C , Dyn1K44A , Dyn2K44A or Dyn2 , and the accumulation of virus particles in the cell culture supernatant was measured by SIV p27 antigen-capture ELISA . Whereas AP180C and Dyn2K44A completely abrogated the resistance of wild-type SIV to rhesus tetherin , as indicated by comparable levels of virus release for SIVmac239 and SIVmac239 Δnef , virus release for wild-type SIV was not significantly affected by Dyn1K44A or Dyn2 ( Figure 7B ) . Western blot analyses of cell lysates confirmed protein expression for tetherin , Nef and the dominant-negative mutants ( Figure 7C ) . None of the dominant-negative mutants inhibited tetherin expression . On the contrary , increased steady-state levels were observed in the presence of AP180C , Dyn1K44A and Dyn2K44A ( Figure 7C ) , consistent with the modest increase in cell surface expression of tetherin in the absence of Nef ( Figure 7A ) . A role for Dyn2 in the anti-tetherin activity of Nef was further investigated by comparing virus replication of wild-type SIV and nef-deleted SIV with or without Dynasore , a chemical inhibitor of dynamin . A Herpesvirus saimiri-immortalized rhesus macaque CD4+ T cell line [47] , was infected with SIVmac239 and SIVmac239 Δnef , treated with IFNα to upregulate tetherin , and maintained in medium with or without Dynasore . While Dynasore had little effect on the replication of SIV Δnef , which was suppressed relative to wild-type SIV by the IFNα-induced upregulation of tetherin , Dynasore significantly reduced wild-type SIV replication ( Figure 7D ) . Indeed , wild-type SIV replication in the presence of Dynasore was comparable to SIV Δnef replication , suggesting that this compound fully negated the resistance provided by Nef to the antiviral effects of tetherin . However , since Dyn2 is also required for infectivity enhancement by Nef [40] , these results may reflect an additional effect of Dynasore on viral infectivity . Changes in the subcellular distribution of tetherin in the presence of Nef were examined in uninfected and SIV-infected cells by confocal microscopy . 293T cells expressing HA-tagged rhesus tetherin were infected with VSV-G-pseudotyped SIVmac239 Δenv and stained for Nef and tetherin . In uninfected cells , tetherin was observed at the plasma membrane and within the trans-Golgi network ( Figure 8A and Figure S7 ) , as previously reported [48] . However , in SIV-infected cells , the overwhelming majority of tetherin was observed within intracellular compartments ( Figure 8B ) . To better define the subcellular distribution of tetherin , cells were stained for markers of the trans-Golgi network ( TGN46 ) , endosomes ( CD63 ) and lysosomes ( LAMP-1 ) . In some cells , tetherin co-localized with TGN46 ( Figure 9A ) , but did not appear to co-localize with CD63 ( Figure 9B ) . These results suggest that Nef may partially retain tetherin within the trans-Golgi network with little or no sequestration in endosomes . However , in the majority of the SIV-infected cells , tetherin was found to co-localize with LAMP-1 ( Figure 9C ) , but not in uninfected cells ( Figure S7 ) . These observations were supported by quantifying the extent of tetherin co-localization with TGN46 , CD63 and LAMP-1 by calculating the Pearson's correlation coefficients for these markers in twenty SIV-infected cells . Although the distribution of cells exhibiting co-localization of tetherin with TGN46 was heterogeneous ( Figure 9D ) , it was higher than the extent of co-localization with CD63 ( P = 0 . 042 ) . In the case of LAMP-1 , the extent of co-localization with tetherin was significantly higher than for either TGN46 ( P = 0 . 0048 ) or CD63 ( P<0 . 000001 ) . Therefore , similar to the effects of Nef on CD4 and MHC class I trafficking [49] , Nef appears to direct tetherin to lysosomes .
The primate lentiviruses have evolved to use at least three different proteins to counteract tetherin; Nef , Vpu and Env [5] , [13] , [15] , [16] , [32] , [36] , [50] . Although a number of studies have addressed the mechanisms of tetherin antagonism by HIV-1 Vpu and HIV-2 Env [4] , [5] , [13] , [18] , [23] , [27] , [29] , [30] , [31] , [51] , relatively little is known about the mechanism of tetherin antagonism by Nef–the viral gene product used by most SIVs to counteract the tetherin proteins of their respective hosts . In accordance with the species-dependent activity of Nef in opposing restriction by tetherin [15] , [16] , we show for the first time that Nef selectively binds to rhesus macaque tetherin , but not to human tetherin . We identify residues in the N-terminus , globular core and flexible loop of Nef that are required for anti-tetherin activity , and demonstrate that , despite substantial overlap with sequences required for CD4 downregulation , tetherin antagonism is genetically separable from this activity , as well as from other Nef functions including MHC class I downregulation and infectivity enhancement . We also show that dominant-negative mutants of AP180 and Dyn2 impair tetherin antagonism by Nef , indicating that this activity is dependent , at least in part , on clathrin-mediated endocytosis . Co-immunoprecipitation and surface plasmon resonance assays revealed a selective physical interaction between SIV Nef and rhesus tetherin . The specificity of this interaction is determined by binding of the core domain of Nef to the cytoplasmic domain of tetherin , since a truncated form of the SIVmac239 Nef protein , containing the globular core of the protein , was sufficient for binding to a peptide corresponding to the cytoplasmic domain of rhesus tetherin . However , the affinity of this interaction was low ( Kd∼400 µM ) , suggesting that additional Nef sequences , and perhaps one or more cellular co-factors , contribute to the stability of this interaction in virus-infected cells . In support of this , an analysis of alanine-scanning substitutions identified sequences in the N-terminal , globular core and flexible loop domains of Nef that participate in binding to rhesus tetherin . Although the N-terminal domain and flexible loop were dispensable for binding by SPR , these sequences were required to detect an interaction by co-immunoprecipitation . The contribution of the N-terminus of Nef to interactions with tetherin may reflect an indirect effect on membrane association , since the targeting of Nef to cellular membranes is dependent on the myristyolation of a glycine residue at position 2 , and structural studies suggest that the N-terminus of Nef is disordered in the absence of phospholipids [52] , [53] , [54] . Substitutions at positions 116–117 , 174–175 in the globular core domain , and positions 181–182 , 193–194 , 195–196 and 199–200 in the flexible loop region also reduced binding to rhesus tetherin . Since the flexible loop contains a di-leucine and a di-acidic motif ( E190xxxLM and D204D ) required for binding to the AP-2 subunits ( α2-σ2 and μ2 , respectively ) [42] , [55] , [56] , and substitutions in these sites disrupt tetherin antagonism [37] , it is conceivable that AP-2 stabilizes the binding of Nef to tetherin . Indeed , Nef was recently shown to form a trimolecular complex with the μ1 subunit of AP-1 to stabilize an otherwise low affinity bimolecular interaction with the cytoplasmic tail of MHC class I molecules [57] , [58] . In support of a possible trimeric complex with AP-2 , Nef and rhesus tetherin both co-immunoprecipitated with the μ2 and α2 subunits of AP-2 [24] , [44] . Taken together , these results suggest a model in which the specificity of SIV Nef for rhesus tetherin is driven by a direct physical interaction between the core domain of Nef and the N-terminus of tetherin , which is stabilized by residues in the N-terminal domain and flexible loop region , either through direct contacts or indirect effects on membrane association and/or the recruitment of additional cellular co-factor ( s ) . A systematic analysis of alanine-scanning substitutions throughout the SIVmac239 Nef protein identified a total of 43 mutations that impaired anti-tetherin activity . Substitutions in the C-terminal domain were not tested , since deletion of these sequences did not affect tetherin antagonism . Most of the mutations that disrupted the anti-tetherin activity of Nef also disrupted CD4-downregulation , MHC I-downregulation or infectivity enhancement . In some cases , both CD4− and MHC I-downregulation were lost due to effects on the association of Nef with cellular membranes , such as the G2A mutation and probably also the adjacent substitutions at positions 5–6 . In other cases , these activities were lost due to a decrease in Nef expression or stability . While this was most evident for the changes at positions 178–179 and 180–181 , some of the substitutions in the globular core domain also had partial effects on steady-state levels of Nef that may account for their reduced activity in CD4 and MHC class I downregulation assays . Although there was substantial overlap with sequences required for CD4-downregulation , five mutations were identified that disrupted the anti-tetherin activity of Nef , while retaining nearly wild-type levels of CD4-downregulation , as well as MHC class I-downregulation and infectivity enhancement . These mutations included alanine substitutions at positions 106–107 in the core domain and at positions 181–182 , 193–194 , 199–200 and 209–210 in the flexible loop region . Thus , tetherin antagonism by Nef is genetically separable from other functional activities of the protein . In addition to the sequences identified by Zhang et al . [37] , we identified residues in the flexible loop region outside of the known AP-2 binding sites that separate tetherin antagonism from CD4-downregulation . Substitutions at positions 181–182 and 199–200 ( residues N181V182 and Q199T200 in SIVmac239 Nef ) specifically impaired the anti-tetherin activity of Nef without affecting CD4-downregulation . These residues are well conserved among Nef alleles of SIVsmm/mac and HIV-2 isolates , with identities of 61 . 1% for N181 , 75% for V182 , 94 . 4% for Q199 and 84 . 7% for T200 ( Los Alamos database; http://www . hiv . lanl . gov/content/index ) . Co-immunoprecipitation assays further demonstrated that these residues contribute to AP-2 binding . Our mutational analysis also identified residues in the N-terminal and globular core domains of Nef that are important for tetherin antagonism . Thus , our results reveal that the anti-tetherin activity of Nef is dependent on complex interactions involving multiple residues in the N-terminus , globular core and the flexible loop regions of the protein . Nef is a multifunctional accessory protein that interacts with a number of different cellular factors to modulate cellular trafficking [55] . Nef reroutes MHC I molecules from the trans-Golgi network to lysosomes via AP-1 and promotes the internalization and lysosomal degradation of CD4 via AP-2 [55] , [59] , [60] , [61] . Nef also enhances virus infectivity by an undefined mechanism that depends on a physical interaction with Dyn2 [40] . We previously demonstrated that Nef downmodulates tetherin from the surface of SIV-infected and transfected cells [15] , [36] , and this activity was later shown to occur by an AP-2-dependent pathway [37] . Our experiments with dominant-negative mutants of AP180 and Dyn2 confirm that the internalization of tetherin by Nef , and the capacity of Nef to rescue virus release in the presence of tetherin , depends , at least in part , on clathrin-mediated endocytosis . A role for Dyn2 was further demonstrated by showing that Dynasore , a chemical inhibitor of dynamin , suppressed wild-type SIV replication to an extent comparable to nef-deleted SIV under conditions of interferon-induced upregulation of tetherin . Since Dyn2 is also required for Nef-mediated infectivity enhancement , the inhibition of virus replication by Dynasore may reflect an additional effect of this compound on virus infectivity . Although tetherin antagonism and infectivity enhancement are genetically separable , 10 of the 12 Nef mutants that lost the ability to enhance virus infectivity , and to bind to Dyn2 , also lost the ability to counteract tetherin . The concordance of these activities suggests that a physical interaction with Dyn2 may be necessary for both Nef functions . However , two of the mutations in the globular core disrupted infectivity enhancement and binding to Dyn2 without impairing tetherin antagonism . Moreover , three of the mutants with impaired infectivity enhancement and binding to Dyn2 ( mutants 82–83 , 146–147 and 168–169 ) did not lose binding to tetherin , suggesting that Nef uses distinct protein surfaces to bind to Dyn2 and to tetherin . Therefore , unlike infectivity enhancement , the anti-tetherin activity of Nef does not depend on a physical interaction with Dyn2 . Consistent with previous studies demonstrating the downmodulation of rhesus tetherin by Nef [15] , [36] , [37] , SIV infection resulted in a striking redistribution of tetherin from the plasma membrane to compartments within the cell . An analysis of the distribution of tetherin in SIV-infected cells revealed co-localization with TGN46 and LAMP-1 , but not with CD63 , suggesting that in the presence of Nef , tetherin accumulates in the trans-Golgi network and in lysosomes . Localization of tetherin to the trans-Golgi network in uninfected cells has previously been reported [48] . Thus , the contribution of Nef to directing tetherin to that compartment is unclear . The trafficking of tetherin to lysosomes raises the possibility that , similar to the effect of HIV-1 Nef on CD4 and MHC class I molecules [49] , SIV Nef may direct rhesus tetherin for lysosomal degradation . In summary , we show that the mechanism of tetherin antagonism by SIV Nef; ( 1 ) involves a direct physical interaction between the core domain of Nef and the cytoplasmic domain of rhesus tetherin , ( 2 ) requires sequences throughout the N-terminal , globular core and flexible loop domains , yet is genetically separable from other functional activities of Nef , and ( 3 ) depends , at least in part , on clathrin-mediated endocytosis . These results begin to reveal the molecular interactions and cellular pathways by which the majority of the primate lentiviruses counteract the tetherin proteins of their non-human primate hosts .
293T cells were co-transfected with wild-type or nef-deleted SIV proviral DNA ( 100 ng ) and pcDNA3-tetherin or pcDNA3-tetherin mutants ( 50 ng ) . Differences in the amount of plasmid DNA in each transfection were offset by the addition of empty pcDNA3 vector ( 50 ng ) . Either pCGCG , pCGCG-Nef or pCGCG-Nef mutants ( 100 ng each ) were provided in trans to assess the ability of the Nef mutants to rescue virus release . All transfections were performed in duplicate in 24-well plates seeded the day before at 5×104 cells per well , using GenJet Lipid Transfection Reagents ( SignaGen Laboratories , Gaithersburg , MD ) . Forty-eight hours post-transfection , the amount of virus released into the cell culture supernatant was measured by SIV p27 antigen-capture ELISA ( Advanced Bioscience Laboratories , Inc . , Kensington , MD ) , and virus release was expressed as the percentage of maximal particle release in the absence of tetherin , as previously described [15] , [36] . Forty-eight hours post-transfection , 293T cell lysates were prepared by harvesting in 2× SDS sample buffer . Virions were recovered from the cell culture supernatant by centrifugation at 13 , 000 rpm for 2 hours at 4°C , and resuspended in 2× SDS sample buffer . Samples were boiled for 5 minutes , and separated by electrophoresis on 12% SDS-polyacrylamide gels and transferred to polyvinylidine fluoride ( PVDF ) membranes using a Trans-Blot SD transfer cell ( BioRad , Hercules , CA ) . The membranes were then blocked with 5% non-fat dry-milk in PBS containing 0 . 05% Tween-20 for 1 hour , and probed overnight at 4°C with one of the following primary antibodies . Tetherin/BST-2 was detected with a mouse polyclonal antibody ( abcam cat #ab88523 , Cambridge , MA ) at a dilution of 1∶500 . The SIV Gag proteins p27 and p55 were detected with the mouse monoclonal antibody 183-H12-5C ( AIDS Research and Reference Reagent Program , Division of AIDS , NIAID , NIH ) at a dilution of 1∶1000 . SIV Nef was detected using the mouse monoclonal antibody 17 . 2 ( AIDS Research and Reference Reagent Program , Division of AIDS , NIAID , NIH ) at a dilution of 1∶1000 . Endogenous β-actin was detected with the monoclonal antibody C4 ( Chemicon , Billerica , MA ) at a dilution of 1∶1000 . HA-tagged Dyn1K44A was detected with the HA-specific mouse monoclonal antibody HA . 11 ( Covance , Princeton , NJ ) at a dilution of 1∶1000 . The GFP fusion proteins Dyn2 and Dyn2K44A were detected using an anti-GFP mouse monoclonal antibody ( Sigma-Aldrich , St Louis , MO ) at a dilution of 1∶1000 . The dominant-negative mutant AP180C was detected with a mouse monoclonal FLAG-specific antibody ( Sigma-Aldrich , St Louis , MO ) at a dilution of 1∶1000 . After rinsing the PVDF membranes three times for 10 minutes in PBS 0 . 05% Tween-20 , the blots were probed with an HRP-conjugated goat anti-mouse secondary antibody ( Pierce , Rockford , IL ) at a dilution of 1∶2000 for 1 hour at room temperature . The blots were then rinsed three more times in PBS 0 . 05% Tween-20 , treated with SuperSignal West Femto Maximum Sensitivity substrate ( Pierce , Rockford , IL ) , and imaged using a Fujifilm Image Reader LAS 3000 ( Fujifilm Photo Film Co . , Japan ) . 293T cells ( 6×105 cells ) were co-transfected with constructs expressing wild-type and mutant forms Nef ( 2 µg ) along with rhesus tetherin , human tetherin , tetherin mutants , Dyn2-GFP or empty vector ( pCDNA3 ) ( 2 µg ) . Twenty-four hours later , cells were lysed with 400 µl of Lysis buffer ( Thermo Scientific , Rockford , IL ) and incubated on ice for 30 minutes . Lysates were transferred to a 1 . 5 ml tube and insoluble cell debris was removed by centrifugation at 3 , 000 rpm . Cell lysate ( 200 µl ) was set aside to confirm tetherin and Nef expression by western blot analysis , and the rest of the sample ( 200 µl ) was used for immunoprecipitation . Samples for immunoprecipitation were incubated on a rotating platform for 1 hour at 4°C with 1 µg of the anti-tetherin mouse monoclonal antibody 3H4 ( Sigma-Aldrich , St Louis , MO ) . Protein A sepharose beads or Protein A sepharose magnetic beads ( 50 µl ) ( GE Healthcare , Piscataway , NJ ) were then added , and the incubation was continued overnight at 4°C . The beads were washed ten times in Lysis buffer ( 500 µl ) and boiled in 2× SDS sample buffer . Denatured proteins were separated on 12% SDS-polyacrylamide gels and transferred to PVDF membranes . The blots were probed with the monoclonal antibody 17 . 2 to detect Nef , a mouse monoclonal to detect GFP ( Sigma-Aldrich , St Louis , MO ) , a monoclonal antibody to detect the α2 subunit of AP-2 ( Sigma-Aldrich , St Louis , MO ) , a rabbit monoclonal to detect the μ2 subunit of AP-2 ( abcam , Cambridge , MA ) at a dilution of 1∶1000 , or a polyclonal antibody against tetherin ( abcam , Cambridge , MA ) at a dilution of 1∶500 . Membranes were next probed with an HRP-conjugated goat anti-mouse antibody ( Pierce , Rockford , IL ) , a goat anti-mouse heavy chain specific antibody ( abcam , Cambridge , MA ) , or goat anti-rabbit secondary antibody ( abcam , Cambridge , MA ) , developed in SuperSignal West Femto Maximum Sensitivity substrate and imaged using a Fujifilm Image Reader LAS 3000 as described above . Quantification of the association between Nef and rhesus tetherin , Nef and AP-2 or Nef and Dyn2 was performed by determining the band density from western blots using ImageJ software ( Rasband , W . S . , Image , US . NIH , Bethesda , MD , http://rsb . info . nih . gov/ij , 1997–2008 ) . Synthetic peptides ( 21st Century Biochemicals ) corresponding to the cytoplasmic domain of rhesus macaque ( MAPILYDYRKMPMDDIWKEDGDKRCK ) and human ( MASTSYDYSRVPMEDGDKRCK ) tetherin were biotinylated via stable thioester bonds at conserved cysteine residues ( underlined in sequences shown above ) . In human tetherin cysteine at position 9 was replaced by serine ( italics ) in order to avoid multiple labeling . Additional tetherin peptides ( rA14-A18 , hDDIWK and rΔ10 ) were generated to further define the binding interface with Nef . These peptides were also biotinylated at the conserved cysteine residues mentioned above . Biotinylation was carried out for 2 hours at room temperature and pH 6 . 5 to 7 . 0 with a 10-fold molar excess of maleimide-PEG2-Biotin ( Pierce ) , followed by reverse-phase HPLC purification using a C18 Vydac 4 . 6×250 mm analytical column ( Vydac , Hesperia , CA ) with linear acetonitrile gradient ( 0–72% ) in 0 . 1% TFA ( 1 mL/min ) . Fractions containing the biotinylated peptides were identified by mass-spectrometry , pooled and lyophilized . A nearly full-length SIVmac239 Nef protein ( residues 4–263 ) , the core domain ( residues 96–237 ) and different Nef deletion mutants lacking residues in the flexible loop were expressed in E . coli BL21 ( DE3 ) as 6-His-thioredoxin fusion proteins and purified as described previously by Sigalov et al . [54] . After cell lysis with phosphate/Tris buffered 8 M urea solution ( pH 8 ) , the fusion protein was purified by affinity chromatography using NiNTA ( Qiagen ) under denaturing conditions ( 8 M urea ) , and refolded by dialysis against 20 mM Tris pH 8 . 0 , 150 mM NaCl , 0 . 1 mM DTT . The soluble fusion protein was digested with thrombin ( MP Biochmicals ) leaving both the full-length protein and the core domain of SIVmac239 with two additional N-terminal residues ( GS ) . Further purification was performed by anion-exchange chromatography ( POROS 20 HQ , Applied Biosystems ) and size-exclusion chromatography ( Superdex 200 , GE Healthcare ) . Surface plasmon resonance experiments were carried out on a BIAcore 3000 instrument at 25°C . Neutravidin ( approximately 30 , 000 resonance units ( RU ) ) were coupled to a CM5 sensor chip ( GE Healthcare ) in 10 mM acetate buffer pH 5 . 0 , and 0 . 005% ( v/v ) surfactant P20 at 10 µl/min using standard amine coupling protocols . Excess activated dextran carboxylate groups were capped with ethanolamine . Biotinylated peptides of tetherin ( rBST-21–26 and hBST-21–21 ) were captured ( 3 , 000 RU ) in different neutravidin-coupled experimental flow cells leaving one flow cell unmodified as a neutravidin-only control surface . Nef binding was studied at 25°C in PBS under reducing conditions ( 5 mM DTT ) . Purified samples of full-length , core domain or flexible loop deletion mutants of SIVmac239 Nef protein were injected at a flow rate of 5 µl/min over each experimental and control flow cell generating SPR sensorgrams . The sensorgram from the control cell was subtracted from the sensorgram of each experimental flow cell to correct for any nonspecific interaction with the CM5 or neutravidin surface . No regeneration step was required . Experiments were run in triplicate . For equilibrium analysis RU binding levels at equilibrium were extrapolated from each sensorgram ( corrected for nonspecific interaction ) in the concentration series , and plotted against concentration to derive a binding curve that was fit to a hyperbolic equation y = RUmax*x/ ( Kd . app+x ) , where y is the observed RU value , x is the concentration of Nef , and adjustable parameter RUmax and Kd . app are the RU value at saturation and the apparent binding constant ( Kd . app ) , respectively . Ten million Jurkat cells were electroporated with bicistronic pCGCG constructs ( 40 µg ) that express wild-type Nef , or Nef mutants , and green fluorescent protein ( GFP ) from a downstream internal ribosomal entry site . Twenty-four hours later , cells were stained with a PerCP-conjugated monoclonal antibody to CD4 ( BD Pharmingen , Billerica , MA ) and an APC-conjugated monoclonal antibody to MHC-I ( HLA-ABC , Dako , Carpintería , CA ) . After gating on the GFP+ cells , the mean fluorescence intensity ( MFI ) of CD4 and MHC I expression was determined . Data was collected using a FACSCalibur flow cytometer ( Becton Dickenson ) and analyzed using FlowJo 8 . 8 . 7 software ( TreesStar ) . 293T cells were co-transfected with nef-deleted SIV proviral DNA ( 100 ng ) and either pCGCG , pCGCG-Nef or pCGCG-Nef mutants ( 100 ng each ) . All transfections were performed in duplicate in 24-well plates seeded the day before at 5×104 cells per well , using GenJet Lipid Transfection Reagents ( SignaGen Laboratories , Gaithersburg , MD ) . Forty-eight hours post-transfection , the amount of virus released into the cell culture supernatant was measured by SIV p27 antigen-capture ELISA ( Advanced Bioscience Laboratories , Inc . , Kensington , MD ) . Next , 50 ng of p27 equivalents for each virus were inoculated overnight onto GHOST X4/R5 cells seeded the day before in 12-well plates at 2 . 5×104 cells per well . Twenty-four hours later , cells were washed and kept in fresh media . Forty-eight hours post-infection , cells were fixed and analyzed by flow cytometry as described above . The amount of infected cells was determined by calculating the percentage of GFP+ cells , and the infectivity of each Nef mutant relative to wild-type Nef was determined . 293T cells stably expressing HA-tagged rhesus tetherin ( 5×104 cells ) were transfected with 200 ng of pCGCG-Nef ( or Nef mutants ) or empty vector . In the case of experiments with dominant-negative mutants of endocytic pathways , cells were also transfected with 300 ng of each of the expression vectors coding for the dominant-negative mutants or empty vectors . Twenty-four hours post-transfection , cells were briefly trypsinized and stained for the surface expression levels of tetherin with a primary mouse monoclonal anti-HA antibody ( Covance , Princeton , NJ ) at a dilution of 1∶4 and a secondary donkey anti-mouse APC-conjugated antibody ( BD Pharmingen , Billerica , MA ) at a dilution of 1∶40 . Cells were gated on the GFP+ population and the levels of tetherin at the cell surface were determined by calculating the MFI . The percentage of tetherin present at the plasma membrane was calculated by dividing the MFI obtained in each transfection by the MFI obtained in transfections with empty vectors . Data was collected using a FACSCalibur flow cytometer ( Becton Dickenson ) and analyzed using FlowJo 8 . 8 . 7 software ( TreesStar ) . Two million 221 T cells , a Herpesvirus saimiri-immortalized rhesus macaque CD4+ T cell line [47] , were infected in duplicate with 20 ng p27 of SIVmac239 and SIVmac239 Δnef . After 3 h of incubation at 37°C , cells were washed three times and resuspended in 5 ml of R20+IL-2 ( 100 U ) . One day post-infection cells were treated with 100 U of IFNα , and 8 hours later one of the replicates was treated with 20 µM of Dynasore . Virus replication was monitored at selected time points by p27 antigen-capture ELISA of the culture supernatant . 293T cells stably expressing HA-tagged rhesus macaque tetherin ( 2×104 cells in a 8-well slide ) were infected with VSV-G pseudotyped SIVmac239 Δenv ( 50 ng p27 eq . ) . Twenty-four hours later , cells were washed and fixed for 10 minutes in acetone/methanol and blocked for 20–60 minutes with 100 mM glycine diluted in 10% normal goat serum in PBS with 0 . 2% fish skin gelatin , 0 . 1% Triton ×100 and 0 . 02% sodium azide ( 10% NGS-PBS-FSG-Tx100-NaN3 ) . The cells were then washed three times in 10% NGS-PBS-FSG-Tx100-NaN3 , and stained . The mouse monoclonal antibodies 17 . 2 ( IgG1 ) and 3H4 ( IgG2a ) were used at a dilution of 1∶250 to stain for Nef and tetherin , respectively . The cells were subsequently stained with Alexa-488- and Alexa-568-conjugated goat anti-mouse secondary antibodies specific for IgG1 and IgG2a , respectively ( Invitrogen , Grand Island , NY ) ( 1∶1000 ) , and with TO-PRO3 ( Invitrogen ) ( 1∶5000 ) to visualize cell nuclei . To stain intracellular compartments , rabbit polyclonal antibodies specific for TGN46 ( Sigma-Aldrich , St Louis , MO ) , CD63 ( Santa Cruz Biotechnology , Santa Cruz , CA ) and LAMP-1 ( abcam , Cambridge , MA ) were used at a dilution of 1∶50 . Next , an Alexa-568 goat anti-rabbit ( Invitrogen , Grand Island , NY ) was used to detect these cellular markers . In this case , Nef staining was performed by using a secondary Alexa-633-conjugated goat anti-mouse IgG1 . After staining , the cells were washed and mounted on slides with antiquenching mounting-medium ( Vector Laboratories , Inc ) . Images were acquired using a Leica TCS SP5 II confocal microscope .
|
Tetherin ( BST-2 , CD317 or HM1 . 24 ) is an interferon-inducible cellular restriction factor that prevents the release of enveloped viruses from infected cells . Human and simian immunodeficiency viruses have evolved to use different viral proteins to overcome the anti-viral effects of tetherin . Whereas HIV-1 Vpu and HIV-2 Env counteract human tetherin , most SIVs use the accessory protein Nef to counteract tetherin in their non-human primate hosts . Here we show that the mechanism of tetherin antagonism by SIV Nef involves a direct physical interaction between the core domain of Nef and the cytoplasmic domain of tetherin , which results in the removal of tetherin from sites of virus assembly and release on the cell surface by a mechanism that depends on clathrin and dynamin 2 . The Nef-mediated internalization of tetherin leads to the accumulation of tetherin within lysosomal compartments , suggesting that , similar to CD4− and MHC I-downregulation , Nef promotes the lysosomal degradation of tetherin .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"medicine",
"infectious",
"diseases",
"immunology",
"biology",
"microbiology",
"molecular",
"cell",
"biology"
] |
2013
|
Tetherin/BST-2 Antagonism by Nef Depends on a Direct Physical Interaction between Nef and Tetherin, and on Clathrin-mediated Endocytosis
|
Macrophages display flexible activation states that range between pro-inflammatory ( classical activation ) and anti-inflammatory ( alternative activation ) . These macrophage polarization states contribute to a variety of organismal phenotypes such as tissue remodeling and susceptibility to infectious and inflammatory diseases . Several macrophage- or immune-related genes have been shown to modulate infectious and inflammatory disease pathogenesis . However , the potential role that differences in macrophage activation phenotypes play in modulating differences in susceptibility to infectious and inflammatory disease is just emerging . We integrated transcriptional profiling and linkage analyses to determine the genetic basis for the differential murine macrophage response to inflammatory stimuli and to infection with the obligate intracellular parasite Toxoplasma gondii . We show that specific transcriptional programs , defined by distinct genomic loci , modulate macrophage activation phenotypes . In addition , we show that the difference between AJ and C57BL/6J macrophages in controlling Toxoplasma growth after stimulation with interferon gamma and tumor necrosis factor alpha mapped to chromosome 3 , proximal to the Guanylate binding protein ( Gbp ) locus that is known to modulate the murine macrophage response to Toxoplasma . Using an shRNA-knockdown strategy , we show that the transcript levels of an RNA helicase , Ddx1 , regulates strain differences in the amount of nitric oxide produced by macrophage after stimulation with interferon gamma and tumor necrosis factor . Our results provide a template for discovering candidate genes that modulate macrophage-mediated complex traits .
At the cellular level , innate immune cells , such as macrophages , are central to the development and prevention of infectious diseases . On engagement of surface signaling receptors or pattern recognition receptors ( PRRs ) such as toll-like receptors ( TLRs ) , RIG-I-like receptors ( RLRs ) and the cytosolic NOD-like receptors ( NLRs ) , by immune factors such as cytokines or conserved microbial products , macrophages can assume different activation states . The most extreme of these states are the classical ( M1 , M ( IFNG ) ) and the alternative ( M2 , M ( IL-4 ) ) states , separated by several intermediate activation states [1–3] ( We are following the recently described macrophage activation phenotype nomenclature [1] ) . Ultimately , macrophage activation results in pathogen clearance by downstream antimicrobial effector mechanisms , such as inflammasome activation , or activation of adaptive immune responses [4–6] . Although the outcome of macrophage activation is dependent on the stimulus engaged by the PRRs , emerging empirical data , from both human and mouse studies , indicate that the macrophage genetic background also plays a significant role . The initiation of immune responses by macrophages can occur in the presence of pro-inflammatory cytokines such as interferon gamma ( IFNG ) , while anti-inflammatory cytokines such as interleukin ( IL ) -4 , and IL-13 , prime macrophages for the resolution of immune responses and tissue repair [7–9] . This macrophage ability to initiate and resolve immune responses , while important in regulating immunopathology , can be exploited by pathogens to evade macrophage-associated immunity [10] . Indeed , to disseminate in their hosts most pathogens circumvent macrophage-mediated microbicidal mechanisms by modulating macrophage signaling pathways and activation phenotypes [11–16] . In addition to destroying pathogens , activated macrophages are important mediators in several inflammatory pathologies , including atherosclerosis , diabetes and cancer [17] . Studies in mice have linked several macrophage- or immune-related genes , such as Nramp1/Slc11a1 , Icsbp1/Irf8 , Csfgm , and Nos2 , with the development of several infectious diseases , including salmonellosis , toxoplasmosis , and leishmaniasis [18–20] . Although the compendium of macrophage- or immune-related genes that modulate infectious disease pathogenesis is broad , the role of individual differences in macrophage activation phenotypes in determining individual differences in susceptibility to infectious disease is just emerging [21–25] . Furthermore , the genetic basis for individual differences in macrophage activation phenotypes has not been identified . Macrophage activation is likely modulated by complex gene and metabolite networks that cannot be defined one gene at a time , thus the difficulty in unraveling the genetics of macrophage activation . This hypothesis is reinforced by results from genetic perturbation experiments that have revealed multiple genes that individually modulate macrophage activation phenotypes , including IRF8 , PPARG and AKT2 [26 , 27] . Empirical data show that macrophages display distinct transcriptional programs in response to infectious and inflammatory stimuli [22] and that this macrophage response differs between genetically segregating individuals [22–25] . Our hypothesis is that inter-individual differences in susceptibility to infectious disease are partly due to genetic differences in the macrophage response to pathogens . Quantitative trait locus ( QTL ) analyses have been used to elucidate the complex genetic basis for many traits in humans and model organisms [28 , 29] . However , the region spanned by individual QTL is often large and encompasses multiple genes , making the transition from QTL to individual genes influencing disease ( quantitative trait gene , QTG ) difficult . It has been shown that differences in the abundance of certain transcripts can explain phenotypic variations between individuals [30 , 31] . Forward genetics approaches that combine traditional QTL mapping with expression quantitative trait mapping ( eQTL; in which case transcript abundance is the quantitative trait ) [32] are increasingly being used to successfully transition from QTL to QTG [33–35] . Traditional QTL analysis will identify the genomic regions affecting trait variation , while eQTL analysis can help in understanding which genes , pathways , and biological processes are also under the influence of a given QTL . By examining the relationship between transcript location , the location of the eQTL and the pleiotropic effects of the eQTL , tools of systems biology such as network and functional analysis can be used to further delineate the complex genetic interactions modulating complex traits and reconstruct genetic pathways that underlie such traits [33–35] . In this study , we used the AXB/BXA recombinant inbred ( RI ) mice to investigate the relationship between the macrophage genotype and their response to inflammatory stimuli or infection with Toxoplasma , an obligate intracellular protozoan parasite . The AXB/BXA mice are derived from an initial reciprocal cross of AJ ( A ) and C57BL/6J ( B ) mice followed by multiple rounds ( ≥20 ) of inbreeding resulting in a stable mosaic of blocks of the parental alleles in their genomes [36] . These mice have been used to investigate the development and susceptibility to a variety of infectious and inflammatory diseases [19 , 37–40] . Importantly , the parental strains , AJ and C57BL/6J , differ at key loci that regulate immune responses , including the complement 5 a ( C5a ) [41] and interleukin 3 receptor alpha ( Il3ra ) [25 , 42] . These mice also exhibit differences in the amount of IL-10 and tumor necrosis factor ( TNF ) produced in response to bacterial infection [43 , 44] . Furthermore , the parental AJ and C57BL/6J vary in susceptibility to a variety of pathogens , including Staphylococcus aureus , Toxoplasma gondii and Trypanosoma cruzi [38 , 44–47] . By linking QTL analyses of defined macrophage phenotypes and macrophage transcriptional profiles , captured by high-throughput RNA-sequencing , we have identified many loci that affect the macrophage response to inflammatory stimuli and infection . These loci could provide the foundations for further studies in identifying the genetic basis for the differences in susceptibility to inflammation and infection in these mice . As an example , we report that differences in nitric oxide ( NO ) production , in AJ and C57BL/6J macrophages is due to differences in the expression of the RNA helicase Ddx1 .
Although correlations between genetic variations in immune-related genes and the response to infectious and inflammatory stimuli have been reported in AJ and C57BL/6J ( B6 ) [19 , 43 , 44] , the role of specific immune cells in these phenotypic differences are mostly equivocal . Therefore , we stimulated AJ and B6 bone marrow-derived macrophages ( BMDM ) with interferon gamma ( IFNG ) and tumor necrosis alpha ( TNF ) ( IFNG+TNF ) or interleukin 4 ( IL4 ) . IFNG+TNF induces the classical ( M ( IFNG ) ) while IL-4 induces the alternative ( M ( IL-4 ) ) macrophage activation phenotypes . Additionally , to mimic activation by bacteria and pathogen-associated molecular patterns ( PAMPs ) , we stimulated the macrophages with lipopolysaccharide ( LPS ) ( a component of gram-negative bacteria ) , or CpG ( a synthetic oligonucleotide ) , respectively . A summary of the stimulation regimen and the corresponding phenotypes measured is shown in S1 Table . Next , we captured the macrophage response to the individual stimulus by measuring M ( IFNG ) and M ( IL-4 ) markers . For M ( IFNG ) markers , we measured the amount of nitric oxide ( NO ) and IL-12 , while for M ( IL-4 ) markers , we quantified the amount of urea ( a by-product of Arginase I enzyme activity ) , IL-10 and chemokine ( C-C motif ) ligand 22 ( CCL22 ) ( Fig 1 ) . Finally , because: 1 ) IFNG is indispensable in the resistance to Toxoplasma gondii [48] , 2 ) Toxoplasma is a master regulator of macrophage signaling pathways [14] , and 3 ) AJ and B6 mice segregate for susceptibility to Toxoplasma [38 , 49] , we infected non-stimulated or IFNG+TNF-stimulated BMDM with a strain of Toxoplasma engineered to express firefly luciferase [50] and assessed parasite growth by measuring luciferase activity , which is often used to approximate Toxoplasma burden in in vitro or in vivo infection models [49–52] . We observed high amounts of NO and CCL22 in AJ BMDM , while the B6 BMDM produced higher amounts of IL-12 , IL-10 and urea ( Fig 1A–1E ) . Despite producing high amounts of M ( IL-4 ) markers ( urea and IL-10 ) , the B6 BMDM also produced high amounts of an M ( IFNG ) marker ( IL-12 ) relative to AJ BMDM . Similarly , the high amount of NO , an M ( IFNG ) marker , produced by AJ BMDM was accompanied by high amounts of CCL22 , an M ( IL-4 ) marker . This dual expression of M ( IFNG ) and M ( IL-4 ) markers is perhaps indicative of complex molecular modulation of macrophage activation or the heterogeneity of macrophage activation phenotypes . Consistent with the known resistance of AJ mice to Toxoplasma relative to B6 mice [38 , 49] , and the divergent BMDM activation phenotypes , Toxoplasma growth in the IFNG+TNF-stimulated AJ BMDM was significantly reduced compared to its growth in B6 BMDM ( Fig 1F ) . Thus , the variable susceptibility to Toxoplasma in AJ versus B6 observed in vivo [38 , 49] can be recapitulated in vitro using AJ and B6 BMDM . As such , the AJ and B6 BMDM can be used to gain insight into the molecular mechanisms that underlie the differences in AJ versus B6 susceptibility to Toxoplasma . Together , the difference in macrophage response to IFNG+TNF stimulation , as evidenced by NO ( a key toxostatic effector [53 , 54] ) , and the differences in parasite growth in IFNG+TNF-stimulated BMDM , posit that the variable susceptibility to Toxoplasma between AJ and B6 mice is due to innate differences in the macrophage response to the parasite and/or IFNG+TNF . The differences in response to infectious and inflammatory stimuli in AJ and B6 mice have a genetic component [19 , 38 , 43] . Therefore , having established differences in parasite growth and activation phenotypes in AJ and B6 BMDM , we sought to establish the genetic basis for the differences in macrophage activation and toxoplasmacidal activity between AJ and B6 using the AXB/BXA mice . We stimulated BMDM obtained from 26 age-matched female AXB/BXA RI mice with IFNG+TNF , IL-4 , CpG , or LPS and measured the amount of NO , urea , IL-12 , IL-10 , and CCL22 produced . We also infected IFNG+TNF stimulated BMDM with Toxoplasma that express luciferase and measured relative parasite growth by measuring luciferase activity . These phenotypes exhibited a continuous distribution in the RI mice , which is characteristic of quantitative traits ( S1A–S1F Fig ) . Due to the stable and unique combination of blocks of parental alleles in their genomes , the AXB/BXA RI mice are particularly suited for quantitative trait locus ( QTL ) mapping [55] . Therefore , we used the AXB/BXA genetic map ( containing 934 informative genetic markers [56] ) in a genome-wide scan in R/qtl [57] to identify the genomic regions that modulate differences in AJ and B6 BMDM phenotypes . To correct for multiple testing on the 934 genetic markers , we performed 1000 permutation tests on the individuals relative to their phenotypes [58] to obtain adjusted p-values for each QTL ( Table 1 ) . Parasite growth in the IFNG+TNF-stimulated BMDM mapped to chromosome 3 ( 147 . 7 Mb ) , proximal to the Guanylate binding protein ( Gbp ) locus ( 142 . 6 Mb ) that is known to modulate the murine macrophage response to Toxoplasma [59–62] . Even though AJ and B6 BMDM display distinct polarization states following IFNG+TNF or IL-4 stimulation , except for IL-12 and CCL22 , we did not observe statistically significant QTL peaks for any of the activation phenotypes . Instead we observed a suggestive QTL for NO on chromosome 12 ( Table 1 ) and a second QTL for NO on chromosome 4 ( S2A Fig ) . The additive QTL for NO also mapped to chromosome 4 ( S2B Fig ) . Similar to observations in the parental BMDM , the AJ allele at the chromosome 12 QTL was associated with high amounts of NO ( S2C Fig ) . However , the AJ allele at the chromosome 4 QTL was associated with low amounts of NO ( S2D Fig ) . Although mapping in cis to Arg1 on chromosome 10 , the QTL for the amount of urea did not reach statistical significance . Nevertheless , the allele effect at the suggestive urea QTL was consistent with the parental allele effect on urea . Finally , we selected the top 2 QTL for each phenotype ( where there was marginal difference between the LOD scores for the second and third largest QTL , we picked both ) and grouped the BMDM based on the genotypes at each QTL . We then estimated the QTL inheritance pattern by comparing the values for the corresponding phenotype amongst the genotypes using a one-way ANOVA with Tukey’s-Post-test ( StatPlus , AnalystSoft Inc ) [63] ( Table 2 ) . Discrete transcriptional programs modulate the response of immune cells , including macrophages , to stimuli such as pathogens and immune factors [24 , 31 , 33] . As such transcriptional profiles can be used to gain insight into the intricate and incipient molecular networks that modulate complex traits , such as macrophage activation [24 , 34 , 64 , 65] . Consequently , we investigated whether specific transcriptional programs modulate the differential activation of AJ and B6 BMDM . To do this , we performed high throughput RNA-sequencing ( RNA-seq ) on BMDM obtained from the same 26 age- and sex-matched AXB/BXA mice described above and their progenitors ( 28 samples in total ) , before ( resting , controls ) and after infection with Toxoplasma or stimulation with IFNG+TNF , or CpG . For each sample we generated at least 100 million paired-end reads , except for the IFNG+TNF-stimulated BMDM samples that were sequenced on a single end . Although all the samples were sequenced once , due to the unique recombination of the parental alleles and homozygosity at each of the informative genetic markers ( 934 ) , there are at least 4 replicates ( the minimum number of mice having the same allele at each marker ) for each marker . Thereafter , we processed the RNA-seq data as previously described [25] . Briefly , we aligned the RNA-seq reads to the mouse reference genome ( NCBI build GRcm38 , downloaded from Illumina iGenomes; https://support . illumina . com/sequencing/sequencing_software/igenome . html ) using TopHat [66] . To avoid read alignment bias due to sequence polymorphisms between AJ and B6 genomes , we made a synthetic reference genome in which all the polymorphic nucleotides between AJ and B6 were converted to a neutral nucleotide [67] . However , and consistent with a recent report [68] , allele bias did not significantly affect read alignment to the genome . On average , about 70% of reads in each sample uniquely mapped to the synthetic genome , which was about 1% less than the number of reads uniquely aligned to the iGenome . Henceforth , unless otherwise stated , all RNA-seq data presented herein were processed using the synthetic genome . Transcript abundance was estimated using Cufflinks [69] and reported as fragment per kilobase exon per million reads ( FPKM ) . Next , using the FPKM values from each of the 26 RI BMDM and the corresponding AXB/BXA genetic map , we mapped the genomic loci that modulate gene expression ( expression QTL , eQTL ) [32] using R/qtl [57] . As previously described [25] , we performed 1000 permutations to correct for multiple testing across the 934 informative genetic markers in the AXB/BXA cross . Next , we used a false discovery rate ( FDR ) ≤ 10% , calculated in the qvalue package [70] , to correct for multiple testing on the transcripts and to nominate significant eQTL . Finally , to allow for meaningful comparisons , we only included in the downstream analyses , for each stimulation condition , eQTL for transcripts with an average FPKM ≥ 5 across the 26 RI BMDM . The linkage analyses and the subsequent filtering steps were performed separately in resting , IFNG+TNF-stimulated , CpG-stimulated , and Toxoplasma-infected BMDM and identified , 131 , 367 , 688 and 1008 significant eQTL , respectively , dispersed throughout the genome ( Fig 2A and 2B and S1A–S1D Dataset ) . Thus , consistent with previous studies [24 , 71] , the genetic background influences macrophage gene expression profile . Importantly the different stimuli induced transcriptional programs that were modulated by distinct eQTL hotspots , which can be exploited to unravel the complex molecular networks that regulate macrophage activation states . Variable gene expression can be due to sequence or structural variations close to ( cis ) or further removed from ( trans ) the gene itself , such as polymorphism in the promoter regions or at a distal transcription factor , respectively . Consequently , relative to the physical location of the corresponding gene , an eQTL can be categorized as either cis or trans . Thus , we designated eQTL that co-localized within a 10 Mb genomic window with the corresponding gene as cis and all other eQTL as trans [72] . Except for the Toxoplasma-infected BMDM ( 407 cis vs . 601 trans ) , most of the eQTL were located in cis in: resting ( 99 cis vs . 32 trans ) , IFNG+TNF-stimulated ( 194 cis vs . 173 trans ) , and CpG-stimulated ( 482 cis vs . 206 trans ) BMDM ( S1A–S1D Dataset ) . Suppose a common locus was to modulate the expression of multiple genes in trans , then in linkage analysis , we should expect the eQTL for these genes to co-localize in the vicinity of the common locus forming a trans-eQTL hotspot ( trans-band ) . Indeed , similar to previous studies [23 , 24 , 73] , and indicative of a common variant regulating the expression of multiple genes in trans , we detected trans-eQTL hotspots , within a 10 Mb window , in all the samples ( Fig 2A and 2B and S1A–S1D Dataset ) . Because it is possible for eQTL to co-localize by chance alone , we used Bonferroni-corrected p-values and Poisson distribution to compute the number of trans-eQTL that can co-localize in a 10 Mb genomic window by chance . Using these cutoffs , we identified 2 , 3 , 5 , and 15 trans-eQTL hotspots in the resting , IFNG+TNF-stimulated , CpG-stimulated , and Toxoplasma-infected BMDM , respectively ( S1A–S1D Dataset ) . Previously , it was reported that eQTL that localize close to the physical location of the relevant gene ( cis-eQTL ) are a consequence of single nucleotide polymorphisms ( SNPs ) [72] or larger structural variants ( SV ) , such as insertions and deletions . Thus , we investigated the nucleotide sequence in a 2000 bp window upstream and downstream of the transcription start site ( TSS ) of all significant eQTL . Consistent with these reports , we found that most cis-eQTL were associated with genes reported to have structural or sequence variations within 2000 bp upstream or downstream of their TSS [74 , 75] . That is , in control BMDM we found 89 out of 99 ( Hypergeometric test P≤2 . 3e-7 ) ; in the IFNG+TNF-stimulated BMDM we found 156 out of 196 ( Hypergeometric test P≤3 . 3e-10 ) ; in the CpG-stimulated BMDM we found 422 out of 482 ( Hypergeometric test P≤3 . 8e-23 ) ; and in the Toxoplasma-infected BMDM we found 369 out of 407 ( Hypergeometric test P≤1 . 6e-49 ) cis-eQTL with polymorphisms within 2000bp upstream or downstream of TSS . These included genes with known immunological functions such as Gbp1 , Gbp2 , Irak4 and Srebf1 . As indicated above , trans-eQTL hotspots can be due to transcriptional regulation of several genes by a common genetic variant [76] , such as a polymorphic or differentially expressed transcription factor , enhancer or repressor . Alternatively , a trans-band can be a result of a differentially expressed or polymorphic signaling protein , such as a cell surface receptor , that could lead to differential activation of transcription factors and the genes downstream of these transcription factors . Consequently , the eQTL localizing at the trans-eQTL hotspot are likely to be enriched for binding sites for a common transcription factor ( s ) physically located at the trans-eQTL-hotspot , biological function , or signaling pathway . Therefore , we used gene ontology ( GO ) [77] and rVISTA [78] to functionally characterize and search for transcription factor binding site enrichment in each of the eQTL in the different trans-hotspots ( Table 3 ) . We found that the three largest trans-bands in the Toxoplasma infected BMDM mapped to loci containing genes with either known or putative roles in macrophage inflammatory or metabolic processes , which are important host responses against intracellular pathogens . For instance the trans-band on chromosome 13 ( 117 . 1–127 . 8 Mb ) overlapped docking protein 3 ( Dok3 ) , chemokine ( C-X-C motif ) ligand 14 ( Cxcl14 ) , and AU RNA binding protein/enoyl-coenzyme A hydratase ( Auh ) , which are known to regulate various aspects of cellular inflammatory and metabolic processes [79–81] . Indeed , this trans-band was enriched for , among others , “natural killer cell mediated immunity” . Similarly , the trans-band in the IFNG+TNF-stimulated BMDM on chromosome 15 ( 89 . 6–96 . 6 Mb ) , which was enriched for “regulation of leukocyte mediated immunity” , overlapped the interleukin-1 receptor-associated kinase 4 ( Irak4 ) , known to regulate the immune response to a variety of infectious and inflammatory stimuli [18 , 82 , 83] . Furthermore , Irak4 contains genetic insertions and deletions in AJ relative to the reference B6 mouse strain [84] . Thus , the trans-bands are functionally enriched in biologically relevant processes and can potentially reveal novel regulators and insight in the complex gene interactions that modulate macrophage response to exogenous stimuli . To identify putative regulators for the trans-bands , which can be variable transcription factors or signal transducers , we identified genes that were expressed ( average FPKM ≥5 ) , were physically located within 10 Mb on either side of the trans-eQTL hotspot , and were differentially expressed or had non-synonymous ( NS ) polymorphisms in AJ versus B6 mice . As an example , the chromosome 15 trans-band ( between 79 . 7–99 . 7 Mb ) contained 81 expressed coding and non-coding genes , 5 of which contained non-synonymous SNPs and 8 exhibited differential expression that mapped in cis . Of these genes , we considered Plxnb2 , Irak4 , and Apobec3 to be good candidates since they are known to be involved in immune response pathways [82 , 85 , 86] , similar to the functional enrichment observed for the chromosome 15 trans-band . Additionally , these genes are polymorphic in AJ compared to B6 . Due to insertions and deletions [87] and its function as an immune signaling adaptor , we considered Irak4 to be a strong candidate regulator for the chromosome 15 trans-band . Hence , we used shRNA to knockdown Irak4 in the IFNG+TNF-stimulated macrophages . Enrichment analysis on the perturbed genes following Irak4 knockdown revealed an overrepresentation ( p = 0 . 005 ) of several chromosome 15 trans-eQTL ( Table 4 ) . shRNA-knockdown of the other putative candidate genes identified in this study did not perturbed most of the chromosome 15 trans-band eQTL ( S1E Dataset ) , indicative of a specific effect of Irak4 knockdown on this trans-band . Transcriptional networks capture the connectivity between genes modulating complex phenotypes and may provide a means to unravel the molecular mechanisms underlying complex traits . Because cis genetic variants account only for the phenotypes related to the gene they modulate , we reasoned that they are not good prototypes to illustrate how transcriptional , linkage and network analyses can be leveraged to systematically elucidate the genetic basis of a complex trait . Therefore , we used trans genetic variants , which potentially modulate multiple phenotypes , and followed a step-wise procedure [34] to identify the relationship between transcript levels , QTL , and BMDM phenotypes . First , to gain insight into the transcriptional architecture that modulate macrophage response to stimuli , we constructed gene co-expression network modules for each macrophage stimulation condition using the topological overlap matrix ( TOM ) in the weighed gene co-expression network analysis ( WGCNA ) program [88] . Next , using the eigenvalues for each module , we made correlations between each module and macrophage phenotypes . As proof of principle , we used this approach to nominate candidate genes that modulate the amount of NO produced in IFNG+TNF-stimulated BMDM . Because the amount of NO varied in the BMDM after IFNG+TNF , we correlated the co-expression modules with the amount of NO produced in the IFNG+TNF-stimulated BMDM ( S3 Fig ) . Subsequently , we identified 4 modules ( identified as white , light yellow , blue , and tan ) that showed significant ( P ≤ 0 . 01 ) correlation with NO levels ( S3 Fig ) . It is important to note that , apart from being arbitrary identifiers for each module , the color code used to name each module reveals no further information . Each module , however , contains co-expressed genes ( i . e . genes that exhibit transcriptional correlation across the 26 IFNG+TNF-stimulated BMDM ) . Of the 4 modules , the “white” module showed the greatest association with NO , hence we used it to illustrate our approach . Because the module-trait relationship is based on the correlation of the module eigenvalue with the amount of NO , not all the genes in the module will show significant association with the trait ( amount of NO produced in the BMDM ) . Therefore , we further filtered the genes in the “white” module based on their individual relationship with NO ( gene significance ) ( S1F Dataset ) . Expectedly , the eQTL for most of the genes in the module were localized on chromosome 12 and chromosome 4 , the locations for NO QTL . To identify the potential regulator for cellular amounts of NO , we reasoned that if two genetic traits are both modulated by the same genetic variant , then the QTL for the two traits will co-localize at the common genetic variant [34] . Therefore , to further narrow down the significant genes and to identify the transcriptional network that likely modulate genetic differences in cellular amounts of NO produced after IFNG+TNF-stimulation , we searched for genes with eQTL that overlapped with the NO QTL on chromosome 12 ( 4 . 4–13 . 1 Mb ) and found 11 eQTL ( S1F Dataset ) . These eQTL were functionally enriched for “oxidoreductase activity” ( p = 8 . 322e-04 ) and “nitrogen compound metabolism” ( p = 1 . 345e-03 ) in gene ontology . Because the common regulator for both NO levels and the chromosome 12 trans-band can be a polymorphic or differentially expressed transcription factor or signaling receptor physically located at the chromosome 12 trans-band locus , we narrowed the putative regulators by searching for cis-eQTL or expressed polymorphic genes at the chromosome 12 locus . Ddx1 , a transcriptional regulator of cell cycle , maps in cis at this locus , and was considered a putative candidate for NO . Additionally , we used the Transcriptional Regulation Inference from Genetics of Gene Expression ( Trigger ) program [89] to establish the causal relationship between NO and the genes on chromosome 12 , including Ddx1 . Trigger utilizes randomized genetic backgrounds and phenotypes to test for causality between phenotypes that are linked to the same locus e . g . Ddx1 and NO . Of the chromosome 12 genes , only Ddx1 exhibited a minimum p-value ( ≤0 . 05 ) for causal relationship . As expected , the reciprocal analysis did not show NO or any of the chromosome 12 trans-eQTL as causal for Ddx1 differential expression . Indeed , shRNA-mediated knockdown of Ddx1 in immortalized B6 macrophages resulted in an increase in the amount of NO produced after IFNG+TNF-stimulation ( Fig 3 ) . Similar to NO , Ddx1 transcript abundance in IFNG+TNF-stimulated macrophages mapped to two loci on chromosome 4 and 12 . However , Ddx1 transcript abundance exhibited an inverse relationship with the amount of macrophage NO at both loci ( S4A–S4E Fig ) . Thus , we concluded that the expression of Ddx1 , which is higher in B6 compared to AJ BMDM , inhibited NO production in the B6 macrophages . Due to the important role macrophages play in the pathology of various intracellular pathogens [90] , we investigated whether there were overlaps between our trans-eQTL hotspots and other disease QTL in the AXB/BXA RI mice available in webQTL [91] and found several ( S1A–S1D Dataset ) . For example the QTL for Listeria monocytogenes proliferation , which is attributed to macrophage inflammatory response [92] , and susceptibility to hepatitis virus [93] overlapped the trans-eQTL hotspots on chromosome 15 and 7 , respectively , in the IFNG+TNF-stimulated macrophages . Considering that IFNG+TNF is important in the pathogenesis of Listeria and Hepatitis infections [94–96] , it is likely that these trans-bands harbor genes that modulate the outcome of infection with these and other pathogens .
The activation of macrophages , in response to stimuli such as microbial components or immune factors , into the broadly defined classical , M ( IFNG ) , and alternative , M ( IL-4 ) , phenotypes , determines whether they initiate or resolve immune responses . As gatekeepers against invading pathogens , the macrophage activation phenotype is essential in determining the persistence or resolution of infection . Thus , the hypothesis is that for infection to persist , the pathogen has to “trick” host macrophages to assume the “wrong” activation state and that variation in susceptibility to infection between hosts is due to genetic differences in macrophage response to the pathogen . Indeed empirical evidence shows that many pathogens can alter the macrophage activation to a phenotype that is favorable for its replication and persistence . For example , virulent strains of the intracellular parasite Toxoplasma gondii , which is vulnerable to M ( IFNG ) macrophages , induces the M ( IL-4 ) phenotype in macrophages [12 , 14 , 97] . Similar observations have been made in leishmaniasis , in which the extent of macrophage modulation is dependent on the host [98] . On the other hand , macrophages from genetically segregating hosts have been shown to exhibit differential activation states , as captured by transcriptional and cytokine profiles [21 , 71 , 99–101] . Despite these documented pathogen-macrophage interplays and the variable inter-host macrophage activation phenotypes , the genetic basis for individual differences in macrophage polarization is just emerging . We describe differential activation of macrophages from the genetically segregating AJ and C57BL/6J ( B6 ) mouse strains , which we have linked to the variable response to the obligate intracellular parasite , Toxoplasma gondii . Our in vitro model , which obviates the interference by other immune cells and involves naïve bone-marrow derived macrophages as opposed to elicited peritoneal macrophages , indicates that when stimulated with equal amounts of cytokines , AJ-derived macrophages produce a stronger M ( IFNG ) phenotype relative to B6 ( as measured by NO ) . Additionally , we used a panel of genetically diverse recombinant inbred ( RI ) mice , derived from AJ and B6 mice , to investigate the specific loci responsible for this variable macrophage activation and the associated phenotypes . In mice , nitric oxide ( NO ) production and L-Arginine metabolism are often used to define the M ( IFNG ) and M ( IL-4 ) macrophage activation phenotypes , respectively [100 , 102] . While the cellular levels of these factors vary between genetically divergent individuals and are known to contribute to the differential response to infection [101 , 103 , 104] , the genetic loci that predispose macrophages to either the M ( IFNG ) or M ( IL-4 ) phenotypes are not known . The general assumption is that a single locus will determine whether individual macrophages assume the M ( IFNG ) or M ( IL-4 ) state . Results from the current study contradict this assumption and suggest that macrophage activation , as measured by levels of NO , IL-12 , CCL22 , IL-10 and Arginase I activity , is modulated by several loci , some with antagonistic effects . While the limited number of mice available from AXB/BXA RI line may partly explain the lack of statistically significant QTL for most of the macrophage phenotypes in this study , we submit that the genetic factors that modulate macrophage polarization form complex interaction networks that are not linked to a single locus . This conclusion is supported by the identification of two loci with contrasting effect on NO levels , the high levels of IL-12 and urea in B6 BMDM , and the high levels of NO and CCL22 in AJ BMDM . Furthermore , our observation of significant QTL peaks for IL-12 and CCL22 portends that the complex regulation of other phenotypes , such as NO , rather than the few number of mice , maybe the reason for the insignificant QTL peaks . Apart from the individual macrophage activation phenotypes , we did not observe the convergence of all the M ( IFNG ) ( IL12 , NO ) at a single locus and the M ( IL-4 ) phenotypes ( IL10 , CCL22 and urea ) at another locus , instead each phenotype localized to a unique locus , again reinforcing the complex molecular circuits that modulate the M ( IFNG ) and M ( IL-4 ) macrophage states . Although , the B6 mice carry a dominant negative mutation in the Slc7a2 gene , which is involved in L-Arginine transport and is postulated to contribute to the differential metabolism of L-Arginine in B6 relative to other mouse strains such as BALB/c [105] , the QTL for L-Arginine metabolism did not localize to chromosome 8 , which is the physical location of Slc7a2 . Instead , the QTL for urea ( a measure of Arginase activity ) mapped to chromosome 10 , proximal to Arg1 , while the QTL for NO mapped to chromosome 12 , respectively . It is plausible that both the transport and metabolism of L-Arginine contribute to the difference in urea and NO production in AJ and B6 , hence the lack of significant QTL for these traits . Modulation of host cellular signaling and transcriptional pathways by Toxoplasma is known to aid in immune evasion by the parasite and is achieved via the secretion of polymorphic effector proteins localized in the rhoptry and dense granule organelles [37 , 106–108] . Specifically , the secretion of the polymorphic dense granule protein ( GRA15 ) by the avirulent type II , and rhoptry kinase ( ROP16 ) by the virulent type I Toxoplasma strains , is known to elicit classical and alternative macrophage activation , respectively [12] . Additionally , the secretion of these two Toxoplasma effector proteins is known to modulate intestinal pathology in the susceptible B6 mice [37] . However , to the best of our knowledge , the potential role of macrophages in Toxoplasma-induced intestinal pathology has not been shown . Furthermore , even though AJ and B6 are known to diverge in their response to in vivo Toxoplasma infection phenotypes , there has been no study showing that this variable response is due to a differential response of their macrophages to either IFNG or to the parasite itself . Together , the current study and our previous work [49] , provide compelling evidence that macrophages may play an important role in Toxoplasma pathology . We postulate that alternative macrophage activation by Toxoplasma [12] , and the differential AJ and B6 macrophage response to Toxoplasma and IFNG+TNF , provide the intersection of host-parasite interaction that harbors candidate genes mediating murine toxoplasmosis . Toxoplasma virulence appears to be related to its ability to skew macrophages towards alternative activation , which is abetted in susceptible animals , such as B6 mice . In conclusion , our findings provide an extensive genetic analysis of the macrophage signaling processes in response to exogenous stimuli . Because activation of macrophages by IFNG and/or TNF confers resistance to a wide range of intracellular pathogens and human diseases , and because susceptibility loci for some of these phenotypes overlap , it is expected that this study will provide a framework to help identify candidate genes that mediate some of these disease phenotypes .
All animal experiments were performed in strict accordance with National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Animal Welfare Act . The Massachusetts Institute of Technology Committee on Animal Care ( assurance number A 3125–01 ) approved all protocols . All mice were maintained in specific pathogen-free conditions and euthanasia was performed in controlled CO2 chamber as approved by the MIT Animal Care Committee . Bone marrow-derived macrophages ( BMDM ) were obtained from 6–8 weeks old AJ , C57BL/6J and 26 female AXB/BXA recombinant inbred mice ( Jackson Laboratories ) . Marrow cells were obtained from each mouse by flushing the femur and tibia with cold phosphate buffered saline ( PBS; GIBCO-Invitrogen ) . The cells were then centrifuged at 500 x g for 5 minutes at 4°C and re-suspended in 4 ml red cell lysis buffer ( Sigma ) and incubated on ice for 5 minutes . Next , the cells were passed through a 70 μm cell strainer ( BD Biosciences ) and centrifuged at 500 x g for 5 minutes at 4 °C . The cells from each mouse were subsequently grown on four 10 cm non-tissue culture petri dishes ( Corning ) in Dulbecco's modified Eagle's medium ( DMEM; GIBCO-Invitrogen ) supplemented with 10% heat-inactivated fetal bovine serum ( FBS; HyClone ) , 2 mM L-glutamine , 1 mM sodium pyruvate , 1X MEM nonessential amino acids , and 50 μg/ml each of penicillin and streptomycin . To differentiate the cells into macrophages , the DMEM was conditioned with 20% L929 cell supernatant ( containing GM-CSF ( 40 ng/ml ) , hereafter 20% L929 ) . After incubating the cells at 37°C and 5% CO2 for 3 days , the non-adherent cells were pipetted into 50 ml tubes and centrifuged at 500 x g for 5 minutes at 4°C and seeded in new 10 cm petri dishes . Simultaneously , the old 10 cm petri dishes were topped up with fresh media supplemented with 20% L929 i . e . in the end , for each mouse there were 8 petri dishes . After further incubation at 37°C and 5% CO2 for 4 days , the BMDM were harvested and stored in liquid nitrogen ( 5 million cells/ aliquot ) . The BMDM yield for each mouse strain ranged between 60–150 million cells , with the B6 mouse consistently producing more cells . This protocol has previously been shown to yield pure ( >99% ) macrophages [12] . A Pru ( type II ) Toxoplasma gondii strain engineered to express firefly luciferase and GFP ( Pru ΔHXGPRT A7 ) [109] , maintained in the laboratory by serial passage on Human Foreskin Fibroblasts ( HFF ) , was used for all infections . To immortalize macrophages , we used J2 recombinant retrovirus [110] produced from ψCREJ2 cells ( a generous gift from John MacMicking , Yale University School of Medicine ) . The J2-expressing cells were grown to confluency in DMEM medium supplemented with 10% FBS ( D10 ) . The medium containing retroviral particles was collected and passed through 0 . 45 μM low protein-binding filters ( Millipore ) . In parallel , 5 x 106 primary BMDM from AJ and C57BL/6J , obtained as described above , were thawed and grown for 2 days in DMEM medium supplemented with 10% FBS and 20% L929 . After 2 days , the medium was replaced with the filtered J2-retrovirus-containing medium supplemented with 50% L929 . After 24 hrs , the media was replaced with fresh D10 medium supplemented with 25% L929 . Media was subsequently changed after every 24hrs with concomitant reduction in L929 until 10% L929 concentration was reached . Immortalized cells were harvested and stored in liquid nitrogen until use . Unless otherwise stated , a total of 104 /well immortalized BMDM ( iBMDM ) or 105 /well primary BMDM were used in all the in vitro assays . Unless stated otherwise , before stimulation or infection , the primary BMDM or iBMDM were seeded overnight in D10 supplemented with 20% L929 . Cellular nitric oxide levels were measured using the Griess reagent procedure on supernatant from non-stimulated or stimulated cells . Arginase activity was measured by quantifying the amount of urea as previously described [111] . Briefly , L-arginine was added to cell lysates and incubated at 37°C . After 1 hour , 175μl of an acid mixture containing sulfuric acid/phosphoric acid/water ( H2SO4/H3PO4/H2O ) in a 1:3:7 ratio , was added to each well to stop the enzymatic reaction . Urea was quantified calorimetrically at 540 nm after adding 1 . 25 μl of 1-phenyl-1 , 2-propanedione-2-oxime ( ISPF ) and heating at 95°C for 30–60 minutes . This procedure abrogates interference from other metabolites generated , such as L-citrulline [111] . IL10 , and IL12 were measured on the relevant cell supernatants using ELISA kits as previously described [12] . For parasite growth assay , cells were either left unstimulated ( control ) or stimulated with recombinant mouse IFNG ( 100 ng/ml , Peprotech ) and TNF ( 100 ng/ml , AbD serotec ) for ~18 hr . The supernatant was removed for nitric oxide assay and replaced with D10 containing Toxoplasma at an MOI ~1 . The parasites were allowed to infect and replicate for 24 hrs before luciferase activity was measured using a luciferase assay kit ( Promega ) according to the manufacturer recommendations . Primary BMDM were plated ( 3 x106 ) overnight before stimulation or infection . For the stimulated samples , IFNG ( 100 ng/ml ) and TNF ( 100 ng/ml ) were added to each well for 18 hrs , while for the infected samples , a type II strain of Toxoplasma ( Pru ) was added to the confluent BMDM at an MOI of 1 . 3 for 8 hrs . Total RNA ( Qiagen RNeasy Plus kit ) was then isolated from the non-stimulated and non-infected cells ( controls plated overnight ) , stimulated , and infected cells and the integrity , size , and concentration of RNA checked ( Agilent 2100 Bioanalyser ) . The mRNA was then purified by polyA-tail enrichment ( Dynabeads mRNA Purification Kit; Invitrogen ) , fragmented into 200–400 base-pairs , and reverse transcribed into cDNA before Illumina sequencing adapters were added to each end . Libraries were barcoded , multiplexed into 4 samples per sequencing lane in the Illumina HiSeq 2000 , and sequenced from both ends resulting in 40 bp reads after discarding the barcodes . Our preliminary RNA-seq experiments with infected BMDM have shown that with 4 samples per lane , we still obtain enough read coverage for reliable gene expression analysis . Reads were initially mapped to the mouse genome ( mm9 ) and the Toxoplasma ( ME49 v8 . 2 ) genome using Bowtie ( 2 . 0 . 2 ) [112] and Tophat ( v2 . 0 . 4 ) [66] . We then estimated gene expression levels in cufflinks ( v2 . 0 . 0 ) [113] using the Illumina iGenomes refseq genome annotation ( NCBI build 37 . 2 ) with the multi-read , compatible-hits corrections and upper quantile normalization options enabled . Because the reference genome to which we mapped the RNA-seq reads is based on the C57BL/6J genomic sequence , and due to the known polymorphisms between the AJ and C57BL/6J , we suspected that biases introduced at the read mapping stage might affect our expression results . To mitigate this potential bias towards the reference allele , we created a copy of the mouse genome in which all the known single nucleotide polymorphisms ( SNPs ) between AJ and B6 , as annotated by the Wellcome Trust Sanger Institute sequencing ( ftp://ftp-mouse . sanger . ac . uk/current_snps/ ) , were converted to a third ( neutral ) nucleotide that is different from both the reference and AJ allele [67] . However , this did not substantially change the average proportion of uniquely mapped reads or expression profiles of individual genes in all the samples . In the end we used the mapping data generated from the synthetic genome to quantify gene expression levels . To map QTL , we used 934 AXB/BXA genetic informative markers obtained from http://www . genenetwork . org . For all the in vitro measurements and gene expression linkage analysis , a genome-wide scan was performed using R/qtl [57] . Significance of QTL logarithm-of-odds ( LOD ) scores was assessed using 1000 permutations of the phenotype data [114] and the corresponding p-values reported . For the cellular phenotypes , QTL significance was reported at a genome-wide threshold corresponding to p < 0 . 05 . However , for eQTL mapping , we further corrected for multiple testing on the multiple transcripts by using the p-values to estimate false discovery rate ( FDR ) in the qvalue package [115] and reported significant eQTL at FDR ≤ 10% . To identify cis- and trans- eQTL , we computed the distance from the position of the eQTL and the start of the physical location of the corresponding gene and designated any eQTL located <10 Mb from the corresponding gene as cis , otherwise trans eQTL . The procedure used to determine trans-bands has previously been described [25] . We used shRNA to probe for functional or regulatory significance of some of the candidates identified in our analysis . To do this , we used C57BL/6J immortalized bone marrow-derived macrophages , described above . One day after plating , we added shRNA constructs containing a puromycin resistance marker ( RNAi Platform , Broad Institute ) in the presence of 8 μg/ml polybrene , followed by centrifugation at 800 x g for 2 hrs at 37°C . At the end of the spinfection , the cells were incubated for an additional 24 hrs at 37°C in 5% CO2 . The cells were then grown in fresh cell culture medium for an additional 24 hrs before adding 4 μg/ml puromycin . Transcript knockdown was measured by quantitative reverse transcriptase polymerase chain reaction ( qRT-PCR ) using the KAPA SYBR FAST Universal 2X qPCR Master Mix ( KAPA Biosystems ) on a LightCycler qPCR instrument ( Roche ) . Fold knockdown was measured using the 2 delta-delta method [116] relative to LacZ-shRNA transduced cells . The puromycin-selected cells were either left stimulated or stimulated with IFNG+TNF , as described above , and the cell supernatant collected for nitric oxide assay . The microarray data is available at the NCBI Gene Expression Omnibus archive under accession number GSE47046 .
|
Macrophages provide a first line of defense against invading pathogens and play an important role in the initiation and resolution of immune responses . When in contact with pathogens or immune factors , such as cytokines , macrophages assume activation states that range between pro-inflammatory ( classical activation ) and anti-inflammatory ( alternative activation ) . Even though it is known that macrophages from different individuals are biased towards one of the various activation states , the genetic factors that define individual differences in macrophage activation are not fully understood . Additionally , although macrophages are important in infectious disease pathogenesis , how individual differences in macrophage activation contribute to individual differences in susceptibility to infectious disease is just emerging . We used macrophages from genetically segregating mice to show that discrete transcriptional programs , which are modulated by specific genomic regions , modulate differences in macrophage activation . Murine macrophages differences in controlling Toxoplasma growth mapped to chromosome 3 , proximal to the Guanylate binding protein ( Gbp ) locus that is known to modulate the murine macrophage response to Toxoplasma . Using a shRNA-mediated knockdown approach , we show that the DEAD box polypeptide 1 ( Ddx1 ) modulates nitric oxide production in macrophages stimulated with interferon gamma and tumor necrosis factor . These findings are a step towards the identification of genes that regulate macrophage phenotypes and disease outcome .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"Methods"
] |
[] |
2015
|
Transcriptional and Linkage Analyses Identify Loci that Mediate the Differential Macrophage Response to Inflammatory Stimuli and Infection
|
Human Endogenous Retrovirus type K ( HERV-K ) is the only HERV known to be insertionally polymorphic; not all individuals have a retrovirus at a specific genomic location . It is possible that HERV-Ks contribute to human disease because people differ in both number and genomic location of these retroviruses . Indeed viral transcripts , proteins , and antibody against HERV-K are detected in cancers , auto-immune , and neurodegenerative diseases . However , attempts to link a polymorphic HERV-K with any disease have been frustrated in part because population prevalence of HERV-K provirus at each polymorphic site is lacking and it is challenging to identify closely related elements such as HERV-K from short read sequence data . We present an integrated and computationally robust approach that uses whole genome short read data to determine the occupation status at all sites reported to contain a HERV-K provirus . Our method estimates the proportion of fixed length genomic sequence ( k-mers ) from whole genome sequence data matching a reference set of k-mers unique to each HERV-K locus and applies mixture model-based clustering of these values to account for low depth sequence data . Our analysis of 1000 Genomes Project Data ( KGP ) reveals numerous differences among the five KGP super-populations in the prevalence of individual and co-occurring HERV-K proviruses; we provide a visualization tool to easily depict the proportion of the KGP populations with any combination of polymorphic HERV-K provirus . Further , because HERV-K is insertionally polymorphic , the genome burden of known polymorphic HERV-K is variable in humans; this burden is lowest in East Asian ( EAS ) individuals . Our study identifies population-specific sequence variation for HERV-K proviruses at several loci . We expect these resources will advance research on HERV-K contributions to human diseases .
Endogenous retroviruses ( ERVs ) are derived from infectious retroviruses that integrated into a host germ cell at some time in the evolutionary history of a species [1–5] . ERVs in humans ( HERVs ) comprise up to 8% of the genome and have contributed important functions to their host [6–8] . The infection events that resulted in the contemporary profile of HERVs occurred prior to emergence of modern humans so most HERVs are fixed in human populations and those of closely related primates . However some HERVs are still transcriptionally active and capable of causing new germline insertions so that individuals differ in the number and genomic location occupied by an ERV , a situation termed insertional polymorphism [9–11] . Among all families of HERVs , HERV-K is the only one known to be insertionally polymorphic in humans . However , HERV-K genomes are closely related and as with many repetitive elements , they are difficult to accurately assign to a genomic location using standard mapping approaches [12 , 13] . The DNA form of a retrovirus is called a provirus and minimally encodes the structural gag and env gene , and genes for a protease and polymerase , termed pol . Viral genes are flanked by long terminal repeats ( 5’ or 3’ LTR ) . While there are several HERV-K that are full length , none are infectious and most contain mutations or deletions that affect the open reading frames or truncate the virus . Further , the LTRs are substrates for homologous recombination , which deletes virus genes while retaining a single , or solo , LTR at the integration site [14–16] . Insertional polymorphism typically refers to the presence or absence of a retrovirus at a specific locus [17 , 18] . However an occupied site can contain a provirus in some individuals and a solo LTR in others and hence still display polymorphism . Thus HERV-K and other HERVs have contributed to genomic diversity in the global human population in several ways [19] . The presence of antibodies to HERV proteins or HERV transcripts has spurred a quest to determine if HERVs from multiple families have a role in either proliferative or degenerative diseases in humans [20–26] . Although there are known mechanisms by which a HERV can cause disease; for example , by inducing genome structural variation through recombination [27–31] , affecting host gene expression [32] , and inappropriate activation of an immune response by viral RNA or proteins [23] , it has been difficult to establish an etiological role of a HERV in any disease . HERV-K specifically has been associated with breast and other cancers [3 , 33–37] , and autoimmune diseases , such as rheumatoid arthritis [38 , 39] , multiple sclerosis [22 , 40] and systemic lupus erythematosus [8 , 22 , 41] without definitive evidence of causality or of specific loci involved . Recently , a HERV-K envelope protein was shown to recapitulate the clinical and histological lesions characterizing Amyotropic Lateral Sclerosis [42 , 43] , providing an important mechanistic advance of a role for a HERV-K protein in a disease . Despite growing evidence for a contribution of HERV-K transcripts or proteins to the pathogenesis of human disease , it is difficult to distinguish among HERV-K loci to investigate potential roles and , in particular , to determine if a loci that is polymorphic for presence or absence of a provirus could be involved . In this paper , we focus on characterizing the genomic distribution of known insertionally polymorphic HERV-K proviruses in the 1000 Genomes Project ( KGP ) data . We present a data-mining tool and a statistical framework that accommodates low depth whole genome sequence data characteristic of the KGP—and often patient—data to estimate the presence or absence of a provirus at all loci currently known to contain a HERV-K provirus . Using these data , we determine the number of known polymorphic HERV-K proviruses per genome because HERV-Ks can affect genomic stability [44] contributing to the pathogenesis of a disease . We also provide a tool to visualize HERV-K co-occurrence in global populations to facilitate exploration of synergy that might exist among specific polymorphic HERV-K in disease [45] . Our results provide a reference of global population diversity in HERV-K proviruses at all currently known polymorphic loci in the human genome and demonstrate that there are notable differences in the prevalence of HERV-Ks in different global populations and in the total number of HERV-Ks currently known to be polymorphic within a person’s genome .
The goal of this research was to develop a computationally efficient and easy to use tool that could accurately report the status of all reported insertionally polymorphic HERV-Ks with coding potential ( provirus ) from whole genome sequence ( WGS ) data . We use the KGP database , which represents individuals in five super-populations and 26 populations , to establish the diversity in global populations at each known polymorphic HERV-K proviral locus and the total number of these polymorphic HERV-K in individual genomes to provide a foundation to study the role of HERV-K in human disease . Our reference set consists of all HERV-K sequences that are available in public databases and that can be unambiguously assigned a location in hg19 . Sequences of HERV-K that are not present in hg19 but that were generated by PCR primers to the host flanking regions are included in the reference HERV-K set . From these HERV-K reference sequences , we generate a set of k-mers ( see S2 Fig for optimizing k ) that are unique to all HERV-Ks at each locus . The analysis of subject data starts with a data mining step that recovers all whole genome sequence reads that map to identified HERV-K elements in hg19 . The rationale here is that polymorphic HERV-K that are not present in hg19 are greater than 80% homologous to those in the human reference genome and will map on existing elements . The recovered reads from a query WGS data set are then reduced to k-mers and mapped , requiring 100% match , to the reference set of k-mers ( T ) , which represents all unique sites for HERV-K at each locus . The output is a ratio ( n/T ) of subject k-mers ( n ) that are 100% match to the reference k-mers ( T ) ( see Methods for full details; the value of T for each HERV-K is in S1 Dataset:virus ) . Our preliminary analysis of the KGP data demonstrated that our k-mer-based approach is sensitive to sequence depth; some HERV-K loci are represented by an almost continuous range of n/T values from 0–1 ( S1 Fig ) , making presence/absence classification difficult . However , the majority of the KGP data is approximately 6x depth and thus to make use of this important resource , we developed a mixture model to statistically assign the n/T values from genomes to a cluster considering the sequence depth . K was optimized to 50 because this value improved our model computational efficiency and output ( Fig 1B , S1 Text , S2 Fig ) . The affect of sequence depth on n/T can be seen by comparing the sequence data of 28 individuals in the KGP data that have both low and high sequence depth data ( Fig 1 shows a subset of eight individuals for clarity ) . If read depth is greater than 20 , there is less dispersion of n/T values , most likely because more reads from the query WGS data are recovered from the mapped intervals . The states , ‘provirus’ , ‘solo LTR’ , and ‘absent’ are preliminarily assigned to each cluster based on the high depth data ( data in Fig 1B used for description below ) . Individuals with n/T = 1 have the reference allele ( represented by the yellow cluster of low depth data ) and n/T = 0 ( red cluster ) indicates that the HERV-K is absent ( no k-mers to unique sites in the HERV-K at this locus were recovered from mapped sequence reads ) . The k-mers derived from persons with low ( green ) and intermediate ( blue ) n/T values were mapped to the HERV-K reference for this locus to determine whether they localized only in the LTR ( assign ‘solo LTR’ to green cluster ) or in the coding region ( assign ‘provirus’ to blue cluster ) ( S3 Fig ) . The WGS data of each individual in the KGP dataset were evaluated using our optimized analysis workflow . HERV-Ks on chrY were not considered . Twenty sites , omitting one at chr1:73594980 [see Methods] that have been reported to be polymorphic for presence/absence [10 , 11 , 34 , 46] were identified as polymorphic for a HERV-K provirus by our analysis ( S1 Dataset:virus ) . Polymorphic HERV-Ks greater than 6 kbp in length cluster together in a phylogenetic analysis indicating that they are closely related ( S4 Fig ) . The prevalence ( proportion of individuals in a given population with a provirus present at a given locus ) of the 20 polymorphic HERV-K proviruses varied from 0 . 9% to 99 . 5% when averaged across the entire KGP dataset ( Table 1 ) . However , there were notable differences in prevalence at each HERV-K site among the five super-populations ( AFR , EAS , AMR , EUR , SAS; see Methods for key to abbreviations ) . Of the 20 , the prevalence of seven polymorphic HERV-Ks was greater than 90% and the difference between populations with the lowest and highest prevalence was less than 6 . 5% ( Table 1 ) . There was 100% occupancy for six of the seven high prevalence polymorphic HERV-Ks ( 98 . 8% for the seventh ) , indicating that the rate of conversion to solo LTR is low for viruses at these sites ( see S1 Text for occupancy and S2 Dataset:KGP ( absence , solo , presence ) for model prediction of solo LTR prevalence ) . Two polymorphic HERV-Ks had an overall prevalence of less than 10% in any population ( Table 1 ) and were found in individuals of AFR origin; we found no evidence of a solo LTR at these two sites . Nine of the remaining 11 HERV-Ks are of interest because the difference between super-populations with the highest and lowest prevalence is between 28 and 80 percentage points ( Table 1 ) . Of note , for the three HERV-Ks with the largest difference among super-populations , the prevalence is lowest in EAS populations . Individuals from African populations differ significantly from the other four super-populations in the prevalence of ten of the polymorphic HERV-K , three of which occur in close proximity on chr19 . ( Table 1 , S2 Dataset:compare_prevalence ) . EUR and AFR super-populations are significantly different in the prevalence at all but one of the 20 polymorphic HERV-K based on adjusted p-values ( S2 Dataset:compare_prevalence ) . The HERV-K genome is close to 10 kbp . As there are 20 known HERV-K loci with the potential to encode a provirus that are polymorphic in human populations , we asked if there is a difference in the burden of these repetitive , and potentially functional , viral elements among individuals . This was indeed the case . Of the 20 polymorphic HERV-K proviruses assessed , the number per person’s genome ranges from 7–18 ( Fig 2 , S2 Dataset:HERV-K per person ) . More than 63% of individuals from all super-populations except EAS carry 12 to 14 proviruses in their genome . Individuals from EAS have a lower burden with 69% of individuals carrying 9–11 of the 20 polymorphic HERV-K proviruses . 7% of AFR individuals have 16 or 17 proviruses compared to a maximum of 2% in other groups ( S2 Dataset:HERV-K per person ) . These data suggest that a comprehensive investigation of polymorphic HERV-Ks may be a more productive means to advance studies of their potential disease impact . Our data provide a comprehensive picture of sites occupied by HERV-K provirus in each genome . Although most previous studies investigating a role of HERV-K in human disease assessed the prevalence of the HERV-K at a given locus , it is possible that , for example , two HERV-Ks each at 40% prevalence in a population rarely co-occur in an individual genome . By providing the status of all known polymorphic HERV-K in the genome , our tools facilitate such assessment and can advance investigation of HERV-K and human disease . We assessed combinations of three , four and five polymorphic HERV-Ks in KGP data and found that there are many combinations of co-occurring viruses that are population-specific ( S3 Dataset ) . To facilitate exploration of HERV-K combinations among KGP populations , we developed a D3 . j visualization tool ( see Methods ) that allows a user to choose any combination of the 20 polymorphic HERV-K proviruses and display the co-occurrence prevalence among the 26 populations represented in the KGP data . As an example , we show a combination of four HERV-Ks to represent the variation that occurs in KGP individuals , which in this case ranges from 3% in EAS to 59% in EUR ( Fig 3A ) . We also determine that the three polymorphic HERV-Ks found on chr19 co-occur only from three AFR populations and in less than 2% of individuals ( Fig 3B ) . Because there are clearly population-specific differences in both individual HERV-K prevalence and in the prevalence of HERV-K co-occurrence , we explored whether the presence or absence of these 20 documented polymorphic HERV-Ks is sufficient to distinguish populations using Fisher’s linear discriminant analysis ( LDA ) [47] . Based on the status ‘provirus’ , ‘solo LTR’ , or ‘absence’ , there is little resolution of AFR , EUR , and EAS super-populations ( Fig 4A ) . However , there is sufficient signature to separate AFR , EUR , and EAS if we utilize the n/T ratio of the 20 polymorphic HERV-Ks ( S5 Fig ) and we further improve population separation if we use the n/T ratio for all 96 HERV-Ks ( Fig 4B ) . This indicates that we are losing information by reducing the data to three states and that fixed HERV-K also contain signal for population of origin . An n/T = 1 indicates that the query set contains all k-mers that map to the reference set T for a specific HERV-K . If there is a HERV-K allele that has not been reported in any database but that is common in a population , we expect n/T <1 because we require 100% match to reference set T and k-mers covering allelic sites will be excluded ( see Fig 1B , blue cluster for an example ) . We assessed the density distributions of n/T plots for each of the 96 HERV-Ks for evidence of population-specific alleles ( S1 Text , S7 Fig ) . Five HERV-Ks have some indication of population specific distributions ( S1 Dataset:virus ) . The HERV-K at chr1:155596457–155605636 , which we report as fixed , is notable because the reference allele ( n/T = 1 ) is only found in AFR ( Fig 5A , S7 Fig ) . Individuals from other populations have n/T near 0 . 5 . We mapped k-mers from individuals with n/T near 0 . 5 to the reference HERV-K sequences and confirmed that there is a loss of k-mers at several sites covered by the unique reference k-mers for this virus ( S8 Fig ) . There are also cases where the reference allele is found in all populations except AFR ( Fig 5B and see S7 Fig for additional examples ) .
Our research provides a tool to mine whole genome sequence data to collectively evaluate the status of HERV-K provirus at known polymorphic and fixed sites in the human genome . The tool incorporates a statistical clustering algorithm to accommodate low depth sequence data and a visualization tool to explore the co-occurrence of known polymorphic HERV-K in the global populations represented in the KGP data . There are numerous significant differences in the prevalence of individual and co-occurring known polymorphic HERV-K among the five KGP super-populations . It is notable that individuals from EAS carry a lower total burden of the 20 polymorphic HERV-K than other represented populations . These data provide a comprehensive framework of genomic diversity among 20 documented polymorphic HERV-K proviruses to advance studies on potential roles for HERV-K in human disease , which have been alluring yet difficult to establish [21 , 22 , 24] . Tools developed to interrogate ERV insertional polymorphism typically exploit the unique signature created by the host-virus junction [11 , 48 , 49] . These approaches indicate that a site is occupied by an ERV but not whether there is a provirus associated with the site , which is more difficult to accomplish with short read sequence data . Our analysis tool provides an efficient means to detect occupancy and provirus status in one step . We decrease computational time by analyzing only the set of reads that map to existing HERV-K loci in the reference genome . This approach is justified because the known polymorphic HERV-K that are missing from the human reference are closely related to those in the reference genome assembly ( see S4 Fig ) and hence reads derived from them map to a related HERV-K in the reference . We employ k-mer counting methods , which also increase computational efficiency . A reference set of k-mers that is unique to each HERV-K is generated for each location in the genome and the proportion of reads ( n/T ) from the query set that maps to the k-mer reference set is reported as a continuous variable; there is no threshold of read count or coverage imposed for classification . Instead we utilize a mixture model to statistically cluster values based on n/T and sequence depth and assign the same HERV-K status to all individuals in a cluster . Clusters representing n/T of 1 consist of individuals from whom all the unique k-mers identified in the HERV-K reference set were recovered from their mapped WGS data . We classify other clusters by determining if k-mers mapped on the reference allele are distributed at sites in the coding portion of the genome or only in the LTR; reads mapping only in the LTRs are classified as solo LTR . This approach demonstrated that the k-mers derived from some individuals only covered a subset of the unique sites and led to the interesting finding that several HERV-K loci could have population specific alleles . Wildschutte et al [11] have conducted the most comprehensive study of HERV-K prevalence in the KGP data to date . The goal of that paper was to identify new polymorphic insertions , either provirus or solo LTR , based on detecting reads containing the host virus junction . However , they implemented an additional step to detect provirus and provide the prevalence of some polymorphic HERV-K provirus for comparison with our results ( see S1 Dataset:virus for comparison of prevalence values reported in Wildschutte et al [11] ) . There are five HERV-K previously reported in Subramanian et al 2011 [10] that were not included in Wildschutte et al [11]; all are polymorphic in our analysis ( range 43–99% , see Table 1 and S1 Dataset:virus-column N ) . Seven polymorphic HERV-K , which Wildschutte et al [11] indicate occur in greater than 98% of KGP individuals , are fixed in our study . Our estimated prevalence for 14 HERV-K differs from that reported in Wildschutte et al [11] by 5% or more . Of these 14 , the prevalence estimates at chr1:155596457–155605636 are most divergent . Our data show this site is fixed for provirus and Wildschutte et al [11] report that only 14% of the KGP data , all from AFR , have a HERV-K provirus integration . Our plots for chr1:155596457–155605636 show that AFR individuals carry the reference allele at this site ( n/T near 1 , Fig 5A ) and all other individuals have n/T near 0 . 5 . The k-mers from individuals with low n/T values for chr1:155596457–155605636 map to only a subset of sites marked by unique k-mers in the coding region ( S8 Fig ) , which is consistent with sequence polymorphism or a deletion at these positions . The reference set T is small for this HERV-K and therefore overall coverage of the genome is low . Because Wildschutte et al [11] used a minimum coverage threshold for their k-mer mapping method , it is possible that alleles present in non-AFR populations do not meet their inclusion criteria . There is a similar signal for alleles , represented by lower n/T values , at the other 13 HERV-K sites although the differences between our prevalence estimates and those of Wildschutte et al [11] are small ( S1 Dataset:virus ) . In most cases these putative alleles are found in all populations at different frequencies but in five there is some degree of population specificity ( Fig 5 , S7 Fig , S1 Dataset:virus ) . Our results indicate that there could be considerably more sequence variation in HERV-K among human populations than previously appreciated . These data also suggest that using a HERV-K consensus sequence to study pathogenic potential could miss important features of HERV-K proviral polymorphism , which can be characterized by both the site occupancy status ( presence/absence ) and , when present , by sequence differences among individuals . HERV-Ks are the youngest family of endogenous retroviruses in humans and consequently they share considerable sequence identity . This has the effect of limiting the number of unique sites associated with some HERV-K , which decreases the size of the reference set T ( S1 Dataset:virus ) . The set T is small for near identical HERV-K such as HERV-Ks involved in a duplication event . The HERV-Ks at chr1:13458305–13467826 and chr1:13678850–13688242 are identical and cannot be distinguished . We report n/T for only one of these HERV-K ( see S1 Dataset:virus , column M ) . We treat the two HERV-K proviruses spanning chr7:4622057–4640031 as a single virus with n/T = 1 reflecting the tandem arrangement found in the hg19 . In this case , n/T<1 can mean either that both proviruses are present but with substitutions at a unique k-mer site or that one provirus converted to a solo LTR . Thus although an n/T ratio of 0 or 1 reliably indicates absence and presence of reference HERV-Ks , respectively , when T is small , sequence polymorphism and a deletion event can be difficult to distinguish from a solo LTR . However , because our mixture model statistically clusters similar n/T values based on sequence depth , all individuals in a cluster have the same status ( e . g allele or solo LTR ) even if we do not know what that state is . The ability of our tools to resolve the status of closely related HERV-K provirus sequences will improve as more empirical sequence data becomes available . Our approach provides researchers with a rapid means to determine if the prevalence , and overall burden of the 96 HERV-K proviruses evaluated differ between a patient data set and the population represented in KGP to which they trace ancestry . The visualization tool will facilitate investigation of combinations of HERV-Ks in certain clinical conditions . The potential that HERV-K has multiple allelic forms in different populations is worthy of further analysis because a sequence allele could also contribute to a disease condition .
The 96 HERV-K proviruses previously reported [10 , 11 , 34 , 46] were supplemented with HERV-K alleles present in the NCBI nt database ( November 2016 release ) ( 92 in hg19 , and 4 from the NCBI nt database ) . We required that any allele of a HERV-K from the nt database have at least 2kb of hg19 reference-matching host flanking sequence to confirm genome location . In total , 234 alleles were collected at the 96 known HERV-K loci . The location information and virus features are summarized in S1 Dataset: virus . We identified the k-mers that correspond to unique sequence characterizing each HERV-K . K-mers are substrings ( subsequences ) of length k that exist in a string ( DNA sequence ) . The length k is determined empirically ( S1 Text ) . Each k-mer is labeled with the corresponding viruses in which it is observed . Only those k-mers referring to a single virus locus , unique k-mers , are selected for the set T . Where multiple alleles of a HERV-K are available , k-mers unique to all alleles at that location comprise T . Multiple 2bps different k-mers ( such as SNPs ) corresponding to the same location on the virus , are merged into a single entry for the purposes of computing T . We map unique k-mers back to the corresponding alleles to determine coverage of the HERV-K and whether k-mers are located in LTRs ( S3 Fig; S1 Dataset: virus ) . To develop a method to recover sequences containing information on HERV-K we leverage the fact that HERV-Ks are closely related . Thus , most sequence reads obtained from an individual with a polymorphic HERV-K that is absent in the human reference , hg19 , will map to the location of a closely related HERV-K that is present the human genome reference . ( As we show in S4 Fig , the known polymorphic HERV-K proviruses are closely related . ) A file with the coordinates for all reported HERV-K insertions is used to extract mapped reads from a genome sequence file ( S1 Dataset:bed , which provides the coordinates for both hg19 and hg38 ) . Note that the KGP data were mapped to GRCh37 , which includes the decoy sequence hs37d5 . This decoy contains the HERV-K at chr1:73594980_73595948 , which is not present in hg19 . Thus , we did not recover any reads for this HERV-K , which is polymorphic but reportedly at high prevalence in most populations [11] . The KGP data were downloaded in aligned Binary Alignment/Map ( BAM ) format ( ftp://ftp . ncbi . nlm . nih . gov/1000genomes/ftp/data/ ) . It contains data for 2 , 535 individuals ( S1 Dataset:KGP ) sequenced via low-depth whole-genome sequencing ( mean depth = 6 . 98X ) . The individuals represent 26 populations , derived from 5 super-populations , including African ( AFR ) , Admixed America ( AMR ) , East Asian ( EAS ) , European ( EUR ) , and South Asian ( SAS ) [50 , 51] . Of 2 , 535 individuals , 28 also have high-depth DNA sequences ( mean depth = 48 . 06X ) , which we use as a pilot dataset to develop the mixture model , described below and in Supplementary Text . Our computational framework to indicate the status of each known HERV-K provirus is based on the n/T ratio , which is the proportion of k-mers in the data mined from WGS of each individual that are identical to the reference set T for each HERV-K provirus . Sequence reads are extracted from a mapped file of whole human genome sequence data based on coordinates corresponding to each annotated HERV-K . The reads are k-merized and mapped to the set T , which represents all unique k-mers assigned to each HERV-K in the reference set . We use exact match to map the k-mer data set to the unique k-mer references . The n/T ratio is an indicator of the presence of each HERV-K; n/T = 1 indicates that the individual has the HERV-K in our reference dataset documented to be at that locus while n/T = 0 indicates that no k-mers unique to a HERV-K locus were recovered ( see Fig 1 for more explanation ) . Using a hash table ( S1 Text ) , it takes 15 minutes to generate the n/T matrix for 100 files . The source code for the entire process is at https://github . com/lwl1112/polymorphicHERV We utilized a statistical model to account for the dependency of the number of k-mers obtained from a person’s sequence data ( denoted by nik for the ith subject and kth HERV-K , with i = 1 , … , I , k = 1 , … , 96 ) that maps to the reference set T for each HERV-K on sequencing depth . Thus for each HERV-K we could statistically cluster those nik/T values for i = 1 , … , I based on the sequence depth of the WGS data for each individual for subsequent biological classification ( provirus , solo LTR , absence , see Fig 1 ) . More specifically in our analysis , for each k HERV-K , k = 1 , … , 96 , consider a sample of size I measurements xi ( i = 1:I ) , where each xi is a vector of length 2 xi = ( xi1 , xi2 ) with xi1 being the nik/T measurement and xi2 the log function of depth . Here , for notation simplification , we use xi instead of xik . To perform clustering analysis , we utilize the mixture model approach , which is arguably the most widely used statistical method for clustering . Specifically , we follow the work proposed by Lin et al . [52] that employs a Gaussian Mixture Model ( GMM ) with density function given by f ( xi|θ ) =∑j=1MπjN ( μj , Σj ) , fori=1:I ( 1 ) where all relevant and needed ( unknown ) parameters are represented by θ = ( π{1:M} , μ{1:M} , Σ{1:M} ) . N ( μj , Σj ) is the Gaussian density for the jth component parameterized by the 2-dimensional mean vector μj and 2x2 covariance matrix Σj . π{1:M} are the mixture components prior probabilities summing to 1 . To allow a flexible modeling approach , we employ the standard Bayesian ( truncated ) Dirichlet Process prior for the parameters θ = ( πj , μj , Σj , j = 1:M ) [53 , 54] . The idea is that some of the mixture probabilities ( πj ) can be zero , hence the actual number of mixture components needed may be smaller than the upper bound M . This mechanism allows automatic determination of the number of mixture components needed by the data set at hand . For model estimation , a latent indicator Zi ∈{1 , 2 , … , M} with P ( Zi = j ) = πj is used , for i = 1:I . Specifically , Zi = j if , and only if , xi comes from component j . Given a fitted model via the Bayesian expectation–maximization algorithm , in terms of estimates of all parameters θ , instead of interpreting the fitted Gaussian mixture components as clusters , we identify clusters by aggregating Gaussian components so that non-Gaussian type of clusters can be flexibly represented . Merging components into clusters can be done by associating each of the Gaussian components to the closest mode of f ( x1:I|θ ) = ∏i = 1:If ( xi|θ ) . Hence , the number of modes identified is the realized number of clusters . [S1 Text for additional detail] We consider that both the individual prevalence of a HERV-K and the co-occurrence of multiple HERV-Ks could differ among populations . The time of a brute-force approach for finding all combinations Cm of size m from p polymorphic HERV-K is ( ∑m=1p ( pm ) =2p−1 ) , which is not efficient and is redundant . We employed the Apriori algorithm [55] , which is commonly used for finding frequent pattern sets; in our case indicating which of the known polymorphic HERV-K frequently appear together . It first generates combinations Cm ( initialized to 1 ) . In the optimization , frequent combinations Fm are returned from candidates Cm when prevalence exceeds the minimum threshold of co-occurrence . Fm are then self-joined to generate combinations Cm+1 of size m +1 and out of which Fm+1 satisfy the minimum co-occurrence . In each pass , candidate combinations are pruned so as to avoid generating all combinations , which reduces running time significantly . We made statistical comparisons across 5 super-populations for the following three problems . For each problem , there are ( 52 ) = 10 families of 1-to-1 comparisons conducted . The ‘prop-test’ function in R is used to test whether the proportions for two super-populations are the same . Therefore , multiple hypotheses would be conducted on frequencies F across super-populations P1…5 as follows: Null hypothesis , H0:FPi=FPj , where i≠j; Alternative hypothesis , HA:FPi≠FPj , where i≠j . A separate P-value is computed for each test and the Benjamini-Hochberg procedure [56] is used to account for multiple comparisons . We utilized D3 . js ( Data Driven Documents ) [57] , an open-source java script library to create an interactive visualization to display co-occurrence of polymorphic HERV-Ks in human populations . Our visualization system includes two modules , a welcome page and a result page . Input JSON data include locations of polymorphic HERV-K , population information , and the 0/1 ( absence / presence ) matrix . ( See S1 Text ) . Source code is available at: https://github . com/lwl1112/polymorphicHERV/tree/master/visualization and a searchable tool with the data reported here is at: http://pages . iu . edu/~wli6/visualization/
|
Human Endogenous Retrovirus type K ( HERV-K ) is the youngest of retrovirus families in the human genome and is the only group of endogenous retroviruses that has polymorphic members; a locus containing a HERV-K can be occupied in one individual but empty in others . HERV-Ks could contribute to disease risk or pathogenesis but linking one of the known polymorphic HERV-K to a specific disease has been difficult . We develop an easy to use method that reveals the considerable variation existing among global populations in the prevalence of individual and co-occurring polymorphic HERV-K , and in the number of HERV-K that any individual has in their genome . Our study provides a reference of diversity for the currently known polymorphic HERV-K in global populations and tools needed to determine the profile of all known polymorphic HERV-K in the genome of any patient population .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"linear",
"discriminant",
"analysis",
"statistics",
"human",
"genomics",
"data",
"mining",
"genomic",
"databases",
"mathematics",
"genome",
"analysis",
"mammalian",
"genomics",
"information",
"technology",
"research",
"and",
"analysis",
"methods",
"sequence",
"analysis",
"computer",
"and",
"information",
"sciences",
"bioinformatics",
"mathematical",
"and",
"statistical",
"techniques",
"biological",
"databases",
"statistical",
"methods",
"animal",
"genomics",
"sequence",
"databases",
"data",
"visualization",
"database",
"and",
"informatics",
"methods",
"genetics",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"genomics",
"computational",
"biology"
] |
2019
|
A computational framework to assess genome-wide distribution of polymorphic human endogenous retrovirus-K In human populations
|
During microbial evolution , genome rearrangement increases with increasing sequence divergence . If the relationship between synteny and sequence divergence can be modeled , gene clusters in genomes of distantly related organisms exhibiting anomalous synteny can be identified and used to infer functional conservation . We applied the phylogenetic pairwise comparison method to establish and model a strong correlation between synteny and sequence divergence in all 634 available Archaeal and Bacterial genomes from the NCBI database and four newly assembled genomes of uncultivated Archaea from an acid mine drainage ( AMD ) community . In parallel , we established and modeled the trend between synteny and functional relatedness in the 118 genomes available in the STRING database . By combining these models , we developed a gene functional annotation method that weights evolutionary distance to estimate the probability of functional associations of syntenous proteins between genome pairs . The method was applied to the hypothetical proteins and poorly annotated genes in newly assembled acid mine drainage Archaeal genomes to add or improve gene annotations . This is the first method to assign possible functions to poorly annotated genes through quantification of the probability of gene functional relationships based on synteny at a significant evolutionary distance , and has the potential for broad application .
Gene function prediction is currently one of the fundamental problems in microbiology [1] . The improvement in DNA sequencing technologies has allowed for the sequencing of hundreds of full Bacterial and Archaeal genomes . However , in the dataset of full Bacterial and Archaeal genomes from NCBI , 874 , 583 genes out of 2 , 668 , 809 ( ∼33% ) are annotated as hypothetical proteins , and 25% of the protein families in the PFAM database have unknown functions [1] . In addition to these un-annotated genes , many of the genes in these databases only have general function predictions or may have incorrect function predictions . Thus , improved protein functional prediction methods are urgently needed . It has been proposed that correlations between synteny and evolutionary distance , in concert with homology , can be used for predicting protein function [2] , [3] , [4] , [5] . Synteny has been used to predict the functional interaction of proteins , where interaction is defined as direct physical interaction , the regulation of one protein by the other , membership in a protein complex , or the sharing of a metabolic ( or non-metabolic ) pathway [4] , [5] , [6] . Various protein function prediction methods make use of synteny , as reviewed by Rogozin et al . in 2004 [3] , [7] , [8] , [9] , [10] , [11] , but do not consider evolutionary distance between genomes in their predictions . Preservation of synteny over large evolutionary distances should be weighted strongly in gene function prediction because it is likely the result of selection against rearrangements . Huynen and Snel noted the importance of finding the evolutionary distance at which gene order conservation becomes significant [12] . Snel et al . simulated random genome shuffling to determine the probability of conserved gene order in a specific number of genomes [13] , and Von Mering et al . assessed the likelihood of protein relatedness based on the number of times gene order is conserved in the STRING database of genomes [4] . Here , we link the probability of syntenous protein relatedness and evolutionary distance so that we can determine which genomes are distant enough to accurately utilize synteny-based gene annotation . An overview of our method is provided in Figure S1 . Our analyses included genomes of coexisting Archaea reconstructed from metagenomic sequence from biofilms growing in an extreme acid mine drainage ( AMD ) environment as well as published genomes . The inclusion of AMD Archaea allowed us to apply the method to newly assembled genomes from uncultivated organisms , and to show the utility of the method for comparative genomics and for improving annotations of proteins of unknown function .
We reconstructed four new genomes of uncultivated Archaea: A- , E- , G- , and I-plasma ( Archaea; Euryarchaeota; Thermoplasmata; Thermoplasmatales ) . Ferroplasma acidarmanus ( Ferroplasma Type I , Fer1 ) and Ferroplasma Type II ( Fer2 ) have previously been described [14] , [15] , [16] . Only Fer1 has been isolated [14] . The phylogenetic placement of these organisms based on 16S rRNA gene sequences is shown in Figure 1 . E- and G-plasma are most closely related , whereas I-plasma is distantly related and may not actually belong to the Thermoplasmatales lineage . Data describing the manually curated and binned composite genomes of these Archaea are listed in Table 1 . Note that the estimates of the sizes of all genomes are similar . We used standard measures to evaluate genome completeness: a full suite of tRNAs , rRNAs , and orthologous marker genes in all genomes [17] . All of the genomes of the AMD Thermoplasmatales organisms except for A-plasma are near complete , according to our analysis ( Table S1 ) . In order to carry out regression analysis on genome rearrangements and evolutionary distance , we used gene order conservation ( GOC ) as a measure of whole genome rearrangement . This metric is described by Rocha [18] . Figure 2 shows the relationship between GOC and evolutionary distance as measured by average normalized BLASTP bit score , a proxy for evolutionary distance . Figure 2 includes results for genomes reconstructed for uncultivated AMD Archaea from metagenomic data . These genomes are incomplete ( Table 1 ) , and remaining gaps may affect our analyses . Thus , we investigated the effect of a limited amount of fragmentation on trends by shearing the genome of the Ferroplasma acidarmanus ( Fer1 ) isolate into fragments that corresponded to the lengths of the fragments from our environmental datasets . The fragmented Fer1 pairwise comparisons followed the trend defined by all other genomes with a slight downward shift ( Figure S2 ) . Based on the clear relationship between evolutionary distance and synteny , we explored an improved neighborhood approach to protein functional prediction . We developed a method that involves an evolutionary distance-weighting for each pairwise comparison and incorporates the high probability of synteny due to chance in closely related organisms . We assumed that genes that remain syntenous in organisms separated by large evolutionary distances do so because of selective pressure to maintain function . Genes in predicted operons have previously been shown to rearrange at a slower rate than genes that are never found in operons [18] . We quantified the statistical significance of the difference between the populations for operon and non-operon genes using the phylogenetic pairwise comparison method [19] and the Wilcoxon signed-rank test . We used the phylogenetic pairwise comparison method to choose independent pairs of genomes for comparison and the Wilcoxon signed-rank test to test the hypothesis that there is a significant difference between the values of GOC for populations of pairwise comparisons that include operon genes versus those that include genes not in operons in the two genomes that were compared . The Wilcoxon test indicated a significant difference with a p-value of 1 . 017×10−13 . We posit that this difference is due to stronger selection against the rearrangement of genes in operons because of co-regulation and functional linkage . As an approximation , we also assumed that genes that are not in operons and retain synteny do so solely by chance , that is , selection against rearrangements on non-operon genes is negligible for the purposes of our analysis . We used the trend between gene order conservation ( GOC ) and gene sequence divergence in genes not found in operons between the two genomes being compared ( non-operon genes ) to determine the degree of evolutionary divergence necessary to ensure that genes that retain synteny do not do so by chance . Because a GOC value approximates the probability of that any two genes retain synteny in a pairwise comparison at a given evolutionary distance , to estimate the probability that genes retained synteny due solely to chance , PGOCn , we modeled the relationship between the gene order conservation of non-operon genes ( GOCn ) and evolutionary distance ( Figure 3 ) . We calculated a measure of goodness of fit with the sum of squared errors ( SSE ) and the total sum of squares ( TSS ) ; 1-SSE/TSS = 0 . 9282 . PGOCn values were then compared to the percentage of syntenous genes that were functionally related in genomes included in the STRING database . We modeled this relationship as well , and interpreted the response variable as the probability that any two syntenous genes are functionally related , Pr , ( 1-SSE/TSS = 0 . 7648 , Figure 4 ) . Both models were chosen from a set of models , using Akaike's information criterion ( Table 2 ) . We combined the models to predict Pr from measurements of evolutionary distance ( Figure S1 ) . Thus , for pairwise comparisons below a certain evolutionary distance threshold ( a bit score value of 0 . 3129 ) , Pr was statistically significant; syntenous genes have a 95% or greater probability of being functionally related ( Pr>0 . 95 ) . A gene of unknown function in one such comparison is likely functionally related to its syntenous orthologs . In these cases , functional information for syntenous orthologs that would otherwise be disregarded due to low sequence similarity was used to improve annotations of genes of unknown function . Alternatively , if functional information was available for neighboring genes in a block for which synteny was preserved , the gene of unknown function was annotated as related to its neighbor . We applied this evolutionary distance-weighted method to improve protein functional annotation in AMD Archaea for genes involved in the following pathways and processes i ) cobalamin biosynthesis ii ) molybdopterin guanine dinucleotide ( MOB-DN ) synthesis and MOB-DN-utilizing enzymes iii ) ether lipid biosynthesis and iv ) CRISPR-related proteins . We improved the annotation of 25 genes involved in cobalamin salvage in A , G , and I-plasma as well as Fer1 and Fer2 ( Figure 5 and Table S2 ) . An additional 34 genes were annotated with our method as part of the de novo cobalamin synthesis pathway or as cobalamin-related , including several cobalamin-binding proteins . We inferred a cobalamin-related function for two genes with very general annotations due to their synteny-based annotations ( Table S2 ) . The near complete de novo cobalamin synthesis pathway was found only in the Ferroplasma genomes , indicating a possible difference in these organisms' growth requirements . The synteny-based annotation of molybdopterin synthesis genes also differentiates the various AMD Archaea . Our synteny-based method improved annotations or provided annotations for seventeen genes in A-plasma , eleven genes in I-plasma , ten genes in Fer1 , and six genes in Fer2 that were involved in molybdopterin synthesis , utilization or molybdate uptake ( Figure 6 and Table S3 ) . In silico protein structure modeling supported the functional annotation of a number of these genes ( Table S3 ) . The A , I , Fer1 , and Fer2 genomes have full pathways for the synthesis of molybdopterin guanine dinucleotide ( MOB-DN ) , a molybdopterin cofactor that is used by proteins involved in anaerobic energy metabolism , while E and G-plasma have very few annotated molybdopterin synthesis genes of any kind . Formate dehydrogenase subunit genes are found in A-plasma , I-plasma , Fer1 , and Fer2 genomes within molybdopterin synthesis gene clusters . Formate dehydrogenase is a MOB-DN-utilizing enzyme . In silico protein modeling provided additional evidence for the formate dehydrogenase annotation of these genes ( Table S3 ) . Ether lipid biosynthesis genes were found in all of the AMD Thermoplasmatales Archaea , as expected . Synteny-based annotation improved or provided annotations for a number of genes involved in ether lipid biosynthesis and its feeder pathway , the mevalonate pathway ( Figure 7 and Table S4 ) . This included five genes in A-plasma , seven genes in E-plasma and in G-plasma , ten genes in I-plasma , eight genes in Fer1 and eight genes in Fer2 . A hypothetical protein was identified in all of the AMD Archaea studied that appeared to be associated with the mevalonate pathway based on synteny . Manual curation indicated that it may encode a nucleic-acid binding protein . All of the AMD Archaeal genomes except for that of A-plasma contained most or all of the genes involved in the mevalonate and ether lipid biosynthesis pathways . A-plasma is missing key genes in the mevalonate pathway likely due to its incomplete genome assembly . A-plasma is also the only genome missing genes involved in the ether lipid synthesis pathway found in Archaeoglobus fulgidus [20] . Two genes that maintained synteny with the ether lipid synthesis genes were investigated for possible involvement in the final steps of ether lipid biosynthesis ( e . g . , polar head group attachment and side chain modifications ) . However , BLASTs against all available NCBI Bacterial genomes , indicated that these genes were also found in a number of Bacteria and were thus unlikely to be involved in Archaeal ether lipid synthesis pathways . All of the AMD Thermoplasmatales Archaeal genomes contain some CRISPR-associated proteins that occur in gene clusters with CRISPR spacer regions . A number of the CRISPR proteins in the AMD Archaea are syntenous with distant relatives , allowing us to improve annotations and annotate hypothetical proteins at these loci for twenty-seven CRISPR-associated proteins ( Figure 8 and Table S5 ) . All of the Archaeal genomes contained Cas1 genes , which are generally thought to be in all Cas systems as well as Cas2 genes that are found in most Cas systems [21] . In order to test this new synteny-based method , we compared four well characterized , very distantly related Bacteria and Archaea to one another . We made two comparisons , one between the two Bacteria and one between the two Archaea . We examined a total of 175 unique genes and their syntenous orthologs in the four organisms . Of these 175 genes we found that our method correctly annotated the genes in one organism ( we chose the better characterized one in both cases ) 97 . 1% of the time ( Table S6 ) . In five cases , the annotation appeared to be correct , but one organism had only the general annotation of ABC transporter with a likely substrate specificity instead of a specific ABC transporter protein . In three other cases , the annotations in the well characterized organism did not concur with our manual curation of the gene's function . In only two cases was the annotation method clearly incorrect , in this case substituting two very closely related protein functions that are sometimes found in the same bidirectional enzyme , fumarate reductase and succinate dehydrogenase subunits A and B . The method was also able to reconstruct parts of the Trp operon for E . coli and H . volcanii . This is significant because not only are the functions of the genes in this operon well characterized , but their associations and regulatory systems are also well understood . In the case of E . coli , the method correctly predicted the functions of TrpA and TrpB ( Table S8 ) , while in the case of H . volcanii , the method correctly predicted the functions of TrpD , TrpE , and TrpG ( Table S8 ) .
The trends reported here between synteny and sequence divergence and between protein functional relatedness and synteny were determined based on the phylogenetic pairwise comparison method . This method takes into account phylogeny in order to assign pairs of genomes for comparison that do not share recent phylogenetic history with other pairs . This produces phylogenetically independent data points and allows regression analysis to be carried out without pseudoreplication . We found high measures of goodness of fit for synteny and sequence divergence , as well as for synteny and percent of syntenous genes with related functions . Because there is no known mechanistic link between point mutations and gene rearrangements , these results indicate similar selective pressures on rearrangements and mutations . Despite the advantages of the phylogenetic pairwise comparison method , we recognize that it has inherent biases . Specifically , picking the maximal number of pairs for analysis results in the choice of many closely related pairs . Pairs that are clustered in one portion of the tree may have similar levels of synteny and sequence divergence , but this correlation may be due to some third unknown trait that is also present in this clade . We chose to use all available data as opposed to more evenly spaced taxa in order to obtain enough information for regression analysis . For our analysis , we are interested in more distantly related organisms , thus partially resolving the problem of bias in close relatives . We also recognize that the use of all of the Bacterial and Archaeal genomes available in the NCBI and STRING databases has resulted in a bias in our data towards certain clades and organism types that are overrepresented ( e . g . , pathogens ) . Inclusion of genomes reconstructed from metagenomic sequence data from the natural environment slightly reduces this bias . However , this method could be greatly improved in the future when more fully sequenced genomes are available . In a few cases , two unrelated blocks of syntenous genes were conserved adjacent to one another at significant evolutionary distances . This problem can be avoided by enforcing a stricter evolutionary distance cutoff in the cases where it can be observed that two syntenous blocks are sometimes , but not always conserved next to one another . The mechanism resulting in this type of synteny conservation is unknown . It is important to note that the model we developed for synteny-based annotation assumes that all genes in operons are rearranged slowly compared to those that are not . This is consistent with data analysis shown in Figure 3 and with Rocha's analyses [18] . We also assumed that genes that are rearranged rapidly are not under significant selective pressure so that the trend for the non-operon genes could be used to estimate the probability that any two genes stay together due to chance . Deviation from this assumption in a subset of cases could contribute to scatter in the trend ( thus poorer regressions and weaker correlations ) and lead to a higher value of GOC for significance of synteny for annotation purposes . Thus , our method provides a conservative estimate of the evolutionary distance necessary for functional predictions and the probability of functional relatedness is higher than stated . Understanding the relationship between gene order and evolutionary distance is essential for accurate synteny-based gene functional annotation . In the case of the AMD Archaea , the weighting of conservation of gene order at large evolutionary distances resulted in improved annotations for genes involved in a number of processes , including cobalamin biosynthesis , molybdopterin guanine dinucleotide ( MOB-DN ) synthesis and MOB-DN-utilizing enzymes , ether lipid biosynthesis , and CRISPR-based immunity . Synteny-based annotation of cobalamin biosynthesis genes indicated a clear difference between the nutritional requirements of A , E , G , and I-plasma versus those of the Ferroplasma species . Both of the Fer1 and Fer2 genomes contained full de novo anaerobic cobalamin synthesis pathways , while the other Archaeal genomes contained nearly complete cobalamin salvage pathways [22] . This difference may be important in differentiating the niches of the various types of AMD Archaea . It may allow the Ferroplasmas to compete better with other Archaea in low nutrient conditions , i . e . , in early growth stage biofilms . The synteny-based annotation of molybdopterin biosynthesis and molybdopterin-binding proteins in the AMD Thermoplasmatales Archaea also helped to differentiate their respective physiologies . The molybdopterin guanine dinucleotide synthesis protein ( MobA ) in A-plasma , I-plasma , Fer1 , and Fer2 makes a specific type of molybdenum cofactor that is only used by dimethyl sulfoxide ( DMSO ) reductase family enzymes . These genomes also include a gene for a formate dehydrogenase protein ( a member of the DMSO reductase family ) in their molybdopterin synthesis gene clusters , indicating that they may be able to use this enzyme for nitrate reduction , mixed acid fermentation , or anaerobic carbon fixation . Previously published proteomic data demonstrate that some of these MobA genes are expressed and suggests that some AMD Archaea use one of these anaerobic energy or carbon metabolisms . E and G-plasma's genomes contain only one molybtoperin-related gene ( moeB ) , which may be a misannotation , and thus likely do not use a MOB-DN cofactor . In silico protein structure modeling supported the functional annotation of certain molybdopterin synthesis genes of interest ( Table S3 ) . Specifically , structural modeling suggested that the potential MobA genes in A-plasma and I-plasma do in fact make MobA . Interestingly , the I-plasma homolog for MoaB fits a protein model for MogA . This is intriguing because no Archaea to date have been shown to have true MogA homologs , but MoaB is thought to play the same role in molybdopterin biosynthesis for Archaea as MogA does for E . coli [23] . Structural modeling also supported the functional annotation of the FdhF alpha subunit genes found in the A-plasma , I-plasma , Fer1 , and Fer2 genomes ( Figure S3 ) . These proteins fit the FdhF of the hydrogenase-linked formate dehydrogenase model from Escherichia coli , suggesting a potential involvement of these genes in a formate hydrogen lyase complex and mixed acid fermentation . The synteny-based method identified two new genes that may be involved in MOB-DN synthesis . These are a thioredoxin family gene and a SurE: 5′/3′-nucleotidase . SurE is of particular interest , as it functions in E . coli to remove a phosphate group from nucleotides [24] . SurE has the highest affinity for AMP among nucleotides tested by Proudfoot et al . [24] . An intermediate in molydopterin biosynthesis , adenylated molybdopterin , contains a covalently-bound AMP . Thus , this SurE homolog is potentially involved in dephosphorylation related to molybdopterin biosynthesis or modification . Ether lipid biosynthesis is a pathway common to all Archaea . The mevalonate pathway precedes ether lipid biosynthesis [20] . Thus , we looked for mevalonate pathway genes as well as ether lipid biosynthesis genes . Of the twenty-five mevalonate pathway genes annotated via our synteny-based approach , one hypothetical protein has orthologs in all of the AMD Archaea analyzed . This gene contains the PFAM domain of unknown function 35 that is hypothesized to bind nucleic acids . A possible ortholog of this gene is found in all Archaeal genomes available on NCBI , further supporting some role of this gene in the mevalonate pathway . The CRISPR-related proteins annotated with synteny-based annotation included a number of genes previously annotated as hypothetical proteins . I-plasma and Fer1 included the typical operon configuration of Cas module family I [25] , while the other genomes included novel Cas system arrangements . These annotations provide a starting point for further investigation of the biochemistry of the CRISPR/Cas system . The Ruby scripts used for our analyses are open source and are available at https://github . com/pyelton/Synteny-based-annotator .
For a detailed explanation of sampling , DNA extraction , sequencing and assembly protocol see Text S1 . The completeness of the Archaeal genomes was evaluated based on the number of tRNA , rRNA , and other orthologous marker genes [17] . Binning accuracy was also evaluated by analysis of fragment clustering in emergent self-organizing maps ( ESOM ) created based on tetranucleotide frequencies of consensus contig sequences [26] . Genes missing from the pathways that we analyzed were searched for in the overall AMD DNA dataset . BLAST hits to these genes were then binned via tetranucleotide frequency , using ESOM , and assigned to organisms if possible . Because the objective of this work was to analyze lineage divergence and develop a gene functional prediction method applicable to all Archaea and Bacteria , our analyses included all publicly available Archaeal genome sequences downloaded from NCBI . All available published Bacterial sequences were also added to the analysis for a more comprehensive comparison of genome rearrangements . Note that this consisted of 634 genomes because we used only the full genomes published on the NCBI website that also had full 16S rRNA sequences available on NCBI . These genomes were selected from across all major lineages of Archaea and Bacteria . We identified orthologs and syntenous genes using pairwise comparisons between 638 organisms ( Table S7 ) for Figures 2 and 3 and using pairwise comparisons between all 118 prokaryotic organisms from the STRING database for Figure 4 ( Table S8 ) . Orthologs were operationally defined as those genes that were reciprocal best BLASTP hits that shared 30% or greater amino acid identity over 70% or more of the gene length or BLASTP hits that shared 20% or greater amino acid identity over 50% or more of the gene length and maintained synteny . Synteny was initially defined as conservation of two or more adjacent genes in two genomes . Subsequent analyses defined synteny as the conservation of genes separated by no more than one intervening gene . Trends in synteny versus evolutionary distance did not differ substantially between these two definitions ( data not shown ) . Thus , we generally refer to synteny in this paper as conservation of a gene pair with no more than one intervening gene . We used an established measure of synteny , the fraction of orthologous genes that are syntenous based on at least one shared neighbor ( allowing for a specified number of gene insertions ) in the two genomes compared ( gene order conservation; GOC ) as described by Rocha [18] . For our measure of genome sequence divergence we chose average normalized BLASTP bit score normalized to the maximum possible bit score between two genes . Normalization consisted of dividing the bit score of the alignment by the average of the two maximum possible bit scores of the alignments of self to self for each respective gene ( for details see Text S1 ) . We chose this measure for two reasons . Firstly , previous work has shown that whole genome amino acid identity is a robust measure of evolutionary distance even between close relatives [27] , while sequence insertions and deletions are important in sequence divergence for distant relatives [28] . Average normalized bit score is a measure that captures both insertions/deletions and amino acid identity . 16S rRNA gene sequence divergence was also considered in the analysis because it is a standard measure and for comparison to previous studies . Trends between GOC and 16S rRNA divergence were similar to those using average normalized bit score as a measure of evolutionary distance , but were more variable ( data not shown ) . All genes used as examples in this analysis were manually curated according to the following criteria: Genes were aligned against the interpro and nr databases with a BLASTP algorithm . Genes were then annotated if they had a TIGR or Pfam domain hit that predicted a specific function with greater than 70% amino acid identity , an e-value of at least 1×10−10 and coverage of more than 70% of the protein . Genes were given a “putative” annotation if they met the previous criteria except they had an amino acid identity of 30–70% , an e-value between 1×10−4 and 1×10−10 , and matched 50–70% of the protein , or if their domain-based hits provided only general functional information . In these cases , additional evidence from hits from the nr database was used if possible to provide a specific functional annotation . Genes were given a “probable” annotation if they had annotated hits in the nr database with greater than 30% amino acid identity over 70% of the length of the gene . In order to determine the rates of synteny loss over different evolutionary distances , we looked for correlations and trends between average normalized bit score , GOC , and the percentage of syntenous genes that are known to be functionally related . Our initial regressions compared genome pairs from NCBI and our dataset and regressed GOC on average normalized bit score . The regression of percent syntenous genes that are related on GOC used genome pairs from the STRING database . Genes were considered related if they had a predicted association in STRING based on fusion events , experimental evidence , co-expression , database information ( involvement in the same pathway or complex ) , and text mining information ( co-occurrence in multiple papers ) . To avoid circularity in our method genome context was not used in predicting functional relatedness , that is , neither co-occurrence in genomes nor synteny was used to predict protein functional relatedness . Because of the inherent non-independence of pairwise comparisons between different taxa , we made use of a method to select phylogenetically independent pairs [19] , [29] . For details on this method see Text S1 . Genes were predicted to be in operons when they had the same transcription direction and no more than thirty bases between the two . We compared genes that were in predicted operons in one or both of the two genomes in a pairwise comparison , “operon genes” , and genes that were not found in predicted operons in either genome of the pairwise comparison , “non-operon genes” . For a more detailed explanation of the operon prediction method see Text S1 . In order to show that the GOC of pairwise comparisons of operon genes were significantly different from comparisons of non-operon genes , we chose to use a non-parametric test because of the unknown distribution of the data . In order to test the validity of this method to near-complete environmental genomes , the Ferroplasma acidarmanus ( Fer1 ) isolate genome was sheared into fragments . For information on genome shearing see Text S1 . Open reading frames for the Archaeal genomes were identified using the Prodigal software [30] . Annotations were automatically generated through a pipeline that includes homology searches against KEGG and Uniref90 , and domain/motif homology searches using InterProScan . Annotations were ranked in order of increasing confidence of a match: Rank A annotations are the most confident and Rank G annotations represent gene predictions with no functional assignment . For an explanation of rankings , see Text S1 . All annotations specifically mentioned in this paper were manually curated based on conserved domains in InterProScan and similarity to the nr sequence database from NCBI . The remainder of the functional annotation and physiological inferences for the genomes of the AMD Archaea will be reported separately ( Yelton et al . , in preparation ) . Our weighted synteny-based annotation approach is related to a previously published approach [18] . Rocha noted that GOC is an estimate of the probability that genes remain unshuffled over a certain evolutionary distancet . He also noted that genes in operons in either organism experience much slower rates of gene rearrangements than other orthologs [18] . We calculated the probability that genes are syntenous due solely to chance at a given evolutionary distance ( PGOCn ) by assuming that the GOC for rapidly shuffling genes ( those not in operons; GOCn ) was due entirely to chance . GOCn was plotted against evolutionary distance and was fitted to the data by nonlinear regression . The regressions were based on the following functions: Average normalized bit score regression:Where a , b , and c are constants , e is Euler's number , and t is the average normalized BLASTP bit score between two genomes . Percent of syntenous genes that are related regression on Pchance:Where c is constant , e is Euler's number , and Pchance is the value of GOC calculated from the bit score of the comparison based on the GOCn regression . These functions were chosen for each regression based on comparison of the following types of models using Akaike's information criterion: for the GOC on average bit score regression we looked at linear models , log models , exponential models , and sigmoidal models . AIC indicated that a gompertz model fit the GOC and average bit score data the best ( Table 2 ) . This was not surprising because the data appears to be sigmoidal and asymmetrical . For the percent of functionally related syntenous genes on GOC regression we considered linear models , exponential decay models , and quadratic models . These models were forced through the point ( 0 , 1 ) because at a Pchance of zero , where the probability that two genes retain synteny due to chance is zero , the proability that sytnenous genes are functionally related must be equal to one . AIC indicated that the exponential model fit the data the best in this case ( Table 2 ) . We found the t at GOCn = PGOCn = 0 . 05 , an average normalized bit score of 0 . 3129 , the evolutionary distance at which there is a 95% probability that syntenous genes are functionally related according to the STRING database information ( Figure S1 ) . At this evolutionary distance , there is at least a 95% probability that genes that retain synteny have done so for a reason ( presumably selective pressures ) . In fact , the probability that syntenous genes have related function is likely higher than 95% because the STRING database does not have exhaustive protein interaction data . Based on the derived values of t , we chose genomes that were sufficiently distant relatives that genes are not likely to be syntenous by chance so that synteny could be used to annotate genes in the AMD Archaea . For genome comparisons with a bit score of less than 0 . 3129 , we assigned or improved annotations of genes that are found in syntenous blocks in AMD Thermoplasmatales Archaea . Each gene was then annotated with the annotation of its ortholog if that gene had an annotation , or as “functionally related to gene X” where gene X is its syntenous neighbor gene . If the orthologous genes in these pairs had the same annotation but one was poorly annotated , the poorly annotated gene was given an additional score that indicated a synteny-based annotation improvement . Our annotation method was tested against 175 genes of known function in four genomes , two Bacterial genomes and two Archaeal genomes . The organisms used were Escherichia coli K12 MG1655 , Chlamydia trachomatis D/UW-3/CX , Haloferax volcanii , and Sulfolobus solfataricus P2 . These organisms were chosen for three reasons: 1 . They are all very well experimentally characterized and have more than 600 articles on each of them in the ISI Web of Science database 2 . They are sufficiently distant relatives that they pass the significance threshold for using our synteny-based method . It was particularly hard to find a well-characterized Bacterium that was sufficiently distant to E . coli K12 . 3 . With the exception of Chlamydia trachomatis , they all have genetic systems that have been used for a number of years , allowing for genetic confirmation of gene function . We chose Chlamydia trachomatis because it is very distant from E . coli and there have been recent advances in the development of a genetic system for this organism [31] that may lead to future confirmation of our findings . The method was tested in the following manner . Syntenous orthologs were found between Escherichia coli K12 MG1655 and Chlamydia trachomatis D/UW-3/CX , and between Haloferax volcanii and Sulfolobus solfataricus P2 . 88 syntenous orthologs were found between the two Bacteria and 117 syntenous orthologs were found between the two Archaea . Of these , we determined that 145 were unique to one or the other pairwise comparison based on KEGG identifiers and E . C . numbers . 30 genes were potentially shared between the two pairwise comparisons . We were able to analyze a total of 145 unique syntenous orthologs and 30 shared syntenous orthologs , thus 175 genes overall . For these 175 syntenous orthologs , we chose to mask their function in one of the organisms in each pairwise comparison , reannotating the genes as “hypothetical proteins” . We chose to hide the functions of the genes in E . coli K12 in the first comparison and in S . solfataricus in the second comparison . We chose these organisms because they are better characterized than their pair in each case . We then took these “hypothetical proteins” and applied our synteny-based annotation method to them , determining their function solely based on the function of their counterpart in the given comparison . Then we compared the new function attributed to the “hypothetical protein” by our method to the original annotation of the protein . We considered the functions the same if they had the same KEGG identifier [32] or gene name and E . C . number in the cases where the gene did not have a KEGG identifier .
|
Based on trends between gene sequence divergence and gene order divergence over time , we developed a new synteny-based method to refine functional annotation . This method uses these trends to determine the probability that any two syntenous genes ( genes that are sequential in two organisms ) are functionally related . Organisms that are distant relatives have few syntenous genes , but these syntenous genes have a very high probability of functional relatedness . We applied this method to newly assembled genomes of co-occurring , uncultivated acid mine drainage Archaea in order to improve their gene annotations . This application revealed important physiological differences between the co-occurring organisms in this clade , including the ability of some but not all of the Archaea to manufacture vitamin B12 and to carry out anaerobic energy metabolism . We also used this method to identify new genes possibly involved in vitamin B12 synthesis , ether lipid synthesis , molybdopterin synthesis and utilization , and microbial immunity through the CRISPR system .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genomics",
"biology",
"computational",
"biology",
"microbiology",
"genetics",
"and",
"genomics"
] |
2011
|
A Semi-Quantitative, Synteny-Based Method to Improve Functional Predictions for Hypothetical and Poorly Annotated Bacterial and Archaeal Genes
|
Substitutions that disrupt pre-mRNA splicing are a common cause of genetic disease . On average , 13 . 4% of all hereditary disease alleles are classified as splicing mutations mapping to the canonical 5′ and 3′ splice sites . However , splicing mutations present in exons and deeper intronic positions are vastly underreported . A recent re-analysis of coding mutations in exon 10 of the Lynch Syndrome gene , MLH1 , revealed an extremely high rate ( 77% ) of mutations that lead to defective splicing . This finding is confirmed by extending the sampling to five other exons in the MLH1 gene . Further analysis suggests a more general phenomenon of defective splicing driving Lynch Syndrome . Of the 36 mutations tested , 11 disrupted splicing . Furthermore , analyzing past reports suggest that MLH1 mutations in canonical splice sites also occupy a much higher fraction ( 36% ) of total mutations than expected . When performing a comprehensive analysis of splicing mutations in human disease genes , we found that three main causal genes of Lynch Syndrome , MLH1 , MSH2 , and PMS2 , belonged to a class of 86 disease genes which are enriched for splicing mutations . Other cancer genes were also enriched in the 86 susceptible genes . The enrichment of splicing mutations in hereditary cancers strongly argues for additional priority in interpreting clinical sequencing data in relation to cancer and splicing .
As the cost of sequencing technologies is declining , the number of genomes and exomes sequenced is increasing , resulting in an expanding archive of genetic variation in both diseased and healthy individuals [1 , 2] . To keep pace with the ever growing variant archive , in silico tools are being created to determine the functional impact of variants discovered [3–6] . However , most tools used to determine the pathogenicity of variants rely on in silico methods aimed at deciphering protein features associated with the variant and fail to take into account the potential regulatory functions of sequences in gene processing mechanisms and expression [7] . The sequences that encode for proteins ( exons ) and the intervening , noncoding sequences ( introns ) are known to have an important regulatory role in an RNA processing mechanism known as precursor messenger RNA ( pre-mRNA ) splicing . Variants that alter the regulatory regions necessary for splicing typically result in the deletion of large portions of the coding sequence and generally result in a non-functional protein [8] . Among the reported sequence variants , splicing mutations located at the 5′ and 3′ canonical exon-intron boundaries , or splice sites , make up 13 . 4% of the disease-causing mutations reported in the Human Gene Mutation Database ( HGMD ) [9] . However , in addition to splicing variants located at the splice sites , splicing variants within the exonic sequences can also modulate splicing by altering the multitude of exonic splicing enhancers ( ESE ) and silencers ( ESS ) present in exons . Due to the difficulty in classifying exonic mutations as splicing mutations , it is becoming evident that new methods and tools will need to be implemented to correctly and thoroughly identify exonic splicing mutations ( ESM ) . An ESM is a hereditary disease allele that falls within the exon and was originally annotated as a protein coding mutation . For the purpose of this analysis , a splice site mutation ( SSM ) falls within the 5`splice site ( i . e . -3 to +6 position 5`end of the intron ) or the 3`splice site ( i . e . -20 to +3 position of the 3`end of the intron ) . Recently , studies have been aimed at re-analyzing reported sequence variants for splicing defects [10 , 11] . Much of this work suggests that splice-altering variants are more common than previously anticipated . For example , a recent re-analysis of 20 coding mutations located in exon 10 of MLH1 , reveal a high proportion of previously uncharacterized ESM ( 17 of the 20 or 77% ) [11] . In fact , using the position dependence of splicing elements as a measure to infer disruptive splicing , it has recently been predicted that one-third of all disease-causing variants lead to aberrant splicing [12] . Here , we present a comprehensive analysis of splicing mutations in human disease . We report 86 genes enriched for SSM , in patients that present with hereditary disease ( see Materials and Methods ) . Of these 86 SSM-prone genes , three were the main causal genes of Lynch Syndrome ( MLH1 , MSH2 , and PMS2 ) , which account for 32% , 39% , and 14% of Lynch Syndrome cases , respectively [13] . Lynch Syndrome , a cancer-susceptibility disorder caused by autosomal dominant germline mutations in the mismatch repair ( MMR ) genes above , accounts for ~5% of all colorectal cancers . In addition , individuals with Lynch Syndrome have an elevated risk of developing early-onset colorectal and endometrial cancers [14] . With colorectal cancer being the second leading cause of cancer death in the United States [15] , it will be imperative to understand the disease mutational mechanisms underlying Lynch Syndrome to aid in the development of therapeutic strategies . However , not only were Lynch Syndrome genes members of the 86 SSM-prone genes , but it was also found that the COSMIC set of cancer genes were overrepresented [16] . This work highlights the importance of allocating additional priority to investigating splicing defects in a described set of genes , many of which have been associated with some feature of cancer risk or progression .
A recent analysis of coding mutations located in exon 10 of MLH1 revealed a high level of coding mutations ( 17/22 or 77% ) altered the splicing of exon 10 [11] . To see if the results of this survey of MLH1 exon 10 was indicative of high levels of splicing phenotypes in exonic mutations across all genes , a larger pool of exonic variants ( outside canonical splice sites ) was analyzed using a high-throughput reporter assay , MaPSy [10] . MaPSy was used to screen variants in five additional MLH1 exons . Of the 36 pathogenic MLH1 exonic mutations surveyed with MaPSy , 11 ( 30 . 5% ) affected splicing ( Fig 1A and 1B , S1 Table ) in an in vivo minigene assay and in an in vitro splicing assay . On average , disease causing point mutations disrupt splicing 10% of the time ( MaPSy 5K panel , n = 4 , 964 alleles ) [10] . In other words , the rate of splicing misregulation in MLH1 disease alleles was almost three times higher than the background rate of splicing disruption in disease alleles . Mapping potential exonic splicing regulatory sequences ( ESRs ) [17] in the MLH1 exons analyzed in MaPSy revealed exon mutations that altered splicing resulted in a greater difference in wild type ( wt ) –mutant ( mt ) ESR scores than mutations not resulting in a splicing defects ( average ∆ESR score 1 . 845 and 0 . 8583 respectively , P = 0 . 0280 Mann-Whitney , S1 Fig , S1 Table ) . MLH1 missense and nonsense mutations were found to frequently disrupt splicing in vitro and in vivo: 6/22 ( 27% ) missense mutations and 5/14 ( 36% ) nonsense mutations . Taken together , this data a ) confirms the previous report that exonic mutations in MLH1 frequently disrupt splicing b ) exonic mutations that alter ESR signals are more likely to result in a splicing defect , and c ) suggests that the rate of splicing disruption is not homogenous across genes ( i . e . MLH1 is an outlier ) . Interestingly , ESMs were also disproportionately distributed among the exons within the MLH1 gene . Of the five exons that were included in this study , three had no ESMs . However , all the exonic mutations in exon 8 ( 6/6 ) and 71% ( 5/7 ) of the mutations in exon 15 significantly altered splicing ( Fig 1A and 1B ) . Thus , it appears that certain exons in MLH1 are more prone to splicing disruption . To investigate the possibility that certain exons may be more prone to ESMs , a permutation approach was used to identify exons that exceeded the expected number of ESMs discovered ( see Materials and Methods ) . 11 of the 2 , 061 exons analyzed using MaPSy were predicted with a P < 0 . 01 to have more ESM than expected ( S2 Fig ) . Remarkably , two of these 11 exons identified in the simulation as being enriched for ESMs were MLH1 exon 8 and exon 15 , further confirming the previous finding . To mechanistically investigate the defective splicing of MLH1 mutations , the representation of MLH1 alleles in the fractions of the in vitro spliceosomal assembly assay was examined ( see Materials and Methods and S3 Fig ) . Here , the accumulation of an allele in intermediate complexes was interpreted as an indication that the allele blocked the next stage of spliceosome assembly [10] . In general splice site recognition is thought to occur early in spliceosome assembly [8 , 18] , however for the ESMs in MLH1 , the disruption occurred later . 63% of exonic splicing mutations were primarily blocked at the A complex in transition to the B complex and 37% were blocked at the B complex ( Fig 2 ) . Several mutants reduce more than one step in the assembly ( Fig 2 ) . As expected , adjacent mutations that were close enough to fall within the same cis-element shared a similar pattern of disruption . In effect , these clusters of variants mutationally defined a particular cis-elements required for particular spliceosomal transitions ( e . g . Fig 2 , CM045463 and CM082944 ) . The surprisingly high fraction of disease-causing splicing mutations both reported in the splice-sites and unreported in exonic positions of MLH1 ( as shown by the MaPSy 5K panel ) may be due to chance or the enrichment for splicing mutations in the gene/disease . To eliminate the null hypothesis , Monte Carlo ( MC ) simulations were used to generate a distribution of SSM frequencies for each gene given the total number of mutations reported in that gene ( see Materials and Methods ) . Of the ~3 , 600 disease genes reported in the HGMD , 86 genes , including the three main casual Lynch Syndrome genes ( MLH1 , MSH2 , and PMS2 ) , had more SSM than expected based on the distribution of SSM in the HGMD dataset ( Fig 3A , S2 Table ) . Although SSM generally have a severe impact on splicing outcome by disrupting the essential interactions with the core spliceosome components , variants located within the exonic sequence can also alter splicing by disrupting the myriad of exonic splicing regulatory ( ESR ) elements [18] . Using the results obtained from the MaPSy 5K panel , we found that the 86 SSM-prone genes not only had a higher proportion of mutations in the canonical splice sites but also contained exonic mutations that were almost twice as likely to disrupt splicing as exonic mutations that occurred in the remaining genes ( 1 . 84-fold effect; P = 1 . 48 x 10−9 , Kruskal-Wallis , Fig 3B ) . These results suggest that the 86 SSM-prone genes are not only prone to SSMs but also to ESMs , with three ESMs in the 86 SSM-prone genes being validated in individual minigene constructs ( Fig 3C ) . We next sought to determine if a certain class of disease genes were overrepresented in the 86 SSM-prone genes ( S2 Table ) . The initial report of an association between MLH1 and splicing mutations also associated other cancer related genes such as BRCA1 , BRCA2 , and NF1 with disrupted splicing . Furthermore , Gene Ontology ( GO ) enrichment analysis [19] of the 86 SSM-prone genes revealed an enrichment of genes associated with the DNA repair pathway ( P = 2 . 53x10-2 , S3 Table ) , a pathway commonly associated with cancer phenotypes [20 , 21] . To determine if cancer genes were overrepresented in the 86 SSM-prone genes , the Catalogue of Somatic Mutations in Cancer ( COSMIC ) was crossed referenced with the HGMD disease genes [16] . Of the 609 cancer genes associated with elevated somatic mutations in tumors ( i . e . the COSMIC gene set ) , 280 were reported with germline mutations in hereditary cancers ( i . e . HGMD ) . These cancer genes were particularly enriched in the SSM-prone genes ( 1 . 5 fold in the upper category 20/86 , P < 0 . 01 , permutation test , Fig 4A ) . Not only were cancer genes overrepresented in the SSM-prone genes , but they also contained 1 . 5-fold more SSM and 1 . 4-fold more ESM than the rest of the genes in the HGMD ( P = 0 . 011 and P = 0 . 0075 , Mann-Whitney , for SSM and ESM respectively , Fig 4B and 4C ) . When further dividing the cancer genes into oncogenes and tumor suppressor genes ( TSG ) , it became apparent that TSG have more SSM and ESM than the rest of the genes in the HGMD ( P = 0 . 0178 and P = 1 . 14 x 10−4 , Mann-Whitney , for SSM and ESM respectively , S4 Fig ) . However , this enrichment for SSM and ESM was not apparent when comparing oncogenes to the rest of the genes in the HGMD ( P = 0 . 4821 and P = 0 . 1914 , Mann-Whitney , for SSM and ESM respectively , S4 Fig ) . Thus , it appears that TSG are more prone to splicing dysfunction most likely due to their loss-of-function disease mutational mechanism . A number of genomic and sequence features have been implicated in the context of splicing [17 , 22–25] . We , therefore , sought to determine if genomic and sequence features existed that would result in the predisposition of a gene to SSM . In fact , multiple features appeared to modulate the predisposition of a gene to SSM . When analyzing 19 genomic features ( S4 Table ) [17 , 23 , 26–28] , we found that the 86 SSM-prone genes contained 2 . 5 fold more introns than the rest of the genes in the analysis ( P = 2 . 54 x 10−14 , Kruskal-Wallis , S5 Fig ) . Thus a trivial explanation for predisposition of the 86 SSM-prone genes is the larger mutational target presented by their higher number of splice sites . To determine if the SSM-prone genes were predisposed due to the number of introns , we repeated the MC simulation normalizing for the number of introns ( see Materials and Methods ) . Surprisingly , this correction did not dramatically alter the result . After normalization , about 74 . 4% ( 64/86 ) of the genes that were significantly enriched for splice site mutations , were present in the recalculated SSM-prone gene list ( S5 Table ) . In addition to having more introns , the 86 SSM-prone genes are generally more haploinsufficient ( HI ) , have shorter and more structured exons ( predicted to have more base-pairing interactions ) , and less conserved variants found in the exomes of ~60 , 000 healthy individuals [26] ( S5 Fig ) . To determine the relative contribution of each feature to the classification , several machine learning approaches were trained on the HGMD mutation dataset . Briefly , the Random Forest ( RF ) [29] and a Logistic Regression ( LR ) predictive models were utilized to predict whether a gene would be associated with a significant excess of SSM ( red dots , Fig 3A; for feature ranking please see Materials and Methods ) . The model indicates that HI genes and genes with less structured exons have a higher risk of being frequently affected by SSM ( Fig 5A ) . In addition to feature prioritization , the classifier was also used to predict additional genes that may be prone to SSM but had not yet been identified as human disease genes . To test the performance of both classifiers , ROC curve analysis was performed . The mean area under the curve was measured for both machine learning models . The RF model was the most predictive ( AUC = 0 . 839 , Fig 5B , see S6 Table for cross-validation ) . A control classifier trained to predict genes that were not prone to SSM ( i . e . Lower-Expected genes , Fig 3A , green ) was considerably less accurate , presumably because this category is lower confidence with fewer associated mutations overall . As haploinsufficiency was an important feature in the prediction of SSM predisposition ( upper category ) and splicing defects generally result in a severe loss of gene function , it is possible that the degree of haploinsufficiency largely determines a genes predisposition to SSM . However , the RF model still performed well with HI removed ( AUC = 0 . 805 ) . Therefore , it does not appear that there is a single dominant feature such as HI or the number of introns that drives the accuracy of the predictor . Instead it is most likely a combination of features that determine a genes predisposition to SSM . This analysis suggests that the prediction of genes predisposed to SSM using a broad spectrum of features is feasible and could potentially be used to identify new disease genes that are prone to splicing mutations . In order to identify new disease genes that are prone to splicing mutations , the predictive model was applied to ~13 , 000 non-disease associated genes ( Fig 5C ) . While the classifier was run at a range of stringencies . Using a probability cutoff of 0 . 6–0 . 86 returned by the classifier , 499 genes were predicted to be SSM-prone ( see Materials and Methods , S7 Table ) . It is possible that these 499 genes were not previously identified as disease genes because their function was required for organismal viability . To explore the degree to which variation can be tolerated in these 499 genes , the aggregated exome sequencing data from 60 , 706 presumably healthy individuals provided by Exome Aggregation Consortium ( ExAC ) [26] was cross referenced with the 499 genes . The 499 predicted SSM-prone genes had significantly fewer reported ExAC splice site ( SS ) region variants than the rest of the testable genomic genes in the analysis ( Fig 5D , P = 6 . 1043e-18 , Mann-Whitney ) . This analysis suggests that the splicing elements in the predicted SSM-prone genes are evolving under a higher level of selective pressure . However , this analysis considers all variations equivalently making no distinction between neutral variants and clear loss of function variants . For the variants that fall within the splice sites , position weight matrix ( PWM ) models can be used to evaluate whether a variant represents a stronger or weaker match to the splice site consensus . In other words , PWM can potentially distinguish loss of function splicing mutants from neutral variation . In this analysis , variants that greatly weaken the match to a splice site model ( e . g . ∆ > 5 , Fig 5E ) and would be expected to result in a loss of function are four fold underrepresented in common single nucleotide polymorphisms ( SNPs ) . This suggests a scenario where loss of function variants are eliminated from the variant pool before the SNP can reach a reasonable frequency in the population . Conversely , variants that fall within the 5′ ss but strengthen the agreement of the site to the consensus tend to accumulate in the high frequency set ( e . g . ∆ < -2 , Fig 5E . ) . The same trend is observed in variants that localize to the 3′ ss ( S6 Fig ) . An independent measure of selection can be found in analysis that maps obvious loss-of function variants to the predicted SSM-prone genes . For example , 3 , 230 genes that were depleted of predicted protein-truncating variants ( PTV’s ) in the exomes of 60 , 706 individuals are a gold standard for genes in which loss of function variants are poorly tolerated [26] . While PTV depletion is unrelated to splicing , there is a four or five-fold enrichment of predicted SSM-prone genes in this dataset ( S7 Fig , P = 7 . 53e-98 Fisher’s Exact , S7 Table ) . The lower proportion of ExAC variants located in the genomic genes predicted to be SSM-prone and the enrichment of PTV-intolerant genes in the SSM-prone genes suggests that they are intolerant to variation and appear to be functionally important genes . It is therefore more likely that splice disrupting variants that map to these genes will be deleterious . To gain more insight into the uncharacterized set of predicted SSM-prone genes , GO Enrichment analysis was performed . Regulation of cell cycle ( P = 2 . 20e-2 ) and mitosis ( P = 5 . 08e-5 ) were the two functions enriched in predicted SSM-prone genes ( S8 Table , for individual GO term associations see S7 Table ) . Since the hallmark of cancer is generally the abnormal growth and division of cells , it is possible that mutations within this set may play some yet undiscovered role in cancer . While a more complete characterization of these genes awaits future study , an online browser has been developed to visualize the splicing results of the exonic mutations assessed in the SSM-prone cancer genes studied using MaPSy ( S9 Table ) .
High rates of splicing disruption were reported in the literature for exonic variations in a panel of exons in medically important genes [10 , 11 , 30 , 31] . As there have been a wide variety of estimates of the degree to which splicing defects accompany disease-causing mutations , this current study was initially intended to perform this analysis at a larger scale . The query was expanded to include both exonic and splice site mutations in the set of human genes known to cause hereditary disease . This analysis confirmed the initial reports of high mutation rates in the genes studied but also demonstrated that the degree to which splicing causes disease varies significantly from gene to gene . Recent analysis of mutations in MLH1 , a mismatch repair gene tied to Lynch Syndrome , indicated a high degree of splicing disruption as a common disease mechanism of exon 10 . Due to Lynch Syndrome’s highly penetrant nature in inherited colorectal cancer predisposition , understanding the pathogenesis of the syndrome will be fundamental in devising treatment methods . To further analyze the disease mechanisms in MLH1 , 36 additional exonic mutations were tested with 31% disrupting splicing ( Fig 1 ) . The degree to which exonic mutations affect splicing also vary across exons . For example , in MLH1 , all of the ESM occurred in two of the five exons tested ( Fig 1A ) . Earlier work on spliceosome assembly suggested a mechanism where the spliceosome ‘commit’ to splice sites early in the process [32] . In contrast , many of these mutations that disrupted splicing fairly late in the assembly of the spliceosome ( Fig 2 ) . Overall , the MaPSy assay demonstrated a three-fold increase in likelihood that a missense mutation in MLH1 would result in a splicing defect . This study confirms earlier findings of high frequency of splicing defects in MLH1 mutants , but also suggests that the Lynch Syndrome genes , MLH1 , MSH2 and PMS2 , and the other tested genes are outliers and are prone to splicing disruption . A major conclusion drawn from this study is the existence of a class of diseases that are often caused by splicing mutations ( i . e . SSM and ESM ) . The role that splicing defects plays in genetic disease varies across disease genes but genes with elevated SSM also have elevated ESMs ( Fig 3 ) . The discovery of a class of genes prone to splicing mutations , led to an exploration of what features and cellular functions that predisposed splicing genes encode . GO term analysis indicated that many of these genes were involved in cancer initiation and progression . Defining a set of ‘cancer’ genes at the intersection of the COSMIC and HGMD dataset revealed a significant elevation of SSM and ESM in cancer genes , including genes involved in Lynch Syndrome ( Fig 4 ) . Cancer genes are enriched in the SSM-prone genes ( Fig 3A , red category ) . Cancer genes in this category have higher predicted haploinsufficiency than cancer genes associated with lower levels of SSMs ( Fig 4D ) . Machine learning was used to determine other features associated with the SSM-prone genes ( Fig 5A ) . In general , no single feature dominated , rather a combination of features determined whether a disease gene was prone to splicing mutations . However , there are certain properties of splicing mutation that warrant further consideration . Splicing disruptions are potent loss of function mutations . This property probably explains the evidence of haploinsufficiency in the SSM-prone genes . Finally , unlike protein coding variants , splicing variants could have tissue specific affects . Consistent with a model of tissue specific affects , Lynch syndrome causes a wide variety of cancer types . While beyond the scope of this work , further studies will be needed to explore tissue specific differences in splicing for Lynch syndrome mutations . As there is a high medical importance in discovering new cancer genes , the random forest classifier that was trained on the set of 86 SSM-prone genes was applied across the entire genome to reveal a set of 499 predicted SSM-prone genes . One possibility is these 499 SSM-prone genes could be targets of splicing factors that contain dominant oncogenic mutations ( e . g . SF3B1 , U2AF1 ) [33–35] . Highly significant enrichment in the overlap between the targets of these driver mutations and SSM-prone genes was observed . However , this enrichment disappeared when a correction for intron number was applied to the analysis . While little is known about this novel set of genes , the mark of purifying selection is evident in the degree of variation tolerated in these genes . Using the ExAC dataset , significantly fewer variants are tolerated within splice site regions in the predicted SSM-prone genes . Stratifying these variants by the degree to which the mutation disrupts the splice site suggests a strong selection against splicing mutations in common SNPs . In other words , variants that significantly decrease the PWM scores at the 5′ ss and 3′ ss are underrepresented in common SNPs implying that they are removed by natural selection before they reach MAF >0 . 01 in the human population ( Fig 5E , S4 Fig ) . The finding that more than half of the 499 predicted SSM-prone genes also do not tolerate premature stop codons is further indication of strong selection ( S5 Fig ) . While it is beyond the scope of this work to define the role and function of each of these genes , there is an indication that many relate to cancer . Of the 12 GO terms enriched in this set , 4 categories were also associated with the original set of cancer genes suggesting the existence of novel cancer genes ( comparison of COSMIC cancer gene GO terms and 499 predicted SSM-prone gene GO terms ) . Taken together these findings suggest a set of genes that should be prioritized in the analysis of clinical sequencing data with a particular emphasis on cancer .
The 36 exonic MLH1 mutations assessed for splicing defects mapped to internal exons and were selected based on their classification of being disease causing ( DM ) with a previously undocumented role in splicing . The splicing efficiency of wildtype and mutant exons was calculated as below: log2 ( spli∕∑i=1nsplinpi∕∑i=1ninp ) where spli is the count for spliced output i , inpi is the count for input i , and n is the number of species that were analyzed in the library pool . MaPSy experiments in vivo and in vitro were performed as previously described [10] . Briefly , solid-phase oligonucleotide synthesis technology was used to generate a 200 nt fragment ( 200-mer ) that included both the wildtype and mutant exons , 15 nt of the downstream intron and ≥55 nt of the upstream intron , and were flanked by 15-mer common primer sequences . The in vivo splicing reporters were generated using overlapping PCR and consists of the Cytomegalovirus ( CMV ) promotor , Adenovirus ( pHMS81 ) exon with part of its downstream intron at the 5′ end , followed by the 200-mer library , and exon 16 of ACTN1 with part of intron 15 and the bGH polyA signal sequence at the 3′ end . The resulting in vivo reporters were transfected into human embryonic kidney hek293T cells . After 24 hours of transfection , RNA was extracted and both the input reporters and spliced species were sequenced . The in vitro splicing reporters have a similar design to the in vivo reporters , but exclude the ACTN1 exon , and the CMV promoter was replaced with the T7 promoter . The in vitro splicing reporters were obtained through in vitro transcription using T7 RNA Polymerase . The resulting RNA was then used for splicing reactions in 40% HeLa-S3 nuclear extract . Pools of the input and spliced RNAs were converted to cDNA and prepped for deep sequencing . The allele ratios between wildtype and mutant exons in the different spliceosomal fractions were obtained as follows: log2 ( mie/miimje/mji ) where mie and mii is the counts for the minor allele in the selected pool and input , respectively , mje and mji is the counts for the major allele in the selected pool and input , respectively . For each wildtype-mutant pair , the allele that splices more efficiently is assigned as the major allele . Wildtype and mutant sequences of exon 15 of MLH1 ( NM_000249 . 3:c . 1684C-T ) , exon 2 of BRCA1 ( NM_007294 . 3:c . 5425G-T ) and exon 12 of OPA1 ( NM_015560 . 2:c . 1199C-T ) were synthesized by Synbio Tech ( Monmouth Junction , NJ ) and incorporated into MaPSy in vivo backbone ( Adenovirus ( HMS81 ) and ACTN1 exon 15 by overlapping PCR [10] . MaPSy constructs were transfected into 293T cells and RNA were extracted after 24 hours . RT-PCR were subsequently performed and ran on 1 . 5% agarose gel , as previously described [10] . Hexamer ESEs and ESSs were downloaded from published data ( 17 ) . A sliding window of 1 nucleotide was used plot the predicted ESEs and ESSs in the MLH1 exons assayed with MaPSy ( S1 Fig ) . The ‘ESR wt/mt difference’ in S1 Table was computed as the wild type-mutant difference in hexamer scores ( 17 ) . Disease causing splicing and coding sequence mutations ( DM–disease mutations ) were selected from the Human Genome Mutation Database ( HGMD ) . The mutations were then classified as SSM , missense , or nonsense mutations . To be considered an SSM , the variant was required to be within the canonical splice-sites ( -3 to +6 positions at the 5′ ss and -20 to +3 at the 3′ ss ) and labeled as a splicing mutation by HGMD . The number of missense , nonsense , and SSM mutations were determined for each intron-containing gene . The list of 86 SSM-prone genes from HGMD and the list of 499 predicted SSM-prone genes were analyzed for the enrichment of specific GO terms using the PANTHER GO-Slim Biological Process annotation data set provided by the PANTHER Classification System . The list of cancer genes provided by the Catalogue of Somatic Mutations in Cancer ( COSMIC ) was downloaded and intersected with the list of HGMD genes . A permutation test was then performed to determine if cancer genes were overrepresented in the SSM-prone genes . ESS , ESE , and ESR’s were downloaded from published data [17] and the density was calculated by dividing the total number of regulatory elements by the length of the exonic sequences and averaging the density per gene . SNP density was calculated using the list of common SNPs ( MAF > 0 . 01 ) provided by exome consortium [26] and dividing by the length of the exonic sequence ( ‘Exon SNP dens’ ) or the length of the gene ( ‘Gene SNP dens’ ) . Conservation was scored using PhastCons46way placental for both the exonic sequences ( ‘Exon Cons’ ) and coding sequence ( ‘Gene Cons’ ) . The free energy estimate ( ∆G ) was computed using RNAfold [27] , with default settings for both the exonic sequences ( ‘Exon ∆G’ ) and the for 70 nucleotides up- and down-stream of the splice-sites ( ‘SS ∆G’ ) . Haploinsufficiency scores were obtained from a previous study that developed a haploinsufficiency prediction model using a large deletion data set ( Wellcome Trust Consortium ) [23] . Splice site strength was calculated using perl scripts from the MaxEntScan [28] . ExAC variant conservation was determined using the intersection of the ‘phastCons100way’ track with ‘ExAC Variant’ locations over each gene reported in the HGMD . The intersection generated an average conservation score for the variant sites in each gene based on a zero to one scale . R implementation of random forest , package ‘randomForest’ [29] , was used to determine the individual contribution of various functional genomic features ( see ‘Random forest predictor variables and features’ methods section ) in distinguishing SSM-prone genes from non-SSM-prone genes and to generate a predictive model . ‘randomForest’ is a nonparametric ensemble learning method where individual trees ( kth trees ) in a forest are constructed based off a different sub-sample ( bootstrap sample ) from the original training set and then averaged to provide unbiased estimates of predicted values . Two-thirds of the training set was used for the construction of the kth trees with the remaining one-third ( out-of-bag data ) used for cross-validation and estimates of variable importance . Default parameters were used to construct the random forest model , with the exception that ‘strata’ was used to sample the majority class ( genes with the expected number of SSM ) to make the frequency of the expected class closer to the frequency of the rarest class ( genes with more SSM than expected ) . Variable importance was measured by the degree of model accuracy decrease with the permutation of a single predictor variable . The larger the mean decease in accuracy , the more important the variable is deemed in the classification of the data . R implementation of logistic regression , ‘glm ( ) ’ function , was used to generate a predictive model for distinguishing SSM-prone genes from non-SSM-prone genes . Logistic regression is a classification method that relies on fitting a regression curve given a set of predictor variables and categorical response variables . Again , two-thirds of the data was used to construct the model with the remaining one-third of the data used for cross-validation . Default parameters were used to construct the logistic regression model , with the exception that ‘family = ‘ was set to binomial . The random forest model generated from the HGMD dataset was then applied to the rest of the testable genes in genome . Testable genes were required to be void of a previously described disease phenotype by HGMD , contain introns , and have sufficient genomic feature data . This resulted in ~13 , 000 genes that could be tested using the random forest predictive model . The ‘predict ( ) ’ function with ‘type = ‘ set to ‘prob’ was used to predict SSM-prone genes based on a probability estimate . A probability threshold of > 0 . 6 was set to select SSM-prone genes , which resulted in 499 predicted SSM-prone gene . All low frequency ( MAF < 0 . 01% ) single nucleotide ExAC variants reported in the splice site regions of genes ( -3 to +6 position at the 5′ ss and -20 to +3 position at the 3′ ss ) were counted for each gene and divided by the number of SS’s . The list of ExAC SS region variants per SS was then intersected with the genomic genes tested using the random forest model . The genes were then divided into genes predicted to be SSM-prone ( n = 497 , after intersection ) and genes predicted with a high probability ( prob > 0 . 6 ) to have the expected number of SSM ( n = 5995 ) . The average ExAC SS region variants per splice site were plotted for genes predicted to be prone SSM and genes with the expected number of SSM . The 499 predicted SSM-prone genes were intersected with RefSeq database and only the ones having RefSeq transcript id were retained for the downstream analysis ( n = 486 ) . All ExAC variants that fall within the splice sites ( both 3′ and 5′ ) of the 486 genes were scored using the Maximum entropy model for splice sites ( PMID: 15285897 ) . The ExAC variants were separated based on their minor allele frequency into rare ( MAF < 0 . 01% ) and common ( MAF > 1% ) . The entire distribution of scores and the two classes of alleles were plotted . The collapsed plots based on splice site score threshold were also generated . The list of 3 , 230 genes depleted of predicted PTV’s in ExAC ( PTV-intolerant ) were intersected with the list of genomic genes analyzed with the random forest model . 1 , 746 PTV-intolerant genes were analyzed using the random forest model . 281 of 1 , 746 were predicted to be prone to SSM . The intersection of the two datasets was plotted as a Venn diagram and significance was determined using the Fischer’s exact test .
|
To understand the extent to which disrupted pre-mRNA splicing causes human disease , we re-analyzed coding mutations in MLH1 , one of the causal genes of Lynch Syndrome . We found that a high fraction of the MLH1 coding mutations resulted in disrupted splicing . To further investigate a more general role of defective splicing across human disease genes , simulation strategies were used to identify 86 disease genes prone to splice site mutations . In these 86 genes , there was an enrichment of cancer genes including the three main casual genes of Lynch Syndrome ( MLH1 , MSH2 , and PMS2 ) . Thus , it appears defective splicing may be the main driver of Lynch Syndrome and other cancers . Genes prone to splicing mutations have certain features that allow for the comprehensive prediction of splicing-prone diseases genes in the human genome . Our findings strongly argue for additional clinical sequencing prioritization in both cancer genes and genes prone to splice site mutations .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"genetic",
"diseases",
"mutation",
"hereditary",
"nonpolyposis",
"colorectal",
"cancer",
"forecasting",
"mathematics",
"statistics",
"(mathematics)",
"genome",
"analysis",
"nonsense",
"mutation",
"molecular",
"biology",
"techniques",
"autosomal",
"dominant",
"diseases",
"research",
"and",
"analysis",
"methods",
"exon",
"mapping",
"genome",
"complexity",
"gene",
"mapping",
"mathematical",
"and",
"statistical",
"techniques",
"statistical",
"methods",
"molecular",
"biology",
"clinical",
"genetics",
"point",
"mutation",
"genetics",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"genomics",
"gene",
"prediction",
"computational",
"biology",
"introns"
] |
2018
|
Hereditary cancer genes are highly susceptible to splicing mutations
|
Chagas cardiomyopathy , caused by the protozoan Trypanosoma cruzi , is characterized by alterations in intracellular ion , heart failure and arrhythmias . Arrhythmias have been related to sudden death , even in asymptomatic patients , and their molecular mechanisms have not been fully elucidated . The aim of this study is to demonstrate the effect of proteins secreted by T . cruzi on healthy , isolated beating rat heart model under a non-damage-inducing protocol . We established a non-damage-inducing recirculation-reoxygenation model where ultrafiltrate fractions of conditioned medium control or conditioned infected medium were perfused at a standard flow rate and under partial oxygenation . Western blotting with chagasic patient serum was performed to determine the antigenicity of the conditioned infected medium fractions . We observed bradycardia , ventricular fibrillation and complete atrioventricular block in hearts during perfusion with >50 kDa conditioned infected culture medium . The preincubation of conditioned infected medium with chagasic serum abolished the bradycardia and arrhythmias . The proteins present in the conditioned infected culture medium of >50 kDa fractions were recognized by the chagasic patient sera associated with arrhythmias . These results suggest that proteins secreted by T . cruzi are involved in Chagas disease arrhythmias and may be a potential biomarker in chagasic patients .
Chagas disease is an important public health problem in Latin America currently affecting an estimated 8 million people in 21 countries and spreading by human migration to a number of non-endemic regions [1] . The protozoan Trypanosoma cruzi is the etiologic agent of the disease in mammals; the parasite is transmitted by blood-sucking triatomine bugs , blood transfusions or trans-placentally [2] . This illness is characterized by an acute phase , which is generally asymptomatic , or oligosymptomatic , an indeterminate phase , which may persist for several years , and a chronic phase in which dilated cardiomyopathy and arrhythmias are primarily observed . Chagas cardiomyopathy has been attributed to an alteration in intracellular ions , an imbalance between adrenergic and cholinergic innervations , to cellular and humoral autoimmunity and to parasitic effects or micro-ischemic disturbances [3] . Cardiac arrhythmias are one of the most important alterations in Chagas heart disease and may be associated with sudden death [4] . The principal disorders reported are atrial , ventricular extrasystoles , intraventricular and/or AV conduction disturbances and primary ST-T wave changes [5] . Classically , arrhythmias have been linked to autonomic dysfunction [6] , anti-adrenergic and anti-cholinergic autoantibodies [7] and to wall motion abnormalities [8] . Although , Chagas patients may present with arrhythmias and sudden death in the absence of ventricular dysfunction ( known as the arrhythmogenic form ) [9] , the causes associated with nonstructural arrhythmias are poorly understood . Notably , ST and T abnormalities , ventricular and supraventricular arrhythmias and low voltage QRS have been reported in a recent acute outbreak characterized by high parasitemia [10] , which suggests that the secreted proteins of the parasite may be involved in arrhythmia generation . The interaction between the parasite and the host cell has gained attention in the pathophysiology of Chagas disease . Parasite surface proteins , such as mucins , trans-sialidases and mucin-associated proteins ( MAPs ) , are adhesion factors involved in the invasion of the host cell [11] . Additionally , these proteins are able to increase the intracellular calcium concentration to facilitate the entry of the parasite [12 , 13] . Interestingly , a recent report described a calcium overload in the ventricular myocytes of chagasic patients [14] that may be related to parasite signaling and could be responsible for the arrhythmias observed in Chagas disease . However , there are few reports that have linked protein secretion by T . cruzi with arrhythmias [15] . Consequently , the aim this study is to demonstrate that proteins present in T . cruzi-conditioned medium are able to produce arrhythmias in an isolated beating rat heart model .
Written consent from all patients involved in this study was obtained prior to processing the samples . The collection of serum samples from adult Chagas disease patients’ was approved by the Bioethics Committee for Human Research of the Universidad Centro Occidental Lisandro Alvarado , Barquisimeto , estado Lara , Venezuela ( IVIC-DIR-0480/09 ) . Data on human subjects was analyzed anonymously and clinical investigations have been conducted according to the Declaration of Helsinki . For animal experimentation the project was also approved by the COBIANIM ( IVIC-DIR-0376/1509/2014 ) . COBIANIM is an advisory body of IVIC , with national reach , in regards to the ethical use of animals in research , in accordance with national and international standards . This committee oversees all research activities at IVIC , requiring the use of animals and wildlife to meet with Venezuelan law and universal ethical values . The Commission assessed the methodological , bioethical and legal aspects of this project ( by resolution IVIC/Nro . 127/November , 4 , 2009 , according to the Código de Bioética y Bioseguridad , Ministerio del Poder Popular para Ciencia , Tecnología e Industrias Intermedias Fondo Nacional de Ciencia , Tecnología e Innovación , 2008 and , Guide Laboratory Animals for care and use , Eighth Edition , www . nap . edu , see details in http://www . ivic . gob . ve/cobianim/ ? mod=proyectos . php ) . Sprague Dawley IVIC female rats ( 300–400 g . ) were obtained from the Bioterio Services at IVIC . The animals were allowed to acclimate for 2 weeks prior to the study . They were housed in a clean wire mesh cages ( 10 rats per cage ) and maintained under standard laboratory condition of 12 hours natural light and 12 hours darkness at ambient room temperature . The rats were fed on pellets and water was made available ad libitum . Vero ( African green monkey kidney cells , ATCC CCL-81; American Type Culture Collection , Rockville , Md . ) were maintained at 37°C and 5% CO2 in complete Minimum Essential Medium ( Mediatech , Herndon , Va . ) containing 10% heat-inactivated fetal bovine serum ( FBS; Gibco-BRL , Gaithersburg , Md ) , 20 mM HEPES , 2 mM L-glutamine , 1 mM sodium pyruvate , and 50 μg of gentamicin/ml ( all Sigma-Aldrich , St Louis , MO ) . Confluent Vero cultures plated in a 75 cm2 Easy Flask were infected with 2x105 EP strain trypomastigotes/ml . EP human strain at a rate of 2 parasites per cell . The EP strain of T cruzi was isolated from a fatal human case in 1967 as describe by Contreras et al [16] . Free parasites were removed after 24 hours and the complete medium was changed at this point to medium FBS-free . The fifth or sixth day post-infection the conditioned serum-free medium was collected . The criteria for harvesting of the conditioned infected medium ( CMi ) were that a minimum of 75% of the Vero cells should remain adhered and that at least 2 . 5 x 106 trypomastigotes/ml should be present in the supernatant . The medium was centrifuged at 3000 x g for 10 minutes to separate the parasites , and the supernatant was subsequently filtered using a 0 . 2 μm membrane filter ( Millipore , Billerica , MA , USA ) before being stored at -20°C until use . Control medium ( CMc ) was collected from uninfected Vero cells cultured under the same conditions . Cells were enzymatically detached with a 1:1 mixture of trypsin ( 0 . 25% v/v ) and EDTA ( 0 . 25% ) , were lysed at 4°C with ten rounds of sonication for 20 s at power level 3 with a 550 Sonic Desmembrator ( Fisher Scientific , Pittsburgh PA , USA ) . The suspension was centrifuged at 3000 x g for 10 minutes at 4°C to separate the cell debris , and the supernatant was subsequently filtered using a 0 . 2 μm membrane filter ( Millipore , Billerica , MA , USA ) before being stored at -20°C until use . The supernatant fluid which had been decanted was concentrated by using low-protein binding membrane Diaflo ( Millipore , AmiconCorp . , Cambridge , Massachusetts ) ultrafiltration cell operated in a cold room ( 4°C ) at 50 psi . The Diaflo Model 50 ultrafiltration cell is provided with internal stirring and has a capacity of 40 ml . Supernatant from T . cruzi-infected cells ( henceforth CMi ) and those from uninfected Vero cells ( henceforth CMc ) SFV-free were passed through a ultrafiltration membranes . In each round , the fraction was concentrated to a volume of 10 ml ( 4x ) and washed three times with milli-Q water to remove the lower molecular weight proteins . The pore size cutoff used were 300 , 50 , and 10 Kda , and the ultrafiltrates obtained were a ) >300 Kda , b ) <300 and > 50 Kda and c ) <50 Kda [17] . After wash , each fraction was reconstituted to a final protein concentration of 5 μg/ml using Minimum Essential Medium ( MEM; 1 . 8 mM CaCl2 , 0 . 81 mM MgSO4 , 5 . 33 mM KCl , 117 . 24 mM NaCl , 1 . 0 mM NaH2PO4 , 5 . 56 mM D-Glucose , with L-Glutamine , Phenol Red and essential amino acids ) . The osmolalities of the reconstituted medium were measured using an osmometer ( model Osmette A , Precision Instruments , Sudbury , MA ) and were adjusted at 295–300 mOsm and pH to 7 . 4 with HCl . Finally , three concentrated fractions , corresponding to three molecular weight ranges , were obtained for the CMc and the CMi as follows: unfiltered , > 50 kDa and < 50 kDa . For beating heart experiments , hearts were removed from adult female Sprague Dawley rats ( weighing 300–400 g ) previously anesthetized via intraperitoneal injection of pentobarbital ( 40 mg/kg ) . The isolated hearts were placed in cold Tyrode’s solution ( 25 mM sodium bicarbonate , 10 mM glucose , 116 mM NaCl , 3 . 3 mM KCl , 2 . 5 mM Ca2Cl , 1 mM MgSO4 . and cannulated through the aorta . The hearts were perfused in a retrograde manner with warm Tyrode’s solution ( 37°C ) for 30 minutes . The heart rate data were collected using the isolated beating heart system ( PowerLab ADInstruments , Sydney , NSW , Australia ) . The presence of a sinus rhythm , a heart rate ( HR ) greater than 160 bpm , a perfusion pressure higher than 30 mmHg , and a flow rate of 2–6 ml/min were considered to be stable values for all experiments ( Fig . 1 , panel A ) . The isolated hearts were perfused with conditioned media in accordance with a protocol consisting of three consecutive recirculation-reoxygenation cycles ( Fig . 1 , panel B ) . During the first recirculation stage , 5 mL of deoxygenated MEM was perfused in a closed circuit at a flow rate of 6 mL/min . To avoid cardiac cell damage due to anoxia , a 2 mL/min parallel flow of oxygenated Tyrode’s solution was added , whereby a standard flow rate of 6 mL/min was achieved with partial oxygenation . During the second recirculation stage , the hearts were perfused with: 1 ) Vero cell lysate or >50 kDa fraction ( 5 μg/mL ) of 2 ) CMc ( n = 5 ) , 3 ) CMi ( n = 8 ) and 4 ) CMi preincubated with serum samples of chagasic patients ( n = 4 ) , henceforth CMi+S . During the third stage , the hearts were perfused with MEM . Hearts in treatment received stabilization perfusion by one reoxygenation stage . The reoxygenation stage consisted of perfusing the heart with oxygenated Tyrode’s solution at a rate of 6 mL/min . All experiments were carried out under pression and perfusion controlled condition . In this study , we used a Bolztmann equation for fitting to HR kinetics during recirculation-reoxygenation protocol as previously described [15] to model the I/R process . A panel of Venezuelan Chagas patients’ samples of a sera collection compiled from 2002 to 2006 by the Laboratory of Parasites Physiology and the Cardiology Service of Instituto de Medicina Tropical , Universidad Central de Venezuela , was used for physiological assays . Inclusion criteria were Chagas disease diagnosed by two distinct Chagas tests , serological diagnosis was made with Cruzi ELISA kit as recommended by the manufacturer [18] and polymerase chain reaction ( PCR ) using a high conservation of DNA kinetoplast sequences ( S35 5’-AAA TAA TGT ACG GG ( T/G ) GAG ATG CAT GA-32and S36 5’-GGG TTC GAT TGG GGT TGG TGT-3′ ) in T . cruzi allows detection of the parasite by means of an amplicon of 330 bp [19] . In this study , we used sera from Chagasic patients’ classified as Class II according to the New York Heart Association ( NYHA ) functional classification system: A serum of Class II Chagasic patients’ without arrhythmias was used as control in western blot analysis . To demonstrate preservation of myocardial function ex vivo during recirculation-oxygenation process , we determined aspartate aminotransferase ( AST ) activity by sampling the effusate . These activities were carried out in 10 μl aliquots of coronary effluent [20] in order to provide indication of cardiac damage . The samples were collected at the end of each recirculation-reoxygenation cycle . Each sample was immediately stored at -20°C until the analysis was performed . The enzyme activity was measured using a commercially available ELISA kit ( Invelab , Caracas , Venezuela ) in accordance with the manufacturer’s protocol . SDS—PAGE was performed in 6–18% polyacrylamide gradient gels according to the method of Laemmli [21] . Proteins separated by SDS—PAGE were transferred to Immobilon-P filters ( Millipore , USA ) . Fifteen micrograms of protein was loaded into each lane . Membranes were stained with Ponceau red to verify the transfer of the proteins . The membrane was incubated one hour with chagasic sera at a dilution 1/100 and washed three for five minutes in PBS and 0 . 05% Tween 20 . The secondary antibody incubation was performed with peroxidase-conjugated goat anti-human immunoglobulin , diluted 1:5000 . Immunoblots were developed by using diaminobenzidine ( Sigma-Aldrich ) . Proteins were quantitatively assayed by the Lowry’s method as modified by Schacterle et al [22] with bovine serum albumin as standard . In order to quantify the relationships between continuous variables , QT , PR intervals and heart rate at the different experimental conditions , a Canonical Analysis of populations ( CAP ) were performed , then we performed a projection into a maximum information subspace which is known as canonical biplot analysis . CAP calculates the highest possible correlation between linear combinations of the values of studied variables , as within as between the a priori generated groups of individuals . Mean squared error ( MSE ) , confidence interval and parameters functions was estimate on the parametric bootstrap methods for bias correction in linear mixed model ( n = 250 ) [23] . Bootstraps were performed using InfoStat professional version ( http://www . infostat . com . ar/ ) and the canonical biplot were carried out with MULTBIPLOT software [24] .
ECG recordings from hearts perfused with the different media are shown in Fig . 2 . The panels A , B , C , D and E represents the ECG recordings of five hearts , basal recording corresponds to the internal control condition of each ECG heart . The hearts were subjected to three-recirculation and reoxygenation cycles ( see fig . 1B ) . The recordings corresponding to the perfusion of hearts with either Vero cell lysate ( Fig . 2A ) or CMc ( Fig 2B ) did not show significant change at any stage of perfusion . In hearts perfused with CMi ( Fig . 2C and 2D ) , a complete AV block with prolonged asystole was observed , together with an episode of ventricular fibrillation . In Fig . 2 , panel E is a representative experiment of four co-incubations tested independently , of T . cruzi infected Vero cells conditioned media plus chagasic patients’ sera ( CMi+S ) . Notably , it should be noted that chagasic patient’s sera preincubation abolished the observed effects when the hearts were perfused with >50 kDa proteins Cmi . Also , a partial reversion was observed in the 3rd cycle , and a total reversion was achieved with Tyrode perfusion ( Fig . 2E ) . We did not observe any abnormal cardiac conduction effects in the ECG recordings taken from hearts perfused with the <50 kDa fractions . It is known that the parasites secrete several bioactive lipids and glycolipids , thus lipid compounds were removed from conditioned medium and tested , and did not observe any abnormal cardiac conductions effect . Fig . 3A shows the percentage of heart rate in three frame times in relation to the recirculation-reperfusion cycles . Three groups were defined by the heart rate in different frame times: Group 1 , from 0 to 170 s , Group 2 , from 171 to 280 s and , Group 3 , higher than 280 s . The adjusted model represents the effect on heart rate kinetics for the three experimental conditions that resulted from perfusing hearts with the different fractions >50 kDa corresponding to CMc , CMi and CMi+S ( S2 Table Rawdata Fig . 3A ) . The adjusted linear mixed models ( LMMs ) shows that significant differences between conditions , times and its interaction ( p≤0 . 05 ) . The lowest heart rate percentage was found for the CMi and the highest for CMc . The average heart rate corresponds to 94 . 21 ± 1 . 16; 135 . 77 ± 1 . 18 and 150 . 61 ± 0 . 78 percent’s , for Groups 1 , 2 and 3 , respectively ( see Fig . 3A ) . The comparison from CMi curve with Boltzmann model shows , in group 1 , a significant decrease of initial HR parameter . These alterations , especially bradycardia , may be related to AV blockade observed in EGC ( Fig . 2C ) . Remarkably , these changes were reversed during the reoxygenation stage of the same heart ( Fig 2C and 2D ) and/or by incubation of CMi+S ( Figs . 2E and 3A ) . Additionally , the Boltzmann analysis shows discrete changes of inflection times in hearts perfused with CMi and CMi+S , suggesting a HR recovery delayed effect during the early re-oxygenation stage probably related to immunogenic proteins ( Fig 3 , Groups 2 and 3 ) . Also , we found difference between the initial HR values in the recirculation cycle in Cmi , this result suggests an involvement of secreted proteins in bradycardia . This bradycardia was independent of the QT interval over 450 s ( Figs . 3A and 3B ) . The Figs . 3B and 3C show the QT and PR segments estimated from EGC of isolated hearts perfused with CMc , CMi and CMi+S . This estimation was carried out over 2700 s . In both , QT and PR segments by LMMs adjusted analysis showed that condition CMi had the most variability observed compared with control condition ( p≤0 . 05 ) . The lowest QT interval was found for the CMi+S and the highest for CMi . The highest PR interval was observed was found for the CMi . This model demonstrates a sinusoidal behavior that cannot be associated with a tendency towards an increased PR interval ( Fig . 3C ) . To evaluate cardiac cellular damage during the recirculation-reoxygenation protocol , we collected the media after each recirculation-reoxygenation cycle and measured AST activity ( Fig . 4 ) . The arrows identify the times at which the samples were taken . Rc1: recirculation 1; Rc2: recirculation 2; Rc3: recirculation 3; Ro1: reoxygenation 1; Ro2: reoxygenation 2; Ro3: reoxygenation 3 . We observed that AST activity remained at basal levels during the entire protocol in hearts perfused with proteins >50 kDa CMi and CMc , which suggests the absence of cardiac cellular damage . Fig . 5 describes the projection at different times of the variables and , the three conditions studied such as CMc , CMi and CMi+S , whose correlation was determined by Mahalanobis distance . The sum of percentages of two axis selected in the canonical biplot explain almost 100% of the variance with a very high quality of representation of the groups CMc ( 91 . 6% ) and CMi ( 99 . 3% ) in the canonical axis 1 and of CMi+S ( 56 . 3% ) in the canonical axis 2 ( S1 Table Rawdata canonic biplot ) . The goodness of fit of the variables into the canonical subspace demonstrate that 16 of 55 variables were represented with quality over 80% ( HR170:93 . 02 , HR120: 91 . 2 , HR180: 90 . 37 , HR130: 89 . 79 , HR160: 89 . 55 , HR60: 88 . 96 , HR150: 88 . 69 , PR2700: 88 . 23 , HR70: 86 . 77 , HR50: 86 . 16 , PR900: 85 . 41 , HR30: 83 . 48 , HR40: 83 . 43 , HR80: 83 . 39 , PR1200: 83 . 1 , HR140: 82 . 18 ) 81 . 25% of the variables are related to the heart rate , and the rest are related to PR . Besides , 48 of 55 ( 87 . 27% ) variables has a quality of representation over 50% , and 100% of the individuals were has a quality of representation above 85% in the canonical subspace . To initiate characterization of the proteins in conditioned media that could be responsible for the observed effects on cardiac conduction , we performed western blotting with serum from chagasic patients’ . As shown in Fig . 6 , chagasic serum recognized various ultrafiltrate proteins . More antigenic proteins are in the range of molecular weight 50–200 kDa . In all cases , proteins in CMc were non-antigenic . Notably , chagasic serum obtained from patient without arrythmias did not reveal any bands .
In developing countries a proportion of chagasic patients die undiagnosed . Developing models to understand the pathophysiology of this disease and test new therapies are necessaries . The role of the secretome of the T . cruzi has been gaining attention . Traditional use of secretome protein in serodiagnosis , i . e . antigenic proteins detection by ELISA test called Trypomastigote Excreted-Secreted Antigens ( TESA ) . There are currently no studies which associate T . cruzi secretome proteins as biomarkers for cardiac arrhythmias in Chagas cardiomyopathy . Recently , Wen et al ( 2012 ) [25] resolved the proteome signature of high and low abundance serum proteins in chagasic patients demonstrating the serum oxidative and inflammatory response profile , and serum detection of cardiac proteins parallels the pathologic events contributing to Chagas disease development . In this manuscript , we established a reproducible model using a recirculation-reoxygenation protocol that resembles the early interactions that occur between parasite secretome with cardiac cells . We found that CMi induces a biological effect on healthy , isolated beating rat heart model similar to those observed in Chagas patients . These effects may be related to the direct interaction of proteins into CM on heart cells . Some authors have demonstrated the persistence of the parasite in chronic lesions in patients [26] , which reinforces the hypothesis that tissue damage is related to cellular parasitism in vivo . The detection of significant neuronal cell loss in the sympathetic and parasympathetic nervous systems of Chagas cases , in the absence of T . cruzi in situ , is the basis for the hypothesis of factors released from the parasite nest hidden somewhere in the host body , producing cell damage . [2] . Additionally , there is good in vitro and in vivo evidence for autoantibodies against neuroreceptors ( beta-adrenergic and muscarinic ) in Chagas disease [7 , 27] . In this study , we evaluated the role of the secretome fractions of T . cruzi co-cultured with Vero cells on cardiac arrhythmias using an isolated beating rat heart model . We observed bradycardia , ventricular fibrillation and complete atrioventricular block in hearts during perfusion with >50 kDa CMi . The antigenicity of the secreted proteins was tested by Western blotting using chagasic patient’s sera . The effects observed in this in vitro heart model are different from the results observed in autoimmunity studies [2] , since we detected that immunogenic T . cruzi-secreted proteins was able alter cardiac function independent of a systemic immune response . It should be noted that chagasic patient’s sera preincubation abolished the observed effects when the hearts were perfused with >50 kDa proteins CMi , confirming the relationship between cardiovascular alterations , the immunogenic T . cruzi-secreted proteins and its correlation with arrhythmias in Chagas disease . We have previously [15] demonstrated that cardiac damage can be estimated based on the amount of AST released into the coronary effluent . In the present study , we carried out serial AST enzymatic measurements as an internal control . The enzymatic activity remained stable through the recirculation-reoxygenation protocol , indicating lack of induced myocardial tissue damage . The T . cruzi invasion has been linked to the interaction of parasite membrane glycoproteins with cellular ligands and their associated signaling pathways [28] . Several members of the T . cruzi-mucin family ( TcMuc ) have been linked to the process of invasion and to the increase in intracellular calcium concentration in particular . The secretion of MASP52 , a member of the mucin associated protein ( MASPs ) family , has been associated with parasite attachment and with parasite invasion of Vero cells [29] . A recent report characterized the T . cruzi secretome obtained from medium conditioned by culturing the epimastigote and metacyclic trypomastigote forms under axenic conditions [30] , in this study , it was demonstrated that proteins are shed in vesicles and that 3 . 8% of the secreted proteins are involved in parasite-cell interactions . The study identified the surface glycoprotein GP90 , MASP52 , trans-sialidase , and an 82 kDa glycoprotein , along with additional proteins , as being secreted by T . Cruzi ( Table 1 ) . However , these reports did not evaluate any pathophysiological role of this group of proteins . Ventricular arrhythmias in Chagas patients’ are related to calcium overload [14] . Accordingly the relationship between proteins secreted by the parasite and the regulation of intracellular calcium levels could provide insight into the arrhythmias observed in isolated beating hearts perfused with high molecular weight T . cruzi proteins . Previously , T . cruzi GP82 , [13] GP90 and GP35/50 proteins [31] have been described as involved in both the modulation of calcium increase in the host cell and in determining the invasiveness of the parasite strain . Our group [15] , demonstrated that T . cruzi-conditioned medium was able to increase the frequency of occurrence of tachyarrhythmia and cause a decrease in the heart rate during post-ischemic recovery . A novel approach recently reported by Elliott et al [32] showed that Trypanosoma brucei cathepsin-L supernatant disturbs the heart electrical activity , leading ventricular premature complex ( which cause palpitations ) and triggers arrhythmias in whole rat heart . In the present work , we observed a reversible ventricular fibrillation and a total AV block associated with bradycardia in a non-damage-inducing protocol , the effect was reversed by incubating the infected conditioned medium with chagasic patient’s serum , confirming that a direct interaction between the parasite secreted proteins and cardiomyocytes exist in the pathophysiology of Chagas cardiomyopathy . It is plausible that pro-arrhythmogenic proteins secreted or released by T . cruzi could act as enhancers causing the cardiac conduction system to cross an arrhythmic threshold in Chagas patients . This is the first report that implicates proteins secreted by T . cruzi with arrhythmias in an ex-vivo model . We obtained a reproducible pattern of antigenic recognition of T . cruzi-secreted proteins by patient’s sera suggesting that immunogenic T . cruzi-secreted proteins are implicated with arrhythmias in Chagas disease . T . cruzi proteins could be used as virulence markers in the prognosis of the arrhythmias in Chagas patients . To support the secretome proteins interaction a statistical analysis was carried out in order to quantify the relationships between continuous variables , QT and PR intervals and heart rates at the different experimental conditions ( CMc , CMi and CMi+S ) . We decided to perform a Canonical Analysis of Populations ( CAP ) . This methodology is used to project in a biplot simultaneously the structure of the a priori generated groups and the variables responsible for the separation between them . The CAP is used to obtain canonical axis that reflects the maximum separation between groups , not between individuals . This statistical method has been used in other fields such as econometrics , and recently it has been used in biological systems was reported [33] . In this type of analysis there are two assumptions that should be tested , 1 ) the mean vectors of the groups must be significantly different . This assumption is evaluated by the study of global contrast based on Wilk’s Lambda ( L ) which turned out to be statistically significant 132 . 0474 with a p-value of 4 . 3475e-31 . 2 ) the variances-covariances of the variables of the groups must be equal . The three variances-covariances matrices turned out to be singular which means that the variables have linear relation between them so it is possible to reduce the dimensionality of the system by eliminating variables . This was corroborated by the fact that with two dimensions we can explain almost 100% of the system variance , but it is important to remark that despite of the high quantity of redundant variables , they were ignored because of the use of Mahalanobis distance . Canonical analysis of populations proved that 87 . 27% of the variables were represented with a quality above 50% and , 100% of the individuals were represented with a quality above 85% . This methodology permits us to elucidate two important aspects , first , we demonstrate that with just 16 variables we can explain over the 80% of the information and 13 of them are heart rate variables and the other three are PR variables . This means that the infection phenomena relative to the secretome proteins generated by the host-pathogen interaction could be successfully followed in time just by studying the heart rate . Besides not all time-points are necessary , just the mentioned above . Second , the group projections in the biplot analysis are in agreement with the results shown in Figs , 2 and 3 , mainly in Fig . 2A which we can see the heart rate kinetics . We can infer that there is a significant recovery from bradycardia when the conditioned infected medium were mixed with chagasic patients serum , in other words the secretome proteins could be responsible for the heart dysfunction observed . We are already working on identifying the proteins of the secretome and how are their relationships with the dysfunction of the heart . The contribution of this study was to evaluate , in an isolated heart model , the arrhythmogenic role of the parasite secretome proteins . This reproducible recirculation-reoxygenation model can be an useful system to investigate new drugs for the treatment of Chagas disease arrhythmias . The dissection and simplification of a complex system as constituting Chagas infection becomes necessary to understand and control the disease , in this sense , we wanted to contribute with a simplistic model which evaluates arrhythmogenic factors . This work represents a heuristic contribution to the study of arrhythmias produced by immunogenic proteins secreted by Trypanosoma cruzi in vitro . The findings of this study provide a glimpse into the role of this parasite’s secretome in the pathogenesis of the Chagas’ disease .
|
Chagas disease , caused by the parasite Trypanosoma cruzi , is an endemic disease of Latin-American countries , affecting an estimated 8 million people in 21 countries . It is spread by the bite of triatomine reduvid bug . Due to immigration towards non-endemic regions , the disease can spread and affect people around the world via blood transfusions . Infection usually occurs in childhood , and some patients may develop acute myocarditis; however , most remain asymptomatic for many years before chronic cardiac and/or gastrointestinal manifestations appear . Chagas disease is characterized by an acute phase , which is generally asymptomatic , or oligosymptomatic , an indeterminate phase , which may persist for several years , and a chronic phase in which dilated cardiomyopathy and arrhythmias are primarily observed and sudden death may occur . Once heart failure develops , death usually occurs within several years . In this work , we demonstrate the pathophysiological role of proteins secreted by T . cruzi on cardiac arrhythmias . The antigenicity of these fractions was tested by an immunological test using chagasic patients’ sera associated with arrhythmias . We showed that perfusion of the proteins secreted by T . cruzi , in an isolated beating rat heart model , induced cardiac arrhythmias such as bradycardia and complete atrioventricular block .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[] |
2015
|
Evidence of Reversible Bradycardia and Arrhythmias Caused by Immunogenic Proteins Secreted by T. cruzi in Isolated Rat Hearts
|
Genome-wide association ( GWA ) analyses have generally been used to detect individual loci contributing to the phenotypic diversity in a population by the effects of these loci on the trait mean . More rarely , loci have also been detected based on variance differences between genotypes . Several hypotheses have been proposed to explain the possible genetic mechanisms leading to such variance signals . However , little is known about what causes these signals , or whether this genetic variance-heterogeneity reflects mechanisms of importance in natural populations . Previously , we identified a variance-heterogeneity GWA ( vGWA ) signal for leaf molybdenum concentrations in Arabidopsis thaliana . Here , fine-mapping of this association reveals that the vGWA emerges from the effects of three independent genetic polymorphisms that all are in strong LD with the markers displaying the genetic variance-heterogeneity . By revealing the genetic architecture underlying this vGWA signal , we uncovered the molecular source of a significant amount of hidden additive genetic variation or “missing heritability” . Two of the three polymorphisms underlying the genetic variance-heterogeneity are promoter variants for Molybdate transporter 1 ( MOT1 ) , and the third a variant located ~25 kb downstream of this gene . A fourth independent association was also detected ~600 kb upstream of MOT1 . Use of a T-DNA knockout allele highlights Copper Transporter 6; COPT6 ( AT2G26975 ) as a strong candidate gene for this association . Our results show that an extended LD across a complex locus including multiple functional alleles can lead to a variance-heterogeneity between genotypes in natural populations . Further , they provide novel insights into the genetic regulation of ion homeostasis in A . thaliana , and empirically confirm that variance-heterogeneity based GWA methods are a valuable tool to detect novel associations of biological importance in natural populations .
Genome Wide Association ( GWA ) analysis is a powerful approach to study the genetic basis of complex traits in natural populations . It is widely used to study the genetics of human disease , but is equally useful in studies of other populations . For example , it has been used to dissect the genetics of traits of importance in agricultural applications ( see e . g . [1] for an example in cattle ) and ecological adaptation using collections of natural accessions in the genetic model plant Arabidopsis thaliana , for example [2–7] . The standard GWA approach screens the genome for loci where the alternative genotypes differ significantly in the mean for the trait or traits of interest . Although hundreds of loci have been found to affect a variety of quantitative traits using this strategy , it has become clear that for most complex traits this additive approach fails to uncover much of the genetics contributing to the phenotypic variation in the populations under study . It is therefore important to explore the genetics of such traits beyond additivity [8] . An alternative way that genetic variation can contribute to the phenotypic variability in a population is via direct genetic control of the variance [9] . To identify an individual locus that makes such direct contributions to the trait variance , a statistical test is used to identify significant differences in the phenotypic variance between the groups of individuals that carry alternative alleles at the locus . When such a variance difference exists between the genotypes at a locus , the locus displays a genetic variance-heterogeneity . These loci are therefore often referred to as variance-heterogeneity loci ( or vQTL for short [10] ) . By performing genome-wide analyses to identify such variance-heterogeneity loci , novel trait associations and alternative genetic mechanisms involved in shaping the total phenotypic variance in the analyzed populations can be identified [8 , 10] . The direct genetic control of the phenotypic variance is a topic that has been studied for many years in quantitative genetics with a primary focus on its potential contributions to adaptation in natural populations and agricultural selection programs . Theoretical and empirical work has increased our understanding of how individual loci that display variance , rather than mean , differences between genotypes might cause phenomena such as fluctuating asymmetry , canalization and genetic robustness [9 , 11] . Empirical work now also supports the general principle that a direct genetic control of the variance is an inherent feature of biological networks and individual genes ( see [12] for a review ) and that it contributes to both capacitation [13 , 14] and maintenance of developmental homeostasis [15] . Although it was already shown in the 1980s that it was possible to map vQTL [16] , this approach has only recently been more widely adopted to explore the role of variance-heterogeneity loci in , for example , environmental plasticity [15] , canalization [17] , developmental stability [18] , and natural variation in stochastic noise [19] . With the advent of GWA analysis , and the later realization that standard additive models leave much of the genetic variance in the analyzed populations uncovered [8] , there has been an increased interest in exploring the contribution of genetic variance-heterogeneity to the phenotypic variability in complex traits [10 , 20] . Several recent studies in , for example , humans [21] , plants [7 , 19 , 22] , Drosophilia melanogaster [23] and yeast [24] have shown that part of this previously unexplored heritable genetic variation , beyond the narrow-sense heritability , can be uncovered by re-analyzing existing GWA datasets using methods to detect differences in trait variance ( variance-heterogeneity GWA or vGWA for short ) between genotypes [20–22] . Previously , we re-analyzed ionomic data from a GWA study based on 93 wild-collected A . thaliana accessions [2] and detected a variance-heterogeneity locus with a genome-wide significant difference for the variance in leaf molybdenum concentrations between the genotypes . This association was found near the MOT1 ( Molybdate transporter 1 ) gene [22] . Importantly , this locus did not affect the mean leaf molybdenum concentrations in this dataset [2 , 22] . Molybdenum is an essential element for plant growth due to its role as a part of the molybdopterin cofactor that is required by several critical enzymes [25] . Both deficiency and excess of molybdenum have an impact on plant development [26] . The ability of plants to acquire minerals from the soil , and regulate their levels in the plant , depends on complex biochemical and regulatory pathways . The genetic architecture of such ionomics traits is thus complex [27] . To date , several studies in A . thaliana have exploited natural variation and QTL analysis to examine mineral content [28–34] , and important insights have been gained into the underlying biological mechanisms by dissecting the molecular determinants for nine of these QTL . These include QTL for the accumulation of Co , Mo , Na , Cd , As , S/Se , Zn , Cu and sulfate [5 , 6 , 35–42] . Further , GWA analysis has also been used to identify both candidate loci and functional polymorphisms contributing to natural variation in these ionomics traits [2 , 3 , 5 , 6 , 43] . Here , we quantified molybdenum concentrations in leaves in a larger collection of 340 natural A . thaliana accessions to replicate and dissect the genetic architecture of the previously detected variance-heterogeneity locus around the MOT1 gene [22] . We uncovered that a complex multi-locus , multi-allelic genetic architecture leads to the genetic variance-heterogeneity at this locus . Several polymorphisms in three closely linked loci were significantly associated with the mean molybdenum concentration in the leaf , and due to an extended LD between the minor alleles at these loci , their joint effects cause the genetic variance-heterogeneity at this locus . By dissecting this variance-heterogeneity locus in detail , we both reveal the genetic complexity of an adaptive locus for molybdenum homeostasis in A . thaliana [37] and uncover a significant amount of novel additive genetic variance that otherwise would remain undetected and contribute to the “missing heritability” .
The first GWA analysis searching for genetic effects on mean leaf molybdenum concentrations [2] did not uncover any genome-wide significant associations for this trait . This was surprising as it was known from earlier QTL studies that a strong polymorphism affecting this trait was segregating in the analyzed population [36] . To investigate this further we measured the molybdenum concentration in leaves from at least six replicate plants of 340 natural A . thaliana accessions ( S1 Table ) that had earlier been genotyped using the 250k A . thaliana SNP-chip [3] . 58 of the accessions used in this study overlapped with those in the previous study [2 , 22] . In this larger dataset , we detected several SNPs associated with the mean leaf molybdenum concentrations in , or near , the MOT1 locus ( Fig 1 ) . The minor alleles for some associated SNPs increased the mean phenotype , whereas others decreased it relative to the major allele ( Table 1; Fig 1B ) . In our earlier study we identified a genome-wide significant genetic variance-heterogeneity for leaf molybdenum concentrations at this same locus containing MOT1 [22] . Here , we therefore aim to functionally dissect this region further to obtain a deeper understanding of the genetic mechanisms controlling the range of leaf molybdenum concentrations observed in A . thaliana [36] . A vGWA analysis of leaf molybdenum concentrations in the 340 accessions , searching for genetic effects on the between accession variance heterogeneity ( S1 Text ) , revealed several SNP markers that displayed a genome-wide significant genetic variance-hetereogeneity in the region of the reported vQTL near the MOT1 gene [22] . The associations were particularly strong ( Fig 1A ) for a number of SNPs in high LD on chromosome 2 ( Fig 1B; vBLOCK ) . By visualizing the genotypes for the analyzed accessions across vBLOCK , we observed that the population contains two distinct multi-locus genotype classes for this segment: one that predominantly contains high-variance associated SNP alleles ( vBLOCKhv ) and another with low-variance associated SNP alleles ( vBLOCKlv; Fig 1C ) . vBLOCK contains in total 20 annotated genes , and the most obvious functional candidate for the association is MOT1 ( 10 , 933 , 061–10 , 934 , 551 ) . MOT1 is an obvious functional candidate gene for the genetic variance-heterogeneity for vBLOCK . A 53 bp deletion in the promoter-region of this gene has earlier been shown to decrease MOT1 expression , leading to low concentrations of molybdenum in the plant [36 , 44] . To complement our SNP-marker dataset with this known , and other potentially functional , structural promoter polymorphisms segregating in the analyzed population , we screened the promoter region of MOT1 using PCR fragment size differentiation ( see Methods for details ) and identified in total six non-coding structural polymorphisms ( Fig 2 , S1 Table ) . These were then genotyped in 283 of the 340 phenotyped accessions . Two of the six segregating MOT1 promoter polymorphisms were significantly associated with mean leaf molybdenum concentration . The first was DEL53 which is located 13 bp upstream from the transcription start-site of MOT1 . Baxter et al . [36] earlier showed that this 53 bp deletion ( DEL53 ) allele lacks the TATA-box in the MOT1 promoter , which leads to a reduced expression of MOT1 and decreased molybdenum concentration in the leaf . We confirm that this allele decreased the mean molybdenum concentrations in the leaf also in this dataset ( Table 1; pnominal = 4 . 2x10-16; Fig 2A ) and found the DEL53 allele only among low molybdenum accessions ( Mo < 3 μg g-1 dry weight ) . We also found a strong association ( pnominal = 5 . 0x10-11; Table 1; Fig 2A ) to a locus ( DUP ) located 263 bp upstream from the translation start site . Here , several accessions share a 330bp long duplication ( Fig 2B ) located inside a transposable element ( AT2TE47050 ) . The duplication exists in two distinct variants ( alleles ) differing by four polymorphisms: three point-mutations and one 4bp insertion ( DUP326 and DUP322 in Fig 2B ) . In our dataset , the DUP326 allele altered leaf molybdenum concentrations and it was found only among accessions with high leaf molybdenum concentrations ( Mo > 10 μg g-1 dry weight ) . To our knowledge , this duplication has not previously been described in the literature . Using qRT-PCR , we tested the MOT1 expression in 5 accessions carrying the low-molybdenum DEL53 allele and found that 4 of these have significantly lower expression than Col-0 in the root ( 95% CI 0 . 2–0 . 6 fold; 2 . 5 × 10−15 < p < 2 . 5 × 10−3 from Fishers method combining p-values for the biological replicates; S3 Table ) . Using the same assay , we tested 6 accessions carrying the high-molybdenum DUP326 allele . All these accessions had higher ( 95% CI 2 . 2–7 . 8 fold; 2 . 5 × 10−23 < p < 2 . 2 × 10−3 from Fishers method combining p-values for the biological replicates; S3 Table ) MOT1 expression than Col-0 in the root . Although these results do not provide direct functional evidence that the DUP326 allele increases the molybdenum concentration in the leaves via an increased expression of MOT1 in the roots , it suggests this as a plausible mechanism worth further explorations . Together , our results provide further evidence that allelic heterogeneity at MOT1 is an important component of the genetic architecture of natural variation in leaf molybdenum concentrations . Multiple associations to loci with either mean- or variance differences between genotypes for leaf molybdenum concentrations were uncovered in the single-locus GWA and vGWA analyses . To confirm the independence of these effects , and evaluate their joint contributions to leaf molybdenum , we fitted all markers ( SNPs and structural variants ) on chromosome 2 in a generalized linear model to the mean leaf molybdenum concentration using the LASSO method [45] . This penalized maximum likelihood regresses the effects of polymorphisms that make no , or only a minor , independent contribution to the trait towards zero and highlights the markers that jointly make the largest contribution to the trait variation . The penalty in the analyses was chosen so that all highlighted polymorphisms in the final model also have a genome-wide significant effect in the earlier GWA or vGWA analyses ( S1 Fig; see Methods section for details ) . In this way , the LASSO method picks up the genome-wide significant polymorphisms that have independent effects on the trait . The MOT1 promoter polymorphisms DEL and DUP were the most strongly associated loci in the LASSO analysis . Two additional SNP markers , one located ~25 kb downstream ( rs347469902; 10 , 909 , 091 bp; SNP1; Table 1 ) and one ~600 kb upstream of MOT1 ( rs347287517; 11 , 528 , 777 bp; SNP2; Table 1 ) , were also highlighted . The minor alleles at SNP1 and SNP2 ( SNP1+ and SNP2+ ) were both enriched among accessions with high leaf molybdenum concentrations . The minor alleles at three of the four associated loci thus increased the mean leaf molybdenum concentrations ( Table 1; DUP326 , SNP1+ , and SNP2+ ) , and one decreased it ( Table 1; DEL53 ) . Under certain conditions , multi-allelic genetic architectures can lead to a genetic variance-heterogeneity in association-analyses based on bi-allelic SNPs ( see e . g . [10] ) . For example , if a locus contain a SNP with two alleles , SNPA and SNPB , where the major SNP allele is completely linked to the major allele at gene M regulating trait T ( i . e . only the SNPA-MWT haplotype exists in the population ) . If now locus M also contains two minor alleles , M- and M+ , that decreases/increases T an equal amount relative to the value of MWT , and that are tagged by the SNPB allele , the SNPA and SNPB genotype-classes will have identical means , but different variances . Here , we will show that the genetic variance-heterogeneity we detected for vBLOCK is due to a multi-allelic genetic architecture that closely resembles this example . There was a strong LD ( D’ ) between three loci ( SNP1 , DEL and DUP ) associated with the mean leaf molybdenum concentration and the SNPs across vBLOCK that displayed a highly significant genetic variance-heterogenity ( Fig 2A; Table 2 ) . All the 20 accessions carrying either the DEL53 or DUP326 alleles also carry the high-variance associated vBLOCKhv . Of the 29 accessions that carry the high molybdenum SNP1+ allele , 19 carried vBLOCKhv ( Fig 1C; see Methods section for further detail ) . The minor alleles at two of these ( DUP326 , SNP1+ ) increased , and at one of them ( DEL53 ) decreased , the leaf molybdenum concentration . This results in a situation similar to that in the example above: multiple alleles with different directional phenotypic effects are unevenly distributed across the two variants of vBLOCK . The fact that one variant ( vBLOCKhv ) tags three different minor alleles ( DUP326 , DEL53 and SNP1+ ) with different effects on the mean molybdenum concentration explains the increased phenotypic variance for this group of accessions . To statistically disentangle the genetic effects on the mean and variance by this multi-allelic , multi-locus genetic architecture , an additional vGWA analysis was performed where we fitted a linear model with separate effects for the mean and variance to the data as outlined by Valdar and Rönnegård [10] . The three mean associated loci that were located within vBLOCK ( DUP , DEL and SNP1 ) were fitted as loci with mean effects when screening chromosome 2 for loci with potential effects on the variance using this method . The entire variance signal to vBLOCK disappears in this analysis ( Fig 3A ) illustrating that the variance-heterogeneity association to vBLOCK is due to the presence of the DEL53 , DUP326 and SNP1+ alleles on the high-variance associated vBLOCKhv ( Fig 3C ) . We estimated the broad-sense heritability of leaf molybdenum concentrations from the within/between accession variances to be H2 = 0 . 80 using an ANOVA across all replicated measurements . This estimate is similar to that reported in earlier studies ( 0 . 56 [43] to 0 . 89 [2] ) . The narrow-sense heritability was estimated to be h2 > = 0 . 63 using a mixed model based analysis where the accession mean phenotypes were regressed onto the genomic kinship matrix . The first GWA analysis for leaf molybdenum concentrations by Atwell et al . [2] was unable to detect any loci contributing to the variation in the trait mean . The later vGWA study by Shen et al . [22] identified a genetic variance-heterogeneity in the MOT1 region that explained 27% of the phenotypic variance where the contribution by mean ( additive ) and variance ( non-additive ) effects were 4/23% of the phenotypic variance , respectively . Using the variance decomposition proposed by Shen et al . [22] , we estimate that the genetic variance-heterogeneity at vBLOCK contributes 3 and 19% to the phenotypic variance via its effect on the mean and the variance . The total amount of genetic variance associated with the vGWA signal here is thus comparable to that of Shen et al . [22] , but in both studies it leaves much of the total additive genetic variance unexplained as it only accounts for about 5% of h2 . The contribution to H2 is , however , larger and between 24 to 28% in these two studies . However , after considering the individual contributions made by the three polymorphisms identified on vBLOCKhv ( DEL53 , DUP326 , SNP1+; Fig 3 ) , much additive genetic variance is uncovered . Nearly all the contribution from vBLOCK becomes additive ( 83% of the total variance ) to explain 45% of h2 and 43% of H2 . By also accounting for the fourth locus ( SNP2; Fig 2 ) , the contribution h2 and H2 increases further to 60 and 50% , respectively . By dissecting the genetic architecture of the vGWA signal into its underlying multi-locus , multi-allelic components , we were thus able to reveal a significant contribution by vBLOCK to the “missing heritability” of molybdenum concentration in the leaf in the original GWA [2] and vGWA [22] analyses . Here , we functionally explore the associations outside of the coding and regulatory regions of MOT1 in more detail to identify additional functional candidate polymorphisms and genes for the regulation of molybdenum homeostasis . Two regions outside of the coding and regulatory region of MOT1 ( chromosome 2 10 , 933 , 061–10 , 935 , 200 bp ) were associated with the mean leaf molybdenum concentrations ( SNP1 and SNP2 in Figs 1B; 3A ) . Genes located in the chromosomal regions covered by SNPs in LD ( r2 > 0 . 4 ) with SNP1 and SNP2 , respectively , were explored as potential functional candidates for the associations using T-DNA insertion alleles ( S4 Table ) . Four T-DNA alleles of five different genes in the region around SNP1 ( 10 , 909 , 091 bp; S2 Fig; S4 Table ) were evaluated for leaf molybdenum concentrations , but in none of these did the leaf molybdenum concentrations differ from that of the wild-type Col-0 . We also evaluated 19 mutants with T-DNA insertions in 14 genes around SNP2 ( 11 , 528 , 777 bp; Fig 4; S4 Table ) , and identified two with significantly altered leaf molybdenum concentrations compared to the wild-type Col-0 ( Table 3 ) . One ( SALK_138758 ) has an insertion covering genes AT2G27020 and AT2G27030 , and the other ( GK-350E02 ) has an insertion in gene AT2G26975 . These T-DNA alleles showed on average 55 and 58% reductions in leaf molybdenum concentrations compared to wild-type Col-0 , respectively ( Table 3 ) . AT2G27020 was also evaluated via another T-DNA insertional allele ( SAIL_760_D06 ) , and this line had wild-type leaf molybdenum concentrations . Thus , AT2G27030 ( ACAM2/CAM5; 11 , 532 , 004–11 , 534 , 333 ) appears to be the most likely functional candidate gene of the two . Calmodulin is a known metalloprotein and a Ca2+ sensor , but no previous connections to molybdenum has been reported . The reduced leaf molybdenum concentration of the T-DNA insertional allele of AT2G26975 ( Copper Transporter 6; COPT6 ) makes this a second functional candidate locus for the association around SNP2 . Interestingly , as well as low molybdenum , the T-DNA knockout allele of this gene has a slightly increased leaf copper concentration compared to wild-type ( 3 . 82 and 3 . 36 μg / g dry weight , respectively , in GK-350E02 and wild-type Col-0; p = 0 . 0018 ) , suggesting a role of COPT6 also in the regulation of copper homeostasis . From the literature it is known that copper and molybdenum homeostasis are related and that copper depleted Brassica napus plants have up-regulated expression of both copper transporter genes and MOT1 [46] .
Common approaches to dissect the genetics of complex traits in segregating populations are linkage mapping and association studies . These studies aim to identify the loci in the genome where genetic polymorphisms control the phenotypic variance in the studied populations . This is achieved by screening for significant genotype-phenotype associations across a large number of genotyped polymorphic markers in the genome . The most common statistical models used in such analyses aim to identify loci with significant mean phenotype differences between the genotypes at individual loci . Although such models are powerful for capturing much genetic variance in populations , they have limited power when challenged with more complex genetic architectures including multiple-alleles , variance-heterogeneity and genetic interactions [8 , 47] . It is therefore important to also develop , and test , methods that explore statistical genetic models reaching beyond additivity when aiming for a more complete dissection of the genetic architecture of complex traits . The genetic architecture of variation in mean leaf molybdenum concentrations has earlier been explored using GWA analyses in a smaller set of 93 wild collected A . thaliana accessions [2] . No genome-wide significant associations were found for leaf molybdenum , which was surprising given that the trait has a high heritability [36 , 43] and that several polymorphisms in MOT1 are known to contribute to natural variation in this trait [36 , 37] . When we re-analyzed this data using a method to detect variance differences between genotypes , a strong genetic variance-heterogeneity was identified near the MOT1 gene [22] . Here , we studied a larger set of 340 A . thaliana accessions to replicate and fine-map the molecular determinant of this genetic variance-heterogeneity , and find that the strongest associations are to an extended region surrounding MOT1 ( vBLOCK ) . This is the first successful fine-mapping and replication of a variance-heterogeneity locus on a genome-wide significance scale and in an independent dataset . In this larger dataset we also identified four loci that independently alter the mean concentration of leaf molybdenum . The minor allele at one of these ( DEL53 ) was a deletion in the promoter region of MOT1 previously identified using an F2 bi-parental mapping population . This deletion allele decreases the concentration of molybdenum in leaves by down-regulating MOT1 transcription [36] . Further , we also identified three previously unknown loci , and the minor alleles at these loci ( DUP326 , SNP1+ and SNP2+ ) increased the concentration of molybdenum in leaves . One allele ( DUP326 ) was an insertion polymorphism in the promoter region of MOT1 , and our analyses revealed that accessions carrying this polymorphism have higher expression of MOT1 compared to the Col-0 accession that does not carry this polymorphism . The other two associations were to SNPs in regions that were not in LD ( r2 ) with the MOT1 gene or its promoter . One of these SNPs was found ~25 kb downstream of MOT1 ( SNP1 ) and the other ~600 kb upstream of the MOT1 transcription start-site ( SNP2 ) . The regulation of molybdenum concentrations in the leaves is hence due to multiple alleles in a gene known to regulate molybdenum uptake , MOT1 , but also alleles at other neighboring loci that have earlier not been found to contribute to molybdenum homeostasis in A . thaliana . These results support and refine earlier results from QTL and functional analyses of the MOT1 region that highlighted the central importance of the MOT1 region in the regulation of molybdenum homeostasis in natural populations and also suggested that the natural variation in this trait might have a multi-allelic background [36 , 37] . As it is well known that major loci affecting traits under selection often evolve multiple mutations affecting the phenotype , and that allelic heterogeneity is an important driver of evolution in natural A . thaliana populations [48] , our finding of multiple polymorphisms in this key locus is not surprising . Striking examples of allelic heterogeneity in natural A . thaliana populations include the large number of different loss-of-function mutants in the GA5 locus leading to semidwarfs [49] , the MUM2 locus leading to altered seed flotation [50] and the FRIGIDA locus leading to an altered flowering-time [51] . Multi-allelic loci are , however , a major challenge in traditional GWA analyses [48] . It is therefore valuable to note that such loci , under certain conditions , can lead to a genetic variance-heterogeneity ( see e . g . [10] ) that can be detected with a vGWA analysis . The following two examples illustrate how genetic variance-heterogeneity can arise under i ) classic allelic heterogeneity where multiple loss-of-function alleles have evolved independently at a locus , and ii ) general multi-allelic architectures where the alleles affect the phenotype to various degree and hence either increase or decrease the phenotype relative to that of the major allele . To illustrate how a genetic variance-heterogeneity can emerge under these scenarios , let us consider an example when looking for associations to a bi-allelic SNP with alleles SNPA and SNPB and where the major SNP allele ( SNPA ) is completely linked to the major allele at the functional gene M ( MWT ) . Below , we illustrate how the distribution of the minor alleles across the SNP genotypes will alter the differences in phenotypic mean and variances between the genotypes , and hence affect the power to detect them in GWA and vGWA analyses . Hence , the vGWA analysis is likely to be useful for identifying loci under a set of different scenarios ranging from classic allelic heterogeneity to loci with multiple alleles having distinct effects on the phenotype . As shown here , the genetic variance-heterogeneity for vBLOCK was detected based on its genetic variance-heterogeneity due to its close resemblance to scenario ( c ) above ( Fig 2A ) . Here , we dissected a locus displaying a genetic variance-heterogeneity for the molybdenum concentration in A . thaliana leaves into an underlying multi-locus , multi-allelic genetic architecture . We find several alleles at MOT1 that contribute to this association , which is consistent with findings in earlier studies reporting that several functional variants of this gene alter the mean molybdenum concentrations in A . thaliana [36 , 37] . Such multi-allelic architectures , where the different genetic variants affect traits under selection to varying degrees , are not unique to this study but have been described also for other traits and species . For example , in A . thaliana the Flowering Locus C ( FLC ) locus has a natural series of alleles with different effects on vernalization that have been identified [52] . Similar examples also exist in , for example , domestic animal populations for both Mendelian traits , such as coat color [53–55] , and complex traits , such as muscularity [56] and meat quality [57] . As illustrated above , the vGWA analysis is a straight-forward and computationally tractable analytical strategy that could be used to identify loci where multi-allelic genetic architectures reduce the additive genetic variance that can be detected by traditional GWA approaches . The examples above suggest that such genetic architectures are likely to be more common than what has been empirically shown to date . We therefore recommend that the vGWA approach be tested on more datasets to reveal how common this type of architecture might be for complex traits . This will also help reveal how large a contribution such multi-allelic genetic architectures contribute to the “missing heritability” . Little is currently known about the genetic mechanisms contributing to variance-heterogeneity between genotypes in natural populations . Ayroles et al . [23] recently reported the first dissection of a locus displaying a genetic variance-heterogeneity in a segregating population and found that mutating a single gene ( Ten-a ) led to a genetic variance-heterogeneity for a behavioral phenotype in Dropsophila melanogaster . A number of other , not mutually exclusive , hypotheses have been proposed to explain the origin of genetic variance-heterogeneity at a locus . These can broadly speaking be divided into two categories: those due to the individual locus itself such as multiple functional alleles , incomplete linkage disequilibrium and developmental instabilities [7 , 10 , 22] , and those due to interactions between the locus and other genetic or environmental factors ( i . e . epistasis or gene-by-environment interactions ) [8 , 10 , 21] . Here , we present the first empirical evidence illustrating how population-wide genetic variance-heterogeneity in a natural population can result from a complex locus involving multiple loci and multiple alleles . We show that this genetic variance-heterogeneity originates from the LD ( D’ ) between multiple functional polymorphisms and the SNP markers defining an LD block around MOT1 ( vBLOCK ) . The high-variance associated version of this LD-block ( vBLOCKhv ) contains three independent polymorphisms ( DEL53 , DUP326 and SNP1+ ) altering the molybdenum concentration in leaves relative to the major alleles at these loci on the low-variance associated version ( vBLOCKlv ) . Two of these polymorphisms increase molybdenum and one decrease it , leading to a highly significant genetically determined variance-heterogeneity amongst the accessions that share vBLOCKhv ( Fig 2A; multi-allelic example c above ) . Our work also illustrates how the use of alternative genetic models in GWA analyses can provide novel insights to complex genetic architectures underlying adaptively important traits in natural populations . The LD ( D’ ) between multiple functional polymorphisms and vBLOCK in this collection of natural A . thaliana accessions is the key genomic feature that facilitated the discovery of this locus in the vGWA . Although the molecular basis for this LD-pattern , as well as the reasons for multiple independent polymorphisms being found almost exclusively with one of the variants of this LD-block , is unknown , it is interesting to note that they could have emerged via the processes discussed in relation with the appearance of synthetic LD in GWA studies [58] . It would therefore be interesting to , in the future , explore whether the same basic genomic processes might drive the emergence of both synthetic and vGWA associations in general , or whether the resemblance between the genetic architecture described here and the mechanism proposed by Dickson et al . [58] is a rare case of where the two overlap . Many GWA studies have found that the total additive genetic variance of associated loci is considerably less than that predicted based on estimates of the narrow-sense heritability , i . e . the ratio between the additive genetic and phenotypic variance in the population . This common discrepancy between the two is often called the curse of the “missing heritability” and is viewed as a major problem in past and current GWA studies [59] . Here , we provide an empirical example of how a vGWA is able to identify a locus [22] that remained undetected in a standard GWA [2] and that , when the underlying genetic architecture was revealed , was found to make a large contribution to the additive genetic variance and narrow-sense heritability . This illustrates the importance of utilizing multiple statistical modeling approaches in GWA studies to detect the loci contributing to the phenotypic variability of the trait , and then also continue to further dissect the underlying genetic architecture to uncover how the loci potentially contribute to the heritability that was “missing” in the original study [2] . By evaluating T-DNA insertional alleles of genes in LD with the SNPs associated to leaf molybdenum concentrations , we are able to suggest two novel functional candidate genes involved in molybdenum homeostasis in A . thaliana . Little is known about the function of one of these , AT2G27030 , and further work is needed to explore the mechanisms by which it may alter molybdenum concentrations in the plant . The second gene ( AT2G26975; Copper Transporter 6; COPT6 ) located ~600 kb upstream of MOT1 is from earlier studies known to be involved in the connected regulation of copper and molybdenum homeostasis in plants . It was recently reported [46] that MOT1 and several copper transporters were up-regulated under copper deficiency in B . napus , suggesting a common regulatory mechanism for these groups of genes . Further experimental work is needed to explore the potential contributions of these genes to natural variation in molybdenum homeostasis , and the potential connection between copper and molybdenum homeostasis . Here , we dissect a complex locus affecting molybdenum concentration in the A . thaliana leaf and find it likely that three closely linked genes contribute to this effect . Clustering of genes with similar function is well known for Resistance ( R ) genes [60] and close linkage between genes important for growth rate has also been evidenced [61] in A . thaliana . How common such functional clustering into complex loci will be for traits of importance for evolution is still largely unknown as the resolution in most complex trait studies does not allow the separation of effects from closely linked loci . Our finding that not only the already known gene in this region , MOT1 , but likely also other novel genes contribute to the diverse range of molybdenum concentrations in the leaf observed in this collection of natural A . thaliana accessions suggest that the clustering of loci has been of adaptive value for this ecologically relevant trait . This makes the locus a highly interesting candidate for future work to better understand the role of gene clustering for the evolution of adapted populations . In summary , here we dissect a locus displaying a genetic variance-heterogeneity for leaf molybdenum concentration in A . thaliana [22] into the contributions from three independent alleles that are in high LD with the high-variance associated version of an extended LD-block surrounding the MOT1 gene . This is the first empirical example of how a multi-locus , multi-allelic genetic architecture can lead to genetic variance heterogeneity at a locus . The dissection of the genetic architecture underlying the vGWA signal allowed the transformation of non-additive genetic variance into additive genetic variance , and hence allowed the detection of a significant part of the “missing heritability” in the variation in leaf molybdenum concentrations in this species-wide collection of A . thaliana accessions . This study also delivers insights into how vGWA mapping facilitates the detection and genetic dissection of the genetic architecture of loci contributing to complex traits in natural populations . It thereby illustrates the value of using alternative statistical methods in genome-wide analyses . Further , it provides an approach to infer multi-allelic loci , which are likely to be both a common , and far too often ignored , complexity in the genetics of multifactorial traits that contributes to undiscovered additive genetic variance and consequently the curse of the “missing heritability” .
The concentration of molybdenum in leaves was measured in 340 natural A . thaliana accessions from the ‘HapMap’ collection ( [3]; S1 Table ) . This dataset contains 58 of the 93 accessions used in the earlier GWA [2] and vGWA [22] analyses of leaf molybdenum concentrations supplemented with 282 newly phenotyped accessions . All accessions were grown in a controlled environment with 6 biological replicate plants per accession , and analyzed by Inductively Coupled Mass Spectroscopy ( ICP-MS ) for multiple elements including molybdenum , as described previously by Baxter et al . [3] . All the ICP-MS data used for the GWA and vGWA is accessible using the digital object identifier ( DOI ) 10 . 4231/T9H41PBV , and data for the evaluation of candidate genes using T-DNA insertional alleles is accessible using the DOI 10 . 4231/T9NP22C0 ( see http://dx . doi . org/ ) . All accessions have previously been genotyped using the 250k A . thaliana SNP chip and that data is publicly available [3] . SNPs where the minor allele frequency was below 5% were excluded from the analyses . Genotypes were available for more than 95% of the SNPs in all accessions , so none were removed due to problematic genotyping . In total , 200 , 345 SNPs passed this quality control and were used in our GWA and vGWA analyses . We evaluated the region upstream of MOT1 for structural polymorphisms in a set of 283 accessions selected to cover the range of leaf molybdenum concentrations ( S5 Table ) . This was done using gel electrophoresis to identify PCR fragment size differentiation using the primers described in S6 Table . The PCR reactions were completed as follows: 1μl DNA + 5X GoTaq Bf , 2 . 5mM dNTP’s , 25mM MgCl2 , 0 . 4μM of each primer , 0 . 3μl Taq polymerase , and 9 . 7μl nuclease free water for a total reaction volume of 25μl . PCR conditions were 94°C for 1 minute to denature , 54°C for 1 minute to anneal , and 72°C for 1 . 25 minutes for extension , repeated for 40 cycles in the Thermo Px2 thermal cycler ( Electron Corporation ) . DNA was prepared for the accessions that displayed suggestive evidence for structural polymorphisms and submitted for sequencing using Macrogen ( dna . macrogen . com ) . The sequences were then compared to the Col-0 reference sequence using DiALIGN ( http://bibiserv . techfak . uni-bielefeld . de/dialign/ ) , which uncovered five loci and six segregating structural polymorphisms ( S2 Table ) that were then genotyped in the 283 phenotyped accessions ( S5 Table ) . All analyses described in the sections below were performed using the R-framework for statistical computing [62] . All figures , except Fig 2 , were prepared using R . The vGWA analyses identify a strong variance-heterogeneity signal across a number of markers on chromosome 2 that contains the functional candidate MOT1 gene . The LD is high among these significant markers that define an extended vGWA associated vBLOCK . Visual inspection of the genotype-matrix of this region , sorted by the genotype of the leading SNP in the vGWA analysis ( Table 1 ) , indicated the presence of two major groups of accessions that carry the same alleles across a large number of the associated markers ( Fig 1C ) .
|
Most biological traits vary in natural populations , and understanding the genetic basis of this variation remains an important challenge . Genome-wide association ( GWA ) studies have emerged as a powerful tool to address this challenge by dissecting the genetic architecture of trait variation into the contribution of individual genes . This contribution has traditionally been measured as the difference in the phenotypic means between groups of individuals with alternative genotypes at one , or multiple loci . However , instead of altering the trait mean , certain loci alter the variability of the trait . Here , we describe the genetic dissection of one such variance-controlling locus that drives variation in leaf molybdenum concentrations amongst natural accessions of Arabidopsis thaliana . The variance-controlling locus was found to result from the contributions of multiple alleles at multiple loci that are closely linked on the chromosome and is a major contributor to the “missing heritability” for this trait identified in previous studies . This illustrates that multi-allelic genetic architectures can hide large amounts of additive genetic variation , and that it is possible to uncover this hidden variation using the appropriate experimental designs and statistical methods described here .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
The Multi-allelic Genetic Architecture of a Variance-Heterogeneity Locus for Molybdenum Concentration in Leaves Acts as a Source of Unexplained Additive Genetic Variance
|
Mathematical modelling has become an established tool for studying the dynamics of biological systems . Current applications range from building models that reproduce quantitative data to identifying systems with predefined qualitative features , such as switching behaviour , bistability or oscillations . Mathematically , the latter question amounts to identifying parameter values associated with a given qualitative feature . We introduce a procedure to partition the parameter space of a parameterized system of ordinary differential equations into regions for which the system has a unique or multiple equilibria . The procedure is based on the computation of the Brouwer degree , and it creates a multivariate polynomial with parameter depending coefficients . The signs of the coefficients determine parameter regions with and without multistationarity . A particular strength of the procedure is the avoidance of numerical analysis and parameter sampling . The procedure consists of a number of steps . Each of these steps might be addressed algorithmically using various computer programs and available software , or manually . We demonstrate our procedure on several models of gene transcription and cell signalling , and show that in many cases we obtain a complete partitioning of the parameter space with respect to multistationarity .
Mathematical models in the form of parameterized systems of ordinary differential equations ( ODEs ) are valuable tools in biology . Often , qualitative properties of the ODEs are associated with macroscopic biological properties and biological functions [1–4] . It is therefore important that we are able to analyse mathematical models with respect to their qualitative features and to understand when these properties arise in models . With the growing adaptation of differential equations in biology , an automated screening of ODE models for parameter dependent properties and discrimination of parameter regions with different properties would be a very useful tool for biology , and perhaps even more for synthetic biology [5] . Even though it is currently not conceivable how and if this task can be efficiently formalized , we view the procedure presented here as a first step in this direction . Multistationarity , that is , the capacity of the system to rest in different positive equilibria depending on the initial state of the system , is an important qualitative property . Biologically , multistationarity is linked to cellular decision making and ‘memory’-related on/off responses to graded input [2–4] . It has been suggested that different stable equilibria of a cell represent different cell types [6 , 7] . Whole-cell modelling provides an opportunity to understand the number and type of the stable equilibria of the cell and could potentially give insight into the different cell types that a particular cell can differentiate into and transition between . Currently , this is an important open question in biology [8] . Moreover , the existence of multiple equilibria is often a design objective in synthetic biology [9 , 10] . Various mathematical methods , developed in the context of reaction network theory , can be applied to decide whether multistationarity exists for some parameter values or not at all , or to pinpoint specific values for which it does occur [11–20] . Some of these methods are freely available as software tools [21 , 22] . It is a hard mathematical problem to delimit parameter regions for which multistationarity occurs . Often it is solved by numerical investigations and parameter sampling , guided by biological intuition or by case-by-case mathematical approaches . A general approach , in part numerical , is based on a certain bifurcation condition [18 , 19 , 23 , 24] . Alternatively , for polynomial ODEs , a decomposition of the parameter space into regions with different numbers of equilibria could be achieved by Cylindrical Algebraic Decomposition ( a version of quantifier elimination ) [25] . This method , however , scales very poorly and is thus only of limited help in biology , where models tend to be large in terms of the number of variables and parameters . Here we present two new theoretical results pertaining to multistationarity ( Theorem 1 and Corollary 2 ) . The results are in the context of reaction network theory and generalize ideas in [26 , 27] . We consider a parameterized ODE system defined by a reaction network and compute a single polynomial in the species concentrations with coefficients depending on the parameters of the system . The theoretical results relate the capacity for multiple equilibria or a single equilibrium to the signs of the polynomial as a function of the parameters and the variables ( concentrations ) . The theoretical results apply to dissipative reaction networks ( networks for which all trajectories eventually remain in a compact set ) without boundary equilibria in stoichiometric compatibility classes with non-empty interior . These conditions are met in many reaction network models of molecular systems . We show by example that the results allow us to identify regions of the parameter space for which multiple equilibria exist and regions for which only one equilibrium exists . Subsequently this leads to the formulation of a general procedure for detecting regions of mono- and multistationarity . The procedure verifies the conditions of the theoretical results and further , calculates the before-mentioned polynomial . A key ingredient is the existence of a positive parameterization of the set of positive equilibria . Such a parameterization is known to exist for many classes of reaction networks , for example , systems with toric steady states [14] and post-translational modification systems [28 , 29] . The conditions of the procedure might be verified manually or algorithmically according to computational criteria . The algorithmic criteria are , however , only sufficient for the conditions to hold . For example , a basic condition is that of dissipativity . To our knowledge there is not a sufficient and necessary computational criterion for dissipativity , but several sufficient ones . If these fail , then the reaction network might still be dissipative , which might be verified by other means . By collecting the algorithmic criteria , the procedure can be formulated as a fully automated procedure ( an algorithm ) that partitions the parameter space without any manual intervention . The algorithm might however terminate indecisively if some of the criteria are not met . Table 1 shows two examples of reaction network motifs that occur frequently in intracellular signalling: a two-site protein modification by a kinase–phosphatase pair and a one-site modification of two proteins by the same kinase–phosphatase pair . These reaction networks are in the domain of the automated procedure and conditions for mono- and multistationarity can be found without any manual intervention . The conditions discriminating between a unique and multiple equilibria highlight a delicate relationship between the catalytic and Michaelis-Menten constants of the kinase and the phosphatase with the modified protein as a substrate ( the kc- and kM-values ) . If the condition for multiple equilibria is met , then multiple equilibria occur provided the total concentrations of kinase , phosphatase and substrate are in suitable ranges ( values thereof can be computed as part of the procedure ) . The paper has three main sections: a theoretical section , a section about the procedure and an application section . We close the paper with two brief sections discussing computational limitations , related work and future directions . In the theoretical section we first introduce notation and mathematical background material . We then give the theorem and the corollary that links the number of equilibria to the sign of the determinant of the Jacobian of a certain function , which is derived from the ODE system associated with a reaction network . In the second section we state the procedure , derive the algorithm and comment on the feasibility and verifiability of the conditions . Finally , in the application section we apply the procedure to several examples . The S1 File has six sections . All proofs are relegated to §1–4 together with background material . In §5 we elaborate further on how the conditions of the procedure/algorithm can be verified . In §6 we provide details of the algorithmic analysis of the examples in Table 1 . Also we include a further monostationary example for illustration of the algorithm .
In this part of the manuscript we present the theoretical results . We start by introducing the basic formalism of reaction networks . Theorem 1 , Corollary 1 and 2 below apply to dissipative networks without boundary equilibria and concern the ( non ) existence of multiple equilibria in some stoichiometric compatibility class . Corollary 2 assumes the existence of a positive parameterization of the set of positive equilibria . Before stating the results these five concepts are formally defined . In the previous subsection we applied Corollary 1 and Corollary 2 to the running example by going through a number of steps corresponding to the conditions of the statements and the calculation of the determinant . In this section we outline the steps formally . Afterwards we discuss the steps and how they can be verified either manually or algorithmically , that is , without user intervention . Finally we devise an algorithm to conclude uniqueness of equilibria or to find regions in the parameter space where multistationarity occurs . We conclude this section with some extra examples that follow the steps of the procedure . We assume the reaction rate functions v ( x ) depend on some parameters κ . The reaction rate functions are further assumed to be polynomials ( as for mass-action kinetics ) or quotients of polynomials ( as for Michaelis-Menten and Hill kinetics with integer exponents ) . The input to the procedure is v ( x ) and N ( the stoichiometric matrix ) and the output is parameter regions for which the network admits multistationarity or uniqueness of equilibria . Procedure ( Identification of parameter regions for multistationarity ) Input: N and v ( x ) depending on κ . 1 . Find f ( x ) , a row reduced matrix W of size d × n such that WN = 0 , and check that v ( x ) vanishes in the absence of one of the reactant species , that is , check that it satisfies Eq ( 2 ) . 2 . Check that the network is dissipative . 3 . Check for boundary equilibria in P c for P c + ≠ ∅ and c ∈ R d . 4 . Construct φc ( x ) , M ( x ) and compute det ( M ( x ) ) . 5 . Analyze the sign of det ( M ( x ) ) . Find conditions on the parameters κ such that sign ( det ( M ( x ) ) ) = ( −1 ) s for all x ∈ R > 0 n , in which case Corollary 1 holds . If Corollary 1 does not hold for all κ , continue to the next step . 6 . Obtain an algebraic parameterization Φ ( x ^ ) of the set of positive equilibria for all κ , as in Eq ( 7 ) , such that the coefficients of the numerator and the denominator of each Φ i ( x ^ ) possibly depend on κ . Compute a ( x ^ ) = det ( M ( Φ ( x ^ ) ) ) . By hypothesis , a ( x ^ ) can be written as the quotient of two polynomials in x ^ with coefficients depending on κ , whose denominator takes positive values . 7 . Analyze the sign of the numerator of a ( x ^ ) . 7a . Identify coefficients with sign ( −1 ) s+1 and coefficients that can have different signs depending on the parameters . 7b . Use the terms corresponding to identified coefficients to construct parameter inequalities such that , whenever these inequalities hold , one has either sign ( a ( x ^ ) ) = ( - 1 ) s for all x ^ ∈ R > 0 m or sign ( a ( x ^ ) ) = ( - 1 ) s + 1 for at least one x ^ ∈ R > 0 m , in which case either Corollary 2 ( A ) or ( B ) holds . There is no guarantee that all steps of the procedure can be carried out successfully , let alone automatically . While step 1 and 4 usually are straightforward ( only computational issues might arise for large networks ) , step 2 , 3 , 5 , 6 and 7 might in particular require case specific approaches . However , there exist computationally feasible sufficient criteria that guarantee the conditions in each step can be checked efficiently . To illustrate several aspects of the algorithm we provide a detailed step-by-step analysis of a collection of examples . The computational complexity of some of the steps in the procedure are demanding . Some conditions can be checked using linear algebra and do not depend on parameter values , others depend on parameter values and require symbolic manipulations . In some situations , the calculation can be done for even large networks at the cost of time , while in other situations symbolic software ( like Mathematica and Maple ) have inherent limits to what it can process . We offer here a few remarks about computational strategies and time complexity . Dissipativity . There are efficient algorithms to check whether the network is conservative and strongly endotactic , using linear algebra or mixed-integer linear programming [21 , 47] . We are not aware of a systematic way to check if Proposition 1 is fulfilled or not . Finding the minimal siphons of a network requires in general exponential time and there might be exponentially many of these [59] . Different algorithms developed in Petri Net theory can be applied to find the minimal siphons; see for example [48 , 49 , 59] and references therein . The complexity of this computation can often be substantially reduced by removing so-called intermediates and catalysts from the network [50] ( see §5 . 1 in the S1 File for details ) . Finding all non-interacting and reactant-non-interacting sets requires in general exponential time . One strategy is the following . We first remove all species Si for which αij > 1 or βij > 1 for some reaction Rj ( the latter constraint is omitted if we are looking for reactant-non-interacting sets only ) . Then we build non-interacting ( reactant-non-interacting ) sets by adding new species recursively until no more species can be added without having an interacting pair of species in the set . Calculation of the symbolic determinant of the matrix M ( x ) , and hence also of a ( x ^ ) , often fails in our experience for networks with more than 20 variables on common laptops [60] . However , this clearly depends on the sparsity of the matrix M ( x ) , that is , on the number and order of the reactions . Strategies to reduce the complexity of the computation by expanding the determinant along the non-symbolic rows ( conservation relations ) were inspected in [60] . Specialized software like Singular [61] and/or better hardware could probably push what is possible to something closer to 50 variables . At this size , however , what might best be called ‘cognitive limitations’ come into play: symbolic software typically has problems with collecting and simplifying terms ‘the right way’ if there are many variables and/or parameters . And if terms are not collected appropriately it might be difficult , if not impossible , to decide on the sign of the polynomial coefficients . Our approach is therefore best suited to systems of moderate size ( say 20-30 variables ) . Furthermore , it is our experience that large non-linear models tend to be multistationary because of the many non-linear dependencies that typically are present [60] . Positive parameterizations: The worst case scenario involves checking ∑ i = 1 d ( n i ) different sets of variables , each with at most d variables . Finding the vertices of the Newton polytope can be done with existing symbolic software , for example Polymake [62] or Maple , as we demonstrate in the S1 File . We stress that it is always beneficial to guide the procedure/algorithm whenever possible in the sense that , if something is known for the network , there is no reason to go through many possibilities .
The main result of this paper , the procedure to identify parameter regions for unique and multiple equilibria , combines Brouwer degree theory and algebraic geometry . In particular , under the assumptions of Corollary 2 , we show that there exist stoichiometric compatibility classes with at least two equilibria if , and only if , a certain multivariate polynomial can attain a specific sign . Discriminating regions of the parameter space where multistationarity occurs is a hard mathematical problem , theoretically addressable by computationally expensive means [25] . Our approach beautifully overcomes these difficulties by building on a simple idea , the computation of the Brouwer degree of a function related to a dissipative network . Additionally , not only closed-form expressions in the parameters are obtained , but , as illustrated in examples , these expressions are often interpretable in biochemical terms , providing an explanation of why multistationarity occurs . The procedure applies theoretically to any choice of algebraic reaction rate functions . However , in practice , the procedure works well with mass-action kinetics . For example , we have considered the two-site phosphorylation cycle depicted in the second row of Table 1 , but now modelled with Michaelis-Menten kinetics instead of mass-action kinetics . This network is known to be multistationary [63] , and the conditions to apply Corollary 1 and Corollary 2 are valid . However , a positive algebraic parameterization does not exist , and hence our approach cannot be used to find parameter conditions for multistationarity . However , Corollary 1 might be used with rational reaction rate functions for monostationary networks . This is the case for example for the one-site phosphorylation cycle S ⇌ S p with Michaelis-Menten kinetics [63] . This network has two species and rank one . The sign of det ( M ( x ) ) is −1 for all parameter values and all x ∈ R > 0 2 . By Corollary 1 , the network admits exactly one positive equilibrium in every stoichiometric compatibility class P c with P c + ≠ ∅ for all parameter values . If a reaction network does not have any conservation relation , then the set of equilibria consists typically of a finite number of points . In this case an algebraic parameterization is an algebraic expression of the equilibria in terms of the parameters of the system . Since m = 0 , then R m consists of a single point and it follows directly that there is a unique equilibrium . Such an expression rarely exists . Therefore the procedure applies mainly to reaction networks with conservation relations . In particular , this rules out reaction networks where each species is produced and degraded . Several natural questions remain outside the reach of our procedure . Firstly one would like to determine the particular stoichiometric compatibility classes for which there are multiple equilibria . As stated in Corollary 2 , if sign ( a ( x ^ ) ) = ( - 1 ) s + 1 , then c : = W Φ ( x ^ ) defines a stoichiometric compatibility class with multiple equilibria . However , this only establishes c indirectly through x ^ . In some situations , it might be possible to find a positive parametrization that uses some of the conservation relations ( ideally , all but one ) and the stoichiometric compatibility classes with multiple/single equilibria would be determined up to a single parameter . Secondly , one could ask for parameter regions that differentiate between the precise number of equilibria ( that is , 0 , 1 , 2 , … ) . This question should be seen in conjunction with the previous question: in typical examples , when there are two equilibria in a particular stoichiometric compatibility class , then there exists another class for which there are three . Hence the number of equilibria cannot be separated from the stoichiometric compatibility classes . A third question concerns the stability of the equilibria , which cannot be obtained from our procedure . It is , however , known that if the sign of the Jacobian evaluated at an equilibrium is ( −1 ) s+1 , then it is unstable [34] . This is in particular the case for an equilibrium fulfilling the condition in Corollary 2 ( B ) . We have shown that for some reaction networks our procedure can be formulated as an algorithm . We consider therefore our research a step in the direction of providing ‘black box tools’ to analyse complex dynamical systems . Such tools would easily find their use in systems and synthetic biology , where it is commonplace to consider ( many ) competing models . A particular problem is to exclude models that cannot explain observed qualitative features , such as multistationarity .
We used Maple for the symbolic computations , such as finding det ( M ( x ) ) , the positive parameterizations , a ( x ^ ) and the vertices of the Newton polytope .
|
Mathematical modelling has become an important tool in biology . As modelling requires separating the essential from the ordinary , there is never just one model but a collection thereof . To understand biology through modelling it is therefore crucial to be able to tell which of these models are capable of reproducing an observed behaviour and which are not . For example , to understand cellular decision making , models allowing multiple equilibria are studied and one asks which models allow for this behaviour . Here we describe a procedure that links the existence of a unique and of multiple equilibria to the sign of a single expression . We demonstrate the usefulness of the procedure by applying it to models of gene transcription and cellular signalling .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"phosphorylation",
"chemical",
"compounds",
"applied",
"mathematics",
"reactants",
"organic",
"compounds",
"dna",
"transcription",
"simulation",
"and",
"modeling",
"algorithms",
"systems",
"science",
"mathematics",
"chemical",
"equilibrium",
"algebra",
"basic",
"amino",
"acids",
"amino",
"acids",
"polynomials",
"physical",
"chemistry",
"research",
"and",
"analysis",
"methods",
"computer",
"and",
"information",
"sciences",
"system",
"stability",
"proteins",
"gene",
"expression",
"chemistry",
"biochemistry",
"histidine",
"organic",
"chemistry",
"post-translational",
"modification",
"genetics",
"biology",
"and",
"life",
"sciences",
"chemical",
"reactions",
"physical",
"sciences"
] |
2017
|
Identifying parameter regions for multistationarity
|
Dynamics can provide deep insights into the functional mechanisms of proteins and protein complexes . For large protein complexes such as GroEL/GroES with more than 8 , 000 residues , obtaining a fine-grained all-atom description of its normal mode motions can be computationally prohibitive and is often unnecessary . For this reason , coarse-grained models have been used successfully . However , most existing coarse-grained models use extremely simple potentials to represent the interactions within the coarse-grained structures and as a result , the dynamics obtained for the coarse-grained structures may not always be fully realistic . There is a gap between the quality of the dynamics of the coarse-grained structures given by all-atom models and that by coarse-grained models . In this work , we resolve an important question in protein dynamics computations—how can we efficiently construct coarse-grained models whose description of the dynamics of the coarse-grained structures remains as accurate as that given by all-atom models ? Our method takes advantage of the sparseness of the Hessian matrix and achieves a high efficiency with a novel iterative matrix projection approach . The result is highly significant since it can provide descriptions of normal mode motions at an all-atom level of accuracy even for the largest biomolecular complexes . The application of our method to GroEL/GroES offers new insights into the mechanism of this biologically important chaperonin , such as that the conformational transitions of this protein complex in its functional cycle are even more strongly connected to the first few lowest frequency modes than with other coarse-grained models .
Protein dynamics plays a key role in describing the function of most proteins and protein complexes . The importance of protein dynamics studies has been increasingly recognized alongside the importance of the structures themselves . Experimentally , protein dynamics can be studied using nuclear magnetic resonance ( NMR ) [1 , 2] , time-resolved crystallography [3] , fluorescence resonance energy transfer ( FRET ) [4] and other single-molecule techniques [5] , etc . Computationally , the study of protein dynamics most commonly relies upon molecular dynamics ( MD ) simulations [6–8] . Normal mode analysis ( NMA ) is another popular and powerful tool for studying protein dynamics and was first applied to proteins in the early 80’s [9–11] . The advantage of normal modes over MD is that they can most efficiently describe protein motions near the native state . To apply NMA , a structure is first energetically minimized . The minimized structure is then used to construct the Hessian matrix , from which normal modes can be obtained from its eigenvectors and eigen-frequencies . This method poses a huge demand on computational resources , especially memory , since some large supramolecules may have hundreds of thousands of atoms . The time spent on computing the eigenvalues/eigenvectors also is large , of the order of the cube of the number of atoms . Consequently , its applications are limited to smaller systems . For this reason , many simplified models [12–33] have been developed for efficient normal mode computations . These models use simplified structural models or simplified force fields or commonly , both . One commonly applied type of coarse-grained model is the elastic network model [13 , 16] , which usually treats each residue as one node , and residue-residue interactions as Hookean springs . It has been demonstrated for a large number of cases that these extremely simplified models can still capture quite well the slow dynamics of a protein [12] . And because of their high level of simplicity , they have been successfully applied to study the normal mode motions of the largest structural complexes such as GroEL/GroES [18 , 34–38] , ribosome [22 , 39–41] , nuclear pore complex [29] , etc . However , along with the significant gains from this simplicity comes also some loss of accuracy , particularly in the accuracy of the normal modes [42 , 43] . The validity of most simplified models was justified a posteriori , by comparing with experimental B-factors or sets of multiple experimental structures for example . How well they preserve the accuracy of the original NMA has rarely been assessed directly [33] . To overcome this problem of accuracy , we built a strong connection between NMA and elastic network models ( ENMs ) through a series of steps of simplification that began with NMA and ended with ENMs , and proposed a new way to derive accurate elastic network models in a top-down manner ( by gradually simplifying NMA ) [33] . Our derivation was based on the realization that the Hessian matrix of the original NMA can be written as a summation of two main terms , the spring-based terms and the force/torque-based terms , with the former contributing significantly more than the latter . By ignoring the latter term , we obtained at a new model , sbNMA ( or spring-based NMA ) , that has high accuracy and closely resembles the original NMA and requires no energy minimization . sbNMA , like the original NMA , is force-field dependent and uses many parameters . By further simplifying it , we arrived at two force-field independent elastic network models , ssNMA ( simplified spring-based NMA ) and eANM ( enhanced ANM ) , both of which use many fewer parameters and yet still preserve most of the accuracy of NMA [33] . For example , the mean square fluctuations predicted by ssNMA for a set of small to medium proteins have an average correlation of nearly 0 . 9 with those predicted with the original NMA [33] . It was shown [42] also that ssNMA modes are more accurate than those from other elastic network models . However , this bridging , as detailed in Ref . 33 , connected NMA only with all-atom elastic network models but not with coarse-grained ones . Both ssNMA and eANM , though strongly resembling NMA , are by nature all-atom models and cannot be directly applied to coarse-grained structures . There is little doubt that for very large biomolecular systems , coarse-grained structure representations are needed , since all-atom normal mode analyses for such systems are computationally often out of reach . Our aim in this work is to extend the idea of bridging between NMA and elastic network models to coarse-grained models while preserving sufficient accuracy to obtain accurate protein dynamics even for very large systems . Is it possible to efficiently construct coarse-grained models whose description of the dynamics of a coarse-grained structure remains as accurate as that given by all-atom models ? Coarse-grained models , such as Cα-based models , obviously do not have all the structural details of all-atom models . But can they produce the dynamics of the Cα atoms as accurately as all-atom models ? Is it possible to have both the simplicity of coarse-grained structures and the accuracy of all-atom interactions ? These questions are the focus of this work . And we demonstrate affirmative answers to these questions by employing a novel iterative matrix projection technique . While our earlier work [33] connects between NMA and all-atom elastic network models and represents a force-field simplification of NMA while maintaining most of its accuracy , the present work presents an additional structural simplification from all-atom elastic network models to coarse-grained elastic network models . Combined together , the two pieces of work provide a bridge between all-atom NMA and coarse-grained elastic network models and should reveal deep insights for how to develop coarse-grained elastic network models that preserve most of the accuracy of all-atom NMA .
It is useful first to perform an operation that separates out the atoms used for the coarse-graining from the remainder of the atoms . Mathematically , it is possible to define a precise interaction model ( in the form of a Hessian matrix ) for the coarse-grained structure by first rearranging the original Hessian matrix Hall into parts for the coarse-grained atoms and the remainder of the atoms in separate subspaces , as was done by Eom et . al . [48] and Zhou and Siegelbaum [49]: H a l l = ( H c c H c r H c r ⊤ H r r ) , ( 1 ) H ˜ c c = H c c - H c r H r r - 1 H r c ⊤ , ( 2 ) where c stands for the atoms used for the coarse-graining , r stands for the residual part of the structure , and ⊤ represents the matrix transpose . It can be shown mathematically [50 , 51] that H ˜ c c maintains the same description of the mean-square fluctuations and cross-correlations of the coarse-grained structure as the original Hessian matrix . All elements in H ˜ c c - 1 are the same as their corresponding elements in H a l l - 1 . A similar idea of using matrix projection to obtain the motions for subsystems was previously used also by Brooks and Zheng and their co-workers [52 , 53] to develop their VSA ( vibration subsystem analysis ) model . However , this mathematical rearrangement in Eq ( 2 ) requires the inversion of Hrr , which appears to be nearly as difficult as computing the inverse of the original all-atom Hessian matrix , assuming the number of atoms in the coarse-grained structure is much smaller than that of the original all-atom model . Therefore , unless H ˜ c c can be computed in an efficient way , the precise interaction model defined in Eq ( 2 ) would be computationally too expensive to apply for very large systems and thus of little practical utility . In the next section , we present a novel way for computing H ˜ c c efficiently , without directly inverting Hall or Hrr . As a result , this permits an efficient construction of coarse-grained models that can represent the dynamics of the coarse-grained structure as accurately as all-atom models . To efficiently obtain the Hessian matrix H ˜ c c from Eq ( 2 ) without having to directly invert Hrr , we take advantage of the fact that the Hessian matrix Hall , the second derivatives of the potential , can be highly sparse for some all-atom models . Hall is not so sparse for the conventional NMA , due to the persistence of electrostatic interactions to long distances . However , it is sparse for ssNMA , an accurate all-atom model that closely resembles NMA as mentioned above . The potential for ssNMA includes most of the same interaction terms as for NMA , except for the electrostatic interactions [33] . As a simplified model of spring-based NMA ( or sbNMA ) , ssNMA uses one single uniform spring constant for all bond stretching terms , one uniform spring constant for all the bond-bending terms , and one for the torsional terms . Its non-bonded van der Waals interactions are truncated near the equilibrium distance to avoid negative spring constants in the Hessian matrix [33] . A single set of van der Waals radii are used for all van der Waals interactions . All the equilibrium values such as bond lengths , bond angles , and torsional angles are taken from the reference structure . Consequently , most of the off-diagonal elements in the ssNMA Hessian matrix are zero . In the following , we use ssNMA to construct the all-atom Hessian matrix Hall and show how a precise interaction model H ˜ c c can be efficiently constructed through an iterative matrix projection procedure . We call this model coarse-grained ssNMA , or CG-ssNMA . CG-ssNMA preserves the same accuracy as the all-atom ssNMA in its description of the dynamics of the coarse-grained structure . The procedure , as detailed below , takes full advantage of the sparseness of the Hessian matrix . Given a protein that has n atoms , one can iteratively reduce its size ( or coarse-grain it ) by removing one atom , or a group of r atoms , at a time without losing accuracy in depicting the motions of the remaining atoms . This can be done by adding a correction term to the interactions among the remaining atoms . Define by H the Hessian matrix with n atoms as follows: H = ( H k k H k r H k r ⊤ H r r ) , ( 3 ) where Hkk is the block matrix of H for the kept n − r atoms , Hrr the block matrix for r atoms to be removed , and Hkr represents the interactions between the group of atoms to be removed and the remaining atoms . The effective Hessian matrix H ˜ k k of the kept atoms after taking into account the correction term can be written as [42 , 48 , 49]: H ˜ k k = H k k - H k r H r r - 1 H k r ⊤ , ( 4 ) with H k r H r r - 1 H k r ⊤ being the correction term . It can be shown that the motions of the remaining atoms as described by H ˜ k k is the same as from the original Hessian matrix H . This numerical preservation is crucial when an all-atom Hessian matrix is gradually coarse-grained by repeatedly removing non-Cα atoms , since it guarantees that the quality of the description of the Cα atoms remains the same while the size of the Hessian matrix is being reduced . Note that each atom interacts only with a few , say m on average , atoms due the sparseness of the Hessian matrix . As a result , Hkr has only a small number ( rm ) of non-zero elements , representing the interactions between the group of atoms to be removed and the kept atoms . Therefore , the term H k r H r r - 1 H k r ⊤ in Eq ( 4 ) can be computed in O ( r3 + r2 m2 ) time . Coarse-graining the whole protein structure takes roughly n/r iterations and thus requires a total time of O ( ( r2 + rm2 ) n ) , which is linear in the protein size n . To further reduce the running time , matrix elements that are near zero ( weak interactions ) are set to zero if their absolute values are less than a predetermined threshold value ξ . A properly chosen ξ can further improve computation speed while preserving the accuracy , by effectively reducing the number of interactions , especially those between the atoms being removed and the remaining atoms . Different ξ values are tested , as detailed in the next section . Fig 1 illustrates how the sparseness of the Hessian matrix is maintained throughout the iterative matrix projection procedure . At the initial step , atoms are shuffled so that Cα atoms are grouped together and placed on the left-most side of the Hessian matrix , as shown in Fig 1 ( A ) , where the grouped Cα and non-Cα atoms are represented by dark and light gray blocks , respectively . Blue dots represent the non-zero elements of the Hessian matrix . The non-Cα atoms can then be further rearranged , for example , using the Cuthill-McKee algorithm [54] , so that the atoms that interact with one another are placed close together in the matrix . As a result , the non-zero elements are relocated near the diagonal of the matrix ( see Fig 1 ( B ) ) . In such a sparse matrix , Fig 1 ( C ) shows the effect of applying one matrix projection using Eq ( 4 ) , where the red dots represent the elements of the matrix whose values are modified . Note that the sparseness of the non-Cα region is mostly unaffected by the projection . The sparseness of the white region ( interactions with Cα atoms ) can be maintained by using an appropriate threshold value ξ mentioned earlier . Algorithm 1 lists the steps that iteratively reduce the all-atom Hessian matrix to a coarse-grained one . The algorithm takes as input the all-atom Hessian matrix H , a set of Cα atom indices {k1 , … , kn} , and a threshold value ξ . All matrix elements whose absolute values are less than ξ are set to 0 . In practice , it turns out that lines 4–11 run more efficiently if each iteration of the coarse-graining process removes not a single atom but a group of atoms ( Ri as in line 2 ) . Removing a group of adjacent atoms reduces the average number of interactions ( m in the above Big-O notation ) with the remaining atoms . These groups of atoms are determined by spatially partitioning the whole structure ( 3-D ) into cubic blocks ( 18~Å for each dimension ) . These blocks represent initial groups of atoms . The reason why atoms are partitioned in this way is to minimize the number of interactions among the different groups . Blocks are then sorted by their sizes ( i . e . , the number of atoms ) in descending order . Next , starting with the smallest one , blocks on the “small” end ( usually blocks on the outsides of a structure ) are iteratively merged together with the next smallest block as long as the size of the merged group does not exceed the size limit ( which is about 500 atoms per group , the number of atoms in a regular cubic block ) . The merging process stops when there are no small blocks left to be merged . In lines 7 and 9 , sparse ( A , b ) returns a sparse matrix of A by setting to zero A’s elements that satisfy |Ai , j| < b , where |Ai , j| is the absolute value of Ai , j . Threshold ξ/m is used in line 9 since the addition ( or subtraction ) in line 10 is accumulated m times . Line 9 prevents very small values from being added to H in line 10 and then removed in line 7 at the next iteration . Algorithm 1 CoarseGrain ( H , {k1 , … , kn} , ξ ) 1: K ← {k1 , … , kn} 2: R ← {R1 , R2 , … , Rm} 3: H ← Hessian matrix of H reshaped in the order of K , R1 , R2 , … , Rm 4: for i = m , m − 1 , … , 1 do 5: k ← | K | + ∑ j = 1 i - 1 | R j | 6: r ← k + ∣Ri∣ 7: B ← sparse ( H1 . . k , k + 1 . . r , ξ ) 8: D ← Hk + 1 . . r , k + 1 . . r 9: E ← sparse ( BD−1 B⊤ , ξ/m ) 10: H1 . . k , 1 . . k ← H1 . . k , 1 . . k − E 11: end for 12: H ← sparse ( H1 . . |K| , 1 . . |K| , ξ ) 13: return H
In this section , we first verify computationally that the coarse-grained ssNMA model constructed according to the proposed procedure indeed not only preserves the accuracy of all-atom models in its description of the motions of the coarse-grained structure but also is computationally efficient . To this end , we first show , by applying it to a dataset of 177 small to medium proteins , that with a properly chosen threshold value ξ , the coarse-grained ssNMA preserves full accuracy . We then extend the same coarse-graining procedure , using the same ξ value , to construct coarse-grained ssNMA Hessian matrices for 80 large superamolecules of different sizes and show that the construction of these ssNMA Hessian matrices requires only a nearly linear time and can thus be carried out quickly , even for large systems . To validate the accuracy of the method , Algorithm 1 is applied to 177 small-to-medium sized proteins whose sizes are greater or equal to 60 residues but less than 150 . This is the same set of proteins that was used in our earlier work [33] . Only small to medium sized proteins are used at this stage due to the high computational costs of running all-atom models , which have also been computed here for comparison purposes . Each protein structure is first energy minimized . From the all-atom ssNMA Hessian matrix , two coarse-grained Hessian matrices , H and H ^ , are computed . H is computed by direct matrix projection ( as in Eq ( 2 ) ) , which is an exact but very expensive computation , while H ^ is computed with the proposed iterative projections as in Algorithm 1 . To show that H ^ preserves the same accuracy as H , we compute the correlations between mean square fluctuations ( MSF ) computed with H and those with H ^ , and the eigenvalue-weighted overlaps between modes by H and those by H ^ . The eigenvalue-weighted mode overlap is defined as: ∑ i = 7 3 n w i w | m i · m ^ i | , ( 5 ) where n is the number of atoms , mi ( and m ^ i ) is the ith mode of H ( and H ^ ) , wi = 1/λi is the relative weight and is set to be the inverse of the ith eigenvalue of H , and w = ∑ i = 7 3 n w i is the normalization factor . The reason why we use the modes with the same indices ( mi and m ^ i ) instead of the best matching modes when computing the weighted-overlap is to measure also how well the order of the modes is preserved . Lower frequency modes are given higher weights in this weighted overlap measure . The intuition behind this weighted mode scheme is that it represents how similar the modes ( including their orders ) are between the two models . Table 1 shows the levels of accuracy that can be achieved when different threshold values ξ are applied to ssNMA [33] . It is seen that ssNMA preserves the full accuracy ( 1 . 0 in correlations and overlaps ) in mean square fluctuations and modes when a threshold value ( ξ ) as large as 0 . 01 is used . Similar results are also seen for the enhanced ANM model ( eANM ) [33] , another all-atom model that closely resembles NMA . Using a large threshold value allows the sparseness of the Hessian matrix to be maintained during the iterative matrix projection process and consequently the construction of the coarse-grained ssNMA Hessian matrix to be carried out quickly . For conventional NMA , however , the iterative coarse-Graining approach as described above does not work nearly as well ( see Table 1 ) . This is due to the slowly-decreasing , long-range electrostatic interactions . Secondly , we look at the efficiency , i . e . , how much time does this iterative coarse-graining procedure require ? To this end , we apply the same iterative coarse-graining procedure to construct coarse-grained ssNMA Hessian matrices for a number of large proteins and protein complexes . The same threshold value , ξ = 0 . 01 , is used , which has been shown in the previous section to preserve the full accuracy . Fig 2 shows the efficiency ( computational time ) of the proposed method as a function of the system size . In the figure , each blue and red point represent respectively , for a protein of that size , the coarse-graining time , i . e . , the time required to construct the coarse-grained ssNMA Hessian matrix ( with ξ = 0 . 01 ) , and the diagonalization time of that coarse-grained Hessian matrix . The dashed lines show the growth rates of the time cost as a function of the system size . The curves are obtained from the least squares fitting to a non-linear function f ( x ) = axb . As shown in the figure , the diagonalization time ( red curve ) grows approximately as the cube , while the coarse-graining time grows approximately linearly . Especially for large complexes , the time needed for coarse-graining the all-atom Hessian matrix using Algorithm 1 becomes increasingly smaller relative to the diagonalization time . As a result , the total time for computing the normal modes for such large protein complexes using the coarse-grained ssNMA Hessian matrices is about the same as for other coarse-grained elastic network models such as ANM . In summary , the results in this section demonstrate that the proposed iterative coarse-graining procedure not only preserves the accuracy in depicting the motions of the coarse-grained structures but is also computationally highly efficient . This result is significant since it means that we can construct coarse-grained models that preserve all-atom accuracy even for very large protein complexes , which was not previously possible . Next , as an application , we apply the proposed procedure to compute and analyze the dynamics of the GroEL/GroES complex . The GroEL/GroES complex [55] is a molecular chaperone that assists the unfolding of partially folded or misfolded proteins , by providing them with the chance to refold . GroEL consists of cis and trans rings , each of which has 7 subunits . Each subunit is 547 residues . GroES also has 7 chains and each chain contains 97 residues . The GroEL cis-ring and GroES form a capped chamber that can hold proteins and facilitate protein unfolding partly through their intrinsic collective motions , such as compressing , stretching , twisting , shearing , and relaxing . Fig 3 shows the GroEL/GroES structure ( pdbid: 1AON ) in top and front views . In Fig 3 ( A ) , the three domains of the cis and trans rings are distinguished with different colors: equatorial ( green ) , intermediate ( yellow ) , and apical ( blue ) domains . To understand its functional mechanisms , it is informative to obtain the intrinsic motions of this complex . However , for large protein complexes such as GroEL/GroES that has over 8 , 000 residues , standard all-atom NMA will take a prohibitively large memory and a long time to run . Consequently , past normal mode studies on this complex were limited to coarse-grained models [18 , 36] , or all-atom models of single subunits [34] . Though a more accurate description of its normal modes is highly desirable and may provide deeper insights into the functional mechanism of the complex , it was lacking due to computational constraints . Here , we apply the proposed iterative procedure to obtaine a coarse-grained ssNMA Hessian matrix for the entire GroEL/GroES complex . This coarse-grained ssNMA ( or CG-ssNMA ) model preserves the same all-atom accuracy in its description of the motions of the coarse-grained structure as the original ssNMA . First , we apply CG-ssNMA to compute mean-square fluctuations . To this end , we use the GroEL-GroES- ( ADP ) 7 complex ( pdbid: 1AON ) [55] as the initial structure . This structure is composed of the co-chaperone GroES , the cis-ring whose subunits are bound with 7 ADPs , and the trans-ring ( see Fig 3 ) . The motion correlation ( or cooperativity ) Ci , j between the i-th and j-th residues can be expressed as follows: C i , j = ⟨ r i · r j ⟩ ( ⟨ r i · r i ⟩ ⟨ r j · r j ⟩ ) 1 / 2 , ( 6 ) where ri and rj are the displacement vectors for the i-th and j-th residues in a given mode , respectively , a ⋅ b is the dot product of two vectors a and b , and ⟨a⟩ is the average value of a within the first k lowest frequency modes . Fig 5 shows the cooperativity of residue motions within each subunit and across the whole protein complex . The cooperativity plot is generated from the first 15 dominant ( i . e . , lowest frequency ) modes given by the coarse-grained ssNMA . Fig 5 ( A ) shows the cooperativity among residue pairs within a single set of subunits: one subunit from the cis ring ( chain A of 1AON ) , one from the trans ring ( chain N ) , and one from GroES ( chain O ) . The cooperativity of residue pairs is color coded: red for strong correlated motions ( Ci , j = 1 ) , cyan for uncorrelated ( Ci , j = 0 ) , and purple/blue for anti-correlated ( Ci , j = −1 ) . The most noticeable difference between the cis and trans rings is the involvement of the intermediate domain in the motions of the apical or equatorial domain . In the cis ring , the red regions indicate that the motions of intermediate domain ( I1 and I2 ) are strongly correlated with those of the equatorial domain ( E1 and E2 ) , while the motions of the apical domain ( A ) are largely independent of them . In the trans ring , however , the motions of intermediate domains ( I1’ and I2’ ) are more correlated with those of the apical domain ( A’ ) than with the equatorial domain ( E1’ and E2’ ) . A similar cooperativity plot for the ANM model is given in Supporting information ( S1 Fig ) . Overall , the two methods give similar correlation patterns . The main noticeable difference is that the relative motions between equatorial ( E1’ and E2’ ) and apical ( A’ ) domains of the trans-ring subunit are more clearly shown as anti-correlated ( i . e . , the region appears to be bluer ) in Fig 5 ( given by ssNMA ) than with ANM shown in S1 Fig . One general role of the intermediate domain is connecting the apical and equatorial domains and facilitating the communication between them . The results in Fig 5 imply that the dynamics or motion partner of the intermediate domain depends on the structural form of the GroEL ring: cis or trans . Considering the structure transitions of cis → trans and trans → cis that take place during the GroEL/GroES functional cycle , it is not surprising that the transition path in the former case may be different from a simple reverse of the latter . Additionally , Fig 5 ( A ) shows that the motions of GroES and the apical domain ( A ) of the cis ring also are highly correlated . The cooperativity of all the residues in the complex is presented in Fig 5 ( B ) . Along the off-diagonal there are four dark blue mesh bands , implying that the apical domains of the subunits that sit on opposite sides across the rings , such as chain C/D and chain A , are strongly anti-correlated . Another interesting observation is that the motions of GroES are strongly anti-correlated to the equatorial domain of the cis ring . The ssNMA model presented in this work , though coarse-grained in structure , maintains an all-atom level accuracy in its description of the interactions and consequently an all-atom level accuracy in its description of the normal mode motions of the coarse-grained structure . Such an accurate description of the normal mode motions is highly desirable but has not been performed before for large protein complexes such as GroEL/GroES that has over 8 , 000 residues . In the following , we will examine closely the first few lowest frequency modes of ssNMA and characterize their motions . The quality of these modes is then assessed . A comparison with Cα-based ANM modes is made at the end . Fig 6 characterizes the slow dynamics of GroEL/GroES in individual modes or pairs of modes . The first lowest frequency mode portrays a rotational motion around the cylindrical axis of the complex . This mode matches with the first mode of ANM nearly perfectly , with a high overlap of 0 . 97 . The third mode is about opening the gate of the trans ring to receive substrates into its chamber , by moving its apical domains to conform its structure to resemble that of the cis ring . The second and fourth modes are mainly about a swing motion of the trans ring . This motion also helps to open the chamber gate of the trans ring . In ssNMA , this gate opening motion in the trans ring is clearly captured by these three distinct modes , especially the third mode , whose importance is manifested also in the conformation transitions during the GroEL/GroES functional cycle that will be described in the next section . In ANM , there is not a single mode that closely matches the third mode of ssNMA . The gating opening motion seems to spread into several modes in ANM and be mingled with other motions . The 5th–6th modes are shearing motions of the GroES cap and the apical domains of the cis ring . This motion causes them to shift significantly relative to the equatorial domains . This motion ( in the 5th/6th modes ) is similar , to some extent , to that in the second and third modes of ANM , which in turn have some resemblance also to the second/fourth modes of ssNMA . The 7th–10th modes display alternating motions of compression and extension of the whole complex . The 11th mode is mainly about stretching/compressing the chamber of the cis-ring . To some extent , this motion ( of the 11th mode ) changes the structure of the cis ring towards the shape of the trans ring . The 12th–13th modes are mainly about tilting the cis/trans rings and the GroES cap . The animations of the top 13 dominant modes ( lowest frequency ) of ssNMA ( and ANM ) are made available at http://www . cs . iastate . edu/~gsong/CSB/coarse . Next , we compare more quantitatively the modes of ssNMA and ANM . In this section , we apply CG-ssNMA to interpret the conformation transitions in the functional cycle of GroEL/GroES . Our hypothesis is that the intrinsic normal mode motions of the complex should facilitate its conformation transitions . To measure how well the modes are related to the conformation transitions , we compute the overlaps between normal modes and a given transition . We then repeat the computations and analysis using ANM and compare the results with those from CG-ssNMA . In total there are six conformation transitions among the five known conformation states of the complex ( see Table 3 ) considered: T → R , T → R′′′ , R ′ ′ → R flipped ′ ′ , R nocap ′ ′ → R nocap , flipped ′ ′ , R′′ → S , and S → R′′ , where “nocap” stands for the absence of the GroES cap . Table 4 summarizes , for these transitions , the top 3 largest overlaps found using CG-ssNMA and ANM . The indices of the modes that give the largest overlaps also are given . The first two cases represent transitions from the apo form to ATP/GroES bound forms . The transitions R ′ ′ → R flipped ′ ′ and R nocap ′ ′ → R nocap , flipped ′ ′ were thought to take place during the functional cycle of GroEL/GroES [58] , in which the two GroEL rings alternate as a functional chaperone . However , recent work [59] suggested that in vivo the GroEL/GroES complex assumes a football shape in the functional process and that both GroELs might work simultaneously as protein unfolding chaperones . For this reason , we consider also the functional transitions between states R′′ and S . Table 4 lists the results . T → R and T → R * ′ ′: Transitions T → R and R * ′ ′ in Table 4 show that these are mostly achieved with a torsional motion along the vertical axis of the structure . Both the CG-ssNMA and ANM models capture this torsional motion , but their mode indices are different . It is the fourth mode in CG-ssNMA that gives the largest overlap while it is the first in ANM . The results clearly show that the motion to R ( as induced by ATP binding ) is along the path to R * ′ ′ , as observed by Roseman et al . [60] from low resolution cryo-EM images . R′′→Rflipped′′: Ranson et al . [58] suggested that the functional process of GroEL/GroES involves alternations to the two GroEL rings as functional units and the complex is bullet-shaped [55] in vivo . Here we consider the transition from a bullet-shaped complex ( R′′ ) to its flipped counterpart . In this transition , one of the GroEL rings goes from the trans form to the cis form , while the other ring changes from cis to trans . Results in Table 4 show that the coarse-grained ssNMA captures well the transition from trans to cis using its fourth mode , which has the second largest overlap , while the 17th mode has the best overlap and characterizes mostly the transition from cis to trans ring , as well as a partial transition from trans to cis . ANM , on the other hand , describes the transition of trans → cis and cis → trans using the 17th and 18th modes , each of which is a mixture of both cis-ring and trans-ring deformations . It is thought that after the binding of the ATPs to the trans ring , the GroES cap is removed and the substrate protein is released . Then the two GroEL rings go through trans → cis and cis → trans transitions , respectively , and another GroES will bind the opposite ring , completing a cycle . The GroES cap stabilizes the cis ring in its conformation and prevents its transition to a trans conformation . However , after the ATP binding at the opposite ring , the GroES cap is removed , which makes the transition from a cis to a trans conformation easier . The larger overlap seen in this transition without the GroES cap ( see Table 4 ) provides evidence that GroES is probably removed first before the cis ↔ trans conformation transitions take place rather than occurring simultaneously . This agrees with the idea that structures facilitate functional transitions . R′′ → S ( opening the trans ring gate ) : Recent work by Fei et al . [59] suggested that the GroEL/GroES complex in vivo should have a football shape . The formation of a football-shaped GroEL/GroES complex was thought to be promoted by substrate protein ( SP ) , and that “SP shifts the equilibrium between the footballs and bullets in favor of the former , consequently making them the predominant species . ” [59] Here , we examine the transitions between a football-shaped complex and a bullet-shaped complex . Transition R′′ → S opens the gate of the trans ring to receive a substrate protein ( unfolded or misfolded ) in its chamber . This is accomplished by conforming the structure of its apical domain to that of a cis ring ( see the third mode in Fig 6 and in S1 Video ) . S2 Fig highlights the conformation change that takes place within a trans-ring monomer in this transition . The overlaps between the transition and normal modes reveal a large contribution by the torsional rotation along the vertical axis ( mode 1 ) , as the trans ring of S is rotated about 8 degree counter-clockwise from that of R′′ [59] . Secondly , this transition is captured by the third ssNMA mode that mainly depicts a chamber-opening motion . In contrast , CA-ANM provides this transition mainly using its 20th mode , which is a mixture of the chamber opening motion and some other deformation of the cis ring and the GroES cap . S → R′′ ( closing the cis ring gate ) : Transition S → R′′ closes the gate of the cis ring to conform its structure to that of a trans ring . Similar to the transition R′′ → S , this transition requires torsional rotations and gate-closing motions . The coarse-grained ssNMA captures this transition using the second and third low frequency modes . CA-ANM captures the torsional rotation properly using the third mode , but has to rely on higher-frequency modes to capture the gate-closing transition ( See Table 4 , last column ) .
Although the proposed iterative coarse-graining procedure can be used to efficiently construct coarse-grained models whose description of the dynamics of the coarse-grained structures preserves all-atom accuracy , it is limited in that it can be applied only to some of the models , such as ssNMA , eANM , or sbNMA ( see S1 Table ) . It cannot be applied to the conventional NMA . This is because the potential of NMA contains electrostatic interactions that decay rather slowly and consequently the NMA Hessian matrix is not sparse; however , there remain some uncertainties about how to best compute the electrostatics . A possible partial solution is to add a switch function to the non-bonded interactions of NMA and make it decay to zero at some cutoff distance , as is commonly done in MD simulations . This will make the Hessian matrix much sparser and make it possible to apply the proposed iterative procedure to NMA . We have shown this to be the case ( see results in S1 Table ) . However , this is only a partial solution since it recovers only the short range part of the electrostatics . The long range electrostatic interactions , which may have a pronounced contribution to long-range collective motions and cooperativity , are still missing . Additionally , the cumbersome energy minimization ( which ssNMA does not require ) becomes necessary , which can be a challenge when working with large biomolecular complexes . One possible future work is to study the effects of electrostatic interactions on normal modes , specifically the extent of contributions by short-range or long-range electrostatic interactions . If the short-range component of the electrostatic interactions dominates the long range component in contributing to normal modes , then the aforementioned partial solution will provide an excellent approximation .
|
Proteins and other biomolecules are not static but are constantly in motion . Moreover , they possess intrinsic collective motion patterns that are tightly linked to their functions . Thus , an accurate and detailed description of their motions can provide deep insights into their functional mechanisms . For large protein complexes with hundreds of thousands of atoms or more , an atomic level description of the motions can be computationally prohibitive , and so coarse-grained models with fewer structural details are often used instead . However , there can be a big gap between the quality of motions derived from atomic models and those from coarse-grained models . In this work , we solve an important problem in protein dynamics studies: how to preserve the atomic-level accuracy in describing molecular motions while using coarse-grained models ? We accomplish this by developing a novel iterative matrix projection method that dramatically speeds up the computations . This method is significant since it promises accurate descriptions of protein motions approaching an all-atom level even for the largest biomolecular complexes . Results shown here for a large molecular chaperonin demonstrate how this can provide new insights into its functional process .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[] |
2015
|
Bridging between NMA and Elastic Network Models: Preserving All-Atom Accuracy in Coarse-Grained Models
|
Germline stem cells in the Drosophila ovary are maintained by a somatic niche . The niche is structurally and functionally complex and contains four cell types , the escort , cap , and terminal filament cells and the newly identified transition cell . We find that the large Maf transcription factor Traffic jam ( Tj ) is essential for determining niche cell fates and architecture , enabling each niche in the ovary to support a normal complement of 2–3 germline stem cells . In particular , we focused on the question of how cap cells form . Cap cells express Tj and are considered the key component of a mature germline stem cell niche . We conclude that Tj controls the specification of cap cells , as the complete loss of Tj function caused the development of additional terminal filament cells at the expense of cap cells , and terminal filament cells developed cap cell characteristics when induced to express Tj . Further , we propose that Tj controls the morphogenetic behavior of cap cells as they adopted the shape and spatial organization of terminal filament cells but otherwise appeared to retain their fate when Tj expression was only partially reduced . Our data indicate that Tj contributes to the establishment of germline stem cells by promoting the cap cell fate , and controls the stem cell-carrying capacity of the niche by regulating niche architecture . Analysis of the interactions between Tj and the Notch ( N ) pathway indicates that Tj and N have distinct functions in the cap cell specification program . We propose that formation of cap cells depends on the combined activities of Tj and the N pathway , with Tj promoting the cap cell fate by blocking the terminal filament cell fate , and N supporting cap cells by preventing the escort cell fate and/or controlling the number of cap cell precursors .
Stem cells retain the capacity for development in differentiated organisms , which is important for tissue growth , homeostasis and regeneration , and for long-term reproductive capability . Stem cells are often associated with a specialized microenvironment , a niche that is essential for the formation , maintenance , and self-renewal of stem cells by preventing cell differentiation and controlling rate and mode of cell division [1 , 2] . The niche for the germline stem cells ( GSCs ) in Drosophila serves as an important model for the analysis of interactions between niche and stem cells [1 , 3–5] . The astounding fecundity of Drosophila females that can lay dozens of eggs per day over several weeks depends on approximately 100 GSCs that are sustained by 40 stem cell niches . To understand the formation and maintenance of these GSCs , it is important to understand how stem cell niches form and how they function . The GSC niche of the Drosophila ovary consists of three somatic cell types: cap cells , escort cells , and terminal filament ( TF ) cells ( Fig 1A ) . GSCs are anchored to cap cells by DE-cadherin-mediated adhesion and require close proximity to cap cells to retain stem cell character [6–8] . Cap cells secrete the BMP homolog Decapentaplegic ( Dpp ) , activating the TGFß signaling pathway in adjacent GSCs [9] , which leads to the repression of the germline differentiation factor Bag-of-Marbles ( Bam ) [10 , 11] . Through Hedgehog ( Hh ) signaling , cap cells also appear to stimulate escort cells to secrete Dpp [12] . The combined pool of Dpp from cap and escort cells , together with mechanisms that concentrate Dpp in the extracellular space around GSCs [13] , promotes the maintenance of 2–3 GSCs , whereas the adjacent GSC daughter cells that have lost the contact to cap cells will enter differentiation as cystoblasts [3 , 4] . In contrast , TFs are not in direct contact with GSCs but serve important functions in the development and probably also in the maintenance and function of GSC niches [14] . Formation of GSC niches begins with the progressive assembly of TFs by cell intercalation during the 3rd larval instar [15–17] . The process of TF cell specification is not understood but might start in 2nd instar when the first TF precursor cells appear to leave the cell cycle [18 , 19] . TF morphogenesis depends on the Bric à brac transcriptional regulators that control the differentiation of TF cells and their ability to form cell stacks [15 , 16 , 20] , and involves the Ecdysone Receptor ( EcR ) [21 , 22] , Engrailed [23] , Cofilin [24] , and Ran-binding protein M ( RanBPM ) [25] . The number of TFs that form at the larval stage determine the number of GSC niches at the adult stage [26–28] , and are regulated by several signaling pathways that control cell division and timing of cell differentiation in the larval ovary , including the EcR [22] , Hippo and Jak/Stat [27 , 28] , Insulin [29] and Activin pathways [19] . Despite the recent advance in elucidating mechanisms that control the number of GSC niches and the temporal window in which they form [14] , relatively little is known about the origin and specification of the somatic cell types of the GSC niche . Notably , the origin and specification of cap cells , the main component of an active GSC niche is little understood . Cap cells ( also called germarial tip cells ) are first seen at the base of completed TFs at the transition from the 3rd larval instar to prepupal stage [16 , 17] . They appear to derive from the interstitial cells ( also called intermingled cells ) of the larval ovary that are maintained by Hh signaling from TFs [14 , 30] . The formation of cap cells is accompanied by the establishment of GSCs [17] . The N pathway contributes to the development of cap cells [3] . A strongly increased number of functionally active cap cells per niche form in response to overexpression of the N ligand Delta ( Dl ) in germline or somatic cells , or the constitutive activation of N in somatic gonadal cells [8 , 22 , 31] . The ability of N to induce additional cap cells seems to depend on EcR signaling [22] . Loss of Dl or N in the germline had no effect on cap cells . However , loss of N in cap cell progenitors or Dl in TF cells caused a decrease in the number of cap cells [8 , 32] . A current model suggests that Dl signaling from basal-most TF cells to adjacent somatic cells together with Dl signaling between cap cells allows for a full complement of cap cells to form [8 , 32] . Furthermore , N protects cap cells from age-dependent loss as long as its activity is maintained by the Insulin receptor [32 , 33] . The Jak/Stat pathway , which operates downstream or in parallel to the N pathway in the niche [34] , is not required for cap cell formation [34 , 35] . As cap cells were reduced in number but never completely missing when the N pathway components were compromised [8 , 31 , 32] , the question remains whether N signaling is the only factor that is important for cap cell formation . Furthermore , no factor that operates downstream of N has been identified that is crucial for cap cell formation . Here , we find that Traffic jam ( Tj ) is both required for cap cell specification and for the morphogenetic behavior of cap cells , enabling them to form a properly organized niche that can accommodate 2–3 GSCs . Tj is a large Maf transcription factor that belongs to the bZip protein family [36] . Its four mammalian homologs control differentiation of several cell types and are associated with various forms of cancer [37–39] . Tj is essential for normal ovary and testis development [36 , 40–42] , and is only expressed in somatic cells of the gonad [36 , 43 , 44] . Interestingly , Tj is present in cap cells and escort cells but not in TFs [36] . We show that Tj is essential for the formation of the GSC niche . First , Tj regulates the behavior of cap cells , enabling them to form a cell cluster instead of a cell stack , which appears to be important for the formation of a normal-sized GSC niche with the capacity to support more than one GSC . Second , cap cells adopt the fate of TF cells in the absence of Tj function , and TF cells develop cap cell-like features when forced to express Tj , indicating that Tj specifies the cap cell fate . Genetic interactions suggest that Tj and N are required together for cap cell formation , but have different functions in this process . For somatic gonadal cells to adopt the cap cell fate , we propose that Tj has to be present to inhibit the TF cell fate and N has to be present to prevent the escort cell fate and/or produce the correct number of cap cell precursors .
To understand the defects in the stem cell niche of tj mutant ovaries , we reviewed the organization of the wild-type GSC niche , confirming and extending previous observations . The three somatic cell types of the GSC niche could be distinguished based on their position , cell and nuclear shape , and marker expression ( Fig 1A–1D; S1 Table ) [3 , 4 , 45] . The TF is a stack of disc-shaped cells ( Fig 1A and 1B ) [46] . The cap cell cluster at the tip of the germarium was either centered ( Fig 1B ) or formed an asymmetric streak that was attached to the base of a TF ( Fig 1C ) [6 , 47] . Cap cells had a rounded shape and were tightly packed in a cluster , with their nuclei in close proximity ( Fig 1A–1D ) . Nuclei of escort cells had an angular ( often triangular ) appearance , and were bigger and more widely spaced than cap cell nuclei ( Fig 1D ) . The anterior and posterior location of cap cells and escort cells , respectively , in relation to GSCs , produced a prominent gap between cap and escort cell nuclei ( Fig 1A and 1D ) . GSCs made extensive contact to cap cells by forming a Bezel set-like rim of plasma membrane around each cap cell ( Fig 1A and 1D–1F' ) [48] . We found that a GSC usually forms at least one prominent cellular protrusion toward cap cells , which distinguishes it from cystoblasts ( Fig 1D , 1E' and 1F' arrowheads ) . These protrusions were seen with germline-specific markers that either label the cytoplasm ( Vasa; Fig 1D ) or the plasma membrane ( nos-Gal4 UAS-Gap43-mEos; Fig 1E and 1F' ) . GSC protrusions were visible at various stages of the cell cycle as indicated by changes in the position of the spectrosome organelle ( Fig 1E and 1F' ) [49] . The described morphological features helped identify cell types in the ovarian stem cell niche in addition to molecular markers . Despite their different morphologies , cap cells have several markers in common with TF cells and some markers with escort cells ( Fig 1B and 1D; S1 Table ) . Very few markers have been identified that seem to be specific for just one of these three cell types , but several markers showed differences in expression level ( Fig 1B and 1D; S1 Table ) [3 , 4 , 45] . Tj is expressed in cap cells and escort cells , which are located within the germarium and in contact to germline cells , but is not detected in TF cells , which form a stalk outside of the germarium ( Fig 1B–1D ) [36] . In addition , the cell that connects the cap cell cluster with the TF also contains Tj although at a considerably lower level than adjacent cap cells . We named this cell , which is disc-shaped similar to TF cells and aligned with TF cells , 'Transition cell' ( Fig 1A and 1D ) . It might correspond to one of the basal cells of the TF that have been mentioned previously [47] . In each ovariole of a wild-type ovary , a bab-lacZ positive TF and cap cell cluster are followed by a string of follicles ( Fig 2A ) . Adult ovaries from tjeo2/tjeo2 null mutant females ( tjnull ) lack germaria and follicles , and appear to mostly consist of TFs and ovariole sheath tissue ( Fig 2B ) [36 , 40] . Although TFs were seen properly oriented and enveloped by ovariole sheaths in some tj mutant ovaries , they were often not fully separated from each other , forming a tangled mass , or protruded from the ovary and adhered to extra-ovarian fat body tissue ( Fig 2B ) . Strikingly , the TFs appeared substantially longer in tjnull than in wild-type ovaries ( Fig 2A and 2B ) . Instead of containing an average of 8 disc-shaped cells as in wild type ovaries ( Fig 2C ) [15] , tjnull ovaries had TFs that contained on average 15 disc-shaped cells ( Fig 2C ) . Moreover , cap cell clusters were not detected . To determine whether there is a connection between the larger stalks and the absence of cap cells , we used tjz4735 , a genetic null allele that produces non-functional but detectable Tj protein to visualize cap cells [36] . The analysis of pupal tjz4735/tjeo2 ovaries showed that Tj-positive cells , which were never seen outside the germarium in wild type ( Fig 2D ) , formed the basal portion of the TFs in mutant ovaries and were disc-shaped similar to normal TF cells ( Fig 2E ) . The Tj-positive cells were often organized in a single file following the Tj-negative TF cells , although some stalks were found to branch or to form knob-like structures ( Fig 2F ) . We conclude that cap cells form a TF-like stalk in the absence of tj function . A similar niche defect was observed in a hypomorphic tj mutant . We isolated a very weak hypomorphic tj allele , tj39 , through mobilization of tj-Gal4 . It contains a P element fragment just upstream of the tj transcription unit and does not affect the tj coding region ( S1A and S1B Fig; see Materials and methods ) . Although tj39 homozygous females had normally looking and functional ovaries , tj39 caused sub-fertility in trans to the tjeo2 null allele . tj39 produces full-length Tj protein , whereas tjeo2 produces a truncated isoform that is predicted to lack the DNA binding and leucine zipper domains due to a premature stop codon ( S1C Fig ) [36] . The amount of full-length Tj in tj39/tjeo2 ovaries was reduced to 40–50% of the wild-type value , whereas it was only reduced to approximately 70% in tjeo2/+ ovaries ( S1C and S1C' Fig ) . Hypomorphic tj39/tjeo2 ( tjhypo ) ovaries had proper ovarioles with a germarium and developing follicles , but developed unusually pear-shaped germaria with age and had abnormal interfollicular stalks ( see S3C and S3D Fig ) . Notably , tjhypo ovaries displayed abnormally long TFs that included Tj-positive cells ( Fig 2C and 2G–2I ) . In some cases , all Tj-positive cells anterior to GSCs were integrated into the TF ( Fig 2H and 2J ) . More frequently , while most Tj-positive cells were part of the TF a few remained clustered at the tip of the germarium ( Fig 2I and 2J ) , explaining the smaller cell number in TFs of tjhypo compared with tjnull mutant ovaries ( Fig 2C ) . Moreover , stalk-forming Tj-positive cells were often disc-shaped and arranged in a single row similar to normal TF cells , and even clustered Tj-positive cells often appeared flatter in shape than regular cap cells ( Fig 2H and 2I ) . The range in cellular behavior suggests that these Tj-positive cells have a hybrid character , having gained TF cell characteristics and lost cap cell features to a variable degree . The hypomorphic tj mutant phenotype supports the notion that Tj is important for niche organization , enabling cap cells to form a cluster inside the germarium where they can contact GSCs . If additional TF cells form at the expense of cap cells , as our data suggest , one would expect the number of cells in the TF of tjnull ovaries to equal the sum of TF cells and cap cells in wild-type ovaries . Indeed , those numbers were similar when we counted the cells of individual stalks using the markers B1-lacZ and Lamin C ( LamC ) that both label TF and cap cells but not escort cells ( Fig 2K , S1 Table ) . In tjhypo ovaries , a combination of the markers LamC , labeling TF and cap cells , and Tj , labeling cap and escort cells , allowed us to clearly distinguish all three cell types . The number of cap cells in our controls was similar to previous reports ( Fig 2L ) [6 , 8 , 32] . A minor increase in the number of cap cells in tjhypo mutant ovaries was observed in two out of three genetic backgrounds , with an average of 6 . 5–8 . 3 cap cells in tjhypo mutants compared with 5 . 8–7 cap cells in controls ( Fig 2L ) . However , there was no significant difference in the number of TF cells or in the combined number of TF and cap cells between control and tjhypo ovarioles ( Fig 2L ) . The total number of stalk-forming cells was lower than the combined count of TF and cap cells in tjhypo ovarioles , which was expected as not all cap cells become part of the stalk in tjhypo ovaries ( Fig 2J ) . Taken together , our quantitative analysis indicates that the number of anterior niche cells remained unaffected in tj mutant ovaries , suggesting that Tj regulates the fate of niche cells but not their numbers . Cap cells adopted the shape and morphogenetic behavior of TF cells in tj mutants . To determine whether a reduction of Tj causes indeed a change in cell fate , we used several markers that differ in their expression in the two cell types ( S1 Table ) . In tjhypo ovaries , cells in the upper portion of the terminal stalks expressed low levels of Bab2 and background levels of 1444-lacZ similar to wild-type TF cells ( Fig 3A and 3B ) . In contrast , the lower portion of the tjhypo mutant stalks expressed high levels of Bab2 and 1444-lacZ ( Fig 3B ) , which is typical of wild-type cap cells ( Fig 3A ) . Furthermore , the markers LamC and B1-lacZ , which stained TF cells more intensely than cap cells in wild type ( S2A Fig ) , showed a stronger signal in the upper than in the lower portion of tjhypo mutant stalks ( S2B Fig ) . This indicates that the additional stalk cells retain the expression profile of cap cells despite the dramatic change in morphology . However , LB27-lacZ , a TF-specific marker that is expressed in a complementary pattern to Tj in wild type ( S2E Fig ) , was sometimes seen at reduced levels in the stalk-forming Tj-positive cap cells in tjhypo ovarioles , ( S2F Fig ) , pointing toward a potential defect in cell specification . To further investigate the function of Tj in cap cell specification , we evaluated the expression of markers in tjnull ovaries , using different allelic combinations , including tjeo2 , tjz4735 , and a newly generated transcriptional null mutation , tjDf1 ( see Materials and methods , S1A Fig ) . The absence of the cap cell-specific marker 1444-lacZ ( Fig 3C and 3D ) , the weak signal of Bab2 ( Fig 3C–3F ) , the strong signals of LamC ( Fig 3E–3G and S2C and S2D Fig ) and B1-lacZ ( S2C and S2D Fig ) , and in particular the presence of the TF-specific marker LB27-lacZ ( Fig 3E–3G and S2E and S2G Fig ) throughout the elongated stalk of tjnull mutants are all indicative of a shift in cell fate . This expression profile is consistent with the TF cell-like disc-shaped morphology and stalk-forming behavior , and we therefore conclude that additional TF cells form at the expense of cap cells in the absence of Tj function . To test whether the effect of Tj depletion on the fate of cap cells is cell-autonomous , we induced tjnull mutant cell clones in the GSC niche during the larval stage . We focused on germaria that contained mutant anterior niche cells ( cap and/or TF cells ) but did not contain mutant escort cells close to cap cells to separate the tj loss-of-function effect on cap cells from that on escort cells ( S2H and S2I Fig ) . Use of the tjz4735 allele allowed us to distinguish between regular TF cells and transformed cap cells , as the latter expressed Tj . In cases of mosaic cap cell groups , tj homozygous mutant cells usually looked like TF cells and had become part of the TF , whereas the non-mutant cap cells were rounded and clustered posterior to the mutant cells in the germarium ( Fig 3H and 3I and S2H and S2I Fig ) . The abnormal behavior of tj mutant cap cells was independent of whether the neighboring bona fide TF cell was a tj mutant cell ( GFP negative , Fig 3H ) or a control cell ( GFP positive; Fig 3I ) . Our clonal analysis shows that Tj is cell-autonomously required for cap cell morphology and behavior . To determine whether presumptive cap cells are abnormally specified at the time of their origin or are not able to maintain the cap cell fate when Tj is depleted , we looked at developing ovaries at the stage of cap cell formation . Cap cells develop gradually during the late 3rd instar larval and early prepupal stage following the formation of TFs [17] . Already at the prepupal stage , the TFs were longer in tjnull ovaries than in wild-type ovaries , consisting of an increased number of cells that were aligned in a single file ( Fig 3J and 3K ) . The expression of niche markers at the prepupal stage is different from the adult stage ( S2 Table ) . All cells of the mutant stalks showed prominent LamC staining in contrast to control prepupal ovaries , where this marker was not detected in cap cells and only weakly expressed in basal TF cells ( Fig 3J–3M; S2 Table ) . Moreover , 1444-lacZ , which was co-expressed with Tj in cap cells of prepupal wild-type ovaries ( although not yet seen in escort cells ) , was not detected in the Tj-positive stalk cells of tjnull ovaries ( Fig 3L and 3M ) . This shows that the defects in cap cell specification already develop at the time of niche formation . Taken together , our data indicate that all the anterior niche cells adopt a TF cell fate in the absence of Tj function , implicating Tj as a crucial factor for cap cell specification . As Tj is required for cap cell specification , we asked whether expression of Tj in anterior niche cells would be sufficient to induce the cap cell fate . We induced Tj-expressing cells in TFs in early 3rd instar , before TFs begin to form by cell intercalation [15 , 16] , and analyzed mosaic TFs in adult ovaries . The frequency of mosaic TFs was similar in ovaries with clonal expression of either Tj or GFP , suggesting that Tj expression did not affect the survival of TF cells ( S3 Table ) . Strikingly , Tj-positive TF cells expressed high levels of 1444-lacZ and low levels of LamC similar to wild-type cap cells ( Fig 4A and 4B ) . In addition , the expression of LB27-lacZ was strongly reduced in Tj-positive TF cells compared with control TF cells , although not completely abolished ( Fig 4C–4E ) . Thus , expression of Tj in TFs resulted in ectopic expression of a cap cell marker and partial suppression of a TF cell marker . Despite the changes in marker expression , Tj-expressing TF cells remained in the TF ( Fig 4A–4C ) , and even formed a stack of aligned cells when all TF cells expressed Tj ( Fig 4D and 4E ) . However , Tj-expressing TF cells appeared rounder than their control neighbours ( Fig 4B and 4C ) . Cell shape analysis confirmed that Tj-expressing TF cells have a significantly increased height and decreased width compared to control TF cells ( Table 1 ) , which is consistent with a rounder , more cap cell-like morphology . Our data indicate that Tj induces TF cells to adopt molecular and morphological characteristics of cap cells . The spread-out cluster of cap cells in a wild-type germarium provides a large contact surface for anchorage of GSCs [25] . In tjhypo ovarioles , however , most cap cells are recruited into the TF , and few remain in the germarium , potentially limiting their availability to GSCs . In cases where all cap cells form a single file stalk , only the basal-most cap cell would offer a physical GSC anchor point . Therefore , we asked whether the abnormal organization of cap cells in tjhypo ovarioles might affect the number or maintenance of GSCs . We noticed that the germaria of tjhypo ovaries were often unusually narrow ( Fig 5A and 5B ) , harbouring only 1–2 GSCs ( Fig 5B , 5C' and 5D ) in contrast to 2–3 GSCs in wild-type germaria ( Fig 5A and 5D ) [6 , 50] . As Tj was found to be neither expressed nor required in the germline [36] , the reduction in GSC numbers is likely caused by the observed defects in the stem cell niche of tj mutants . Only a single GSC was present when all cap cells had joined the TF in tjhypo ovarioles ( Fig 5B ) . The number of GSCs per mutant ovariole increased with the number of cap cells that remained in the germarium ( Fig 5C , 5C' and 5E ) . The majority of ovarioles with 2–3 cap cells within the germarium had still only one GSC , whereas those with four and six cap cells in the germarium had usually two and three GSCs , respectively ( Fig 5E ) . Therefore , the number of GSCs in tjhypo mutant ovarioles correlates with the number of cap cells that remain in the germarium instead of the total number of cap cells . These data are consistent with an approximate 2:1 ratio of cap cells to GSCs that has been observed for wild-type ovarioles [6 , 8] , the finding that GSCs require direct contact to cap cells [7 , 8 , 22 , 51] , and the observation that GSCs partially envelop more than one cap cell with cytoplasmic extensions in wild type ( Fig 5F ) . If , however , a contact to more than one cap cell is required to support a GSC , how could any GSC exist if all cap cells are arranged in a stalk . Interestingly , we discovered that in 79% of those cases ( n = 14 ) , the GSC produced a long cellular protrusion that reached far into the TF of a tjhypo ovariole , allowing it to contact at least two cap cells ( Fig 5B and 5G ) . In comparison , when two or more than two cap cells remained in the germarium , the frequency of unusually long GSC protrusions was only 42% ( n = 12 ) and 0% ( n = 11 ) , respectively . To determine whether the cells that we identified as GSCs based on morphological criteria are bona-fide GSCs , we analyzed the activity of the Dpp signaling pathway by probing for the presence of phosphorylated Mothers against dpp ( pMad ) , the effector of this pathway [52] . In wild-type germaria , nuclear pMad identifies GSCs ( Fig 5H ) [11] . Similarly , in tjhypo germaria , nuclear pMad was restricted to germline cells that abutted cap cells and had the morphology of GSCs ( Fig 5I ) . In most tjhypo ovarioles , only a single germline cell was positive for pMad , consistent with a reduced number of GSCs . The staining intensity of nuclear pMad was comparable between GSCs of wild-type ( n = 23 ) and mutant germaria ( n = 25 ) . Consistent with this finding , bam expression , which is repressed by pMad to prevent differentiation of GSCs [10 , 11] , was not detected in GSCs but was seen in differentiating germline cysts in both tjhypo and wild-type germaria ( Fig 5J and 5K ) . Together , this suggests that Dpp signaling from niche cells is active and confirms the presence of GSCs in the tjhypo mutant . An aging experiment that assessed whether reduced expression of Tj affects the maintenance of GSCs showed that the number of GSCs remained stable over a period of three weeks in wild-type and tjhypo mutant ovarioles ( Fig 5D ) . Presence of germaria and rows of follicles of successive developmental stages in 2–22 day-old tjhypo ovaries similar to wild-type ovarioles confirms the proper maintenance of GSCs ( S3A–S3D Fig ) . Absence of bam-GFP expression ( S3E and S3F Fig ) and presence of nuclear pMad in the anterior-most germline cells that contact cap cells ( S3G and S3H Fig ) are consistent with the conclusion that GSCs , although smaller in number , are maintained normally in tjhypo mutant ovaries . As cap cells are considered essential for GSC establishment and maintenance [3 , 4] , and our analysis indicates that cap cell specification depends on Tj , we expected a loss of GSCs in tjnull mutant ovaries . In ovaries of wild-type prepupae , the anterior-most row of germline cells next to the newly formed niches represented GSCs as indicated by the presence of nuclear pMad ( Fig 6A ) [11 , 17] . In tjnull prepupal ovaries ( n = 15 ) , 68% of the TFs were not associated with any germline cells ( orphan TFs ) . The remaining TFs captured usually no more than one germline cell . Surprisingly , we found that some of the TF-associated germline cells displayed nuclear pMad ( Fig 6B ) , although their number was very low , with a mean of 2 . 3 nuclear pMad-positive cells in a tjnull ovary ( n = 17 ) compared with 13 . 6 in a wild-type prepupal ovary ( n = 11 ) . As orphan TFs in tj mutant ovaries might potentially result from the abnormal distribution of germ cells and interstitial cells [36] , we asked whether the number of pMad-positive cells per occupied TFs was different from wild type . Taking into account that even in wild-type prepupal ovaries only a third of niche-associated germline cells were positive for nuclear pMad , that only 32% of the mutant TFs were occupied by a germline cell , and that the mean number of TFs was reduced by 20% in tjnull ovaries ( 15 and 19 TFs in tjnull ( n = 18 ) and wild-type ovaries ( n = 17 ) , respectively ) , we calculated a 33% reduction of nuclear pMad-positive germline cells per occupied TF in mutant ovaries ( 0 . 48 and 0 . 72 nuclear pMad-positive cells per occupied TF in tjnull and wild-type ovaries , respectively ) . Notably , bam-GFP , which was absent in GSCs of wild-type ovaries was detected in some of the TF-associated germline cells in tjnull ovaries ( Fig 6C and 6D' ) , suggesting entrance into differentiation [11 , 17] . To follow the fate of germline cells in tjnull mutant ovaries , we analyzed ovaries at the pupal and adult stage . We had previously reported that more than 25% of tjnull ovaries from young adult females were devoid of germline cells [36] . Already at the mid pupal stage , when the germaria had matured in wild-type ovaries , and displayed a largely expanded germline cell population ( Fig 6E ) , tjnull ovaries contained only a few small and scattered germline cells or cell clusters and pMad staining was drastically reduced ( Fig 6F ) , suggesting a rapid loss of germline cells during the pupal period . The bam-GFP signal that revealed early germline cysts in control ovarioles ( Fig 6G and 6G' ) ranged from non-detectable to prominent in the remaining germline clusters in tjnull ovaries ( Fig 6H and 6H' ) . In adult ovaries , where 96% of the few remaining germline cell clusters ( n = 45; 10 ovaries ) were associated with TFs ( Fig 6I and 6J ) , pMad-positive cells were rare , and only present in 9 . 5% of the clusters ( n = 21; 18 ovaries ) . Some clusters consisted of individual germline cells , as indicated by spectrosomes ( Fig 6J and 6J' ) , others had undergone transit amplification with incomplete cytokinesis , displaying branched fusomes [36] . Taken together , we infer that the complete loss of Tj activity severely compromises GSC establishment and maintenance . To determine whether the reduction/loss of germline cells in tj mutant ovaries is responsible for the recruitment of cap cells into the TF we analyzed the behavior of cap cells in tudor and oskar maternal effect mutants ( tudmat and oskmat , respectively ) that lack germline cells [53 , 54] . The number of cap cells was reduced in tudmat ovaries ( average of 3 . 7 cap cells; n = 49 ) compared with wild type ( average of 7 . 7 cap cells; n = 11; p<0 . 0001 ) . This indicates that cap cells can form in the absence of a germline , although their number is reduced , which is consistent with previous reports [8 , 51 , 55] . Importantly , however , the number of TF cells was not increased but rather slightly reduced in tudmat ovaries ( average of 7 . 1 TF cells; n = 44 ) compared with wild type ( average of 8 TF cells; n = 11; p<0 . 01 ) , indicating that the reduced number of cap cells is not caused by a change from cap cell to TF cell fate . Furthermore , the remaining cap cells had a rounded shape , were organized into a cluster that resided within the germarium , and expressed Tj similar to wild-type cap cells ( Fig 7A and 7B ) . Similarly , cap cells were organized as a cluster in oskmat mutant ovaries . These findings indicate that the number but not the morphology or spatial arrangement of cap cells depends on the presence of GSCs . Mutants with reduced N activity displayed a decrease in the number of cap cells [8 , 31 , 32] . Similarly , knocking down N by tj-Gal4-driven expression of UAS-NRNAi , which caused a typical N loss-of-function phenotype with fused follicles ( Fig 8A ) [56] , reduced the average number of cap cells per germarium from a normal complement of 6 down to 2 ( Fig 8B and 8J ) . Accordingly , the average number of GSCs dropped from 3 in control germaria to 0 . 8 in NRNAi germaria ( Fig 8J ) . When all cap cells were missing , GSCs were absent , and escort cells were misplaced to the tip of the germarium , where they made contact with the TF and a differentiating germline cyst ( Fig 8C ) . Although the number of cap cells was reduced , the number of TF cells was normal in N depleted ovarioles ( Fig 8J ) , indicating that the loss of cap cells is not due to a cap cell to TF cell fate change . This shows that the tj and N loss-of function phenotypes are different . As both Tj and N are required for cap cell formation , we asked if and how their functions might be related . First , we investigated whether their expression is dependent on each other . To determine whether the expression of Tj depends on N signaling we checked Tj expression in NRNAi ovaries . However , we could not separate a direct effect on Tj from an effect on cap cells . Remaining cap cells in NRNAi ovaries always expressed Tj , and at an apparently normal level ( Fig 8B ) . That existing cap cells remained in the germarium and were not recruited into TFs also suggests that the expression level of Tj is not affected . Although this argues against Tj being a downstream target of the N pathway , it cannot be excluded that remaining N activity in NRNAi ovaries can enable the formation of a few cap cells with full expression of Tj . To test whether Tj influences the expression of N signaling components , we evaluated the expression pattern of N , its ligand Dl , and its target and effector Enhancer of split ( E ( spl ) ) in tjnull ovaries at the prepupal stage when anterior niches have formed ( S4 Fig ) . Dl staining was much stronger in TFs than in cap cells of control ovaries ( S4A Fig ) [8] . Dl staining in the upper half of TFs in tjnull ovaries was as robust as in controls . Interestingly , however , Dl staining in the lower portion of the stalk , which is composed of transformed cap cells in tjnull ovaries , was as weak as in cap cells of controls ( S4B Fig ) . Thus , in contrast to all other tested markers , Dl expression appears not to have changed in the transformed cap cells , indicating that Dl expression in niche cells is not regulated by Tj . Expression of N protein and an E ( spl ) expression reporter ( E ( spl ) mß-CD2 ) , which can be used to detect activity of the N pathway in niche cells [8] , appeared to be rather homogeneous throughout the anterior niche cells of tj mutant ovaries similar to wild-type ovaries ( S4C and S4D Fig ) . Thus , Tj appears not to affect the activity of the N pathway in cap cells . Taken together , our expression analysis is consistent with Tj acting downstream or in parallel to the N pathway in the formation of cap cells . To further analyze the relationship between Tj and N , we looked for genetic interactions by changing their expression level either in the same or opposite direction . Overexpressing Tj in its endogenous pattern , including cap and escort cells , by driving expression of UAS-tj with tj-Gal4 did not cause any obvious defects in the stem cell niche ( Fig 8K ) , although it led to defects at later stages of oogenesis . The NRNAi phenotype prevailed in the presence of increased Tj expression ( Fig 8D–8F and 8J ) , suggesting that Tj cannot rescue cap cells in the absence of N . To achieve a double knockdown of Tj and N , we used a UAS-tjRNAi transgene that strongly reduces Tj expression [57] . When driven with tj-Gal4 , the tjRNAi knockdown was variable but consistently strong and frequently caused a phenotype that was indistinguishable from a tjnull ovary phenotype ( Fig 8N ) . tj N double-knockdown ovaries largely resembled tjnull ovaries , with TFs and ovariole sheaths remaining and all other cell types drastically reduced or missing ( Fig 8G ) . Not surprisingly , cap cells were absent ( Fig 8J ) . Interestingly , however , tjRNAi NRNAi ovaries did not have the extended TFs that are the hallmark of tj mutant ovaries ( Fig 8G ) . With an average of 9 . 1 cells , TFs of tjRNAi NRNAi ovaries were considerably shorter than those of tjRNAi ovaries , which had an average of 16 cells similar to tj mutant ovaries ( compare Fig 8G with 8N and 8Q ) . Although the number of TF cells per stack was highly variable in tjRNAi NRNAi ovaries ( Fig 8J ) , 42% of the TFs had more than 8 cells per stack and were therefore longer than those of control ovaries ( average of 7 . 8 cells ) , resembling more the combined number of TF and cap cells in the niches of NRNAi ovaries ( average of 9 . 9 cells ) ( Fig 8J ) . All cells in these elongated stalks of tjRNAi NRNAi ovaries had a TF cell identity based on morphology and marker expression ( Fig 8H and 8I‴ ) . We propose that cap cells that remained after strong reduction of N expression acquired the TF cell fate due to the loss of Tj . Further analysis of ovaries with tj N double-knockdown revealed that the total number of TF stalks was strongly reduced , with an average of 6 . 2 ( n = 12 ) , compared to 16 . 1 in NRNAi ovaries ( n = 14 ) and 19 . 3 in tjRNAi ovaries ( n = 9 ) , and that most ovaries contained several unusually short TFs with less than 7 TF cells . This suggests , that loss of Tj and/or N does not only affect cap cell formation but that their combined loss affects TF cell formation as well . Next , we investigated the effects of increased or decreased Tj expression on the phenotype caused by Nintra , which constitutively activates N signaling . Overexpression of Tj did not appear to affect the number of cap cells ( Fig 8K ) , whereas driving UAS-Nintra with tj-Gal4 led to a large increase of cap cells ( Fig 8L ) , similar to what had been reported previously for Nintra expression with another somatic driver [8] . Co-expression of Nintra and transgenic Tj resulted in the same phenotype ( Fig 8M ) . For both genotypes , we observed two different types of germaria . Some germaria contained a cluster of undifferentiated germ cells that resembled GSCs ( Fig 8O ) , consistent with a previous report [8] , whereas other germaria were devoid of germ cells despite a large aggregate of cap cells ( Fig 8L and 8M ) . We were particularly interested in the phenotype of Nintra tjRNAi ovaries . If N determines the number of cap cells while Tj controls their identity , one might expect that all Nintra-induced additional cap cells become TF cells in the absence of Tj , causing even longer TFs than when Tj alone is lost . The phenotype of Nintra tjRNAi ovaries was variable , consistent with the variability in the individual tjRNAi and Nintra phenotypes but the defects were considerably more severe ( Fig 8N–8P ) . Most Nintra tjRNAi ovaries were extremely small , not connected to an oviduct and often attached to the gut , and seemed to consist largely of TFs and a few germline cells embedded in fatbody ( Fig 8P ) . In contrast to tjRNAi or Nintra ovaries , the epithelial sheaths were missing and TFs were located side-by-side ( Fig 8N–8P ) . In the most severe cases , Nintra tjRNAi ovaries consisted only of TFs ( Fig 8P ) . To account for the two UAS-constructs in Nintra tjRNAi ovaries , we co-expressed a second UAS-construct together with tjRNAi or Nintra in our controls , which expressed GFP ( Fig 8Q and 8R ) . This did not appear to influence the mutant phenotype but showed that tj-Gal4 is active in the lower half of the extended TFs in tjRNAi ovaries ( Fig 8Q ) , and is excluded from the TF but present in cap cells of Nintra ovarioles ( Fig 8R ) , as expected . Notably , TFs in Nintra tjRNAi ovaries were always considerably longer than in Nintra ovaries ( Fig 8O , 8P and 8R–8T ) , although they were on average slightly shorter than in tjRNAi ovaries ( Fig 8N , 8P and 8T ) . Fig 8S shows a rare example of a particularly long TF in a Nintra tjRNAi ovary . Taken together , these findings suggest that the tjRNAi niche phenotype is epistatic to the Nintra niche phenotype . The effects on cap cells in response to alterations in Tj and N expression are summarized in Fig 8U .
Loss of Tj has a profound negative effect on the establishment , number , and maintenance of GSCs . Effects of Tj on the germline were previously shown to be indirect as Tj is neither expressed nor cell-autonomously required in the germline [36] . Therefore , we propose that the dramatic change in the structure of the somatic niche affects GSCs when Tj function is compromised . An inverse causal relationship , where a reduced number of GSCs would trigger the somatic niche defects was ruled out by showing that cap cells can still look and behave normally in the absence of any germ cells . We conclude that Tj controls GSCs indirectly by controlling somatic cell fate and cell arrangement in the stem cell niche . By controlling the morphology and behavior of the cap cells , Tj regulates the GSC-carrying capacity of the niche . When Tj expression was moderately reduced , the number of GSCs per niche was reduced , with the remaining GSC properly maintained over several weeks . The decrease of GSCs per niche correlated with a decrease of cap cells in the germarium . Two cap cells were on average required to sustain one GSC , similar to what has been proposed for a wild-type ovary [6 , 8] . Our data indicate that the reduced niche capacity is due to a reduction in the available contact surface between cap cells and GSCs . Tj-depleted cap cells that convert from forming a cluster inside the germarium to forming a stalk outside the germarium minimize their availability for GSC attachment . A connection between the GSC-cap cell contact area and niche capacity is similarly reflected in the increased number of GSCs that accompanies an increase in cap cell size due to loss of RanBPM [25] . Here , we show that the spatial arrangement of the cap cells has a crucial impact on the number of stem cells per niche . When Tj function was completely abolished , the number of GSCs was drastically reduced , as expected in the absence of cap cells . The very few pMad-positive GSC-like cells in tj mutant prepupal ovaries were always associated with a TF , suggesting that TFs might temporarily provide enough Dpp to activate Mad in a few germline cells , consistent with the finding that Dpp is expressed in TFs at the late larval stage [17 , 58] . This is not sufficient , however , to maintain GSCs and adult ovaries rarely contain pMad-positive germline cells . This is in agreement with the finding that Dpp is not detected in adult TFs [6] , and corroborates that cap cells are required for GSC maintenance . In addition , the rapid loss of the entire germ cell pool in Tj-depleted ovaries during the pupal stage might be precipitated by loss or defects in escort cells . Escort cell precursors are not properly intermingled with germ cells at the larval stage and differentiated escort cells appear to be missing in adult ovaries that lack Tj [36] . As escort cells are crucial for germ cell differentiation [59–62] , the defect in escort cell differentiation could be responsible for the demise of the germline in tj mutants . We discovered that GSCs have broad cellular protrusions , which they use to reach and tightly ensheath the accessible surface of cap cells . In wild type , relatively short protrusions are sufficient to make extensive contact with more than one cap cell . However , when cap cells formed a stalk , GSCs were often observed to produce unusually long extensions that allowed them not only to contact the immediate cap cell neighbor but also a more distantly located cap cell . This suggests that GSCs respond to a chemotactic signal from cap cells and send protrusions toward this signal . It remains to be investigated whether this is a response to Dpp signaling or signaling through another pathway . The importance of cellular protrusions in signaling events in the stem cell niche has recently come to light with the discovery of nanotubes that mediate Dpp signaling between GSCs and hub cells in the Drosophila testes [63] , and cytonemes that contribute to Hh signaling from cap to escort cells in the ovary [12] . Our analysis shows that Tj is required for the specification of cap cells . In the absence of Tj function , additional TF cells form at the expense of cap cells , resulting in unusually long TFs while the cap cell fate is not established . Whereas the formation of cap cell precursors appears not to require Tj , this transcription factor is essential for the ability of these precursors to take on the cap cell fate and to prevent the TF cell fate that is otherwise adopted as a default state . The following findings support this conclusion: ( i ) In the absence of Tj function , cap cells were missing while additional cells that displayed TF cell-characteristic morphology , behavior and marker expression were integrated into the TF . The number of additional TF cells was comparable to the normal number of cap cells . ( ii ) Prospective cap cells cell-autonomously adopted a TF-specific morphology and behavior in the absence of functional Tj . ( iii ) A hypomorphic tj mutant provided direct evidence for the incorporation of cap cells into TFs , forming the basal portion of these stalks . ( iv ) Ectopic expression of Tj in TF cells caused a change toward cap cell-typical marker expression and morphology . Together , these data demonstrate that Tj promotes cap cell specification . The expression pattern of Tj supports the notion that Tj has a function in cap cells but not in TF cells . Tj is continuously expressed in cap cells [36; this study] . Tj is also present in the anterior interstitial cells of the larval ovary [36 , 44] , which are thought to develop into cap cells [14] . In contrast , Tj is neither detected in the cell population that gives rise to TFs during 3rd larval instar , nor in differentiated TFs [28 , 36] . Interestingly , we found that even in the absence of Tj function , the tj gene remains differentially expressed in the anterior niche , being inactive in regular TF cells but active in the additional TF cells , which form the apical and basal portion of a TF , respectively . This differential expression of Tj indicates that a regionally or temporally regulated mechanism operates upstream of Tj that initiates differences in anterior niche cells . Although it is conspicuous that Tj expression from 3rd instar onwards is restricted to cells that are in direct contact with germline cells , which includes cap cells but excludes TF cells , it has previously been shown that Tj expression is not dependent on the germline [36 , 43] . This suggests that a soma-specific mechanism is responsible for the differential expression of Tj in anterior niche cells . Interestingly , a recent study uncovered the importance of Hh signaling from TFs to neighboring interstitial cells in the larval ovary and proposes that tj is a direct target of the Hh signaling pathway [30] . Our findings suggest the presence of a new cell type in the GSC niche that we named 'transition cell' as it is located between the cap cell cluster and the TF , connecting these two structures of the niche . Notably , the one or occasionally two transition cells have the morphology of TF cells and align with neighboring TF cells despite displaying a cap cell-like marker profile that includes the expression of Tj—although Tj expression is substantially lower than in cap cells . Interestingly , cap cells from ovaries with reduced Tj expression ( tjhypo ) similarly displayed a TF cell-like morphology and behavior while their expression profile remained cap cell-like . A similar , although weaker effect was noted in a tj hemizygous condition , suggesting that Tj function is haplo-insufficient in cap cells . Thus , when Tj levels are reduced , cap cells adopt very similar molecular and morphogenetic properties as the transition cell in a wild-type niche , and might have adopted this cell fate . Together , our findings indicate that Tj has an important role in the establishment of three cell types in the GSC niche: TF cells , transition cells , and cap cells . As lack of Tj function seems to cause a transformation of cap and transition cells into TF cells , and a mild reduction of Tj a cap to transition cell transformation , we propose that different Tj expression levels establish different cell fates and morphogenetic traits . We propose that a high concentration of Tj leads to the formation of cap cells and a lower concentration to the formation of the transition cell , whereas absence of Tj is required for the formation of TF cells ( Fig 9A ) . This model implies that different levels of Tj have different effects on target genes . We predict that Tj has at least one target gene that only responds to high levels of Tj and that specifically controls the morphogenetic behavior of cap cells , allowing them to adopt a round morphology and organize into a cell cluster . Whether this relates to an effect of Tj on the expression of adhesion molecules as observed in other gonadal tissues [36 , 42 , 57 , 64 , 30] awaits further analysis . Our study identifies Tj as essential for cap cell formation . In addition , this process depends on the N pathway [8 , 31 , 32] . Therefore , we wondered how the functions of Tj and N in cap cell formation relate to each other ( Fig 9 ) . A comparison between the loss and gain-of-function phenotypes suggests that Tj and N have different functions in the establishment of cap cells . In the absence of Tj function , cap cell precursor cells are present but take on the fate of TF cells , whereas depletion of N leads to a loss of cap cells but does not cause the formation of additional TF cells . Ectopic activation of N can induce a strong increase in the number of cap cells , whereas overexpression of Tj did not appear to affect the number of cap cells . Therefore , both factors are important for cap cell formation but contribute differently to this process . The questions then are: What is the respective contribution of Tj and N to cap cell formation , and how are their functions related ? The function of N in cap cell formation is still not fully understood . Our observation that depletion of N reduces the number of cap cells confirms previous findings [8 , 32 , 65] . However , neither in our nor any previously published experiments were cap cells lost completely when the N pathway was compromised , and it remains therefore unclear whether N is de facto essential for cap cell formation or primarily functions in regulating the size of the cap cell pool . Interestingly , evidence amounts to a function of the N pathway in a decision between the cap cell and escort cell fate: First , Dl signal from TF cells activates the N pathway in adjacent interstitial cells , inducing them as cap cells , whereas the remaining interstitial cells are thought to develop into escort cells [8 , 32] . Second , escort cells expressing activated N can develop into cap cells [31 , 65] . Third , when we used tj-Gal4 to express active N in interstitial cells , the number of cap cells dramatically increased while the escort cell region became smaller , and some germaria seemed to lack escort cells all together . These germaria also lacked germline cells , although a larger pool of cap cells was expected to increase the number of GSCs [8 , 31] . However , the absence of germline cells is consistent with an absence of escort cells , as escort cells have been shown to be important for maintaining the germline [60] . Together , these observations support the hypothesis that N is involved in a cap cell versus escort cell fate decision , and suggest that the N pathway might promote the formation of cap cells by inhibiting the escort cell fate . To determine how the functions of Tj and N depend on each other we looked for genetic interactions . The N pathway seems to be still functional in tj mutants . First , the expression of N and Dl appeared unaffected and E ( spl ) was activated in the additional TF cells ( = transformed cap cells ) similarly to normal cap cells . Second , the formation of additional TF cells in the absence of Tj depended on the presence of N , as only very few additional TF cells formed in a N compromised background . These findings indicate that the N pathway is still active in cap cell precursors when Tj function is abolished . This together with the observation that constitutively active N cannot suppress the tj mutant phenotype suggests that Tj does not act upstream of N in regulating cap cell fate . Therefore , we asked whether Tj might operate downstream of N . We did not detect a loss of Tj upon N depletion , and this together with the finding that Tj is expressed in all interstitial cells , and not only in those that receive Dl signaling argues against a requirement of N signaling for tj expression . If at all , one would expect tj to be negatively regulated by N as cap cells express a lower level of Tj than escort cells . The maintenance of somatic cell types in N mutant ovaries that are lost in tj mutant ovaries , including the escort cells is also not consistent with a linear relationship . Nevertheless , the ability of Tj to promote the formation of cap cells appears to depend on the activity of the N pathway in cap cell precursors . Again , this is suggested by the finding that when N and Tj were both compromised , the number of additional TF cells were much smaller than when N was fully active . Therefore , we propose that N activity sets aside a pool of percursor cells that in the presence of Tj take on the cap cell fate , and in its absence the TF fate ( Fig 9B ) . Similar to the ovary , N is important for the formation of the GSC niche ( = hub ) in the Drosophila testis [66 , 67] . Interestingly , N contributes to hub cell specification by downregulating the expression level of Tj [42] . Not only is the hub still present in tj mutant testes [36] but additionally , ectopic hub cells form in the absence of Tj [42] . Thus , Tj seems to have opposing functions in testes and ovaries , suppressing the niche cell fate in the testis [42] , while promoting it in the ovary . The interplay between Tj and N seems not restricted to the cap cell fate in the ovary . Whereas neither factor alone is required for TF cell formation , as TF cells formed normally in the absence of either Tj or N , the combined loss of Tj and N led to a strong reduction in the number of TFs and number of TF cells within stalks . This suggests that their combined action is already required at an earlier stage of ovary development , when Tj is still expressed in all somatic cells of the ovary [44] . Moreover , Tj knockdown combined with expression of activated N caused TF cells to be the only cell type remaining of the ovary , indicating that several cell types in the ovary require proper input from both factors . Taken together , our findings support a model , in which both Tj and N operate together to promote the cap cell fate but have separate functions . We propose that Tj and N promote the cap cell fate by blocking the TF cell fate and escort cell fate , respectively , and that the combined actions of Tj and the N pathway are required to establish the cap cell fate ( Fig 9C ) .
tjeo2 ( amorphic ) [36 , 40] , tjz4735 ( amorphic ) [36 , 41] , tj39 ( hypomorphic , see below ) , tj-Gal4 ( hypomorphic , see below ) , tjDf1 ( molecular null , see below ) , and UAS-tjRNAi [57] ( NIG-Fly Stock Center] were used for tj loss-of-function analysis . UAS-tj1 ( 2 ) and UAS-tj6 ( 3 ) ( UAS-tj; full-length tj coding sequence and 3'UTR ) [57] were used for ectopic and overexpression of Tj . UAS-N754 . BF ( UAS-Nintra on 3rd ) [68] and P{TRIP . JF02959}attP2 ( UAS-NRNAi ) [69] ( Bloomington Drosophila Stock Center ( BDSC ) ] were used for N loss- and gain-of-function experiments . tud1 and tudb45 [70] ( gift from M . Van Doren ) , and osk301 [54] were used to generate flies without a germline . We used tj-Gal4 ( P{GawB}NP1624 ) [71 , 72] ( Kyoto Stock Center ) to drive expression in cap and escort cells and their larval progenitors , and the FLPout cassette [73] ( BDSC ) for clonal expression in TF cells . tjz4735 mutant cell clones were induced by mitotic recombination using hs-FLP1 and FRT40A ( BDSC ) . UAS-GFP . S65T and Ubi-GFP ( BDSC ) were used as clonal cell markers . nos-Gal4 ( BDSC ) was used to drive UAS-GAP43-mEos ( BDSC ) in germline cells . The enhancer reporter lines babA128 ( bab-lacZ ) [15 , 74] , P{PZ}1444 ( 1444-lacZ ) [55] , P{A92}LB27 ( LB27-lacZ ) [16] , P{lacW}B1-93F ( B1-lacZ ) [15 , 56] , and bamP702-GFP ( bam-GFP ) [75] were used as cell type specific markers . E ( spl ) mß-CD2 was used to monitor activity of the N pathway [8 , 76] ( gift from D . Drummond-Barbosa ) . We generated the following recombinant chromosomes: tjeo2 1444-lacZ , UAS-tj1 ( 2 ) 1444-lacZ , and UAS-tj1 ( 2 ) UAS-GFP for functional analysis of tj , and UAS-tj6 ( 3 ) UAS-NRNAi , UAS-tj6 ( 3 ) UAS-Nintra , UAS-tjRNAi UAS-NRNAi , UAS-tjRNAi UAS-Nintra , and tj-Gal4 UAS-GFPnls to test genetic interactions between tj and N . Oregon R , w , or y w were used as a genetic background . The copy number of all genetic markers , such as enhancer reporters was identical between control and experimental animals . tjDf1 , a transcriptional null mutation is a genomic deletion of 13 . 3 kb ( 2L:19464294–19477599 ) , beginning 286 bp upstream of the tj start codon and ending 9 . 8 kb downstream of the tj transcription unit ( S1A Fig ) . This mutation deletes the complete coding and 3' UTR sequences of tj , and three predicted RNA coding genes . tjDf1 was generated by FLP-mediated recombination between the FRT elements of the transposable elements P{XP}d06467 and PBac{WH}f02713 [77 , 78] ( Exelixis Collection at the Harvard Medical School ) , using the technique described by Parks et al . [79] . We screened for recombinant flies by eye color as recombinant chromosomes containing a deletion were expected to carry two mini-white genes . Recombination was confirmed by PCR analysis , using genomic DNA from homozygous tjDf1 flies . Primer pair AGCGAATGGTGGCGTTCGAGCTC—ACCACCTTATGTTATTTCATCAT confirmed the presence of the 3' end of P{XP}d06467 , primer pair CCTCGATATACAGACCGATAA—AGCCAAATGAACTGCCCGCT the presence of the 3' end of PBac{WH}f02713 , and primer pair GACCTTTGAAACCACCCACTAAC—GTGGTGTGCGTAAGTCTGAGC the absence of tj-specific sequences . tjDf1 is homozygous viable , but both female and male sterile . tj39 , a weak hypomorphic allele was generated in a P element excision mutagenesis , using tj-Gal4 ( P{GawB}NP1624 ) , which is located in the 5' UTR of tj , 0 . 7 kb upstream of the translation start site [72] , as a starter line . tj39 caused strongly reduced fertility in trans to tjeo2 ( approximately 20% of the fertility of the tjeo2/+ control ) , whereas 29 other excision mutations were fully fertile in trans to tjeo2 . PCR analysis , using genomic DNA of homozygous mutant tj39 flies and four primer ( P ) pairs ( P2: GCTCTTGCACAGTGGTCGAG—P1: ACCACCTTATGTTATTTCATCAT , P1: ACCACCTTATGTTATTTCATCAT—P3: GTGTCGTTTATGGTGGGATC , and P2: GCTCTTGCACAGTGGTCGAG—P4: GAACTCCTGTTGGAAACGTG showed that the genomic sequences flanking the insertion site are still present and revealed a partially excised P element ( the 3' end is still present ) . Sequencing the PCR-amplified tj coding region , using primers described in Li et al . [36] , confirmed that the tj open reading frame is intact , suggesting that the remaining P element impairs tj expression at the transcriptional or translational level . Subsequent tests revealed that tj-Gal4 itself is a weak hypomorphic allele of tj , causing a similar phenotype in trans to a tj null allele as its derivative tj39 . tj39 tested positively for Gal4 activity . Flies were raised and maintained at 25°C on standard Drosophila medium supplemented with yeast pellets . Ovaries were extracted from 1–4 hour old prepupae , two-day-old pupae , or 1–2 day old yeast-fed adult females , which had been kept in the company of males unless indicated otherwise . Staging , dissection , and processing of prepupal and pupal ovaries were done as described in Godt and Laski [15] . For the aging experiment , female flies were collected and separated from males within 24 hours of eclosure , and were transferred every day to a new food vial ( supplemented with yeast pellets ) until they were dissected 2 , 7 , 14 , and 22 days after eclosure . All experiments were independently repeated at least twice . Clonal analysis: ( 1 ) tj mutant cap cell clones were induced in y w hsFlp1/+; tjz4735 FRT40A/ P{Ubi-GFP . D}33 P{Ubi-GFP . D}38 FRT40A larvae by three 2-hour heat shocks at 37°C during early to mid 3rd instar ( at 72–74 , 82–84 , and 90–92 hours after egg deposition ) . Animals were reared at 25°C to adulthood and ovaries dissected from 2-day old females . ( 2 ) To generate Tj-expressing cell clones in TFs we used the following genotypes: y w hsFlp1/+; UAS-tj1 ( 2 ) /+; Act5C>CD2>Gal4/+ , or y w hsFlp1/+; UAS-tj1 ( 2 ) 1444-lacZ /+; Act5C>CD2>Gal4/+ , or y w hsFlp1/+; UAS-tj1 ( 2 ) /+; Act5C>CD2>Gal4/LB27-lacZ , or y w hsFlp1/+; UAS-tj1 ( 2 ) UAS-GFP/+; Act5C>CD2>Gal4/+ . Flies of the genotype y w hsFlp1/+; UAS-GFP/+; Act5C>CD2>Gal4/+ were used as a control . Early 3rd instar larvae ( 72 +/-1 . 5 hours at 25°C after egg deposition ) were heat shocked at 37°C for 11 minutes , cooled down to 25°C for 10 minutes in a water bath , and reared at 25°C to adulthood . Ectopic expression of Tj caused a relatively high degree of lethality in larvae and pupae , and ovaries were extracted from escaper flies . To measure the height and width of a TF cell , a line through the center of the cell was drawn along the anterior-posterior axis and perpendicular to it , respectively , and ßPS integrin was used to recognize the plasma membrane . The following primary antibodies were used: guinea-pig anti-Tj ( G5 or GP6 , 1:5000 ) [57] , rat anti-Bab2 ( R10 , 1:3000; or R7 , 1:2000 ) [20] , rabbit anti-Vasa ( 1:2000 ) [80] , rabbit anti-Vasa ( d-260 , 1:500; Santa Cruz Biotechnology ) , chicken anti-Vasa ( 1:5000; gift from K . Howard and M . Van Doren ) , rabbit anti-α-spectrin ( #254 , 1:1000; gift from D . Branton ) , mouse anti-LamC ( LC28 . 26 , 1:50 ) , mouse anti-Hts ( 1B1 , 1:5 ) , mouse anti-N ( C17 . 9C6 , 1:5; C458 . 2H , 1:5 ) , mouse anti-Dl ( C594 . 9B , 1:5 ) , mouse anti-Engrailed ( 4D9 , 1:5 ) , and mouse anti-ßPS integrin ( CF . 6G11 , 1:10 ) ( Developmental Studies Hybridoma Bank ) , rabbit anti-pMad ( PS1 , 1:250; gift from T . Tabata ) [81] , rabbit anti-pMad ( pSmad1/5 , 41D10 , 1:100; Cell Signalling ) , rabbit anti-ß-galactosidase ( 1:1500; MP Biomedicals ) , and rabbit anti-GFP ( 1:100; BD Biosciences ) . Secondary antibodies ( 1:400 ) were conjugated either to Cy3 , Cy5 ( Jackson Immuno Research Laboratories ) , Alexa-405 , Alexa-555 , Alexa-488 , or Alexa-647 ( Molecular Probes , Life Technologies ) . Ovaries were mounted in Vectashield ( Vector Laboratories ) . All imaging was done with a 40x/1 . 4 Plan-Apo objective , using confocal laser scanning microscopes LSM510 ( Carl Zeiss Microscopy ) and Leica TCS SP8 ( Leica Microsystems ) at RT . A zoom factor of 4–5 was used to image individual stem cell niches . Images represent either individual confocal sections or projections of 2–3 sections that were chosen from Z-stacks ( 1 μm intervals ) , which were routinely acquired of all studied germaria . Image analysis , cell counts and cell shape measurements were done by evaluating Z-stacks , using the LSM 5 Image Browser and ( Carl Zeiss Microscopy ) and Leica LAS X software ( Leica Microsystems ) . Images were processed with Adobe Photoshop and Illustrator CS5 and CS6 ( Adobe Software ) . Unpaired , two-tailed Student’s t-tests or one-way ANOVA tests were used for statistical analysis . Prism 6 ( GraphPad Software ) was used for statistical tests , and Prism 6 and Illustrator CS6 for the generation of graphs .
|
Establishment and maintenance of stem cells often depends on associated niche cells . The germline stem cell niche of the Drosophila ovary has been a long-standing model for the analysis of the interactions between stem cells and niche cells . Surprisingly little is known , however , about the mechanisms that pattern this niche , leading to the specification of different niche cell types and to their distinct arrangement around the stem cells . The observation that Tj is expressed at different levels in the different cell types of the niche motivated us to ask what contribution this transcription factor makes to the formation of the niche . Our data suggest that Tj activity is needed for the presence of escort cells and for the correct specification of cap cells but appears to be dispensable for the formation of terminal filament cells in the germline stem cell niche . Moreover , our analysis indicates that the establishment of the cap cell fate depends on the cooperation between Tj and the N signaling pathway . We conclude that Tj regulates the germline stem cell carrying capacity of the niche by controlling the fate and the spatial arrangement of niche cells .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"cell",
"physiology",
"invertebrates",
"medicine",
"and",
"health",
"sciences",
"reproductive",
"system",
"animals",
"cell",
"differentiation",
"cell",
"formation",
"germ",
"cells",
"animal",
"models",
"developmental",
"biology",
"drosophila",
"melanogaster",
"model",
"organisms",
"stem",
"cells",
"experimental",
"organism",
"systems",
"drosophila",
"research",
"and",
"analysis",
"methods",
"animal",
"cells",
"stem",
"cell",
"niche",
"dpp",
"signaling",
"cascade",
"insects",
"hedgehog",
"signaling",
"arthropoda",
"ovaries",
"signal",
"transduction",
"anatomy",
"cell",
"biology",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"cell",
"signaling",
"organisms",
"signaling",
"cascades"
] |
2017
|
Specification and spatial arrangement of cells in the germline stem cell niche of the Drosophila ovary depend on the Maf transcription factor Traffic jam
|
Virtually every molecular biologist has searched a protein or DNA sequence database to find sequences that are evolutionarily related to a given query . Pairwise sequence comparison methods—i . e . , measures of similarity between query and target sequences—provide the engine for sequence database search and have been the subject of 30 years of computational research . For the difficult problem of detecting remote evolutionary relationships between protein sequences , the most successful pairwise comparison methods involve building local models ( e . g . , profile hidden Markov models ) of protein sequences . However , recent work in massive data domains like web search and natural language processing demonstrate the advantage of exploiting the global structure of the data space . Motivated by this work , we present a large-scale algorithm called ProtEmbed , which learns an embedding of protein sequences into a low-dimensional “semantic space . ” Evolutionarily related proteins are embedded in close proximity , and additional pieces of evidence , such as 3D structural similarity or class labels , can be incorporated into the learning process . We find that ProtEmbed achieves superior accuracy to widely used pairwise sequence methods like PSI-BLAST and HHSearch for remote homology detection; it also outperforms our previous RankProp algorithm , which incorporates global structure in the form of a protein similarity network . Finally , the ProtEmbed embedding space can be visualized , both at the global level and local to a given query , yielding intuition about the structure of protein sequence space .
Using sequence similarity between proteins to detect evolutionary relationships—protein homology detection—is one of the most fundamental and longest studied problems in computational biology . A protein's function is strongly correlated with its 3D structure , and due to evolutionary pressure , protein structures diverge much more slowly than primary sequences . Because protein sequence data will always be far more abundant than high-quality 3D structural data , the computational challenge is to infer evolutionarily conserved structure and function from subtle sequence similarities . When the evolutionary distance is large and the sequence signal faint—so-called remote homology detection—this problem is still unsolved . Stated in purely computational terms , remote homology detection involves searching a protein database for sequences that are evolutionarily related ( even remotely ) to a given query sequence . Most work in this area has focused on developing more sensitive pairwise comparisons between the query and target sequences , including sequence-sequence local alignments ( BLAST [1] , Smith-Waterman [2] ) ; profile-sequence ( PSI-BLAST [3] ) and HMM-sequence comparisons ( HMMER [4] ) ; and , most recently , profile-profile [5] and HMM-HMM ( HHPred/HHSearch [6] ) comparisons . From a machine learning point of view , these recent methods involve building a model of the neighborhood of the query and of the target in protein sequence space and using the local neighborhood models to compute a better similarity measure . However , recent advances in massive data domains such as web search and natural language processing suggest that the global structure of the data space can also be exploited . For example , motivated by the success of Google's PageRank algorithm , we previously developed RankProp [7] , an algorithm that uses graph diffusion on the protein similarity network , defined on a large protein sequence database , in order to re-rank target sequences relative to the query and substantially improve remote homology detection . In the current study , we are motivated by large-scale learning of language models in recent work in natural language processing ( NLP ) [8] . This NLP work exploited large online text data sets ( e . g . , Wikipedia ) to learn an embedding of words into a low-dimensional semantic space , inducing an embedding of sentence fragments . The embedding algorithm iteratively pushes pairs of real sentence fragments together and pulls pairs of real and randomized sentence fragments apart . Thus , at the end of training , words that are near each other in the embedding space are likely to be semantically related . Moreover , the embedding representation can be leveraged to simultaneously train models to solve multiple NLP tasks , using the framework of multitask learning [9] . Here , we present an algorithm called ProtEmbed that learns an embedding of protein domain sequences into a semantic space such that proximity in the embedding space captures homology relationships . After this large-scale training procedure , remote homologs of a query sequence can be detected by mapping the query to the embedding space and retrieving its nearest neighbors . Furthermore , as in the NLP case , we can use multitask learning to incorporate auxiliary information , where available , to improve the embedding , including structural class labels from databases such as SCOP [10] or structural similarity scores for pairs of training examples where both 3D structures are known . It is important to note that our embedding is defined naturally on protein domain sequences rather than multidomain sequences . In particular , inclusion of multidomain sequences in the training data can lead to incompatible distance relationships in the semantic space due to lack of transitivity , resulting in a worse embedding . At testing time , it may be possible to resolve the domain structure of a multidomain query sequence using the learned embedding ( see Discussion ) ; however , we only evaluate performance on domain sequence queries in the current study . We show that ProtEmbed achieves state-of-the-art performance for remote protein homology detection , outperforming our previous algorithm RankProp , which also exploits global structure but uses a fixed weighted similarity network rather than a learned embedding . Our procedure also yields statistical confidence estimates and enables a visualization of the learned protein embedding space , giving new intuition about the global structure of the protein sequence space .
The main idea of our approach is to learn a mapping of protein domain sequences into a vector space that captures their “semantic similarity” , i . e . closeness in the semantic space should reflect homology relationships between sequences . In order to learn an embedding of protein sequences into a semantic space , we need to define ( i ) a feature representation for proteins , ( ii ) a training signal that determines whether a given pair of training sequences are similar and should be pushed together by the algorithm , or dissimilar and should be pulled apart , and ( iii ) an algorithm that learns an appropriate embedding . Let us denote the set of proteins in the database as and a query protein as , where is the set of all possible sequences of amino acids . We then choose a feature map to represent proteins as vectors . This map is necessary so that we can perform geometric operations on proteins . We use the following representation for a protein :where is the E-value returned by a surrogate protein alignment algorithm , such as PSI-BLAST , suitably transformed . Following Rankprop [7] , we use the following transformation:where is the PSI-BLAST E-value assigned to protein given query and where we set the parameter . This transformation yields a stochastic connectivity matrix; i . e . , the value can be interpreted as the probability that a random walk on the protein similarity network will choose to move from protein to protein . Note that , because most protein pairs exhibit no detectable similarity according to an algorithm such as PSI-BLAST , most feature values are zero . ( Specifically , PSI-BLAST assigns a large maximal E-value to all database sequences for which no homology to the query is detected , and the exponential transfer function converts these values to zero . ) The sparseness of the feature vectors will be important for computational reasons . Next , we again use a surrogate protein alignment algorithm , this time as a teacher to provide a noisy training signal . We construct a training set of tuples , where each tuple contains a query , a related protein and an unrelated ( or lower ranked ) protein . The tuples themselves are collected by running PSI-BLAST in an all-versus-all fashion over the database of proteins . Taking any given protein as the query , we consider any protein with an E-value lower than 0 . 1 to be a similar protein ( instance of a ) ; in the current implementation , instances of are chosen randomly from all training examples and with high probability will be dissimilar to . We can then , in principle , construct all possible combinations ( tuples ) from which we sample randomly during online training . Given the feature vectors and the training tuples , our aim is to learn a feature embedding that performs well for protein ranking and classification tasks . We will learn an embedding functionwhere is an matrix , resulting in an embedding . Typically , is chosen to be low dimensional , e . g . . The learning procedure consists of finding a matrix such that similar proteins have close proximity in the embedding space . Specifically , we would like to choose such that , for all tuples , expressing that should be ranked higher than , relative to an appropriate distance measure in the embedding space . We define this distance measure using the -norm ( which is defined as ) : After training , given a query protein , we will rank the database using the ranking score:where we consider smaller values of to be more highly ranked . The training objective employs the margin ranking loss [11] , which has been used successfully in the field of information retrieval to rank documents given a query [12]–[14] . That is , we minimize: ( 1 ) which encourages to be smaller than until a margin constraint of is satisfied . Intuitively , the algorithm tries to push and together while pulling and apart , until the difference in distances achieves a margin of . For an equivalent formulation , we can introduce a slack variable for each tuple and enforce the constraintsfor all tuples while minimizing the objective function This optimization problem is solved using stochastic gradient descent [13]: iteratively , one picks a random tuple and , if , makes a gradient step for that tuple as follows: ( 2 ) where denotes that the sign function is applied componentwise to the vector to yield a vector of values . Pseudocode for training the ProtEmbed embedding is given in Algorithm 1 in Text S1 . One can exploit the sparsity of and when calculating these updates to make them computationally cheap . To train our model , we choose the ( fixed ) learning rate that minimizes the training error , i . e . the loss defined by equation ( 1 ) . We initialize the matrix randomly using a normal distribution with mean zero and standard deviation one . Overall , stochastic training is highly scalable and is easy to implement for our model , and learning can scale to millions of proteins . After training , we precompute the embedding for every protein in the database . At test time , given a query protein , we compute its linear embedding once . Then we are left with only operations per protein in the database to perform when retrieving results for that query . In general , recognizing remote homology relationships among protein structures is easier than recognizing remote homologies based only on protein sequences . Although structural information is available for only a subset of the proteins in the database , we would like to ensure that our embedding captures this structural information in addition to the sequence-based information provided by PSI-BLAST . We consider two sources of structural information: ( 1 ) category labels for a given protein and ( 2 ) similarity scores between pairs of proteins . For the the category labels , we use the Structural Classification of Proteins ( SCOP ) [10] . For pairwise similarity scores , we use pairwise structure alignments of known 3D structures using MAMMOTH [15] . We incorporate this auxiliary information using the framework of multitask learning: in addition to the main embedding task , we simultaneously learn models to solve additional tasks using appropriate subsets of the training data . The tasks share internal representations learned by the algorithm , in this case , the embedding function . In particular , we pose an auxiliary classification task using SCOP categories , and we pose an auxiliary ranking task using either SCOP category relationships or using MAMMOTH similarities . In all cases , the multitask objective function is simply the sum of the original ProtEmbed objective function and of that of the auxiliary task . We consider these two task types in turn . Class-based data . For auxiliary data in the form of a class label for protein we train an auxiliary classification task that is multitasked with the original ProtEmbed objective , sharing the same embedding space . For each fold and superfamily class we create a vector , , which can be thought of as a set of class centroids . We then would like to satisfy the constraints:That is , proteins belonging to some class should be closer to that class centroid than proteins that do not belong to that class . We train this model using the margin ranking loss as before , and multitask this problem with the original objective using the following updates: ( 3 ) Here is a matrix containing the centroid vectors as columns , and ( resp . ) is the bit vector of length whose two non-zero entries are placed at indices for the fold and superfamily of the labeled training example ( resp . ) . Pseudocode for training the ProtEmbed embedding with class-based auxiliary data is given in Algorithm 2 in Text S1 . Ranking-based data . For auxiliary data in the form of similarity scores between pairs of proteins , we simply add more ranking constraints into the set of tuples . That is , we consider additional tuples of the form where and are similar SCOP proteins based on auxiliary data—i . e . , a similarity score comparing these proteins is above a cutoff value—while is chosen at random from all of SCOP and with high probability will be structurally dissimilar to . Then we require these additional tuples to satisfy constraints of the formanalogous to the constraints in the main optimization problem . Two examples of the use of such auxiliary constraints are given by using SCOP superfamily labels or MAMMOTH . For SCOP labels , if two proteins are in the same superfamily , we say they are similar . For MAMMOTH , we choose a cutoff value of 2 . 0 , and a pair of proteins that has a structural alignment scoring above this cutoff is deemed to be similar . Pseudocode for training the ProtEmbed embedding with ranking-based auxiliary data is given in Algorithm 3 in Text S1 . For labeled data—namely , proteins with structural category labels and 3D structures from which to compute pairwise similarity scores—we used proteins from the SCOP v1 . 59 protein database . We used ASTRAL [16] to filter these sequences so that no two sequences share greater than 95% identity . This filtering resulted in 7329 sequences . Our test set consists of 97 proteins selected at random from these SCOP sequences . These test sequences were excluded entirely from the training data . For unlabeled data , i . e . protein domain sequences without category labels or structural information , we used sequences from the ADDA domain database version 4 [17] ( http://ekhidna . biocenter . helsinki . fi/downloads/adda ) . This database contains 3 , 854 , 803 single-domain sequences . We removed from the database sequences comprised entirely of the ambiguity code “X , ” sequences shorter than 6 amino acids and sequences longer than 10 , 000 amino acids . We then randomly selected sequences from the remaining sequences until we had picked 3% of the original sequences . This left us with an unlabeled single domain database of 115 , 644 sequences . We ran PSI-BLAST version 2 . 2 . 8 on the combined SCOP+ADDA database using the default parameters , allowing a maximum of 6 iterations . For a second and more powerful pairwise sequence similarity method based on HMM-HMM comparisons , we also ran HHSearch version 1 . 5 . 0 , using default parameters . HHPred/HHSearch is considered a leading method for remote homology detection [6] . When searching for homologs to the test set domains , we added the HHSearch options “-realign -mact 0 , ” which uses local Viterbi search followed by MAC to realign the proteins globally on a local posterior probability matrix . Similarly , MAMMOTH was run with its default settings . We first trained embeddings on SCOP+ADDA ( with SCOP test sequences held out ) using PSI-BLAST or HHSearch as the pairwise sequence comparison method to serve as “teacher” for producing tuples . In this setting , we did not make use of the category labels or structural information for the SCOP training examples . We then trained embeddings using ADDA as unlabeled data and SCOP as labeled data , where the labeled data was used in ( i ) an auxiliary classification task based on SCOP category labels or ( ii ) an auxiliary ranking task based either on SCOP category relationships or on MAMMOTH similarity scores .
As an initial proof-of-concept test of the ProtEmbed algorithm , we created an embedding of protein domains into a two-dimensional space . This embedding is necessarily underfit , because two dimensions does not provide very much capacity to learn a good embedding . However , a two-dimensional space has the advantage of being easy to visualize . We trained the embedding using the 7329 SCOP proteins from the training set , and then calculated the locations of the all SCOP proteins from all superfamilies with 25 or more members . Figure 1 shows these locations . Proteins are colored and labeled according to their SCOP superfamilies . The embedding generally places members of the same superfamily near one another . To investigate the ability of ProtEmbed to rank homologous proteins above non-homologs , we used a gold standard derived from the SCOP database of protein domain structures . We then used PSI-BLAST , Rankprop , HHSearch and ProtEmbed to rank a collection of 7329 SCOP domain sequences with respect to each of 97 test domains . To provide a rich database in which to perform the search , we augmented the SCOP data set with 115 , 644 single-domain sequences from the ADDA domain database . In our evaluation , protein domains that reside in the same SCOP superfamily as a query domain are labeled positive , and domains in different folds than that of the query are labeled negative . The remaining sequences—from the same fold but different superfamilies—are ignored , because their homology to the query is uncertain . For each query , traversing the ranked list of labeled sequences induces a receiver operating characteristic ( ROC ) curve , which plots the percentage of positives as a function of the percentage of negatives observed thus far in the ranked list . We measured the area under this curve up to the first false positive ( ) or the 50th false positive ( ) . Both scores are normalized such that perfect performance corresponds to a score of 1 . 0 . Before training our embedding , we ran a series of cross-validation experiments within the training set to select hyperparameters; i . e . , parameters that are not subject to optimization . Based on these experiments , we used , for PSI-BLAST , a learning rate of 0 . 05 and an embedding dimension of 250; and for HHSearch , a learning rate of 0 . 02 and an embedding dimension of 100 . In each case , the training was run for 150 epochs , where one epoch corresponds to 20 , 000 tuples . We used the same hyperparameters when training with or without the auxiliary , structural information . Figure 2 compares the performance of PSI-BLAST , RankProp , HHSearch and various versions of the ProtEmbed algorithm . The performance of each algorithm is summarized by the mean or score . To establish the statistical significance of the observed differences , we used a Wilcoxon signed-rank test with a 0 . 05 significance threshold . For both of the performance metrics that we considered , the ranking of the three previously described methods is the same: HHSearch outperforms Rankprop , which outperforms PSI-BLAST . Also , the standard ProtEmbed algorithm , with no auxiliary data , outperforms PSI-BLAST when it is trained using PSI-BLAST and outperforms HHSearch when it is trained using HHSearch , although for the latter comparison , the difference is only significant for the performance metric . Figure 2 in Text S1 , which plots the number of queries for which the or score exceeds a given threshold , shows that the differences among methods are not traceable to queries with particularly high or low ROC values; on the contrary , the improvements from one method to the next span the entire range of ROC values . Figure 2 shows that adding auxiliary , structural information during ProtEmbed training significantly improves the quality of the resulting rankings . Adding structural information to ProtEmbed improves the mean score by 0 . 038–0 . 170 and improves the by 0 . 083–0 . 180 . Perhaps most strikingly , if we consider ProtEmbed trained from HHSearch , the initial embedding is 0 . 154 away from a perfect score , whereas the embedding learned using SCOP rankings is only 0 . 025 away from a perfect score . Thus , in this case , structural information removes 83 . 7% of the residual error . In general , using SCOP information leads to better rankings than using MAMMOTH . This is not surprising , because we are using a gold standard based on SCOP . Between the two modes of representation , the SCOP ranking appears to give better results than using SCOP class-based structural information . This result is somewhat surprising , because our gold standard is based explicitly on SCOP classes and perhaps suggests that the ranking representation is more resistant to overfitting . In evaluations of remote homology detection algorithms , some researchers prefer to ignore members of the same family as the query , since these family members are presumably easy to identify [18] . To ensure that our results are not dependent on family-level information , we repeated the ROC calculations above , but we skipped target proteins that fall into the same family as the query . Figure 3 in Text S1 shows that the conclusions above remain unchanged in this setting: ProtEmbed outperforms HHSearch , RankProp and PSI-BLAST , and using structural information significantly improves ProtEmbed's performance . Next , we evaluated how well ProtEmbed scores are calibrated between queries . We say that our scores are well calibrated if pairs of query and target sequences at similar distances from each other in embedding space also have similar degrees of homology , regardless of where the query embeds . If this property holds , then the scores generated by ranking database sequences relative to different queries can be compared to each other and modeled to assign statistical significance . The experiment reported in Figure 2 , in which ROC scores are computed separately for each query and then averaged , only measures how well the target sequences in the database are ranked relative to each query sequence . To measure the calibration of the scores among queries , we sorted all of the scores from all 97 test queries into a single list . The resulting ROC curves are shown in Figure 3 . The overall ranking of methods is the same as in Figure 2 , in order of improving performance: PSI-BLAST , Rankprop , HHSearch , ProtEmbed . To obtain calibrated scores , PSI-BLAST , Rankprop and HHSearch include specific calibration procedures—calculation of E-values for PSI-BLAST and HHSearch , and calculation of superfamily probabilities for Rankprop . ProtEmbed , in contrast , requires no explicit calibration procedure; instead , the scores are naturally calibrated because they all correspond to distances in a single embedding space . To be useful , a homology detection algorithm must provide scores with well defined semantics . For example , PSI-BLAST reports an expectation value , or E-value , that corresponds to the number of scores as good or better than the observed score that are expected to occur in a random database of the given size [3] . Rankprop reports for each query-target pair the probability that they belong to the same SCOP superfamily [19] . To convert ProtEmbed distances to an interpretable score , we employed a simple empirical null model in which protein sequences are generated by a third-order Markov chain , with parameters derived from the SCOP+ADDA database . We randomly generated decoy protein sequences according to this null model , and we embedded these proteins into the PSI-BLAST ProtEmbed space . Empirical analysis of the resulting sets of scores ( Figure 1 in Text S1 ) shows that the left tail of the null distribution is well approximated by a Weibull distribution . To compute a p-value , we select the null distribution based on the length of the given query sequence . Further details are given in Text S1 . We cannot use these p-values directly , because we must correct for the large number of tests involved in searching a large sequence database . To do so , we employ standard false discovery rate-based multiple testing correction procedures . In particular , for a given query , we first estimate the percentage of the observed scores that are drawn according to the null distribution [20] . We then use the Benjamin-Hochberg procedure [21] to estimate false discovery rates , including the multiplicative factor . Finally , we convert the estimated false discovery rate into a q-value [20] , which is defined as the minimum FDR threshold at which an observed score is deemed significant . For many users of alignment tools such as PSI-BLAST , the multiple alignment produced with respect to a given query is as useful as the rankings and accompanying E-values , because the multiple alignment provides an explanation of the ranking . However , a method like ProtEmbed does not rely solely on multiple alignments . Therefore , although it would certainly be feasible to create , in a post hoc fashion , an alignment of the ranked proteins up to , e . g . , a specified ProtEmbed q-value threshold , such a multiple alignment is not likely to accurately reflect the semantics of the ProtEmbed embedding space . Instead , we propose to use a multidimensional scaling approach to project the top-ranked protein domains into an easy-to-visualize 2D representation . To illustrate how effective such a visualization can be , we systematically generated 2D maps of the neighborhood for all 97 test set domains , using a q-value threshold of 0 . 01 . Thumbnail versions of all 97 neighborhoods are provided in the supplement . Here , we focus on a single example . Figure 4 shows the structure learned by the embedding near a particular query , the C-terminal domain of Staphylococcal enterotoxin B ( PDB ID 3seb ) . Figure 4 ( A ) shows the neighborhood of the query relative to the initial PSI-BLAST based feature embedding of the domain sequences , projected into 2D for easier visualization . This mapping corresponds to the initialization of the embedding algorithm , before any training . We see that the other members of the query's family—the superantigen toxins , C-terminal domain ( SCOP 1 . 75 ID d . 15 . 6 . 1 ) , shown in green—are generally near the query in the initial embedding , but these true positives are intermingled with members of a functionally related but structurally distinct superfamily , the bacterial enterotoxins ( SCOP 1 . 75 ID b . 40 . 2 , shown in blue ) as well as several members of unrelated superfamilies . When we map the query sequence into the final embedding space ( Figure 4 ( B ) ) , we now find that it lands in a tight cluster of its family members , which is near but separated from the cluster of related bacterial enterotoxins . Meanwhile , unrelated superfamilies are appropriately separated into distinct clusters distant from the query . In this example , the homology detection performance improves from an score of 0 . 091 ( of 0 . 716 ) relative to the initial embedding to a perfect ( and perfect ) of 1 . 0 after training .
We have shown that ProtEmbed learns an embedding of protein domain sequences such that proximity in the embedding space reflects homology relationships . Due to efficient stochastic gradient descent methods , the training algorithm can scale to millions of sequences . A flexible multitask framework also enables the use of additional label or ranking information , e . g . protein structural classes or pairwise structural similarity scores , where known , to improve the embedding . Given a test query sequence , its embedding can be computed in the same time that it takes to run the underlying pairwise sequence alignment method . The query's homologs can then be efficiently retrieved by determining the nearby database proteins based on their precomputed embedding coordinates . Moreover , using a faster but less accurate pairwise alignment method , such as PSI-BLAST , together with ProtEmbed , when supplied with labeled data through an auxiliary task , leads to better performance than state-of-the-art but slower pairwise alignments methods , such as HHSearch , used on their own . Moreover , use of more sensitive PSI-BLAST parameters rather than the default choices could potentially further improve the performance of the embedding . While alignment-based pairwise sequence similarity scores are used as features for calculating the embedding , ProtEmbed does not produce multiple sequence alignments for query sequences as an output of its computation . Instead , the embedding neighborhood of the query can be visualized for insight into the relationship between the query and its homologs . For further sequence-based analysis of query-homolog similarities , hits from the ProtEmbed neighborhood could be used to compute an alignment using standard methods [22] or newer graph algorithm approaches [23] . The ProtEmbed algorithm learns its embedding on domain sequences rather than full-length protein sequences , because the embedding only makes sense when transitivity relationships hold . For example , a multidomain sequence will have sequence similarity to its constituent domains , which will typically also be represented as entries in the database; if these domains are dissimilar from each other , then the set of pairwise relationships lead to conflicting constraints during training . Nonetheless , it is possible to process a multidomain query sequences using ProtEmbed by first applying an existing domain decomposition algorithm [24] and then embedding each domain separately . Alternatively , one could potentially use the embedding to help resolve the domain structure: first , one could run a pairwise alignment method such as PSI-BLAST to determine the start and end positions of all the hits , and then these subsequences could be embedded separately as candidate domain sequences . The p-value for the score between the embedded candidate sequence and its nearest neighbor in the database should generally favor candidates with boundaries similar to those of the true domains . Protein sequence analysis is one of the oldest subfields of computational biology , with mature and specialized tools designed to describe the local structure of protein sequence space . By adapting new techniques from massive data domains such as natural language processing and web search , we have demonstrated that the global structural of protein sequence space can be exploited for classical problems like homology detection .
|
Searching a protein or DNA sequence database to find sequences that are evolutionarily related to a query is one of the foundational problems in computational biology . These database searches rely on pairwise comparisons of sequence similarity between the query and targets , but despite years of method refinements , pairwise comparisons still often fail to detect more distantly related targets . In this study , we adapt recent work from natural language processing to exploit the global structure of the data space in this detection problem . In particular , we borrow the idea of a semantic embedding , where by training on a large text data set , one learns an embedding of words into a low-dimensional semantic space such that words embedded close to each other are likely to be semantically related . We present the ProtEmbed algorithm , which learns an embedding of protein sequences into a semantic space where evolutionarily-related proteins are embedded in close proximity . The flexible training algorithm allows additional pieces of evidence , such as 3D structural information , to be incorporated in the learning process and enables ProtEmbed to achieve state-of-the-art performance for the task of detecting targets that have remote evolutionary relationships to the query .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"computational",
"biology/protein",
"homology",
"detection"
] |
2011
|
Detecting Remote Evolutionary Relationships among Proteins by Large-Scale Semantic Embedding
|
As nascent polypeptide chains are synthesized , they pass through a tunnel in the large ribosomal subunit . Interaction between specific nascent chains and the ribosomal tunnel is used to induce translational stalling for the regulation of gene expression . One well-characterized example is the Escherichia coli SecM ( secretion monitor ) gene product , which induces stalling to up-regulate translation initiation of the downstream secA gene , which is needed for protein export . Although many of the key components of SecM and the ribosomal tunnel have been identified , understanding of the mechanism by which the peptidyl transferase center of the ribosome is inactivated has been lacking . Here we present a cryo-electron microscopy reconstruction of a SecM-stalled ribosome nascent chain complex at 5 . 6 Å . While no cascade of rRNA conformational changes is evident , this structure reveals the direct interaction between critical residues of SecM and the ribosomal tunnel . Moreover , a shift in the position of the tRNA–nascent peptide linkage of the SecM-tRNA provides a rationale for peptidyl transferase center silencing , conditional on the simultaneous presence of a Pro-tRNAPro in the ribosomal A-site . These results suggest a distinct allosteric mechanism of regulating translational elongation by the SecM stalling peptide .
The ribosome is a large macromolecular particle that synthesizes polypeptide chains from the substituent amino acid building blocks . The active site for peptide bond formation , the so-called peptidyl transferase center ( PTC ) , is located in a cleft on the intersubunit side of the large ribosomal subunit ( reviewed by [1] , [2] ) . As the nascent polypeptide chain is being synthesized , it passes through a tunnel within the large subunit and emerges at the solvent side , where protein folding occurs . Recently , nascent polypeptide chains have been directly visualized within the ribosomal tunnel extending from the PTC to the exit site on the back of the large subunit [3]–[5] , as originally predicted by Lake and coworkers in the 1980s [6] , [7] . The X-ray structures of bacterial and archaeal ribosomes have revealed that the ribosomal tunnel is predominantly composed of ribosomal RNA ( rRNA ) [8]–[12] , consistent with an overall electronegative potential [13] , [14] . In addition to rRNA , the extensions of the ribosomal proteins L4 and L22 ( L17 in eukaryotes ) contribute to formation of the tunnel wall , and form a so-called constriction where the tunnel narrows [8] , [9] . Near the tunnel exit , the bacterial-specific extension of L23 ( L25 in eukaryotes ) occupies a similar position to the r-protein L39e of eukaryotic and archaeal ribosomes [10]–[12] . Despite its universality , a functional role for the ribosomal tunnel is only beginning to emerge . For many years , the ribosomal tunnel was thought of only as a passive conduit for the nascent polypeptide chain; however , accumulating evidence indicates that , for some nascent chains , the tunnel plays a more active role ( reviewed by [15] ) . In particular , a number of leader peptides have been identified that induce translational stalling in response to the presence or absence of an effector molecule , and in doing so regulate translation of a downstream gene ( reviewed by [16] , [17] ) . Well-characterized examples include the eukaryotic arginine attenuator peptide ( AAP ) and cytomegalovirus gp48 uORF , as well as the bacterial ErmC , TnaC , and SecM leader peptides , for which mutations in the leader peptide sequences , or within the ribosomal tunnel components , can relieve the translational arrest [18]–[21] . The implication of a direct interaction between specific residues of the leader peptide with distinct locations of the ribosomal tunnel has been confirmed by a recent cryo–electron microscopy ( EM ) and single particle reconstruction of a ribosome stalled during translation of the TnaC leader peptide by the presence of high concentrations of free tryptophan [4] . In contrast to stalling by TnaC , translational stalling by SecM does not require an effector molecule [22] . A minimal stalling sequence comprising 17 amino acids ( aa ) ( SecM150–166 ) of the 170-aa SecM leader peptide is sufficient to induce translational arrest [20] . Furthermore , unlike with TnaC , where stalling occurs naturally at the UGA stop codon , i . e . , during termination [19] , stalling of SecM occurs during elongation at a CCU sense codon ( encoding Pro166 ) [20] . The stalled complex has the peptidyl-tRNA ( SecM-tRNAGly ) at the P-site and Pro-tRNAPro at the A-site of the ribosome [23] , and is thus stalled in a pre-translocation state prior to peptide bond formation . Yet , transfer of the SecM nascent peptide from the tRNAGly to the tRNAPro can still occur slowly [23] , and is triggered by the presence of SecA activity to alleviate stalling [24] . Mutational analysis has identified the conserved Arg163 , Gly165 , and Pro166 of SecM as being critical for translational stalling [20] , [25] , with additional contributions from Phe150 , Trp155 , Ile156 , Gly161 , and Ile162 [20] ( Figure 1A ) . Translational arrest is also alleviated by modification of ribosomal components of the tunnel , namely , mutation A2058G , A2062U , or A2503G , or single adenine insertions at A749–A753 of the 23S rRNA [20] , [26] , [27] , as well as mutations , insertions , or deletions within ribosomal proteins L22 and , with lesser effect , L4 [20] , [26] . Despite extensive biochemical characterization , the mechanism by which the PTC of the ribosome is inactivated remains unclear . One structural study on SecM stalling at low resolution purported that the elongation arrest arises from a cascade of rRNA conformational rearrangements [28] . Here we have determined a cryo-EM reconstruction of a SecM-stalled ribosome nascent chain complex ( RNC ) at 5 . 6 Å , enabling the direct interaction between critical residues of SecM and the ribosomal tunnel to be visualized . While we find no evidence for a cascade of rRNA conformational changes , we observe a shift in the position of the tRNA–nascent peptide linkage of the SecM-tRNA . This shift moves the carbonyl carbon of the SecM-tRNA away from the A-tRNA and , thus , is likely to contribute to the impaired activity of the PTC , explaining the SecM-mediated translational arrest .
To generate SecM-stalled RNCs , a construct was prepared that encodes consecutive His- and HA-tags connected by a linker region to the C-terminal 27 aa ( SecM144–170 ) of SecM ( Figure 1A ) . The SecM-stalled RNCs were generated using an Escherichia coli in vitro translation system and purified using Co-NTA affinity chromatography as described previously ( Figure 1B ) [29] . To ensure homogeneity of the RNC sample , 70S monosome fractions of the SecM-stalled RNCs were separated from affinity-purified polysome fractions using sucrose density gradient centrifugation ( Figure 1C ) . An initial cryo-EM reconstruction was generated from 1 . 1 million particles of the monosome fraction , revealing a 70S ribosome with tRNAs occupying A- , P- , and E-sites , very similar to that previously reported [28] . Previous biochemical analysis has shown that the majority of ribosomes stall at position 165 of the SecM ORF , with a glycine as the most C-terminal amino acid bound to the peptidyl-tRNA in the ribosomal P-site , and an obligatory Pro-tRNAPro in the A-site . An additional minor fraction of ribosomes undergo slow transfer of the nascent peptide from the tRNAGly to the tRNAPro after longer incubation times ( Figure 1D ) . We therefore applied an in silico sorting procedure [30] to resolve the conformational heterogeneity within the complex ( Figure 2 ) . Of the 1 . 1 million particles sorted , the largest fraction ( 750 , 000 particles ) had unratcheted ribosomes , with the majority ( 544 , 000 particles; ∼50% ) containing a single peptidyl-tRNA at the P-site . This state was reconstructed at 5 . 6 Å ( 0 . 5 Fourier shell correlation [FSC]; Figure S1 ) and termed the SecM-stalled RNC ( Figure 3A ) . At this resolution , clear density for the SecM nascent polypeptide chain is observed within the exit tunnel of the large subunit ( Figure 3A ) . As expected , a subpopulation of P-tRNA containing unratcheted ribosomes with an additional A-tRNA was also observed , representing SecM-stalled RNCs with Pro-tRNAPro still bound in the A-site . Partial dissociation of the A-site tRNA during the high salt ( 250 mM KOAc ) wash protocol in our RNC preparation may provide an explanation for the low overall occupancy of A-site-bound Pro-tRNAPro ( 9% ) ( Figure 2 ) . Despite low particle numbers , we were able to reconstruct this complex to a resolution of 9 . 3 Å ( Figure S1 ) ; however , the limited resolution does not allow for the direct visualization of the SecM nascent chain ( Figure 3B ) . There is , however , no conformational difference between the two SecM-stalled RNCs , indicating that the presence of the Pro-tRNAPro in the A-site does not trigger any large-scale conformational changes related to stalling ( Figure S2 ) . Computational sorting revealed that another subpopulation ( 350 , 000 particles; 32% ) of ribosomes had undergone a ratchet-like subunit rearrangement of the small subunit relative to the large subunit ( Figure 2 ) . The reconstruction of the ratcheted complex at a resolution of 6 . 0 Å revealed two tRNAs present in A/P and P/E hybrid sites and clear density for the nascent chain in the tunnel ( Figure 3C ) . This peptidyl-tRNA observed in the A/P hybrid site is in accordance with the biochemical studies demonstrating that with incubations longer than 60 min , such as in the RNC purification protocol used here , there is a slow release from the arrested state [23] , i . e . , transfer from tRNAGly in the P-site to the A-site-bound Pro-tRNAPro ( Figures 1D and 3D ) . Following peptidyl transfer , ribosomes are free to ratchet and the associated tRNAs can adopt hybrid states [31]–[34] ( Figure 3D ) . On this basis , we interpret the ratcheted complex as a post-arrest ribosome containing SecM-Pro-tRNAPro in the A/P-site and deacylated tRNAGly in the P/E-site , and thus termed it SecM-Pro-RNC ( Figure 3C ) . The SecM-Pro-RNC hybrid state is similar , in terms of degree of ratcheting , tRNA positions , and L1 stalk movement , to that observed previously with 70S ribosomes containing peptidyl-tRNA mimics fMetLeu- or fMetTrp-tRNA [33] , [34] ( Figure S3 ) . A molecular model for the SecM-stalled RNC was built by rigid-body docking of the ribosomal subunits from the model of the TnaC-stalled RNC [4] . Within the limits of the 5 . 6-Å resolution , we observe an excellent agreement between the ribosome structures of SecM-stalled RNC and TnaC-stalled RNC [4] , as well as with the crystal structures of bacterial ribosomes [11] , [12] . We find no evidence for any cascades of rRNA conformational rearrangements as proposed earlier [28] , suggesting that the purported rearrangements may have arisen due to conformational heterogeneity , which we also observed in the unsorted SecM-stalled RNC sample ( Figures 2 and S2 ) . Taken together , in silico sorting of our dataset resulted in segregation into subpopulations with defined functional/conformational states ( Figures 2 , 3E , and S2 ) that are in agreement with the biochemical data . Moreover , this procedure allowed higher resolution reconstructions to be obtained , enabling the nascent polypeptide to be directly visualized within the ribosomal tunnel , which is not possible at lower resolutions ( Figure S4 ) . The density characteristics indicate that the SecM nascent chain adopts a predominantly extended conformation , similar to that of TnaC [4] ( Figure S5 ) , but with some slight compaction in the upper tunnel ( Figures 4 and S6 ) . A large region of compaction is observed near the tunnel exit , as reported previously for TnaC and Helix RNCs [4] , [5] , but the distance from the PTC indicates that this region is unrelated to the SecM sequence in our construct . Nevertheless , a compacted conformation for SecM between residues 135 and 159 has been reported based on fluorescence resonance energy transfer measurements [35] , which would encompasses SecM in the lower tunnel region . Thus , based on an essentially extended conformation of the SecM nascent chain in the critical region , we have built a polyalanine model that has been used to interpret the observed contacts of SecM with components of the ribosomal tunnel ( Figure 4; Table S1 ) . Because the resolution of the map is limited to approximately 6 Å , all analysis was restricted to the proximity of the Cα atoms of SecM . In the upper region of the tunnel of the SecM-stalled RNC , three connections are observed between the nascent chain and components of the tunnel wall , namely , 23S rRNA nucleotides U2585 , U2609 , and A2062 ( Figure 4 ) . Strong density connects A2062 to the proximity of Arg163 of SecM . This contact is likely to be critical for SecM stalling since scanning mutagenesis with Ser indicates that mutation of only Arg163 of SecM abolishes SecM stalling [20] , [25] . Similarly , the mutation A2062U abolishes both SecM and ErmC stalling [27] . A2062 is highly flexible [36] and appears to adopt a position flat against the tunnel wall in the SecM-stalled RNC , possibly constrained by the close proximity of the bulky Arg163 and Ile162 residues of SecM . Consistent with this , Vazquez-Laslop et al . [27] have recently suggested that this orientation of A2062 triggers a relay through A2503 ( which is also essential for SecM and ErmC stalling [27] ) to inactivate the PTC . In contrast , the interaction of U2585 with SecM in the proximity of Ala164 , and of U2609 with the slightly compacted 160QAQ158 area of SecM , are less likely to be important for SecM stalling ( Figure 4 ) , since mutations of these amino acid residues do not significantly affect SecM stalling [20] , [25] . Within the constriction located in the mid-tunnel region , only one major contact is observed to SecM , namely from the vicinity of A751 towards Trp155/Ile156 of SecM ( Figure 4 ) . Insertion of adenine within the five consecutive adenines A749–A753 of the 23S rRNA , or either mutation Ile156Ala or Trp155Ala , abolishes E . coli SecM stalling [20] . Furthermore , mutations of the neighboring ribosomal protein L22 , specifically Gly91Ala and Ala93Ser at the tip of the β-hairpin that interacts with A751 , also suppress translation arrest due to SecM [20] , [26] , as well as TnaC [37] . Interestingly , TnaC also encodes a tryptophan ( Trp12 ) that is located in a similar position in the tunnel constriction , but which establishes an apparently different interaction with the tunnel that involves directly the loop of L22 as well as A751 ( Table S1 ) [4] . Deeper in the tunnel , the nascent chain establishes contact with K84 of L22 and Q72 of L23 , but predominantly with helix 50 ( H50 ) of the 23S rRNA in the proximity of A1321 ( Figure 4 ) . This region of SecM is poorly conserved and not essential for stalling; however , we note that SecM150–166 is less efficient at stalling than SecM140–166 [20] , consistent with a fine-tuning role of these residues in the placement of the critical Arg163 [25] . At the PTC , density for the ester linkage between the nascent chain and the terminal A76 of the P-tRNA is clearly observable in the SecM RNC map ( Figure 5A ) . The location of the CCA-end of the P-tRNA is also well characterized from a multitude of ribosomal crystal structures and is essentially identical regardless of whether CCA-end mimics or P-tRNAs are bound to bacterial 70S ribosomes or archaeal 50S subunits [12] , [38] , [39] ( Figure 5B ) . Therefore , we were surprised to find that the peptide ester linkage associated with the terminal A76 appears to be shifted in the SecM-stalled RNC , relative to the crystal structures ( Figure 5C ) . In contrast , the position of the CCA-end of the SecM-Pro-tRNA ( Figure 5D ) , as well as that of the TnaC-tRNA [4] ( Figure 5E ) , is not shifted compared to the crystal structures ( Figure 5F ) . Although chloramphenicol was added to reduce peptidyl-tRNA hydrolysis [40] , it is unlikely that it had an effect on the P-site peptidyl-tRNA [41] , since the shift is not seen in the SecM-Pro-tRNA ( Figure 5D ) , nor in a reconstruction of an E . coli RNC with a non-stalling peptide ( Figure S7 ) , both of which were also purified in the presence of chloramphenicol . A direct comparison of the density maps ( Figure 5G ) and models ( Figure 5H ) for the SecM- and TnaC-stalled RNCs [4] suggests that the A76 ester linkage has shifted by approximately 2 Å . Peptide bond formation requires precise positioning of the A- and P-tRNAs to orient the α-amino group of the A-tRNA for nucleophilic attack on the carbonyl carbon of the P-tRNA [2] , [39] ( Figures 5I and 6A ) . Thus , even slight shifts in the relative position of either substrate dramatically reduce the efficiency of peptide bond formation [2] , [39] . Indeed , the 2-Å shift of the ester linkage of the P-tRNA observed in the SecM-stalled RNCs would move the carbonyl carbon further away from the A-tRNA ( Figures 5I and 6B ) and , thus , contribute to the impaired activity of the PTC , explaining the SecM-mediated translational arrest . Together with the available biochemistry , our results support a model for SecM stalling in which there are two main contributors to efficient stalling . First , contacts of the SecM nascent chain with the ribosomal tunnel aid positioning of the critical Arg163 of SecM [25] to interact with A2062 of the 23S rRNA [27] ( Figure 6B ) . We believe that this interaction ultimately leads to a shift in the position of the ester linkage of the P-tRNA , which can be a consequence of a direct constraint on the SecM nascent chain and/or can occur through an indirect relay of 23S rRNA nucleotides via A2503 ( Figure 6B ) , as proposed by Vazquez-Laslop et al . [27] . Second , Pro-tRNAPro in the A-site is critical for stalling [20] , [23] , as is evident from the observation that the mutation Pro166Ala leads to a reduction in stalling by three orders of magnitude [20] , [26] , [42] . Therefore , the changed geometry of the PTC appears necessary but not sufficient for stalling . In this respect we note that the strictly required Pro-tRNAPro in the A-site is characterized by steric constraints and lower nucleophilicity of the N-alkyl amino acid proline [43] , compared with the other 19 amino acids . Pro-tRNAPhe is 23-fold slower than Phe-tRNAPhe , and Pro-tRNAbulk is 3- to 6-fold slower during peptide bond formation than Ala-tRNAAla or Phe-tRNAPhe [43] , making proline a particularly poor acceptor . Thus , we suggest that the poor chemical properties of proline are exploited to exacerbate the unfavorable geometry of the PTC , leading to efficient translational stalling ( Figure 6B ) . Alternatively , the requirement of Pro-tRNAPro for stalling could also be explained by the rearrangement at the PTC occurring faster than the rate of peptide bond formation with a proline in the A-site , but slower than that with an alanine . Relief of this conformationally locked inactive state is possible by the residual transferase activity and prolonged incubation time [23] ( Figures 1D and 6C ) , or through the presence of SecA [24] . It is conceivable that the physiological relief provided by the SecA ATPase is triggered by unlocking of the inactive PTC geometry via disruption of SecM interactions with the tunnel . In general , perturbations of the PTC are also evident in other stalling sequences , such as TnaC [4] , AAP , and CMV [44] , but without a significant shift in the Pro-tRNA , indicating that each stalling sequence appears to utilize a distinct allosteric mechanism .
The SecM construct was generated by PCR using forward T7_RBS_6xHis ( 5′-TAATACGACTCACTATAGGGCCTCTAGAAATAATTTTGTTTAACTTTAAGAAGGAGATATACATATGTCTCATCATCATCATCATCAT-3′ ) ;and reverse DP_SecM ( 5′-AATGGATTAGTGACAATAAAATTGAATTTACCCCACAAGCAAAATTCAGCACGCCCGTCTGGATAAGCCAGGCGCAAGGCATCCGTGCTGGCCCTCAACGCCTCACCTAATAA-3′ ) primers with DP120 ( without signal anchor ) construct as the template ( Figure 1A ) . Uncapped transcripts were then synthesized from the PCR fragments using T7 RNA polymerase . SecM RNCs were generated using an E . coli in vitro translation system ( Promega ) programmed with SecM mRNA . For in vitro translation , two 500-µl reactions were incubated at 30°C for 20 min ( Figure 1B , lane 1 ) . Chloramphenicol ( 1 µg/ µl ) was added to reduce peptidyl-tRNA hydrolysis [40] during the prolonged purification procedure that followed . Each reaction was spun through 500 µl of a high salt sucrose cushion ( 50 mM HEPES [pH 7 . 0] , 250 mM KOAc , 25 mM Mg[OAc]2 , 5 mM 2-mercaptoethanol , 0 . 75 M sucrose , 0 . 1% Nikkol , 500 µg/ml chloramphenicol , and 0 . 2 U/ml RNAsin; Promega ) and 0 . 1% pill/ml ( 1 pill complete protease mix per 1 ml H2O; Roche Diagnostics ) at 70 , 000 g for 150 min in a TLA 120 . 2 rotor ( Beckman Coulter ) at 4°C . The supernatant ( Figure 1B , lane 2 ) was discarded , and the ribosomal pellet ( Figure 1B , lane 3 ) was resuspended in 500 µl of ice-cold 250 buffer ( 50 mM HEPES [pH 7 . 0] , 250 mM KOAc , 25 mM Mg[OAc]2 , 5 mM 2-mercaptoethanol , 250 mM sucrose , 0 . 1% Nikkol , 500 µg/ml chloramphenicol , 0 . 2 U/ml RNAsin , and 0 . 1% pill/ml ) for 45 min at 4°C , transferred onto 500 µl of Talon Metal Affinity Resin ( Clontech ) pre-equilibrated with 250 buffer supplemented with 10 µg/ml of tRNAs and incubated for 5 min at room temperature . The resin was washed ten times with 1 ml of ice-cold 250 buffer . RNCs were eluted with 2 . 5 ml of 250 buffer supplemented with 100 mM imidazole ( pH 7 . 0 ) . The eluted RNCs were spun through 200 µl of a high salt sucrose cushion at 70 , 000 g for 150 min in a TLA 110 rotor at 4°C , and the resulting RNC pellet was resuspended in 1 ml of grid buffer ( 20 mM HEPES [pH 7 . 0] , 50 mM KOAc , 6 mM Mg[OAc]2 , 5 mM DTT , 500 µg/ml chloramphenicol , 0 . 05% Nikkol , 0 . 5% pill/ml , 0 . 1 U/ml RNAsin , and 125 mM sucrose ) for 30 min at 4°C . The resulting SecM RNC ( Figure 1B , lane 4 ) typically had a yield of approximately 2 . 5 OD260 . An affinity-purified 1 ml of RNCs ( 2 . 5 OD260 ) was further applied to 10 ml of sucrose on a 10%–40% gradient in 250 buffer in order to separate the monomeric SecM-stalled RNCs from the polysomes . Gradients were then centrifuged in a Beckman Coulter SW40-Ti rotor at 20 , 000 rpm for 4 h ( 4°C ) . In parallel , 1 ml of crude 70S ribosomes ( 2 . 5 OD260 ) prepared from the same extract used for translation was also applied on the sucrose gradient as a control ( Figure 1C ) . The monosome SecM RNC fractions were pooled and concentrated by ultra-centrifugation . The yield of isolated monosome SecM RNCs was typically approximately 0 . 5 OD260 . Concentrated monosome SecM RNCs were aliquoted in small volumes , flash frozen in liquid nitrogen , and stored at −80°C until needed . As described previously [45] , 3 . 5 µl of SecM RNCs ( 2 . 5 OD260/ml ) was applied to 2-nm carbon-coated holey grids . Micrographs were then recorded under low-dose conditions ( 25 electrons/Å2 ) with a magnification of 38 , 900 on a Tecnai F30 field emission gun electron microscope at 300 kV in a defocus range of 1 . 0–4 . 0 µm . Micrographs were scanned on a Heidelberg Primescan D8200 drum scanner , resulting in a pixel size of 1 . 24 Å on the object scale . The data were analyzed by determination of the contrast transfer function using CTFFIND software [46] . The data were further processed with the SPIDER software package [47] . After automated particle picking followed by visual inspection , 1 . 1 million particles were selected for density reconstruction . The dataset was first sorted semi-supervised into ratcheted ( 350 , 000 particles; hybrid A/P- and P/E-t-RNAs ) and unratcheted ( 750 , 000 particles; A- , P- , and E-tRNAs ) sub-datasets [30] , using reconstructions of programmed and unprogrammed ribosomes as initial references , respectively ( Figure 2 ) . The unratcheted dataset of A- , P- , and E-tRNAs was further sorted into 544 , 000 particles of P-tRNA , 65 , 000 particles of A- and P-tRNA , and 40 , 000 particles of P- and E-tRNA using reconstructions of programmed and unprogrammed ribosomes as references . All sorting steps were performed at a pixel size of 2 . 44 Å/pixel , and reference volumes were filtered from 15 Å to 20 Å . Sorting processes were continued ( normally six to ten rounds of refinement ) unless the particle numbers in each sub-dataset reached a constant number , in which case the initial references were offered only in the first round . It is also noteworthy here that at no point was any ratcheted reference used for sorting , and therefore the ratcheted sub-dataset segregated itself from the non-ratcheted sub-dataset in an unsupervised fashion . This clearly indicates that the result of the sorting is indeed due to intrinsic characteristics of the particles and not an artifact due to reference bias . Densities for the 40S , 60S , and tRNAs were isolated using binary masks . Models were generated as described previously [5] , adjusted manually with Coot [48] , and minimized with VMD [49] . The CCA-Pro and CCA-Gly positions of the nascent chains were modeled based on an alignment with the Haloarcula marismortui 50S subunit in complex with CCA-pcb [39] , [50] . Initial docking of X-ray structures of ribosomal particles [8] , [11] , [12] , [51] and cryo-EM maps was performed using Chimera [52] , whereas alignment of pdbs utilized PyMol ( http://www . pymol . org ) . All figures were generated using Chimera [52] . The cryo-EM maps of the SecM-stalled RNC and SecM-Pro-RNC have been deposited in EMDataBank ( http://www . ebi . ac . uk/pdbe/emdb/ ) under accession numbers EMD-1829 and EMD-1830 , respectively .
|
In all cells , ribosomes perform the job of making proteins . As the proteins are synthesized they pass through a tunnel in the ribosome , and some growing proteins interact with the tunnel , leading to stalling of protein synthesis . Here , we used cryo-electron microscopy to determine the structure of a ribosome stalled during the translation of the Escherichia coli secretion monitor ( SecM ) polypeptide chain . The structure reveals the path of the SecM peptide through the tunnel as well as the sites of interaction with the tunnel components . Interestingly , the structure shows a shift in the position of the transfer RNA ( tRNA ) to which the growing SecM polypeptide chain is attached . Since peptide bond formation during protein synthesis requires precise placement of the substrates , namely , the peptidyl-tRNA and the incoming amino acyl-tRNA , it is proposed that this shift in the SecM-tRNA explains why peptide bond formation cannot occur and translation stalls .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Materials",
"and",
"Methods"
] |
[
"biophysics/macromolecular",
"assemblies",
"and",
"machines",
"biochemistry/rna",
"structure",
"molecular",
"biology/rna-protein",
"interactions",
"molecular",
"biology/translation",
"mechanisms",
"molecular",
"biology/translational",
"regulation",
"biophysics/rna",
"structure",
"biochemistry/transcription",
"and",
"translation",
"biophysics/transcription",
"and",
"translation",
"cell",
"biology/gene",
"expression"
] |
2011
|
SecM-Stalled Ribosomes Adopt an Altered Geometry at the Peptidyl Transferase Center
|
Drosophila telomeres are sequence-independent structures that are maintained by transposition to chromosome ends of three specialized retroelements ( HeT-A , TART and TAHRE; collectively designated as HTT ) rather than telomerase activity . Fly telomeres are protected by the terminin complex ( HOAP-HipHop-Moi-Ver ) that localizes and functions exclusively at telomeres and by non-terminin proteins that do not serve telomere-specific functions . Although all Drosophila telomeres terminate with HTT arrays and are capped by terminin , they differ in the type of subtelomeric chromatin; the Y , XR , and 4L HTT are juxtaposed to constitutive heterochromatin , while the XL , 2L , 2R , 3L and 3R HTT are linked to the TAS repetitive sequences; the 4R HTT is associated with a chromatin that has features common to both euchromatin and heterochromatin . Here we show that mutations in pendolino ( peo ) cause telomeric fusions ( TFs ) . The analysis of several peo mutant combinations showed that these TFs preferentially involve the Y , XR and 4th chromosome telomeres , a TF pattern never observed in the other 10 telomere-capping mutants so far characterized . peo encodes a non-terminin protein homologous to the E2 variant ubiquitin-conjugating enzymes . The Peo protein directly interacts with the terminin components , but peo mutations do not affect telomeric localization of HOAP , Moi , Ver and HP1a , suggesting that the peo-dependent telomere fusion phenotype is not due to loss of terminin from chromosome ends . peo mutants are also defective in DNA replication and PCNA recruitment . However , our results suggest that general defects in DNA replication are unable to induce TFs in Drosophila cells . We thus hypothesize that DNA replication in Peo-depleted cells results in specific fusigenic lesions concentrated in heterochromatin-associated telomeres . Alternatively , it is possible that Peo plays a dual function being independently required for DNA replication and telomere capping .
Telomeres are nucleoprotein complexes that counterbalance incomplete replication of terminal DNA and protect chromosome ends preventing both activation of cell cycle checkpoints and fusion events . In most organisms , the end replication problem is solved by telomerase , which mediates the addition of short GC-rich repeats to chromosome ends . These repeats specifically bind a discrete number of proteins , which recruit a series of additional factors to form large protein assemblies that ensure proper telomere function and homeostasis ( reviewed in [1–4] ) . Drosophila telomeres are not elongated by telomerase but by targeted transposition of three specialized retroelements called HeT-A , TART and TAHRE ( collectively abbreviated with HTT ) . However , Drosophila telomere formation does not require the HTT arrays; abundant evidence indicates that fly telomeres are epigenetically-determined structures that can assemble at the ends of the chromosomes independently of their terminal DNA sequence ( reviewed in [5–7] ) . With the exception of budding yeast , in organisms with telomerase telomeres are protected by the conserved shelterin complex . Human shelterin is a six-protein complex ( TRF1 , TRF2 , POT1 , TIN2 , TPP1 and Rap1 ) that specifically associates with the telomeric TTAGGG repeats . TRF1 , TRF2 directly bind the TTAGGG duplex and POT1 the single stranded overhang . TIN2 and TPP1 do not bind DNA and interconnect TRF1 and TRF2 with POT1 . TRF2 interacts with hRap1 , a distant homologue of S . cerevisiae Rap1 . The shelterin subunits share properties that distinguish them from the non-shelterin telomere-associated proteins: they are specifically enriched at telomeres throughout the cell cycle and appear to function only at telomeres [1] . The human non-shelterin proteins , which are not telomere-specific in localization and function , include the conserved CST complex , HP1a , and proteins involved in DNA repair and/or replication such as the ATM kinase , the Ku70/80 heterodimer , the MRE11/RAD50/NBS1 ( MRN ) complex , Rad51 , the ERCC1/XPF endonuclease , the Apollo exonuclease , the FEN1 nuclease , the RecQ family members WRN and BLM; the RTL1 helicase , RPA70 , the Timeless component of the replisome , and the subunits of the conserved ORC prereplication complex . Depletion of any one of the shelterin subunits or the shelterin-associated proteins leads to visible telomere defects ranging from altered packaging of telomeric chromatin ( multi telomere signals after FISH ) , telomere loss and telomere fusion ( reviewed in [1–4]; see also [8 , 9] ) . Most of the Drosophila telomere-capping proteins have been identified by molecular cloning of genes specified by mutations that cause telomeric fusions ( TFs ) in larval brain cells . Genetic and molecular analyses have thus far identified 11 loci that are required to prevent TF ( henceforth they will be designated as TF genes ) . These are effete ( eff; also called UbcD1 ) that encodes a highly conserved E2 enzyme that mediates protein ubiquitination [10 , 11] , Su ( var ) 205 that encodes Heterochromatin Protein 1a ( HP1a ) [12] , the Drosophila homologues of the ATM , RAD50 , MRE11 and NBS1 DNA repair genes [13–19] , without children ( woc ) that specifies a transcription factor [20]; caravaggio ( cav ) , modigliani ( moi ) , verrocchio ( ver ) and hiphop that encode the components of the terminin complex [21–25] . HOAP ( the cav product HP1/ORC-Associated Protein [21–26] ) , Moi and HipHop are fast evolving non-conserved proteins that do not share homology with any known telomere-associated protein [22–24]Ver is also a fast evolving protein; however , it contains an OB-fold motif that is structurally homologous to the OB fold of the Stn1 protein of the conserved CST complex [23] . Although the structural characterization of terminin is still incomplete , the extant data suggest that HOAP and HipHop are primarily bound to the DNA duplex while Ver is associated with the single-stranded overhang [6 , 22 , 23 , 26] . In contrast with the other telomere capping proteins ( Eff , HP1a , ATM , Mre11 , Rad50 , Nbs , and Woc ) that have multiple localizations and functions , HOAP , HipHop , Moi and Ver localize only at telomeres and appear to function only in telomere maintenance . These properties are similar to the shelterin properties , suggesting that terminin is a functional analog of shelterin [25] . Furthermore , the findings that shelterin subunits are not conserved in flies , that terminin components have no homologues ( with the possible exception of Ver ) outside Drosophilidae , and that terminin subunits are encoded by fast evolving genes have suggested a hypothesis on terminin evolution . We proposed that the transition between a telomerase-driven and a transposon-driven telomere elongation mechanism generated a divergence in terminal DNA sequences , which exerted a strong selective pressure towards the evolution of sequence-independent telomere binding proteins such as those that comprise terminin . We also hypothesized that non-terminin Drosophila telomere-capping proteins with multiple localizations and functions correspond to ancestral telomere components that did not evolve as rapidly as terminin because of the functional constraints imposed by their participation in diverse cellular processes [23–25] . In addition to their peculiar telomere elongation mechanism , Drosophila telomeres are also characterized by striking variations in their subtelomeric regions . Recent work has shown that subtelomeric regions play important regulatory roles in mammalian telomere behavior . For example , it has been reported that most human telomeres are replicated by forks progressing from subtelomere to telomere [27] and that the timing of telomere replication depends on the type of subtelomeric DNA; the telomeres associated with satellite-like subtelomeric sequences replicate later than telomeres that are not associated with this type of subtelomeric DNA [28] . Furthermore , recent work has shown that chimpanzee telomeres carrying subtelomeric heterochromatin replicate later than telomeres devoid of heterochromatic subtelomeres [29] . However , the fusigenic properties of mammalian telomeres carrying different subtelomeres have never been investigated . Drosophila is an ideal model organism for investigating the influence of subtelomeric regions on telomere behavior . All Drosophila telomeres terminate with HTT arrays that are capped by terminin; these HTT arrays are juxtaposed to different types of chromatin: canonical constitutive heterochromatin ( the Y , XR , and 4L telomeres ) , clusters of repetitive telomere-associated sequences ( designated as TAS , and present at the XL , 2L , 2R , 3L and 3R telomeres ) , or sequences with both euchromatic and heterochromatic features ( 4R telomeres ) . Here we describe a Drosophila gene , pendolino ( peo ) , identified by mutations that preferentially induce TFs between telomeres associated with constitutive heterochromatin . The Peo protein binds terminin but does not have the typical terminin properties , as it is conserved in mammals and associates with several chromosomal sites . In addition , Peo is required for PCNA recruitment and for general DNA replication . However , both the present and previous results strongly suggest that telomere lesions generated by general defects in DNA replication are unable to induce TFs in Drosophila cells . We thus propose that loss of peo function results in specific fusigenic lesions concentrated in heterochromatin-associated telomeres , and that these lesions might be generated during telomere replication .
The pendolino1 ( peo1 ) mutation was isolated by a cytological screen of 120 late lethal mutants mapping to the second chromosome , recovered after I element mobilization by I-R dysgenic crosses ( see Materials and Methods ) . Mitotic cells of DAPI-stained brain preparations from peo1/peo1 larvae displayed very frequent telomeric fusions ( TFs; Fig 1A and 1B ) , often resulting in multicentric linear chromosomes that resemble little “trains” of chromosomes . The pendolino gene was named after this phenotype just as caravaggio , modigliani and verrocchio , which are all names of Italian trains . Recombination analysis with visible markers and deficiency mapping placed peo1 in the 46B-C polytene chromosome interval uncovered by Df ( 2R ) X3 and Df ( 2R ) B5 but not by Df ( 2R ) X1 ( Fig 1C ) . Previous studies mapped to this interval 3 lethal complementation groups [30] . peo1 failed to complement the 1527 and 2723 mutant alleles ( henceforth designated as peo1527 and peo2723 ) of group III for both the lethality and the TF phenotype but complemented representative alleles of groups I and II ( Fig 1D ) . peo1 also failed to complement the P element insertion p112 ( henceforth peop112 ) and the small p221 and p520 deficiencies ( henceforth peo∆221 and peo∆520 ) all generated by the remobilization of the P{w+ , ry+}AJN2 insertion [31] , originally localized proximally to the Df ( 2R ) X3 breakpoint ( Fig 1C and 1D ) . Finally , we identified another peo mutant allele ( peoh ) by a cytological screen of a collection of 193 late lethal mutants that arose in the Zucker’s collection of heavily mutagenized viable lines ( see Materials and Methods for details ) . peoh is homozygous viable but male and female sterile , and it is lethal in combination with peo1 ( peo1/peoh ) ; the lethal phases of selected combinations of peo mutant alleles and deficiencies are reported in Fig 1D . To identify the peo gene at the molecular level we exploited the peop112 allele that carries a P{w+ , ry+}construct inserted into the gene [30] . Using inverse PCR we found that the P construct is inserted into the 5’ UTR of the longest transcript of the CG10536 gene ( Fig 2 ) . CG10536 was originally named ms ( 2 ) 46C [32] an then renamed crossbronx ( cbx ) [33] . However , CG10536 does not correspond to ms ( 2 ) 46C/cbx . The phenotype associated with the ms ( 2 ) 46C/cbx mutation was attributed to the P{PZ}05704 insertion that maps just proximal to the UTR region of CG10536 ( Fig 2 ) . However , this attribution was only tentative because the male sterile phenotype was neither mapped over deficiency nor reverted by P element excision [32] . We found that males homozygous for the P{PZ}05704 insertion are sterile and show the spermatid abnormalities previously described [32] . In contrast , males bearing the same insertion over Df ( 2R ) B5 that uncover the 46B-C interval were fully fertile and displayed normal spermatids ( Fig 1D ) . We thus conclude that the ms ( 2 ) 46C/cbx mutation maps outside the 46B-C region , and that CG10536 actually corresponds to pendolino . The organization of the peo locus is rather complex . The gene encodes three different transcripts; the long introns of two of these transcripts contain genes ( Ntmt and CG18446 ) with an opposite transcriptional orientation to peo/CG10536 ( FlyBase; Fig 2 ) . The three putative peo transcripts encode proteins of 183 , 214 and 244 aa and are identified by 2 cDNAs ( FlyBase , Fig 2 ) . These proteins share homology with the E2 variant ubiquitin-conjugating enzymes , which are devoid of the catalytic cysteine that mediates ubiquitin transfer [34] . Interestingly , peo has an intronless paralogue ( CG16894 ) that maps to the 56F9 polytene chromosome region . The CG16894 protein shares 35 . 4 identity and 51 . 6 similarity with Peo and its function is currently unknown ( FlyBase ) . Sequencing of peo1 did not revealed nonsense , frameshift , splice-site or missense mutations in the protein-coding sequences of CG10536 with respect to the FlyBase sequences . In addition , sequencing of approximately 2000 bp upstream of the gene ATG and in situ hybridization experiments did not reveal I element insertions . Nonsense , frameshift or splice-site mutations were also absent from the protein coding sequences of peo1527 and peoh mutant genes . All peo mutant alleles displayed several genomic single nucleotide polymorphisms ( SNPs ) with respect to the Fly Base sequence . However these SNPs were all present in DPGP natural populations ( www . dpgp . org/dpgp3 ) , suggesting that they are associated with little if any phenotypic consequences . Collectively , these results suggest that peo1 , peo1527 and peoh are regulatory mutations that lower the intracellular amount of the Peo protein ( see below ) . However the molecular lesions in these mutant alleles remain to be identified . To unambiguously determine the identity of peo we performed rescue experiments using the LD08052 cDNA clone ( Fig 2; see also FlyBase ) . Sequencing confirmed that this cDNA contains the entire coding sequence of the CG10536-RB transcript , which produces the larger protein isoform encoded by the locus ( Fig 2 ) . The LD08052 cDNA was fused in frame with the heat shock ( hsp70 ) promoter or used to make a construct containing the tubulin promoter and a 3HA sequence at the 3’ of the gene . Both constructs were then used to make transgenic flies and both rescued the lethality and the TF phenotype of the peo1 mutant flies . In hs-CG10536; peo1/peo1 flies exposed to heat shocks ( 1h at 37°C every 24 h throughout development ) , survival of peo1/peo1 individuals with respect their peo1/CyO siblings was 18% of the Mendelian expectation , and the TF frequency dropped to less than 1 per cell from more than 5 per cell ( Fig 1B ) . In the presence of the Tub-CG10536-3HA construct , the survival rate of peo1/peo1 homozygotes with respect to peo1/CyO heterozygotes was 10% and the TF frequency less than 1/cell ( Fig 1B ) . To confirm these rescue data we generated peoh/peoh larvae bearing the Tub-CG10536-3HA construct . The brains of these larvae displayed a 5-fold reduction in TF frequency compared to those of peoh/peoh larvae ( 0 . 2 vs 1 TF/ cell ) . Thus , our results collectively indicate that CG10536 corresponds to peo . In most colchicine-treated metaphases from peo1/peo1 , peo1/ peo1527 , peo1/peo2723 , peo1/peop112 , peo1/ peo∆221 , peo1/peo∆520 and peo1/Df ( 2R ) BSC298 ( henceforth Df ( 2R ) BSC298 will be abbreviated with Df ) , the majority of telomeres were involved in fusions , often forming tangles of chromosomes difficult to resolve ( Fig 1A and 1B ) . However , a careful examination of these tangles led us to estimate the average number of TFs per cell and to determine the relative frequencies of single telomere associations ( STAs ) and double telomere associations ( DTAs ) . STAs involve a single telomere that fuses with either its sister telomere or another single nonsister telomere . In DTAs , two sister telomeres fuse with another pair of sister telomeres . STAs are likely to be generated during the S-G2 phase , while DTAs are thought to result from the replication of TFs generated during G1 [10 , 35] . In peo mutants , DTAs were much more frequent than STAs ( Fig 1B ) just as in the other TF mutants in which the relative frequencies of STAs and DTAs have been determined , namely eff , Su ( var ) 205 , cav , mre11 , rad50 , nbs , tefu ( ATM ) , woc , moi and ver [10 , 12 , 15 , 16 , 21 , 20 , 23 , 24] . This bias towards DTAs may reflect the proximity of Drosophila telomeres during G1 that would be progressively lost as cells proceed through S and G2 [10 , 36] . It has been also suggested that telomere fusion occurs primarily during G1 , because the DNA repair pathways that join chromosome ends are more active in G1 than in S or G2 [37] . Although peo mutants show a high DTA/STA ratio as the other TF mutants , they exhibit a specific pattern of TFs . The analysis of the peoh mutant that shows ~1 TF/cell allowed a very precise definition of the telomeres involved in fusion events . In peoh homozygous brains of males and females , nearly all TFs involved the telomeres associated with the heterochomatic regions of the chromosomes , namely those of the entirely heterochromatic Y chromosome ( YS and YL ) , the telomere of the right arm of the X chromosome ( XR ) and the fourth chromosome telomeres ( Fig 1A ) . Of the two telomeres of the fourth chromosome only one is associated with constitutive heterochromatin ( 4L ) but we were not able to distinguish between 4L and 4R , as the DAPI-stained fourth chromosomes appear as brightly fluorescent dots in which the chromosome arms are not discernible . High frequencies of TFs between heterochromatin-associated telomeres ( henceforth abbreviated with Ha-telomeres ) were also observed in peoh/Df and peoh/peo1 brains ( Fig 3 ) . We note that we classified as DTAs all TFs between Ha-telomeres , because the close apposition of the sister chromatids in heterochromatin does not allow a distinction between STAs and DTAs . Thus , the apparent lack of STAs observed in weak peo mutants ( Fig 1B ) might not reflect a real absence of this type of TFs . The high incidence of TFs between Ha-telomeres is not due to an allele-specific effect of peoh , as a high frequency of TFs involving the Ha-telomeres was also observed in peo1/peo1 mutant cells bearing the Tub-peo+-3HA rescue construct ( Fig 3 ) . This suggests that the Ha-telomeres are preferentially affected by an impairment of peo function and that this effect is partially masked in strong mutants in which most telomeres are fused . The TF pattern observed in peo mutants is highly specific , as none of the TF mutants we characterized in the past showed a prevalence of TFs between the Ha-telomeres . To precisely compare the patterns of fusions we re-examined all extant Drosophila TF mutants ( eff , cav , Su ( var ) 205 , mre11 , rad50 , nbs , tefu , woc , moi and ver ) . All these mutants displayed an inverse TF pattern compared with peo mutants , with lower than expected frequencies of fusions involving the Y , the XR , and the 4th chromosome telomeres and higher than expected frequencies of TFs between the telomeres of the major autosomes ( Fig 3 and S1 Table ) . We also confirmed that eff mutants do not exhibit a prevalence of fusion between Ha-telomeres ( Fig 3 ) [10] . This finding strongly suggests that the canonical E2 ubiquitin conjugating enzyme encoded by eff and the E2 variant encoded by peo do not function in the same telomere protection pathway . In interphase nuclei , the heterochromatic regions of the chromosomes aggregate to form an irregular mass of chromatin called chromocenter [38] . Thus , the specific pattern of TFs observed in peo mutants could depend either on an abnormal organization and compaction of the chromocenter or on a specific chromatin composition of heterochromatic telomeres . To discriminate between these possibilities we introduced in a peoh mutant background a Bs w+ y+ Y chromosome that carries a euchromatic fragment marked with Bs w+ and y+ at its YL end [39] ( see also FlyBase ) . We found that in peoh mutant brains the frequency with which the Bs w+ y+ Y chromosome is involved in TFs is more than 5-fold lower than that of a normal Y ( Fig 4 ) . Specifically , the formation of Y rings , which in peoh/peoh mutants represents ~90% of the total fusions involving the Y ( that together are more than 40% of the observed TFs ) , was drastically reduced in the presence of a Bs w+ y+ Y . Because the cells bearing a Bs w+ y+ Y chromosome do not exhibit detectable morphological variation in the chromocenter compared to wild type , these results strongly suggest that the preferential involvement of the Ha-telomeres in peo-induced TFs is a consequence of their association with heterochromatin and not of their arrangement within the interphase nucleus . The particularly high frequency of Y rings in peo mutants is another peculiar feature of their TF pattern . Indeed , the Y rings are rare or virtually absent in mutants in genes such as ver , tefu , woc , and nbs ( Fig 4 ) . On the assumption of a random involvement of telomeres in TFs , the expected frequency of Y ring is 1/15 of the total fusions involving the Y chromosome and 1/6 of the fusions involving the Y and either the XR , or the 4th chromosome telomeres . Thus , formation of Y rings in peo mutants is specifically and strikingly frequent . This could reflect a particularly high frequency of fusigenic lesions on the Y telomeres , or the contemporary presence of these lesions on the opposite Y telomeres , or both . We have recently observed that brains from larvae heterozygous for both Su ( var ) 20505 and cav exhibit ~ 0 . 1 TFs/cells , while in wild type , Su ( var ) 20505/+ , cav/+ brains the TF frequency was virtually zero ( 300 metaphases analyzed in each case ) . This finding suggests that Su ( var ) 205 and cav genetically interact and that a simultaneous reduction of HP1a and HOAP results in a low level of TFs . We thus asked whether peo exhibits similar genetic interactions with other TF mutants . Double heterozygotes for peo1 and either cav1 , moi1 , ver2 , Su ( var ) 20504 or Su ( var ) 20505 displayed TF frequencies ranging from 0 . 05 to 0 . 1 per cell , whereas heterozygotes for peo1 and either effΔ73 , wocB111 , mre11DC , rad50Δ5 . 1 , nbs1 or tefuatm6 did not exhibit TFs ( in all cases , we examined at least 300 metaphases ) . We next asked whether the Peo protein physically interacts with the terminin components . A preliminary experiment using the yeast two-hybrid assay suggested that Peo directly interacts with HOAP but not with HP1a ( S1 Fig ) . To confirm this result we performed a GST pulldown assay using bacterially expressed 6His-Peo and GST-tagged HOAP polypeptides of different length . As shown in Fig 5A and 5B , the intact HOAP protein and the HOAP fragments containing the N-terminal region of the protein ( aa 1–145 ) precipitated Peo , while a larger HOAP fragments including the 3 repeated segments of the protein ( aa 109–343 ) failed to bind Peo . We then investigated whether Peo interacts with Moi and Ver . We performed GST pulldown experiments with extracts from human 293T cells expressing Peo-FLAG and either GST alone , GST-Moi , GST-Ver or GST-HOAP . We chose to express Drosophila tagged proteins in human cells because a heterogeneous cellular environment is likely to reduce the probability of indirect interactions among fly proteins . As shown in Fig 5C , Peo-FLAG is precipitated by GST-Moi , GST-Ver and GST HOAP but not GST alone . Collectively , these results provide strong evidence that Peo directly binds HOAP; in addition , they suggest that Peo directly interacts with Moi and Ver . peo encodes an E2 variant ( UEV ) enzyme; the UEV proteins are similar to the E2 ubiquitin conjugating enzymes ( UBCs ) but lack the catalytic cysteine residue that mediates the interaction between ubiquitin and E2 [40] . We elaborated a three-dimensional model of Peo exploiting a series of bionformatic analyses ( Fig 6A; see also Material and Methods and S2 Fig ) . We confirmed that Peo lacks the catalytic cysteine of E2 enzymes . In addition , 8 residues before the catalytic cysteine site , Peo exhibits an HPH tripeptide ( S2 Fig ) instead of HPN , which is a canonical signature of the E2 superfamily [41] . Peo contains a UEV domain of ~150 amino acids resembling the canonical E2 fold in its hydrophobic core and active site region . This domain consists of three helices packed against a four-stranded antiparallel β-sheet . Next to the UEV domain , Peo contains two C-terminal helices that are present in all E2 proteins but missing in other E2 variant enzymes such as Tsg101 and Mms2 . Finally , prediction of potentially disordered regions revealed that Peo also contains a long ( 50 aa ) disordered region at the C-terminus ( Figs 6A and S2; see also Materials and Methods ) . To map the Peo sites that interact with the terminin proteins , we subdivided Peo into 3 GST-tagged fragments: Peo 1 ( aa 1–31 ) that contains the Peo N terminal region which is absent in the short isoform ( see Fig 2 ) ; Peo 2 ( aa 31–180 ) that includes the central UEV domain of the protein; and Peo 3 ( aa180-244 ) that contains C-terminal disordered region of Peo ( Fig 6B ) . These bacterially expressed GST-tagged Peo fragments were then used to probe Drosophila S2 cell extracts expressing HOAP-HA , Moi-HA or Ver-FLAG ( Fig 6C–6E ) . GST pulldown showed that both Moi and Ver interact with the entire Peo protein ( GST-Peo ) , GST-Peo 1 and GST-Peo 2 but not with GST-Peo 3 or GST alone ( Fig 6D and 6E ) . HOAP did not display the same interaction pattern as Moi and Ver . It was precipitated by GST-Peo and GST-Peo 1 but not by GST-Peo 2 and GST-Peo 3 and thus failed to interact with the Peo UEV domain ( Fig 6C ) . The finding that Peo binds HOAP , Moi and Ver prompted us to ask whether Peo is required for terminin localization at chromosome ends . Although Peo does not appear to interact with HP1a , we also asked whether loss of the wild type function of peo affects HP1a localization at telomeres . Of the latter proteins , only HOAP is clearly detectable at both mitotic and polytene chromosome telomeres; HP1a , Moi and Ver can be easily detected at polytene chromosome ends but not at mitotic telomeres [23 , 24 , 42] . We thus analyzed HOAP localization in both mitotic and polytene chromosomes of peo1 homozygous mutants; HP1a , Moi and Ver localization was instead studied only in peo1 polytene chromosomes . An analysis of mitotic chromosomes immunostained for HOAP revealed that mutations in peo do not substantially affect HOAP localization at telomeres , as the frequency of HOAP-stained telomeres in peo mutants and the intensity of the signals were similar to those observed in wild type controls ( Fig 7A ) . Consistent with these results , immunostaining with anti-HOAP and ant-HP1a antibodies showed that peo1/Df mutants exhibit normal concentrations of these proteins at polytene chromosome telomeres ( Fig 7B ) . Because antibodies to Moi or Ver are not currently available , to analyze the localization of these proteins in peo mutants we constructed flies expressing GFP-tagged forms of Moi or Ver in a peo1 mutant background . The analysis of unfixed polytene chromosome nuclei from peo1/peo1 flies expressing either GFP-Moi or Ver-GFP revealed that they exhibit 6 discrete GFP signals ( Fig 7C ) , which we have previously shown to correspond to the 6 euchromatic telomeres of polytene chromosomes [23 , 24] . In addition , we found that the intensities of these signals were comparable to those observed in peo1/+ heterozygotes ( Fig 7C ) . Collectively , these results indicate that the wild type function of peo is not required for telomeric localization of HOAP , Moi , Ver and HP1a , and that the strong telomere fusion phenotype observed in peo mutants is not due to the absence of any of these proteins . As pointed out in the introduction , terminin subunits are non-conserved fast-evolving proteins that localize and function only at telomeres . Peo is not a terminin-like protein because it does not share any of these properties . Peo is well conserved in Drosophila species ( S3 Fig ) and has homologues in mouse and humans ( Ft1 and AKTIP , respectively ) . In addition , a three-dimensional model of Drosophila Peo is rather similar to an AKTIP model [43] . To determine the subcellular localization of Peo we raised a rabbit polyclonal antibody against the entirety of Peo and affinity purified it against a GST-Peo fusion protein ( see Material and Methods ) . Western blotting analysis showed that this antibody recognizes 3 bands of ~ 32 , ~28 and ~ 25 kDa . In extracts from peo1/Df , and peo1527/Df larval brains , these 3 bands were reduced by approximately 70% compared to +/Df controls ( Fig 8A and 8B ) . Thus , the band with the highest molecular weight is likely to correspond to the longest Peo isoform , while the other two bands might correspond to the shorter isoforms ( see Fig 2 ) . In peoh/Df mutant brains , the Peo bands were reduced by approximately 20% with respect to +/Df brains , consistent with the relatively low TF frequency observed in peoh mutants ( Fig 1 ) . These results indicate that peo1 and peo1527 are strong hypomorphs compared to the weaker peoh allele . Indirect immunofluorescence experiments on wild type polytene nuclei showed that the anti-Peo antibody stains the nucleolus and many bands along the polytene chromosomes ( Fig 8C and 8D ) . In peoh/Df and peo1/Df nuclei , both the nucleolus and chromosome staining were significantly reduced compared to either +/Df or wild type nuclei , confirming the specificity of the antibody . In addition , the polytene chromosomes of peoh/Df larvae were more intensely stained than those of peo1/Df larvae ( Fig 8C and 8D ) , confirming that the peo1 allele is stronger than peoh . Although Peo is enriched at numerous polytene bands and interbands , we were not able to detect clear Peo accumulations at chromosome ends . Given that Peo interacts with terminin and it is required to prevent telomere fusion , the most likely explanation for this finding is that Peo is present at the telomere caps in amounts that are not detected by the antibody and the immunostaining technique used here . We have recently found that AKTIP/Ft1 is required for proper telomeric DNA replication [43] . This finding suggested that peo mutations could also impair DNA replication and specifically affect heterochromatic telomeres , which are likely to replicate at the end of the S phase together with the bulk of heterochromatin ( reviewed in [38] ) . To test this possibility we examined DNA replication in brain cells by analyzing the incorporation of the EdU ( 5-ethynyl-2’-deoxyuridine ) analog of thymidine . Brains were incubated in saline containing 10 mM EdU for 1 h , immediately fixed and then stained with the Click-It Alexa Fluor method to detect EdU ( see Methods ) . In wild type brains , 10% of the nuclei were actively replicating their DNA and incorporated EdU , whereas in peoh/peoh and peo1/peo1 brains the frequency of EdU-labeled nuclei dropped to 7% and 5 . 5% , respectively ( Fig 9A and 9B ) . In contrast , in woc and ver mutants , the frequency of EdU-positive nuclei was not significantly different from wild type controls , suggesting that the DNA replication defect observed in peo mutants is a specific outcome of the reduced peo activity and not a general consequence of impaired telomere protection . EdU staining also allowed subdivision of the S phase according to the incorporation pattern . We distinguished nuclei in early/mid S ( S1 ) in which the nucleus was partially or completely stained with the exception of the heterochromatic chromocenter , nuclei in mid/late S ( S2 ) in which both the chromocenter and the less compact nuclear areas were stained , and nuclei in late S ( S3 ) in which only the chromocenter displayed EdU incorporation ( Fig 9 ) . We found that wild type and peo mutant brains do not differ in the relative frequencies of nuclei showing S1 , S2 and S3 EdU incorporation patterns . Thus , we conclude that the wild type function of peo is required for general DNA synthesis and not for completion of specific sub-phases of DNA replication . To gather additional information on the role of Peo in DNA replication we analyzed the distribution of PCNA ( proliferating cell nuclear antigen ) in peo1/peo1 and peoh/peoh mutant nuclei . PCNA is a processivity factor for DNA polymerases; in nuclei PCNA is either present in a soluble form that can be extracted by detergent treatment or in a detergent-resistant form tightly associated with DNA replication forks ( chromatin-bound PCNA ) [44] . The analysis of Triton X-extracted brain preparations immunostained for PCNA showed that in wild type 7% of the nuclei were PCNA-positive , while in peo1/peo∆520 and peoh/peoh brains the frequencies of PCNA-stained nuclei were 1 . 2% and 0 . 8% , respectively ( Fig 10 ) . These results are consistent with those on EdU incorporation and show that in peo mutants the frequency of nuclei with chromatin-bound PCNA is much lower than in control . Collectively our results suggest three possibilities: ( i ) that any impairment of Drosophila telomere replication results in telomere fusion , ( ii ) that Peo is required for a specific step of telomere replication and that an incorrect execution of this steps results in fusigenic lesions , or ( iii ) that Peo plays a dual function being independently required for telomere replication and telomere capping . To discriminate between these possibilities we treated wild type and peoh/peoh mutant brains with the DNA polymerase inhibitor aphidicolin ( APH ) , which is known to impair human telomere replication [45–47] . As shown in Fig 11 , control and mutant brains were treated in three different ways . Dissected brains were incubated for 1 . 5 hours in 110 mM APH in NaCl 0 . 7% . They were then washed , transferred into 3 ml of APH-free saline and fixed 1 , 2 or 3 hours after the end of the APH treatment; in all cases , 1 hour before fixation , we added colchicine to the saline to collect metaphases . None of the APH treatments caused TFs in wild type brains or increased the TF frequency in peoh/peoh mutants . We note that the APH treatment was highly effective as it caused a strong reduction in the frequency of mitoses in brains fixed 1 h after the end of the APH treatment ( Fig 11B ) . Our previous work has shown that mutations in Drosophila timeless2 ( tim2 ) result in chromosome breakage but not in TFs [48] . Tim2 is the Drosophila ortholog of mammalian TIM , a replisome component that facilitates DNA synthesis [49 , 50] and is required for telomere replication [8] . These results prompted us to examine the cytological phenotype of brains from mutants in the Blm ( formerly mus-309 ) gene [51] , the Drosophila homolog of the Bloom syndrome gene , which encodes a DNA helicase required for mammalian telomere replication [46 , 47] . We found that BlmD2/BlmD3 mutant brains exhibit chromosome breaks ranging from 4 to 7% but not TFs ( 500 metaphases scored from 7 brains ) . These results suggest that a general impairment of Drosophila telomere replication does not result in TFs .
We have shown that strong peo mutants exhibit an average of 5 TFs/cell; these TFs appear to involve all the telomeres of the Drosophila chromosome complement making difficult a reliable assessment of the relative involvement of individual telomeres in fusion events . However , in weak peo mutants or in strong mutants bearing a peo+ rescue construct , which exhibit ~ 1 TF /cell , the majority of fusions involved the XR , the Y and the 4th chromosome telomeres . These results suggest that these telomeres are preferentially affected by mutations in peo and that this effect is partially masked in strong mutants where most telomeres are fused . In all the other Drosophila mutants we characterized ( eff , Su ( var ) 205 , cav , mre11 , rad50 , nbs , tefu , woc , moi and ver ) individual telomeres were engaged in TFs with frequencies that are different from those expected for a random involvement . The telomeres of the major autosomes were involved in TFs more frequently than expected , while participation of the XL telomere in fusions was either slightly lower or conformed to the expected frequency . In contrast , the Y , the XR , and the 4th chromosome telomeres were engaged in TF less than expected ( Fig 3 and S1 Table ) . peo mutants displayed an inverse TF pattern , with higher than expected frequencies of fusions involving the Y , the XR , and the 4th chromosome telomeres ( Fig 3 and S1 Table ) . To the best of our knowledge , this is the first case in which individual telomeres of an organism exhibit different fusigenic capabilities in response to the genetic background . We believe that this finding reflects the peculiar structural differences between the telomeric regions of the different Drosophila chromosomes . All Drosophila chromosomes of wild type strains terminate with HTT arrays of variable length ( made of complete and incomplete HeT-A , TART and TAHRE elements ) ( reviewed in [5 , 52 , 53] ) and are capped by the multiprotein terminin complex ( reviewed in [6] ) . However Drosophila telomeres differ from each other in both the type of subtelomeric chromatin and in the properties of their HTT arrays ( Fig 12 ) . One of the most straightforward features that contradistinguish some of the Drosophila telomeres is their association with constitutive hetrochromatin . Approximately one-third of the Drosophila genome is made of constitutive heterochromatin; the entire Y chromosome , the short arm ( XR ) and the proximal 40% of the long arm ( XL ) of the X chromosome , the short arm ( 4L ) and the proximal 70% of the long arm ( 4R ) of the 4th chromosome , and the centric 25% of chromosomes 2 and 3 are heterochromatic [38 , 54] . Thus the HTT arrays of YL , YS , XR and 4L are linked to constitutive heterochromatin ( Fig 12 ) . The HHT blocks of XL and those of the major autosomes are not directly associated with euchromatin but are instead juxtaposed to divergent clusters of subtelomeric repeats , known as telomere associated sequences ( TAS ) ( reviewed in [55] ) . The TAS are not only different in sequence but are also occasionally absent from the subtelomeric regions , suggesting that their presence is not essential for proper telomere function [55] . Finally , the 4R telomere is joined to a special type of chromatin that has peculiar features , as well as features shared with both euchromatin and heterochromatin; for example the 4R distal chromatin is enriched in the 4th chromosome-specific Painting of four ( Pof ) protein and the heterochromatic markers HP1a and histone 3 methylated at lysine 9 ( H3K9 ) ( reviewed in [56] ) ( Fig 12 ) . Additional features that differentiate Drosophila telomeres are the silencing properties of their HTT arrays ( Fig 12 ) . Several studies have shown that white+ transgenes inserted within or next to the TAS are partially silenced leading to a variegated eye phenotype ( a phenomenon known as telomere position effect or TPE ) ( reviewed in [5] ) . However , white+ transgenes inserted into the HTT arrays behave differently depending on the insertion site . Insertions into the HTTs of telomeres not directly joined to heterochromatin such as those of the 2L , 3L and 3R arms do not appear to be subject to TPE [57] . In contrast white+ transgenes inserted within the 4R ( only one tested ) and YS ( 6 tested ) HTTs lead to a variegated eye phenotype [57 , 58] . Furthermore , the white+ transgenes inserted into the HTT array of YS and those embedded into or near the TAS respond differently to genetic modifiers . For example , mutations in Su ( var ) 205 ( HP1a ) , which suppress variegation of genes relocated next to constitutive heterochromatin ( position effect variegation or PEV; reviewed in [59 , 60] ) do not affect the TAS-associated TPE [5 , 55] but strongly suppress the TPE of transgenes inserted into the HTT of YS ( 58 ) . These results indicate that the chromatin of the YS HTT shares some features with constitutive heterochromatin and has different properties from the chromatin that includes the autosomal HTT arrays and the TAS . These results suggest that the HTT arrays of XR , YL , and 4th chromosome telomeres have properties similar to those of YS , namely they are subject to a “heterochromatinization” process related to their particular location . Numerous studies on PEV have shown that proteins that are typically enriched in the heterochromatin can spread into neighboring euchromatic genes changing their chromatin composition and packaging and downregulating their expression [59 , 60] . Thus , although the properties of the HTT arrays of YL , XR and 4th chromosome have not been tested , it is quite likely that they are similar to those of YS . This would suggest that in a peo mutant background the preferential involvement of these telomeres in TF is a consequence of their “heterochromatinization” . This in turn implies that the heterochromatic markers of the HTT regions of the YL , XR and 4h chromosome telomeres extend to the terminin-coated chromosome ends . However , “heterochromatinization” does not extend to the HTT arrays at the end of the left arm end of the BSw+y+Y chromosome , because in peo mutants this marked Y forms much fewer Y rings than a normal Y . Our findings on ring Y formation in peo mutants also suggest that the different “heterochromatic” telomeres respond differently to Peo depletion . The higher than expected frequency of ring Ys observed in peo mutants could be a consequence of the particularly high fusigenic properties of the Y telomeres . We would like to note that the “heterochromatinization” of the YL , YS , XR and 4th chromosome termini is the natural condition of these telomeres and that we only know that this condition makes them more fusigenic than the other Drosophila telomeres . However , we have no information on the actual properties of these telomeres . Namely , we do not know the properties of their terminin-associated regions . For example , we do not know whether the terminin-coated chromosome ends of YL , YS , XR and 4L have different silencing properties and different responses to PEV and TPE modifiers compared to the terminin-bound regions of the other chromosome ends . We have shown that mutations in peo genetically interact with mutation in genes that encode the terminin subunits . Consistent with these results , we have also shown that Peo directly binds terminin and mapped the Peo-HOAP interacting domains . Thus , although Peo does not share the properties of the terminin proteins , it is clearly a component of the Drosophila telomere capping machinery . What is then the role of peo in telomere protection ? The finding that mutations in peo do not affect HOAP , Moi and Ver localization at telomeres strongly suggest that loss of peo function does not cause TFs by affecting terminin recruitment at telomeres . Similarly , the normal accumulation of HP1a at the telomeres of peo mutants suggests a telomere fusion mechanism independent of HP1a . More in general , the fact that peo , but not the other TF genes ( eff , Su ( var ) 205 , cav , mre11 , rad50 , nbs , woc , moi and ver ) , preferentially affect “heterochromatic” telomeres suggests that peo might either act upstream to these genes or function in a telomere protection pathway that does not involve them . The findings that mutations in peo impair DNA replication and preferentially affect late replicating “heterochromatic” telomeres raise the possibility that defective telomere duplication might be fusigenic in Drosophila . However , there are several reasons that lead us to exclude this possibility . Should it be correct , one would expect that impairment of some of the many factors that mediate DNA replication would results in TFs . However , in addition to the aphidicolin treatment and Blm mutations described here , several additional DNA replication factors have been described whose loss fails to induce TFs . Hydroxyurea ( HU ) , which blocks DNA replication by inducing a deoxyribonucleotide triphosphate ( dNTP ) pool depletion , does not induce TFs in brain cells [48] . In addition , a number of mutations ( or RNAi treatments ) disrupting different aspect of DNA replication do not results in TFs in Drosophila brain cells or S2 tissue culture cells , although they cause more or less extensive chromosome breakage . These include lesions in the genes/RNAs encoding the origin recognition ( ORC ) and the minichromosome maintenance ( MCM ) prereplication complexes , DNA primase , the Cul4 replication licensing factor , the replisome components DNA polymerase alpha , Rpa70 , Tim2 and PCNA , as well as the chromatin assembly factor Caf1 that assists in loading the histone tetramer after DNA replication [21 , 48 , 61–64] . We thus hypothesize that the peo-dependent TFs are generated by a peculiar defect in telomeric DNA replication that creates specific fusigenic lesions . Heterochromatin replication is likely to be different from that of euchromatin , as it requires not only DNA duplication but also reinstallment of specific epigenetic markers such as heterochromatin-associated proteins and histone modifications ( reviewed in [65] ) . It is thus possible that in the presence of weak peo mutations replication of “heterochromatic” telomeres is preferentially affected leading to specific Peo-dependent fusigenic lesions concentrated in the XR , the Y and the 4th chromosome ends . In strong peo mutants , these fusigenic lesions would be generated also in “euchromatic” telomeres resulting into a more general involvement of telomeres in fusion events . However , we cannot exclude the possibility that peo plays a dual function being independently required for DNA replication and telomere capping . In conclusion , we propose that Peo is a Drosophila telomere-capping protein that preferentially protects chromosome ends associated with heterochromatic markers . Our results also indicate that Peo is required for general DNA replication and most likely also for telomere replication . However , while it is possible that loss of Peo generates specific fusigenic lesions during telomere replication , it is unlikely that a general impairment of Drosophila telomere duplication leads to telomere fusion . In this respect Drosophila telomeres are similar to human telomeres , which fail to fuse following defective replication [8 , 46 , 47 , 66] .
The pendolino1 ( peo1 ) mutation was isolated by a cytological screen of 120 late lethal mutants mapping to the second chromosome , recovered after I element mobilization by I-R dysgenic crosses [67] . The late lethal mutations l ( 2 ) 1527 , l ( 2 ) 2723 , the insertion lines p112 , p221 and p520 were described previously [30] and were kindly provided by P . Taghert ( Washington University , MO ) . The peoh allele was isolated from a cytological screen of a collection of 193 late lethal mutants that arose in the Zucker’s collection of heavily mutagenized viable lines [68] . The Df ( 2R ) X3 , Df ( 2R ) B5 , Df ( 2R ) X1 and Df ( 2R ) BSC298 deletions , the insertion line cbx05704 , were obtained from the Bloomington Stock Center and are described in FlyBase . All peo alleles and the deficiencies of 2R were balanced over CyO Tb balancer [69] . The eff , Su ( var ) 205 , cav , mre11 , rad50 , nbs , woc , moi and ver mutations were previously described [10 , 12 , 15 , 16 , 21 , 23 , 24] . The DNA sequences flanking the peop112 insertion were obtained by inverse PCR using standard procedures . To obtain the rescue constructs , the LD08052 full lenght peo cDNA was cloned into pCASPER4 transformation vectors . Germline transformation was carried out by the BestGene Company . A pCasper4 [w+ , peo+] insertion ( with a Hsp70 promoter ) on the X chromosome was used to establish the [w+ , peo+]; peo1/CyO Tb stock . Animals from this stock were heat-shocked for 1h at 37°C every day starting from the embryonic stages; we then examined the brains of non-Tb third instar larvae for the presence and frequency of TFs and the adult flies for the presence of w+ , peo+; peo1/peo1 individuals . For the rescue experiments we also employed a pCASPER4 [w+ , peo+-HA] insertion ( with a Tubulin promoter ) on the third chromosomes . We established [w+ , peo+-HA]/TM6C; peo1/CyO Tb and [w+ , peo+-HA]/TM6C; peoh/CyO Tb stocks and examined them for TFs in non-Tb larvae . The [w+ , peo+-HA]/TM6C; peo1/CyO Tb stock was also examined for the presence of peo1/peo1 non Cy flies . Information on the genetic markers and balancers used in this study is available at FlyBase ( http://flybase . bio . indiana . edu/ ) . Stocks were maintained and crosses were made on standard Drosophila medium at 25°C . To generate anti-Peo polyclonal antibodies , rabbits were immunized with bacterially expressed 6His-Peo . Immunization and production of anti-Peo antibodies were carried out by the Agro-Bio Company ( France ) . To purify the anti-Peo antibody we used a bacterially expressed GST-Peo protein . About 1 mg of this tagged protein was run on a polyacrylamide gel and then blotted onto a nitrocellulose membrane . The membrane strip containing GST-Peo was cut out , washed in 100 mM glycine/HCl pH 2 . 5 for 5 min , washed for 5 min in TBS , blocked by incubation with 3% BSA for 1 hour , and washed again for 4 min in TBS . The membrane was then cut into small pieces and incubated overnight with rotation at 4°C in 2 ml of serum diluted 1:5 in TBS . After centrifugation and removal of supernatant the membrane pieces were washed for 15 min in 50 mM Tris/HCl pH 7 . 5 , 500 mM NaCl , for 5 min in 50 mM Tris/HCl pH 7 . 5 , 100 mM NaCl , and for 10 min in TBS . The membrane was then incubated for 30 min in 1ml of 100 mM glycine/HCl pH 2 . 5 to elute the antibody . The eluate was then mixed to a proper volume 1M Tris pH 8 . 8 to bring the final pH to 8 . 0 , and then kept at 4°C before use . DAPI-stained colchicine-treated larval brain chromosomes were prepared according to [10] . Preparation and immunostaining of mitotic and polytene chromosomes were carried out as described previously [10 , 20] , with minor modifications . For anti-PCNA immunostaining , dissected larval brains were incubated in PBS with 0 . 5% Triton X-100 for 3 min , and then fixed in 3 . 7% formaldehyde according to [70] . Before immunostaining , brain squash preparations were Triton-X extracted by incubating the slides in 0 . 1% Triton X-100 containing PBS ( PBT ) , 2 times for 10 min . The primary antibodies used for immunostaining were: the rabbit anti-Peo described above diluted 1:10; rabbit anti-HOAP ( 1:100 ) , mouse anti-HP1a ( C1A9; 1:10 ) , and mouse anti-PCNA ( 1:20; Abcam ab29 ) . The anti-HP1a antibody C1A9 was obtained from the Developmental Studies Hybridoma Bank , created by the NICHD of the NIH and maintained at The University of Iowa , Department of Biology , Iowa City , IA 52242 . After an overnight incubation at 4°C with the primary antibody , slides were washed twice in TBS-Tween 0 . 1% for 15 min and then incubated for 2 h at room temperature with FITC-conjugated goat anti-mouse ( 1:20; Jackson Laboratories ) , or AlexaFluor 555-conjugated donkey anti-rabbit ( 1:200; Invitrogen ) antibodies . All slides were then mounted in Vectashield medium H-1200 with DAPI to stain DNA . in vivo detection and immunostaining of GFP-tagged proteins on polytene chromosomes were carried out as previously described [24] . Chromosome preparations were analyzed using a Zeiss Axioplan epifluorescence microscope ( CarlZeiss , Obezkochen , Germany ) , equipped with a cooled CCD camera ( CoolSnap HQ , Photometrics , Woburn , MA ) . Gray-scale digital images were collected separately , converted to Photoshop format , pseudocolored , and merged . To quantify the polytene chromosome fluorescence intensity after Peo immunostaining , we used the ImageJ software ( National Institute of Mental Health , Bethesda , Maryland , USA ) . Given that the distribution of fluorescent bands along the chromosomes was rather uniform , for each polytene nucleus we selected 3–4 different chromosome regions of similar length , and measured both their fluorescence and the fluorescence of a close chromosome-free region to correct for background fluorescence . For each genotype ( wild type , +/Df , peoh/Df and peo1/Df ) shown in Fig 8 we measured at least 30 polytene regions from at least 10 nuclei . The coding sequences of Peo , Hp1a and HOAP were PCR-amplified and cloned into pGBKT7 or pGAD-T7 vectors ( Clontech ) . The S . cerevisiae AH109 strain was transformed with the indicated combinations of plasmids and assayed for growth on SD/–His/–Trp/–Leu selection plates supplemented with 20 mM 3-amino-1 , 2 , 4-triazole ( 3-AT ) , according to the manufacturer’s instructions . To obtain the GST-Moi , GST-Ver , GST-HOAP and GST-Peo fusion proteins , the corresponding full length cDNAs were cloned in either pGEX-6P1 or pGEX-3X , expressed in bacteria and purified as described previously [24] . The GST-Peo1 , GST-Peo2 , GST-Peo3 , GST-HOAP ∆3 , GST-HOAP ∆2 , 3 , GST-HOAP ∆1–3 and GST-HOAP ∆N-TERM truncated proteins were obtained by cloning the corresponding PCR-generated sequences in pGEX-6P1; bacterially expressed GST fusion proteins were then purified by incubating crude lysates with glutathione sepharose beads ( QIAGEN ) as recommended by the manufacturer . To generate 6His-Peo , the LD08052 peo full-length cDNA was cloned into the pQE32 expression vector ( QIAGEN ) , and expressed in bacteria; 6His-Peo was affinity purified with a Ni-NTA resin using standard procedures . To obtain extracts for Western Blot analysis , 50 dissected third instar larval brains were lysed in an ice-cold buffer containing 20 mM Hepes KOH pH 7 . 9 , 1 . 5 mM MgCl2 , 10 mM KCl , 420 mM NaCl , 30 mM NaF , 0 . 2 mM Na3VO4 , 25 mM BGP , 0 . 5 M PMSF , 0 . 1% NP40 , and 1X protease inhibitor cocktail ( Roche ) . To obtain HOAP-FLAG , Ver-FLAG and Peo-FLAG expressing S2 cells , cav , ver or peo cDNAs were cloned in the pAWF vector ( DGRC ) in frame with the FLAG-coding sequence . For the expression of Moi-HA , the moi full lenght cDNA was fused in frame with the HA-coding sequence and then cloned into a pCASPER4 vector . All constructs were transfected in S2 tissue culture cells using Cellfectin ( Invitrogen ) , and cells were harvested 72 h after transfection . Extracts were lysed in 20 mM Tris pH 8 . 0 , 420 mM NaCl , 1 mM MgCl2 , 1 mM DTT , 0 . 1% NP40 , and 1X protease inhibitor cocktail ( Roche ) . For preparation of human cell extracts , HeLa cells expressing Peo-FLAG cloned in the pCDNA vector were harvested after 72hr transfection and lysated in 20 mM Tris pH 8 . 0 , 420 mM NaCl , 1 mM MgCl2 , 1 mM DTT , 0 . 1% NP40 and 1X protease inhibitor cocktail ( Roche ) . For GST-pulldown assays , protein extracts were incubated with 2 μg of each GST fusion protein bound to sepharose beads in a buffer containing 20 mM Hepes KOH , 20 mM NaF and 0 . 8% NP40 for 1h at 4°C . Sepharose-bound GST proteins were collected by centrifugation , washed several times with 20 mM Hepes KOH , 20 mM NaF and 1 . 8% NP40 , and resuspended in Laemli buffer in a 30μl final volume for Western Blot analysis . For immunoblotting , protein samples were run into SDS polyacrilammide gels and electro-blotted on a nitrocellulose membrane ( Bio-Rad ) in a phosphate buffer ( 390 mM NaH2PO4 /610 mM Na2HPO4 ) . For the detection of HOAP-FLAG , Ver- FLAG , Peo- FLAG , and Moi-HA , membranes were probed with anti-FLAG HRP-conjugated ( 1:1000; Roche ) , and anti-HA HRP-conjugated ( 1:500; Roche ) antibodies; Peo and His-Peo were detected with our rabbit anti-Peo ( 1:100 ) , and Giotto with our anti-Giotto ( 1:5000; [71] ) . Secondary antibodies were sheep anti-mouse IgG HRP-conjugated ( 1:5000 ) , or donkey anti-rabbit IgG HRP-conjugated ( 1:5000 ) ( both from Amersham Biosciences ) . The blots were developed using the ECL or ECL Plus method ( Amersham Biosciences ) and signals were detected with the ChemiDoc scanning system ( BioRad ) . Band intensities were quantified using the image acquisition and analysis Image lab 4 . 0 . 1 software ( Biorad ) . EdU labeling was performed as per the manufacturer's instructions ( Invitrogen , ClickiT Alexa Fluor 488 Imaging kit ) . Larval brains were cultured with 10μM EdU in 1 X PBS for 60 min prior to fixation and detection . Brains from third instar larvae were dissected in saline ( 0 . 7% NaCl ) , incubated in saline with 110 μM Aphidicolin ( APH ) for 1 . 5 hours , rinsed in saline , and then transferred into a 33 mm Petri dish containing 3 ml of Schneider’s medium ( SIGMA ) supplemented with 10% fetal bovine serum ( FBS , Gibco BRL ) for 1 , 2 or 3h . Colchicine at a final concentration of 10-5M was added to the medium 1 hour before fixation according to [10] . Control wild type and peoh/peoh brains were incubated in saline without APH for 1 . 5 , and processed like the APH-treated brains; they were fixed after 2 hours incubation n the medium . As a first step towards the construction of a three-dimensional model of Peo , we used its full-length sequence ( Accession code: Q7K4V4 ) as a query to search the UniProtKB database ( http://www . uniprot . org/ ) using CSI-BLAST [72] , with an Expectation ( E ) value threshold of 10–5 . Iterative searches of the database yielded 59 unique sequences . The retrieved sequences were aligned with the CLUSTALW software [73] with default parameters . The multiple sequence alignment ( MSA ) was next used as a seed to construct a Hidden Markov Model ( HMM ) of the family . The HMM was employed to search the Pfam database ( http://pfam . sanger . ac . uk/ ) via the HHpred server [74] . The highest scoring hit was the ubiquitin-conjugating enzyme ( UBC ) family also known as the E2 enzyme family [34] ( probability to be a true positive more than to 99% , E-value equal to 1 . 2 x10-42 ) . The third scoring hit was the ubiquitin E2 variant ( UEV ) family ( Pfam ID: PF05743 ) that includes UBC homologs such as Tsg101 , Mms2 and UEV1 ( probability to be a true positive equal to 96 . 44% , E-value equal to 1 . 3 x10-4 ) . Peo belongs to the UEV family because it contains an aspartic acid residue ( at position 106 , according to SwissProt numbering ) in place of the E2 active site cysteine , and it is unable to catalyze ubiquitin transfer as it lacks the cysteine that forms a transient thioester bond with the C-terminus of ubiquitin ( Ub ) . Prediction of potentially disordered regions using the GeneSilico MetaDisorder server ( http://iimcb . genesilico . pl/metadisorder/ ) revealed that at the C-terminus of Peo there is a stretch of ~ 70 aa ( from residue 177 to 244 , according SwissProt numbering ) that has the tendency to be intrinsically disordered ( i . e . lack a unique three dimensional structure at least in the absence of a binding partner ) , while the region including residues 16–176 shows propensity to form a folded globular domain with a well-defined pattern of secondary structures as revealed by the Quick2d web server analysis [75] . Because no homologous structure with sufficiently high sequence identity with Peo is available , we performed the Peo modeling using the composite approach implemented in I-TASSER server ( Iterative Threading ASSEmbly Refinement ) [76] . The Peo sequence from residue 16 to 176 ( predicted to fold in a globular domain ) was submitted to the server and the model with the best confidence score ( C-score = 0 . 5 ) returned by I-TASSER was selected . We added hydrogen atoms in this model using HAAD software [77] and refined it close to the native structure using FG-MD molecular dynamics based algorithm [78] . Our final refined model of Peo was evaluated as a potentially extremely good model ( with a predicted LGscore of 2 . 50 ) by the PRO-Q model quality assessment program [79] . The QMEAN score [80] was 0 . 6 ( the variability range is 0–1 , with 1 being a perfect model ) . Collectively these parameters indicate that the Peo three-dimensional model is sufficiently accurate for making functional inferences .
|
Telomeres are specialized structures that protect chromosome ends from incomplete replication , degradation and end-to-end fusion . Abnormalities in telomere structure or maintenance can promote a variety of human diseases including premature aging and cancer . Although all human telomeres contain the same DNA sequences , they differ from each other in the subtelomeric regions or subtelomeres . Recent work has shown that human subtelomeres control telomere replication and that abnormalities in these structures can lead to localized chromosome instability and disease . However , the relationships between subtelomeres and telomeres are currently poorly understood . Here , we have addressed this problem using the fruit fly Drosophila melanogaster as model system . Drosophila subtelomers are very different from each other as they contain different types of chromatin . We have found that mutations in a gene we called pendolino ( peo ) cause telomeric fusions ( TFs ) and that these fusions preferentially involve the telomeres associated with a tightly packed form of chromatin called heterochromatin . Interestingly , none of the 10 mutants with TFs so far described in Drosophila shows the pattern of TFs observed in peo mutants . Thus , our data provide the first demonstration that subtelomeres can affect telomere fusion . We believe that these results will stimulate further studies on the role of subtelomeres in the maintenance of genome stability .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
The Analysis of Pendolino (peo) Mutants Reveals Differences in the Fusigenic Potential among Drosophila Telomeres
|
Men who have sex with men ( MSM ) were found to be one of the high-risk populations for Entamoeba histolytica ( E . histolytica ) infection . Accompanied by the prevalence of human immunodeficiency virus ( HIV ) among MSM , invasive amebiasis caused by E . histolytica has been paid attention to as an opportunistic parasitic infection . However , the status of E . histolytica infection among MSM has been barely studied in mainland China . Seroprevalance of E . histolytica was determined using an enzyme-linked immunosorbent assay based on a cross-sectional study conducted in Beijing and Tianjin , China . Factors potentially associated with E . histolytica infection were identified by logistic regression analysis . A total of 602 MSM were included in the study and the laboratory data on serostatus of E . histolytica were available for 599 of them ( 99 . 5% ) . 246 ( 41 . 1% ) and 51 ( 8 . 5% ) of the study participants were E . histolytica seropositive and HIV seropositive , respectively . Univariate analyses suggested preferred anal sex behaviors were associated with E . histolytica seropositivity . In multivariate logistic regression analysis , “only has receptive anal sex” ( OR: 2 . 03; 95% CI: 1 . 22 3 . 37 ) , “majority receptive anal sex” ( OR: 1 . 83; 95% CI: 1 . 13 , 2 . 95 ) , and “sadomasochistic behavior ( SM ) ” ( OR: 2 . 30; 95% CI: 1 . 04 , 5 . 13 ) were found to be significantly associated with E . histolytica infection . High seroprevalence of E . histolytica infection was observed among MSM from Beijing and Tianjin , China . Receptive anal sex behavior and SM were identified as potential predictors . Therefore , E . histolytica and HIV co-infection needs to be concerned among MSM due to their sharing the common risk behaviors .
Entamoeba histolytica ( E . histolytica ) has a worldwide distribution and is endemic in most developing countries . Invasive amebiasis ( IA ) caused by E . histolytica is a very common human gastrointestinal parasitic disease which affected 50 million people worldwide and caused greater than 100 , 000 deaths annually . High risk populations for developing IA include infants , travelers from endemic area , and patients who are taking immunosuppressant [1] , [2] . In mainland China , E . histolytica infection was also very popular in general population . The average prevalence of E . histolytica infection was 0 . 95% , ranged from 0 . 01% to 8 . 12% [3] . In 1967 , the association between amebiasis and homosexuality was suggested for the first time [4] . Men who have sex with men ( MSM ) population had already been found to be a high risk population with E . histolytica infection before 1990 . Homosexuality and oral-anal sex have been most frequently reported as potential risk factors for E . Histolytica infection [5]–[13] . Accompanied by the transmission of human immunodeficiency virus ( HIV ) in MSM population , the prevalence of IA caused by E . histolytica are increasing and getting the attention as an important opportunistic parasitic infection . Recent studies from Australia , Japan , Korea and Taiwan reported increased risks for E . histolytica infection and IA among HIV-positive MSM [14]–[18] . Hung CC and colleagues recently reviewed the status of E . Histolytica infection in MSM [19] . By the end of 2011 , China had about 780 , 000 people living with HIV/AIDS and 17 . 4% of them were MSM . The estimated new HIV infections in 2011 are 48 , 000 and 38 . 1% were MSM [20] . Case report of IA suggested the risk of E . histolytica prevalence among Chinese MSM , especially in the HIV positives [21] . However , to our knowledge , the prevalence of E . histolytica infection among MSM population has not been investigated in mainland China . The major aim of this study was to assess the seroprevalence of E . histolytica infection and the potential impact factors among MSM from China .
The study was approved by the Ethics Committees of the Institute of Pathogen Biology , Chinese Academy of Medical Sciences & Peking Union Medical College . All participants were the adults of men who had sex with men . Written informed consent was obtained before the interview and testing . A cross-sectional study was conducted in Beijing and Tianjin , China . Six hundred study participants were recruited between March and July 2010 , though local non-government organizations ( Beijing Rainbow Volunteers Workstation and Tianjin Deep Blue Volunteers Workgroup ) . Participants' recruitment and inclusion criterion had been addressed in previous study [22] . Questionnaire administered by the trained interviewer were performed for each study participant , the data acquired from the questionnaire includes socio-demographic characteristics ( e . g . , age , income , ethnicity , education , employed , and marital status ) , sexual orientation and homosexual act , sexual behaviors in the past 6 months , history of STDs and HIV infection . Anilingus behaviors were defined as sexual stimulation involving oral contact with the anus and sadomasochistic behaviors ( SM ) were defined as behaviors which aimed to enhance sexual gratification from inflicting or submitting to physical and emotional abuse . Blood samples were collected for E . histolytica and HIV serology . Serum samples were stored at −80°C until tested . Each study participant was assigned a unique identification number that was used to link the questionnaire and specimens . The HIV infection status was screened by an enzyme immunoassay ( Wantai Biological Medicine Company , Beijing , China ) , and positive tests were confirmed by HIV-1/2 Western blot assay ( HIVBlot 2 . 2 WB; Genelabs Diagnostics , Singapore ) . Qualitative screenings of serum immunoglobulin G ( IgG ) antibodies to E . histolytica were retrospectively performed using the commercial enzyme-linked immunosorbent assay ( ELISA ) kit ( Shanghai Fengxiang Biological Technology Co . Ltd . Shanghai , China ) . Purified E . histolytica antigen was used to coat microtitration wells , incubated for 30 minutes at 37°C after adding 10 uL serum samples to wells . Washing and removing non-combinative antibody and other components , then combined HRP-conjugate reagent , then incubated and washed again . The substrate solution was added to each well , after 15 minutes at 37°C , stop solution was added to arrest color development and a ELISA reader was used to measure the absorbance at 450 nm . Each sample was tested in duplicate and the average optical density ( OD ) value was calculated . Test validity was evaluated as: the average of the positive controls should ≥1 . 00 and the average of negative controls should ≤0 . 10 . The cut off value was set as the average of negative controls +0 . 15 according to the instruction of the kit . Questionnaires were double entered and compared with EpiData software ( EpiData 3 . 02 for Windows , The Epi Data Association Odense , Denmark ) . After cleaning , the data were then converted and analyzed using Statistical Analysis System ( SAS 9 . 2 for Windows; SAS Institute Inc . , NC , USA ) . Study population was characterized by site with respect to age , ethnicity , education , marriage status , and current status of HIV . Differences between sites in these variables were assessed with Pearson chi-square test . The associations of E . histolytica infection with the characteristics of demographics , sexual behaviors , diagnosed STDs including HIV , and current status of HIV were estimated using Pearson chi-square test . Variables related with E . histolytica serology ( p<0 . 1 ) in the univariate analyses were included in a multiple logistic regression model additionally adjusted for age , site , and HIV infection status . The Cochran-Armitage test was used to evaluate the trend of OD value in E . histolytica antibody test associated with anal sex behaviors categories .
A total of 607 participants were interviewed and signed the informed consent , 6 of them were excluded ( 3 did not complete sample collection and 2 Vietnamese living in Beijing ) . Finally , 602 were used for the analyses ( 302 from Beijing and 300 from Tianjin ) . Laboratory data on serostatus of E . histolytica and HIV were available for 599 ( 99 . 5% ) and 598 ( 99 . 3% ) of the study participants respectively . Age , ethnicity , education , marriage status , HIV-1 serostatus and E . histolytica serostatus of the study population were compared by site ( Table 1 ) . No significant difference was found for any characters ( p>0 . 05 ) . Therefore , participants from the two sites were pooled together for further association analyses . Age of the participants ranged from 16 to 72 years , with a mean age of 27 . 9±8 . 3 years . The participants showed the following characteristics: 95% ( 570/600 ) were Han nationality; 51 . 5% ( 309/600 ) had more than 12 year's education; 77 . 7% ( 467/601 ) were single . Laboratory data suggested two hundred and forty six ( 41 . 1% ) and fifty one ( 8 . 5% ) of the study participants were E . histolytica seropositive and HIV seropositive respectively . Twenty three participants were E . histolytica and HIV co-infected . Age , ethnicity , site , registered residence , education , marriage status , and alcohol use were not found to be related with E . histolytica serostatus ( Table 2 ) . Homosexual men accounted for 71 . 0% ( 427/602 ) , and bisexual men for 24 . 3% ( 146/602 ) of the study population . 73 . 6% ( 181/246 ) of participants with E . histolytica seropositive were homosexuals . Age of the first homosexual act ranged from 5 to 55 years old , with a mean age of 21 . 4±5 . 0 years . The median number of their homosexual partners was eleven and 43 . 1% of participants had stable homosexual partners before baseline survey . In the past one year , sixty one participants ( 10 . 2% ) had group sex , thirty eight participants ( 6 . 3% ) had ever received money for sex with male partners , and twenty three ( 3 . 8% ) had ever provided money for sex with male partners . In the past 6 months , 133 ( 22 . 2% ) participants reported had sexual behavior less than 1 time per month and only 1 . 3% ( 8/602 ) of participants insisted on using condoms in the process of insertive or receptive anal sex and oral sex . 22 . 5% of participants had ever been diagnosed sexually transmitted diseases . The association between sex behaviors and E . histolytica infection were also assessed . Self-reported preferred anal sex behaviors were classified to four types ( only has insertive anal sex , majority insertive anal sex , majority receptive anal sex , and only has receptive anal sex ) . Univariate analyses suggested preferred anal sex behaviors were associated with E . histolytica infection . HIV infection was not found associated with E . histolytica infection ( Table 3 ) . The variables associated with E . histolytica infection in the univariate analyses ( P<0 . 1 ) were included in a multivariate logistic regression model . Age , site , and HIV infection status were fixed in the model as well . In the multivariate logistic regression model , only has receptive anal sex ( OR: 2 . 03; 95% CI: 1 . 22 , 3 . 37 ) , majority receptive anal sex ( OR: 1 . 83; 95% CI: 1 . 13 , 2 . 95 ) , and SM ( OR: 2 . 30; 95% CI: 1 . 04 , 5 . 13 ) were found to be significantly associated with E . histolytica infection after adjusted for the other variables ( Table 4 ) . The OD in ELISA test was used for further analysis . Five hundred and eighty three participants reported their preferred anal sex behavior . As shown in figure 1 , the OD values increased from the group of only had insertive anal sex ( median OD , 0 . 036 ) , majority insertive anal sex ( median OD , 0 . 049 ) , majority receptive anal sex ( median OD , 0 . 117 ) to only had receptive anal sex ( median OD , 0 . 131 ) . The trend of increasing was significant ( p<0 . 001 ) , which was consistent with the result of multivariate analysis .
This pilot study investigated E . histolytica seroprevelance in MSM from China; potential factors associated with E . histolytica infection were evaluated as well . In a total of 602 study participants , E . histolytica seroprositivity was found to be 41 . 1% . Types of preferred anal sex behavior ( only has receptive anal sex and majority receptive anal sex ) and SM were identified as significant predictors for E . histolytica infection . In addition , significant different antibody levels were observed between subgroups with respect to the preferred anal sex behavior . The first observation of a relation between enteric protozoan infections and sexual behavior was reported in 1968 [4] . Epidemiological studies conducted in the developed countries showed homosexuals or MSM had significant higher risk of E . histolytica infection . Using microscopy for diagnosis , the prevalence varied from 20% to 32% among MSM without gastrointestinal symptoms [5] , [7] , [8] , [23]–[27] . However , microscopy is not sensitive or specific enough for the detection of E . histolytica in clinical specimens , especially for the differentiation E . histolytica from E dispar and E moshkovskii in the epidemiological studies with a large sample size . Therefore , serological tests were used to detect the infection though measuring anti-E . histolytica antibodies and seroprevalence ranging from 0 . 2% to 21% using ELISA test among HIV negative MSM were reported in several developed countries [18] , [19] . Our results , for the first time , suggested a high prevalence ( 41 . 1% ) of E . histolytica infection among MSM community from China . A recently published study conducted among general population in seven provinces in China showed that the seroprevalence of E . histolytica infection varied from 6% to 11% [28] . Although antibody test could not distinguish the past or current infection status and maybe overestimated the epidemic status , the fact that amebic liver abscess and latent infection had become one of the common opportunistic infection diseases among Chinese MSM AIDS patients reminds us to pay attention to the co-infection of E . histolytica and HIV . [21] , [29] . In the present study , homosexual behaviors were mostly classified according to participants' tendency . Receptive anal sex behavior was found to be related to higher prevalence of E . histolytica infection . This finding and its underlying mechanisms should be further studied in the future . Homosexuals and history of anilingus had been demonstrated to be the risk factors of E . histolytica infection . In 1978 , a study from the New York city reported that 20% of eighty nine sexually active homosexual men had amebiasis and the presence of infections associated with history of anilingus [5] . Another study from a venereal-disease-clinic population compared the prevalence of E . histolytica infections in homosexual men , bisexual men and heterosexual men . Homosexuality and oral-anal sex were found to be the most important risk factors for E . histolytica infection [7] . However , such an association was not observed in our study population . Interestingly , SM was found to be associated with E . histolytica infection in the present study . The data on the specific behaviors during the process of SM has not been well studied in China due to the potential issues of social culture and discrimination . Several published studies had revealed that people who had engaged in SM were more likely to have experienced oral-anal sex and other sexual risk practice [30]–[32] . In addition , fecal-oral contamination in these sexual behaviors maybe occurs and increases the opportunity of pathogen infection . Keystone JS' study showed cleaning of anus before anal sex was associated with a significant lower prevalence of infection [23] . But it is difficult to explore the factors linked SM behaviors to the infection susceptibility in our present study because only 30 participants ( 5 . 0% ) reported SM behaviors . Developing targeted prevention and control strategies , such as developing sanitary habit before sexual behavior , may decrease the opportunity of pathogen infection . Therefore , it is necessary to further study potential risky behaviors associated with health problems among Chinese MSM . Several limitations of this study should be kept in mind . First , potential bias due to the inaccurate response to the questionnaire , especially to the questions on sexual behaviors , could not be excluded completely . Second , our study participants might not represent the general MSM population due to the potential limitation of enrollment methods . Therefore , potential selection bias should be considered when interpret our results . Third , serology could not clearly identify the infection status as current infection or past infection , potential bias caused by such misclassification could not be excluded . Although statistically significant difference of antibodies levels was observed with respect to the preferred anal sex behaviors ( p<0 . 001 ) , however , the smaller sample size in each subgroups and a broad range of OD value should be considered . Further studies are needed to explore the underlying mechanisms for the observed relation between receptive anal sex behaviors and E . histolytica infection . Fourth , cross-sectional study design has its limitation on association analysis . Therefore , our results need confirmation by further large-scale case-control studies or prospective studies . In conclusion , high prevalence of E . histolytica infection was observed among MSM from Beijing and Tianjin , China . Receptive anal sex behaviors and SM were identified as significant predictors for E . histolytica infection . Prevention and control of E . histolytica infection among Chinese MSM should be concerned because this special population confronted with high risk of HIV infection .
|
Entamoeba histolytica ( E . histolytica ) is a very common human gastrointestinal parasitic disease which affects 50 million people worldwide . Men who have sex with men ( MSM ) have already been found to be one of the high-risk populations with E . histolytica infection . Previous studies have reported an increased risk for E . histolytica infection and invasive amebiasis in HIV seropositive MSM . This pilot study aimed to investigate the serology of E . histolytica among MSM from mainland China . High prevalence of E . histolytica infection ( 41 . 1% , 246/599 ) was observed among the study population , receptive anal sex behavior and sadomasochistic behavior were found to be associated with the E . histolytica serostatus . Although HIV infection was not found to be associated with E . histolytica infection in this pilot study , studies from other countries had reported increased risks for E . histolytica infection and invasive amebiasis among HIV-positive MSM . Our findings suggest E . histolytica infection control needs to be concerned with respect to the increasing HIV prevalence among Chinese MSM population .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"infectious",
"diseases",
"amebiasis",
"infectious",
"disease",
"epidemiology",
"epidemiology",
"neglected",
"tropical",
"diseases"
] |
2013
|
Seroprevalence of Entamoeba histolytica Infection among Chinese Men Who Have Sex with Men
|
The first identified Chikungunya outbreak occurred in Bangladesh in 2008 . In late October 2011 , a local health official from Dohar Sub-district , Dhaka District , reported an outbreak of undiagnosed fever and joint pain . We investigated the outbreak to confirm the etiology , describe the clinical presentation , and identify associated vectors . During November 2–21 , 2011 , we conducted house-to-house surveys to identify suspected cases , defined as any inhabitant of Char Kushai village with fever followed by joint pain in the extremities with onset since August 15 , 2011 . We collected blood specimens and clinical histories from self-selected suspected cases using a structured questionnaire . Blood samples were tested for IgM antibodies against Chikungunya virus . The village was divided into nine segments and we collected mosquito larvae from water containers in seven randomly selected houses in each segment . We calculated the Breteau index for the village and identified the mosquito species . The attack rate was 29% ( 1105/3840 ) and 29% of households surveyed had at least one suspected case: 15% had ≥3 . The attack rate was 38% ( 606/1589 ) in adult women and 25% in adult men ( 320/1287 ) . Among the 1105 suspected case-patients , 245 self-selected for testing and 80% of those ( 196/245 ) had IgM antibodies . In addition to fever and joint pain , 76% ( 148/196 ) of confirmed cases had rash and 38% ( 75/196 ) had long-lasting joint pain . The village Breteau index was 35 per 100 and 89% ( 449/504 ) of hatched mosquitoes were Aedes albopictus . The evidence suggests that this outbreak was due to Chikungunya . The high attack rate suggests that the infection was new to this area , and the increased risk among adult women suggests that risk of transmission may have been higher around households . Chikungunya is an emerging infection in Bangladesh and current surveillance and prevention strategies are insufficient to mount an effective public health response .
Chikungunya is an arthropod-borne disease caused by Chikungunya virus ( Alphavirus family , Togaviridae family ) which was initially identified in Tanzania in 1952 [1] . Chikungunya outbreaks likely happened before the virus was identified because there were many verifiable depictions of epidemic fevers with remarkable arthralgia [2] . Humans can be a reservoir for Chikungunya virus during epidemics . In the past 50 years , Chikungunya has re-emerged in several occasions in both Africa and Asia [3] . Rapid and local transmission of Chikungunya occurred in the Caribbean and the Americas within 9 months during 2013–2014 [4] . Aedes mosquitoes transmit Chikungunya virus . Aedes aegypti mosquitoes are responsible for transmission of both Chikungunya and dengue [5]and in Asia , have been identified as the primary vector in most urban dengue epidemics [6] . Aedes albopictus was identified as the vector in the 2006 Chikungunya outbreak in La Reunion ( an island in the Indian Ocean ) . This newly identified vector caused effective replication and spread the infection beyond previously endemic areas [6] . A . albopictus can prosper in both rural and urban environments [7] and breed in artificial water containers [8] . Since 2005 , Chikungunya has become an emerging public health problem in Southeast Asia , with large numbers of cases reported in Singapore , Malaysia , and Thailand [9] . In 2006 , an increase in the incidence of Chikungunya in India prompted testing of serum samples collected from febrile patients from two different surveillance projects in Dhaka , Bangladesh . One hundred seventy-five serum samples were tested however none had antibodies against Chikungunya virus [10] . In 2008 , the first recognized outbreak of Chikungunya in Bangladesh was identified in the northwest area of the country . Transmission appeared to be geographically limited to two villages bordering India in northwestern Bangladesh [11] . In late October 2011 , an outbreak of fever and severe joint pain was reported by a local health official in Dohar Sub-district in Dhaka District . Limited antibody testing for dengue and blood smears for malaria conducted at the local health clinic suggested that the illnesses were not caused by dengue or malaria . On November 2 , 2011 , an outbreak investigation team comprised of medical epidemiologists , entomologists , field research assistants and laboratory technicians from the Institute of Epidemiology Disease Control and Research ( IEDCR ) , of the Bangladesh Ministry of Health and Family Welfare , and icddr , b ( formerly known as the International Centre for Diarrhoeal Disease Research , Bangladesh ) began an investigation with the objectives of identifying the etiology of the outbreak , describing the clinical presentation of cases , and identifying associated vectors .
This investigation was carried out as a response to an outbreak investigation and thus the protocol was not reviewed by a human subjects committee . However , participants provided verbal informed consent prior to interviews and blood specimen collection and the Government of Bangladesh approved the outbreak investigation plan .
Data collectors surveyed all 897 households in the village and collected information regarding symptoms for all 3 , 840 residents; 1105 ( 29% ) of household members met the suspected case definition . There were no differences in attack rates by gender among children <10 years of age; however , females were more likely to report illness than males for every other age group and the differences were greatest among residents aged 31–40 years ( 28% of males vs 50% of females ) and 41–50 years ( 29% vs 53% ) ( Table 1 ) . Sixty-four percent of households had at least one suspected case , while 15% had three or more ( Table 2 ) . Twenty-two percent of suspected cases ( 245/1105 ) provided blood for testing and 80% ( 196/245 ) of them had IgM antibodies against Chikungunya virus . Suspected cases that selected for testing were similar in age to those who were not tested , with the exception that fewer children aged <10 presented for a blood draw . In addition , cases who sought testing were more likely to be female than suspected cases who were not tested ( Table 3 ) . Patients tested within 1 week of illness onset were unlikely to have IgM antibodies , while 93% of suspected cases tested between 30 and 60 days post illness onset had IgM antibodies against Chikungunya ( Table 4 ) . Confirmed cases had dates of illness onset from August through November , per the suspected case definition . One case patient provided a blood sample within 2 days of onset and this patient had detectable Chikungunya virus RNA with RT-qPCR; this patient did not have IgM antibodies in serum . Thirty-eight percent ( 75/196 ) of laboratory confirmed cases reported having joint pain that persisted for more than one month and 76% ( 148/196 ) reported a rash ( Table 5 ) . Most rashes were macular ( 77% ) and involved the face ( 67% ) and upper extremities ( 53% ) . Among cases reporting joint pain , the median number of joints affected was nine ( range: 1–14 ) . The most commonly affected joint was the knee ( 36% ) . Joint pain was accompanied by swelling in 29% of the cases which subsided with the remission of joint pain . Sixty-three houses were inspected for water containers with mosquito larvae , and 252 artificial and natural containers with water were found and 534 larvae were collected . Eighty-nine percent of larvae ( 449/504 ) that hatched yielded A . albopictus mosquitoes and the remaining yielded Culex quinquefaciatus mosquitoes . No A . aegypti mosquitoes hatched . The overall Breteau index for the village was 35 per 100 .
Laboratory findings confirmed that Chikungunya virus caused this outbreak and the clinical features were consistent with previously described outbreaks [17 , 18] . This investigation provides further evidence that Chikungunya virus has become an emerging public health problem in Bangladesh [11] . Though no recent community seroprevalence studies of Chikungunya have been published from Bangladesh or nearby countries , a 1995 cross-sectional survey carried out in Kolkata , which is approximately 250 km from Dhaka , indicated that the level of previous exposure to Chikungunya infection in that city was low [19] . Chikungunya infection gives life-long immunity [20] , so the consistently high attack rates by age group in our investigation suggest that Chikungunya was new to this geographic area . An abundance of a particular species of mosquitoes during an outbreak is an important condition for determining the vector responsible for transmission [21] and the fact that A . albopictus hatched from 89% of the larvae collected in the village suggests that this vector was likely responsible for transmission during this outbreak . As A . albopictus has a tendency to breed in water compartments close to homes and to feed during the day [22] , persons who are at home during the day time could be at increased risk due to prolonged exposure to these mosquitoes . Adult women , most of whom spend the majority of their day at or very near the home , experienced the highest attack rates in this outbreak . This finding is similar to outbreaks of Chikungunya in rural areas in other countries where higher risk among women was also reported [23 , 24] . According to WHO , places with a Breteau index >20% have a high risk for dengue outbreaks [16] , and this may be true for other outbreaks of mosquito-borne illness as well . In this outbreak , the Breteau index was 35% , suggesting risk of transmission of mosquito-borne disease in Char Kushai was very high . Based on published data on the clinical presentation of Chikungunya , patients’ symptoms usually resolve within a few weeks [25 , 26] . However , the joint pain associated with Chikungunya virus infection can persist for weeks or months , and in some cases for years [24 , 27] , resulting insignificant economic burden due to this disability . In India , the national burden of Chikungunya during the 2006 epidemic was estimated at 25 , 588 disability adjusted life years ( DALYs ) lost , with an overall burden of 45 . 3 DALYs per million ( range 0 . 01 to 265 . 6 per million in different states ) ; persistent arthralgia accounted for 69% of the total DALYs [28] . In this outbreak , 38% of confirmed Chikungunya cases had joint pain lasting more than one month which increased the burden from the outbreak beyond acute febrile illnesses it caused . However , the patients who were tested and had their clinical features assessed during our investigation self-selected for this additional clinical assessment so were more likely to be severely ill than other suspected cases who did not present for assessment . This self-selection is the best explanation for why their clinical profile is more severe compared to other published reports . This investigation was subject to several limitations . First , four other villages in the sub-district also reported cases , but we only investigated one village and our findings from Char Kushai may not be representative of all of the affected areas . Second , we aimed to detect symptomatic illness and did not look for asymptomatic infections . Therefore , we likely underestimated the number of people infected during the outbreak given evidence that 3–25% of Chikungunya virus infections are asymptomatic [29 , 30] . Third , residents may have been unable to reliably recall their symptoms or onset of illness , particularly for milder illnesses , which may have led to an underestimation of suspected cases . Likewise , some suspected cases may have been missed during the survey . Women , who were almost always interviewed , may have recalled their own illnesses more than illnesses of adult men in the home so this could explain some of the variation in attack rates by gender . However , we would expect that women would be able to recall illnesses among children very well and children were also less likely than women to report systems consistent with the suspected case definition . Therefore , we believe that it is unlikely that the gender differences in attack rates are entirely explained by recall bias . Resource constraints limited our ability to test all suspected cases for antibodies , and the diagnostic test we used was imperfect , with 84% sensitivity and 91% specificity during the convalescent phase of illness [31] . Therefore , we likely missed some true cases; our analysis showing that patients with recent onset were less likely to have IgM antibodies also suggests that we preferentially missed true cases if they were tested early in their illness . However , our attack rates were similar to other confirmed Chikungunya outbreaks and the symptoms exhibited by confirmed cases were consistent with clinical descriptions of disease from other studies . We did not collect date of onset data from all suspected cases which limits our ability to describe the timing of outbreak peaks . The cases who did present for blood draws and data collection about their illness history self-selected for this data collection and may have been motivated to participate because they had severe illness . Therefore , our clinical description may overestimate the severity of infections during this outbreak . The entomologic survey was conducted at the end of the outbreak; therefore , it is possible that the mosquito species most abundant at the beginning of the outbreak were different than those that we found at the end . However A . albipictus have been associated in several Chikungunya outbreaks [32] , and we found no evidence of A . aegypti in our survey . It is possible that A . aegypti played a significant role in transmission at the beginning of outbreak and then disappeared; however , the simplest explanation for our findings is that A . albopictus were primarily responsible for the outbreak . This investigation suggests that rural Bangladeshi populations are at risk for emerging mosquito-borne diseases , such as Chikungunya . Efforts to improve surveillance and identify outbreaks more quickly could provide an opportunity for public health action to reduce transmission , such as mosquito control . However , in rural Bangladesh , no public initiatives are currently implemented for mosquito control . WHO guidelines suggest that environmental interventions , such as destroying natural and human-made mosquito breeding sites in and around homes , may be more cost- effective than chemical methods to kill larva and adult mosquitoes [16] . Research to develop and test low-cost methods to identify and respond to outbreaks of mosquito-borne infections in low-income countries should be explored . Ecological studies to better describe the spatial and temporal distribution of vector habitats could help explain why outbreaks in Bangladesh remain geographically limited and could be used to target interventions in populations at the highest risk for vector-borne diseases .
|
Chikungunya virus is transmitted through bites from Aedes mosquitoes and causes outbreaks of fever and polyarthralgia; the geographic range of infection is expanding . An outbreak of fever with prolonged joint pain was investigated in Bangladesh in 2011 , where house-to-house surveys were carried out to identify suspected cases . Twenty-nine percent of the village inhabitants experienced symptoms consistent with Chikungunya during the three months of the outbreak . Eighty percent of suspected cases had evidence of IgM antibodies against Chikungunya suggesting that this virus caused the outbreak . Attack rates were similar for all age groups , which suggests that this population had little pre-existing immunity to the disease . This is consistent with the assumption that Chikungunya is an emerging infection in this part of the world where the majority of people likely remain susceptible to infection . Attack rates were higher among adult females , which may provide clues to where transmission occurs . Since most rural women spend the majority of their time in and around the home , interrupting vector habitat near houses might be a useful way to control epidemics . Given the continued risk for outbreaks , we need more efficient methods for detection and control .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[] |
2015
|
An Outbreak of Chikungunya in Rural Bangladesh, 2011
|
There is significant heterogeneity in reported sensitivities and specificities of diagnostic serological assays for Chagas disease , as might be expected from studies that vary widely according to setting , research design , antigens employed , and reference standard . The purpose of this study is to summarize the reported accuracy of serological assays and to identify sources of heterogeneity including quality of research design . To avoid associated spectrum bias , our analysis was limited to cohort studies . We completed a search of PubMed , a bibliographic review of potentially relevant articles , and a review of articles identified by a study author involved in this area of research . Studies were limited to prospective cohort studies of adults published since 1985 . Measures of diagnostic accuracy were pooled using a Der Simonian Laird Random Effects Model . A subgroup analysis and meta regression were employed to identify sources of heterogeneity . The QUADAS tool was used to assess quality of included studies and Begg's funnel plot was used to assess publication bias . Eighteen studies and 61 assays were included in the final analysis . Significant heterogeneity was found in all pre-determined subgroups . Overall sensitivity was 90% ( 95% CI: 89%–91% ) and overall specificity was 98% ( 95% CI: 98%–98% ) . Sensitivity and specificity of serological assays for the diagnosis of Chagas disease appear less accurate than previously thought . Suggestions to improve the accuracy of reporting include the enrollment of patients in a prospective manner , double blinding , and providing an explicit method of addressing subjects that have an indeterminate diagnosis by either the reference standard or index test .
Chagas disease , or American Trypanosomiasis , is caused by the parasite Trypanosoma cruzi . The World Health Organization estimates that approximately 10 million individuals are currently infected with T . cruzi and are at risk for developing cardiac or gut pathology normally associated with chronic Chagas disease [1] . T . cruzi is transmitted to humans by infected triatomine bugs that infest housing , take blood meals from the inhabitants , and then defecate , leaving the infective metacyclic stages of T . cruzi to be scratched into wounds or mucosal sites . Although Chagas disease was once confined to the Americas , primarily Latin America , migration from endemic countries has led to the appearance of Chagas disease in non-endemic regions as well [1] . In both endemic and non-endemic settings , transmission of T . cruzi is also possible through blood transfusion , tissue transplantation , and congenitally . Control programs and improvements in housing have led to a reduction in the incidence of disease in Latin America , but screening blood donors and diagnosing chronic , often asymptomatic patients , remains a major challenge . T . cruzi infection is generally controlled by a highly effective immune response but is rarely completely cleared , resulting in a persistent , but low level infection . Early in the infection with T . cruzi , parasites may be detected in the bloodstream either by direct observation of blood or by various culture techniques . Unfortunately , infection at this early stage often goes undetected because symptoms are nonspecific or absent . Once the immune response to T . cruzi is established , parasite detection is very difficult and diagnosis of the infection is based largely upon the detection of anti-T . cruzi antibodies by serological techniques . Conventional serological tests include primarily immunofluorescence assays ( IFA ) , enzyme-linked immunosorbent assays ( ELISA ) and indirect hemagglutination assays ( IHA ) . Because there is currently no single reference standard test , the World Health Organization ( WHO ) recommends that diagnosis of an individual utilize two conventional tests based on different principles and detecting different antigens [2] . In the case of ambiguous or discordant results , a third technique should be used [2] . The goal of this study is to summarize the evidence on the accuracy of diagnostic tests for Chagas disease from high quality diagnostic test studies . Previous research has found that use of a case-control design overestimated the diagnostic odds ratio ( DOR ) by 3-fold compared to studies employing a cohort design [3] . In order to more accurately assess the sensitivity and specificity of serological assays used to screen patients for Chagas disease , we limited our review to studies that prospectively enrolled patients using a cohort study design .
Criteria for inclusion in this systematic review were that the study 1 ) be available in English , Spanish , Portuguese , or German; 2 ) be published since 1985; 3 ) use human subjects rather than model organisms; 4 ) include a minimum of 50 samples; 4 ) prospectively enroll patients without knowledge of T . cruzi infection status using a cohort design; 5 ) examine a serologic assay based on measuring antibody levels in the blood ( e . g . ELISA , IHA , IFA , immunochromatographic assay , complement fixation ) , rather than a PCR assay , urine analysis , or saliva test; 6 ) provide enough information to determine sensitivity and specificity of a serologic assay compared with a reference standard of some kind; 7 ) enroll primarily adolescents or adult patients ( 12 years or older ) . The year 1985 was chosen as a cutoff to capture the time period during which conventional tests of IFA , IHA , and ELISA came into wide use in Latin America [4] . Studies involving exclusively immunodeficient adults , patients with HIV or TB , children , infants , neonates , or pregnant women were excluded . We excluded any study designed to evaluate serologic tests as a means to assess cure of Chagas disease , as the purpose of this systematic review is to evaluate the performance of diagnostic tests in patients with unknown disease status . Case-control studies that identified a group of well characterized cases and well characterized normal controls , often from different sources , were excluded because this is an important methodological limitation [3] . Data for patients with the acute stage of T . cruzi infection were excluded from the analysis , given that these data represented only a small proportion of all studies and provided extremely heterogeneous results . Furthermore , it has been shown that the reactivity of IgG antibodies to various antigens varies between patients with the acute and chronic form of Chagas disease [5] . We used several strategies to identify relevant articles , including searching PubMed with multiple search strategies , reviewing bibliographic citations from both included and excluded articles , and reviewing studies known to be relevant by one of the study authors , an expert in the field of Chagas diagnostics . The following strategy was used in PubMed/Medline to identify studies providing a quantitative evaluation of diagnostic tests for Chagas disease: ( “chagas disease/diagnosis”[MeSH Major Topic] ) AND ( “sensitivity and specificity”[MeSH Terms] OR “likelihood ratio” ) . We identified 156 articles , of which 121 were identified as potentially relevant by one or more authors based on a review of the abstracts . These articles were then assessed for inclusion criteria based on a review of the full text of each article . Medline ( PubMed ) was also searched for previous systematic reviews on the diagnosis of Chagas disease . The following strategy was used: ( “diagnosis”[Subheading] OR “diagnosis”[All Fields] OR “diagnosis”[MeSH Terms] ) AND Chagas[All Fields] ) AND systematic[sb] = ( Diagnosis Chagas ) AND systematic[sb] ) . This search identified 12 studies , one of which was considered potentially relevant based on a review of abstracts [6] . This systematic review , however , was excluded after a review of the full text . Eight full text articles were provided by a study author ( RLT ) as being potentially relevant . In addition to reviewing these articles , we reviewed the bibliographies of all included studies , as well as bibliographies of studies excluded based on the criteria of having a case control rather than cohort design . We also performed a search of LILACS to capture articles that may have been missed by other methods . Our search strategy was to use the search term “Chagas , ” limit study type to cohort studies and limit clinical aspect to “diagnosis . ” This identified 15 studies , of which none met our study criteria . We also performed this search using “Trypanosoma cruzi” and “T . cruzi” as the search term . We found no additional studies . Abstracts and full texts were evaluated for inclusion criteria by three blinded reviewers; one assessed all articles and each of the other two reviewers assessed about half of the articles . Information regarding the characteristics of each study and the data needed to create a contingency ( “2×2” ) table comparing each index test with its reference standard were abstracted into a worksheet by two reviewers in a blinded fashion . If a study reported sera that were indeterminate by the index test , we considered indeterminate results as a positive index test , and incorporated them into the assessment of accuracy of the test whenever possible . The methodological quality of each included study was also independently assessed by two reviewers using the QUADAS tool ( Quality Assessment of Diagnostic Accuracy Studies ) [7] . Differences between reviewers were resolved through consensus discussion . Sensitivity , specificity , positive likelihood ratios ( LR+ ) , and negative likelihood ratios ( LR- ) with 95% CI's were calculated for each test and then pooled using a Der Simonian Laird Random Effects Model . For studies with cells in the 2×2 table containing a value of zero , 0 . 5 was added to all cells to avoid division by zero . Heterogeneity of pooled sensitivity and specificity was estimated with the I2 statistic , where a value for I2 of 0 indicates perfect homogeneity ( all of the variance is within study ) whereas a value of 1 . 0 indicates perfect heterogeneity ( all of the variance is between studies ) . The area under a summary ROC curve for pooled results was also calculated using a Der Simonian Laird Random Effects Model . Because there was little variation in specificity estimates , a bivariate method was not used . Calculations were performed with MetaDisC ver 1 . 4 ( Madrid , Spain ) . Sources of heterogeneity were assessed by meta-regression using the metareg command in Stata version 11 . 0 ( College Station , TX ) under a random effects model . The independent variable was the log of the diagnostic odds ratio ( DOR ) , which compares the odds of having a positive test result in those with T . cruzi compared to the odds of having a positive test result in those uninfected with T . cruzi [8] . It is calculated as ( TP/FP ) / ( FN/TN ) [9] . Within study variance was estimated by the standard error of the ln ( DOR ) , calculated as the square root of [ ( 1/TP ) + ( 1/FP ) + ( 1/FN ) + ( 1/TP ) ] [9] . A random effects meta regression does not assume that all variability exist within a study but also takes into account between study variability in the model [10] . Between study variance was estimated by a restricted likelihood method using an iterative procedure . A fixed effects meta-regression was not performed because this model assumes that all the heterogeneity can be explained by the posited covariates [11] while a random effects model allows for unexplained heterogeneity . Publication bias and small study bias were assessed with a Begg funnel plot using the metabias command in Stata Version 11 . 0 ( College Station , TX ) .
Eighteen studies met the inclusion criteria . Thirteen studies were included based on a search of Medline , one more was included from the personal file of a study author , and four more were included based on a review of citations of included studies and non-included case-control studies . The included cohort studies were published between 1988 and 2010 . All but one included study took place in Central America or South America . The one remaining study took place in Switzerland , although participants were Latin American immigrants . Seven studies ( 39% ) were conducted with blood bank donors . The remaining studies came from “field studies” , population surveys , or other settings . Twelve studies ( 67% ) used subjects with unknown symptoms , while the remaining studies classified participants as asymptomatic or with clinical evidence of Chagas disease . Participants consisted mainly of adults with a mean age in the mid twenties; 8 studies did not report any information on the age of subjects , while 10 did not report any information regarding the sex distribution of participants . The characteristics of each study included in this analysis are outlined in Table S1 . A QUADAS score was created by assigning one point to all QUADAS criteria answered positively , 0 . 5 of a point for studies in which it was unclear whether a criteria was met , and zero points when a study clearly did not meet a QUADAS criteria . Figure 1 summarizes the percentage of studies meeting each of the QUADAS quality criteria , and Figure 2 provides an overview of the quality criteria for each study . Of the 61 tests assessed , 44 were ELISA , 3 were immunofluorescence assays , 7 were indirect hemagglutination assays , 4 were immunochromatographic assays , one was a chemilimunescence assay , one was a Dot-ELISA , and one was a multiple antigen binding assay ( MABA ) . Antigens used in each of the assays are described in Table S2 . Twenty six of sixty one assays ( 43% ) did not report the antigen used in an assay; among those that did , 15 assays used a form of recombinant antigen while 28 used a fixed or whole form of the parasite . Of the 61 assays included in this review , 36 were commercial assays . Details regarding type of commercial assay used are also described in Table S2 . We were unable to use information regarding cutoffs to complete a threshold analysis because of the limited number of studies that reported cutoffs used . In fact , 39 ( 64% ) did not report the exact cutoff or even methodology used to determine the cutoff ( i . e . 2 . 5 SD's above the mean of uninfected patients ) . Because there is no single widely accepted reference standard test for assessing Chagas disease , included studies used a wide variety of methods to classify true positives and true negatives . Four of eighteen studies used a single test as the reference standard . Because the reference standard was either positive or negative , there was no issue of having discordant test results for the reference standard . However , a single reference standard test is unlikely to correctly classify the index test [2] . Four studies used a reference standard of 2 or more out of 3 serological tests being positive as a true positive; three studies considered 2 out of 2 tests as a true positive and everything else as a true negative; and two studies considered 3 out of 3 positive tests as a true positive . Two studies used latent class analysis , which uses the results of index tests to approximate the true unobserved disease state [12] to identify true positives and negatives . Three studies used another method or did not describe the method used for determining true positives and negatives . Details of the reference standard methodology are summarized in Table S1 . Only 4 of 18 studies reported the number of samples that were discordant by the combination of serological assays used to determine the true status of a sample . The area under the summary ROC curve ( Figure 3 ) for all assays was 0 . 99 ( SE = 0 . 002 ) . In the summary ROC curve , it appears that there may be a positive correlation between sensitivity and ( 1-specificity ) , suggesting that some of the heterogeneity in sensitivity and specificity estimates is due to the use of differing cutoffs . However , when we performed a Spearman rank correlation , there was little evidence of a threshold effect ( Spearman correlation coefficient = 0 . 04 , p = 0 . 76 ) . An apparent outlier in the summary ROC curve is a study [13] that reported a low sensitivity for nearly all assays . This reduced accuracy is likely due to the fact that each assay included only one recombinant antigen , while most conventional assays use a combination of multiple antigens or whole parasite extracts . Removing this outlier only increased the sensitivity slightly but did not change the specificity ( Table 1 ) . A Forest plot ( Figure 4 ) of sensitivity and specificity estimates for each study ( with 95% CI's ) reveals that specificity is high and consistent between studies , while estimates of sensitivity vary more widely . While we report summary estimates for sensitivity and specificity , they should be interpreted with great caution given the significant heterogeneity and design limitations of the included studies . Table 1 shows pooled results for several predetermined subgroup analyses . Studies with a QUADAS score above 10 ( better designed studies ) had a lower sensitivity than less well designed ( 80% [95% CI:79–82%] vs 96%[95%CI:95–96%] , p = 0 . 07 ) . ELISA tests and non-ELISA tests were similarly sensitive and specific . Commercial tests were more sensitive than non-commercial tests ( 95% [95%CI: 94–95%] vs 81% [95% CI: 80–83%] , p = 0 . 08 ) but had similar specificity ( 99% [95%CI:99–55%] vs 97% [95%CI:97–98%] , p = 0 . 57 ) . Sensitivity and specificity were higher in studies conducted with blood bank samples compared to tests evaluated in field studies ( 96% [95%CI: 94–97%] vs 88% [95%CI: 87%–89%] , p = 0 . 24 sensitivity and 99% [95%CI:99–99%] vs 96%[95%CI:96–96%] , p = 0 . 44 specificity ) . In the stratification by reference standard used , sensitivity was lowest among studies requiring 3/3 positive tests as a reference standard ( 42% [95%CI:35–49%] ) . Overall , sensitivity and specificity were low for studies using a single test as a reference standard ( 75% [95%CI:73–78%] ) and 93%[95%CI: 92–94%] , respectively ) . Results of the metaregression , which examined the independent effect of study design characteristics on the diagnostic odds ratio , are shown in Table 2 . Study design characteristics significantly associated with the diagnostic odds ratio included whether an ELISA or other assay was employed ( relative diagnostic odds ratio [RDOR] = 4 . 22 ) , whether a study utilized latent class analysis as the reference standard ( RDOR = 90 ) , and whether the index test was blinded to the reference standard ( RDOR 0 . 03 ) . Thus , studies that were blinded reported a lower estimate of diagnostic accuracy , while those using latent class analysis and those using an ELISA assay reported a higher estimate of accuracy . One limitation of this analysis is that the sample size for the metaregression is the number of studies , not the number of patients , and thus there may not be sufficient power to detect other important effects [10] . Publication bias is the tendency of smaller studies with null results to go unpublished while studies showing an effect are more likely to appear in the literature [10] . To assess this effect , we built a funnel plot ( Figure 5 ) of the log of the diagnostic odds ratio against the standard error of the log of the diagnostic odds ratio , an indicator for sample size . Each open circle in the funnel plot represents an individual assay and the line in the center represents the summary diagnostic odds ratio . A gap in the plot of missing values below the summary DOR line is an indication of publication bias . In a regression of the standardized effect estimates against their precisions , we found a positive association between smaller studies and a higher reported DOR ( slope = 4 . 47 , 95% CI: 3 . 6–5 . 3 , t = 10 . 47 , p<0 . 001 ) . The intercept of the fitted line can be interpreted as a measure of bias [14] ( Intercept: −0 . 58; 95% CI: −3 . 71 , 2 . 63; t = −0 . 35; p = 0 . 732 ) . A gap in the expected funnel shape of the scatter plot indicates that there may be publication bias , with a lack of smaller studies reporting negative findings .
Previous studies have reported extremely high values for the sensitivity and specificity of serologic tests for Chagas disease . For example , the recent systematic review by Brasil and colleagues found summary estimates of sensitivity and specific for ELISA of 97 . 7% ( 96 . 7%–98 . 5% ) and 97 . 5% ( 88 . 5%–99 . 5% ) respectively . Summary estimates for commercial ELISA were a pooled sensitivity of 99 . 3% ( 97 . 9%–99 . 9% ) and a pooled specificity of 97 . 5% ( 88 . 5%–99 . 5% ) . Despite these supposedly high sensitivity/specificity levels for tests , a number of groups recommend or advise the use of multiple tests for accurate diagnosis [6] , [15] suggesting that many experts are still skeptical of the high accuracy of individual tests reported in the literature . As can be seen in the QUADAS table ( Figure 2 ) , there are several biases inherent in the manner in which the reference standard was applied . In some studies , sera with results borderline by the index test were excluded [16] , [17] , [18] , thereby inflating estimates of sensitivity and specificity . Several studies based the decision to apply the reference standard on the results of the index test , incorporated the index test into the reference standard , or did not apply a uniform reference standard to all sera in the study . In one study , only patients with a positive index test and a random selection of negative samples received the reference standard [19] . In other studies RIPA was only applied as part of the reference standard if the index or reference test was positive , while those initially negative either did not receive this test or only a random sample received the RIPA test [17] , [20] . In another study , only samples positive by the index ELISA test were submitted to another ELISA test , and only those positive by both ELISA tests were submitted to a Western Blot . Those reactive by all three tests were considered positive . However , only those positive by the index test were even eligible to be considered positive; any false negatives misclassified by the index test received no other verification and would be assumed to be true negatives . One study [21] repeated tests for which the index test and reference standard test were discordant . However , when the index test is applied in clinical settings , there will be no such gold standard to monitor the veracity of a test . In other studies [16] , [22] , initially positive or indeterminate samples were evaluated in duplicate and considered positive only when a repeated test was positive . Therefore , it is important to acknowledge that the reported accuracy in many studies is not that of the index test alone , but rather the accuracy of the entire testing strategy that was implemented . An important source of bias in these studies is dealing with samples that are discordant by the reference standard . In four of the eighteen studies , discordant samples were reported to have been discarded from sensitivity and specificity calculations ( as sensitivity and specificity of a test cannot be calculated with knowledge of the true disease status ) . This would tend to inflate estimates of sensitivity and specificity . Although eight studies reported the manner in which they handled samples discordant by the reference standard , only four reported the number of samples which were discordant by the reference standard . In these four cases , the percentage of discordant samples out of the total was 7% ( 23/335 ) [17] , 4% ( 40/1025 ) [18] , 4% ( 17/398 ) [23] , and 0 . 3% ( 3/999 ) [24] . In a study by Zicker and colleagues , results for IFA and HA were reported for all sera without specifying a clear reference standard . We chose to consider all specimens with both serology tests positive as a true positive and all other sera as true negatives . This study reported a much lower sensitivity ( 88%; 95%CI: 84%–91% ) and somewhat lower specificity ( 94%; 95%CI: 93%–95% ) than the pooled estimates ( 97%; 95% CI: 96%–98% and 96%; 95%CI: 95%–96% ) for the other studies in that subgroup ( those using a reference standard of 2/2 positive tests ) . All other studies in this subgroup reported a sensitivity greater than 95% and specificity greater than 95% . Because we created the reference standard groupings , we are certain that there were no discordant samples discarded , most likely accounting for the approximately 10% discrepancy in sensitivity estimate . This raises the question of whether other studies had a significant number of unreported discordant samples that would inflate the sensitivity of their tests . Study design characteristics associated with biased estimates of sensitivity and specificity included different reference tests for those with a positive and negative index tests , failure to blind or mask , and case-control instead of cohort design [3] . In our study , only five of eighteen studies reported that the results of the diagnostic tests and reference tests were both blinded to the other , and only thirteen of eighteen studies applied a uniform reference standard to all or a random sample of subjects . Our findings are consistent with these general principles of diagnostic study design , as we found increased estimates of sensitivity and specificity in studies with lower quality ( QUADAS<10 points ) and in unblinded studies . This highlights the serious limitations in the existing literature . We are concerned that the sensitivity of current tests is lower than generally reported in lower quality studies , leading to the potential for underdiagnosis . This is a particular concern when these tests are used as a screening test for blood bank samples . The use of case control design continues to be the most widespread method to evaluate diagnostic tests for T . cruzi infection . Based on this analysis , we recommend that future studies use a prospective cohort design with clear reporting , masking , and an appropriate reference standard to provide a more accurate estimate of the diagnostic accuracy of tests for Chagas disease . Our results also suggest that better tests are still needed to assure the safety of transfusions and to improve the public health of countries where the disease is endemic .
|
Chagas disease , an infectious disease endemic to Latin America , is caused by the protozoan parasite Trypanosoma cruzi . T . cruzi can be transmitted through blood transfusions , organ transplants , or from mother to fetus , although it is most commonly transmitted through insect vectors . Infections can remain silent for many years before manifesting as potentially fatal damage to the cardiac and/or digestive system . Diagnosis of Chagas disease during its chronic asymptomatic phase is crucial to preventing future infections with T . cruzi and is often performed using serological tests that detect antibodies in the blood . Because there is currently no gold standard for serological diagnostic tests , multiple forms of serologic testing are often used in conjunction . The purpose of this study was to compare reports on the accuracy of serological tests . After limiting studies by certain criteria , the authors found a lower estimate of accuracy than has previously been reported in the literature and suggest quality improvements that can be made to standardize future reports .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"biology"
] |
2012
|
A Systematic Review of High Quality Diagnostic Tests for Chagas Disease
|
Although systemic immunity is critical to the process of tumor rejection , cancer research has largely focused on immune cells in the tumor microenvironment . To understand molecular changes in the patient systemic response ( SR ) to the presence of BC , we profiled RNA in blood and matched tumor from 173 patients . We designed a system ( MIxT , Matched Interactions Across Tissues ) to systematically explore and link molecular processes expressed in each tissue . MIxT confirmed that processes active in the patient SR are especially relevant to BC immunogenicity . The nature of interactions across tissues ( i . e . which biological processes are associated and their patterns of expression ) varies highly with tumor subtype . For example , aspects of the immune SR are underexpressed proportionally to the level of expression of defined molecular processes specific to basal tumors . The catalog of subtype-specific interactions across tissues from BC patients provides promising new ways to tackle or monitor the disease by exploiting the patient SR .
Breast cancer ( BC ) research has largely focused on understanding the intrinsic properties of the primary tumor in order to therapeutically target key molecular components that drive progression within the tumor epithelial cells [1] . For example , tamoxifen and trastuzumab target the estrogen and human epidermal growth factor receptors ( ER , HER2 ) whose expression levels in tumors define the traditional clinical subtypes of BC . The vast majority of BC-related genomic studies have focused on bulk tumor samples that are expected to be enriched for neoplastic epithelial cells [2] . These efforts have produced subtyping schemes that classify patients into groups based on the similarity of expression of diverse molecular markers and processes [3–9] and generated gene signatures that can predict patient prognosis and benefit from therapy [10–13] . Cancers however are much more than an autonomous mass of epithelial cells . They constitute multicellular systems capable of bidirectional interactions with neighboring non-malignant cells and extracellular components i . e . the tumor microenvironment [14–16] . Tumor-microenvironmental interactions are necessary for tumor progression and drug sensitivity [16 , 17] and are becoming better understood [18–21] . In fact , several genomics studies of the BC microenvironment , including our efforts , show that the microenvironment reflects its tumor and harbors prognostic information [22–24] . However , we also recently established that the primary tumor and its microenvironment does not harbor accurate prognostic signals in approximately 20% of BC patients [9] . Specifically , these patients are consistently misclassified by all hallmarks of breast tumors defining tumor epithelial cells ( such as proliferation and ER status ) and their microenvironment ( such as the infiltration of immune cells , angiogenesis and fibroblast activation ) . The systemic response ( SR ) in cancer patients refers here to the perturbations that occur in peripheral blood cells , which include immune effector cells and circulate throughout the body . The fact that a tumor exerts systemic effects ( via eg soluble or exosomal factors ) may provide an explanation for the clinical observation that patients with one tumor have an increased risk of developing several independent tumors , and that removal of primary cancer improves the survival of patients with distant metastases at the time of diagnosis [25] . In addition , since ER positive ( ER+ ) BC tends to recur as long as 10–15 years after surgical removal of the tumor , it is important to understand systemic factors governing late recurrence and therapeutic approaches that target beyond the tumor site . In fact , there is a rapidly increasing understanding of the various means a tumor employs to favor metastasis in distant organs [26 , 27] . For example , an “instigating” BC can exploit the patient SR so that otherwise-indolent disseminated tumor cells become activated [27–32] . The SR has also been investigated in BC at time of diagnosis . Specifically , our recent comparison of blood profiles of BC patients and matched controls yielded a gene signature that reports the presence of BC [33] . This diagnostic signature is specific to BC ( i . e . the test classifies women with carcinoma other than breast as negative ) , and the composition of genes and enriched pathways in the signature suggest that a cytostatic immune-related signal in the SR of patients is associated with the presence of a tumor . Finally , recent evidence demonstrates that engagement of systemic immunity is critical to the process of tumor rejection in genetically engineered mouse models [34] . This study is the first large-scale genomics effort to study the molecular relationships between patient SR and primary tumor . We generated and analyzed expression profiles from peripheral blood and matched tumor cells in 173 BC patients . First , our results highlight how the patient SR is especially relevant to BC immunogenicity . Second , we present a novel tool entitled Matched Interactions across Tissues ( MIxT ) that starts by identifying sets of genes tightly co-expressed across all patients in each tissue . Then , MIxT identifies which of these gene sets and pathways expressed in one tissue are associated with gene sets and pathways in the second tissue by determining if their expression patterns in tumor and in the patient SR are tightly correlated . We find that there are very few such associations when all BC are considered . However , we do identify biological processes with significant associations between tumor and patient SR when we stratify our analysis by BC subtype . That is , we identify molecular processes in the tumor that are tightly co-expressed with ( different ) molecular processes in the SR across patients of a specific subtype . In particular , we detail how several tumor-permissive signals are associated between the tumor and SR of basal BC patients .
The Norwegian Women and Cancer ( NOWAC ) is a prospective population-based cohort that tracks 34% of all Norwegian women born between 1943–57 . In collaboration with all major hospitals in Norway , we collected blood samples and matched tumor from women with an abnormal lesion , at the time of the diagnostic biopsy or at surgery , before surgery and any treatment ( N ~ 300 , S1 Text ) . RNA preservation for blood samples obtained followed our methodology previously described [33 , 35] and detailed in S1 Text . RNA profiles from blood and tumor cells were generated using Illumina Beadarrays and data were processed following careful procedures ( S1 Text , S1A Fig ) . After quality control , our study retained matched blood ( SR ) and tumor profiles of 173 BC patients diagnosed with invasive ductal carcinoma , and blood profiles of 282 control women ( ie . women with no history of cancer with the exception of basal-cell and cervical carcinoma , which are both very common; Fig 1A ) . The controls are used to determine what constitutes a “normal” SR . BC patients and controls are comparable in terms of age , weight and menopausal status ( Fig 1B ) . Several groups including ours have defined intra- and inter- individual variability of blood gene expression in healthy individuals [35–38] . All together , these studies demonstrate that intra-individual changes that can occur between blood draws are strikingly smaller than the variation observed among samples collected from different individuals . In this study , most women were 50 year-old or older and postmenopausal at time of sampling . Each profile measures the expression of 16 , 782 unique genes ( S1 Text , S1A Fig ) . Almost all BC ( 95 . 4% ) are early-stage disease ( stage I or II ) . Several tumor RNA-based subtyping tools were applied including PAM50 [5] that defines the intrinsic subtypes including luminal A ( lumA ) , luminal B ( lumB ) , normal-like ( normalL ) , basal-like ( basalL ) , and her2-enriched ( her2E ) . The hybrid subtyping scheme partitions ER+ tumors according to their intrinsic subtype and partitions ER- tumors according to their HER2 status [9] ( S1 Text , S1B and S1C Fig ) . In our dataset , all intrinsic luminals ( lumA and lumB ) and most normalL tumors ( 85 . 2% ) are ER+; however , ~40% of basalL and ~50% of her2E BC are ER+ ( Fig 1C , S1 Table ) . We also applied the Cartes d’Identité des Tumeurs ( CIT ) [8] subtyping scheme , which includes a ‘molecular-\ apocrine’ ( mApo ) subtype enriched for ER-/HER2+ tumors ( 78 . 6% ) and the highly immunogenic ER+ luminal C ( lumC ) subtype enriched for ER+/basalL ( 39 . 1% ) . Fig 1C and S1 Table depict the relationships between these three schemes . Although the IntClust ( IC ) subtyping scheme [6] is based on gene expression and DNA copy number profiles simultaneously , subtypes can be inferred using a reported RNA-based surrogate algorithm [7 , 39] . S1 Table reports when subtypes from other schemes are enriched in each IC subtype . Most notably , IC1 and IC9 are enriched for CIT lumB; IC3 , IC7 and IC8 are enriched for lumA; IC4+ is enriched for normalL and at lesser extent CIT lumC , IC5 enriched for mApo-her2E-HER2+ , and IC10 enriched for basalL and ER-/HER2- . IC2 , IC4- , and IC6 include very few patients ( n < 10 ) and were therefore not further considered in our downstream analyses . Restricting our attention to tumor profiles , we performed sparse hierarchical clustering with complete linkage using a permutation approach to select the tuning parameter that weights each gene to compute the dissimilarity matrix [40] . The resulting clusters were strongly associated with BC subtypes for all three RNA-based schemes ( Fig 1D upper ) , which confirms that the transcriptional fingerprint of BC subtypes are also ubiquitous in our tumor samples . When restricting our attention to SR profiles , this unsupervised analysis does not identify patient clusters enriched for any given subtype across the three schemes ( Fig 1D lower ) , suggesting that the transcriptional fingerprint of BC subtypes is not the predominant signal in the patient SR . We then asked if there are genes in the patient SR whose expression covaries with the state of the pathological variables ER and HER2 measured in the primary tumor . Although both are key drivers in BC , neither was found to be associated with individual gene expression changes in the patient SR ( limma , linear models for microarray data , false discovery rate , fdr ≤ 0 . 2 , Fig 1E; S1 Text ) . Similarly , we asked if there are genes in the SR that are markers of tumor subtype ( n patients > 10 ) . For the intrinsic , hybrid , and IntClust subtypes , only the ubiquitin ligase RFWD3 is highly expressed uniquely in the SR of lumA patients , and TIMP3 , an inhibitor of matrix metalloproteinases , is highly expressed uniquely in ER+/her2E patients ( Fig 1E , S2 Fig ) . For the CIT subtypes [8] , we found 70 univariate gene markers in the SR of patients of the lumC subtype . The genes are primarily involved in general cellular processes such as protein processing or transcription in blood cells ( fdr ≤ 0 . 2 , Fig 1E , S3 Fig ) . The lumC subtype is defined by strong activation of several immune pathways at the site of ER+ tumor ( i . e . antigen presentation and processing pathway , hematopoietic cell lineage , NK cell mediated cytotoxicity , T-cell receptor signaling and Toll-like receptor signaling ) [8] , suggesting that the SR is informative in cases where the primary tumor exhibits strong immune properties . To compare genome-wide molecular changes in tumor and SR across patients , we used WGCNA-based clustering to define sets of tightly co-expressed genes ( termed modules ) in tumor and blood , respectively [41] ( S1 Text ) . Briefly , we opted for a distance measure based on topological overlap , which considers the correlation between two genes and their respective correlations with neighboring genes [42] ( S1 Text ) . The WGCNA cut and merge routine [43] after clustering identified 19 and 23 modules in the patient tumor and SR , respectively ( S4 Fig; S1 Text ) . Each of these modules can be considered as a unique and stable pattern of expression shared by a significant number of genes . Modules of the primary tumor are enriched for genes from a broad range of BC hallmarks including angiogenesis ( salmon module ) , extracellular matrix reorganization ( greenyellow ) , proliferation ( green ) , and immune response ( brown and darkturquoise ) ( S2 and S3 Tables , S1 Text ) . For example , the proliferation tumor module is enriched for mitotic cell cycle-related genes ( green , n = 1064 genes; weight01 Fisher test [44] , p-value < 2e-17 ) including the well-known marker of proliferation MKI67 , 12 serine/threonine kinases that are used in the calculation of the mitotic kinase score ( MKS ) [45] , and several components of the Minichromosome Maintenance Complex ( MCM ) . Modules of the patient SR are often enriched for genes involved in either general cellular processes such as translation ( black ) and transcription ( grey60 ) , or immune-related processes such as inflammatory response ( brown , green ) , B-cell response ( saddlebrown ) , innate immune response ( greenyellow ) ( S4 and S5 Tables ) . Thus , seven SR modules are enriched in genes that are specifically expressed in immune cells [46] ( “iris” signature set in S5 Table; Fisher’s Exact Test FET fdr < 0 . 05 ) . We constructed a web-based system to visualize gene expression networks , heatmaps and pathway analyses of the modules in each tissue at http://mixt-blood-tumor . bci . mcgill . ca . In a network , genes are represented by nodes ( colored by their module membership ) that are connected by edges whose length corresponds to their level of co-expression across patients [47] . When selecting only strong gene-gene correlations ( topological overlap > 0 . 1 ) and removing isolated nodes , the SR network has ~20% more genes than the tumor network ( Fig 2A and 2B ) . Moreover , the SR network has approximately twice as many edges ( 89 , 465 connections between genes ) than the tumor network ( 50 , 617 connections between genes ) . Thus , the underlying patterns of expression of the tumor genes ( and modules ) are more dissimilar from each other than the patterns of expression of the SR genes ( and modules ) . In both tissues , the edges that span between modules reflect natural overlaps between cellular process ( Fig 2A and 2B ) . For example in tumors , angiogenesis-related genes of the salmon module are strongly co-expressed with genes of the greenyellow module involved in extracellular matrix remodeling . In blood , modules enriched for genes involved in general cellular processes such as translation ( black ) , RNA processing ( violet ) , and RNA splicing ( darkred ) are also heavily connected to each other . We first investigated the relationships between the expression pattern of each module and patient clinicopathological attributes . Towards this end , each gene of a module is used to rank the patient samples ( S1 Text ) . In particular , the sum of gene ranks ( ranksum ) for each patient provides a linear ordering of the patient samples . Association tests then compare the ranksum values of patients with the attribute of interest eg tumor subtype ( S1 Text ) . When we consider tumor modules , the expression pattern of the green module ( S5A Fig ) , previously established to be enriched for proliferation-related genes ( S2 Table ) , ranks basalL , her2E and lumB tumors significantly higher than lumA and normalL tumors ( ANOVA p-value < 1e-34 , S5B Fig ) . In fact , we observe that the expression pattern of nearly every module is associated with BC subtype ( 15 of 19 modules , Fig 2C , fdr ≤ 0 . 15 ) . Moreover , many tumor modules are associated with the proliferative state of the tumor encoded into the MKS score [45] ( Pearson correlation , fdr ≤ 0 . 15 ) or with ER status ( ER+ vs ER- , t-test , fdr ≤ 0 . 15 ) , two variables that are strongly embedded in the definition of BC subtypes ( Fig 2C ) . These results are consistent with our previous claim that patient subtype is a predominant signal in the primary tumor . Several tumor modules are associated with HER2 status of the tumor , however there are fewer such modules ( n = 6 ) when compared with the proliferative state or ER status of the tumor ( Fig 2C ) , suggesting that transcriptional fingerprint of HER2 is not as ubiquitous in tumor samples . A small number of modules are associated with the lumC subtype , including the brown module enriched for T-cell and inflammatory response genes ( S2 Table ) . This is again consistent with the fact that this is a highly immunogenic subtype [8] ( lumC versus not lumC , t-test , fdr ≤ 0 . 15 , Fig 2C ) . HER2 status , the lumC subtype and tumor size are all associated with modules of the patient SR ( Fig 2D , t-test fdr ≤ 0 . 15 ) . Although we did not find univariate gene markers in blood associated with HER2 status , the saddlebrown SR module is significantly underexpressed in patients with HER2+ tumors compared to other BC subtypes and controls ( fdr = 0 . 07 , S6A Fig ) and is enriched for genes involved in B-cell receptor signaling and proliferation ( including BLK , CXCR5 , CD19 , CD79A , CD79B and FCRL5; S4 and S5 Tables ) . Four SR modules are associated with the immunogenic lumC subtype; one of these modules are also associated with tumor size ( Fig 2D , S6B and S6C Fig ) . Among the 70 univariate gene markers in blood of lumC tumors identified earlier , 31 are included in the darkgreen SR module predominantly underexpressed in lumC patients in comparison to other BC subtypes ( fdr = 0 . 02 , S6B Fig ) . In fact , all four SR modules associated with the lumC subtype are underexpressed compared to other BC subtypes and control samples ( S6B and S6C Fig ) . This includes the purple module highly enriched for genes involved in T-cell ( thymus ) homing ( CCR7 , LTA , LTB , VEGFB , HAPLN3 , SLC7A6 , SIRPG , BCL11B0 ) and activation ( CD47 , TNFRSF25 , MAL , LDLRAP1 , CD40LG ) which are underexpressed in lumC patients ( fdr = 0 . 04 , S6B Fig ) . Genes in the cyan modules are also found underexpressed in patients with large ( > 2cm ) tumors compared to other BC patients and controls ( Fig 2D , S6C Fig ) . Finally , specifically for patients with large tumors , both the darkgrey module , which is enriched for MYC target genes , and the greenyellow module , which is enriched for genes involved in the lymphoid cell-mediated immunity ( including GZMH , GZMB , GZMM , KLRD1 , PRF1 , KLRG1 , and GNLY; S4 and S5 Tables ) , are underexpressed compared to the remaining BC patients and controls . Together these results indicate that distinct SR are detected in BC patients with HER2+ , lumC and/or large tumors , and that overall the patient immune response is underexpressed compared to patients of other subtypes and controls . These results also highlight the importance of distinct immune components for each of these disease groups . In particular , patients with HER2+ tumors exhibit low expression of genes specifically expressed in B-cell compared to patients with other BC subtypes . Patients with lumC tumors exhibit low expression of genes involved in T-cell homing and function compared to patients with other BC subtypes . Patients with large tumors ( >2cm ) exhibit low expression of genes involved in lymphoid cell-mediated immunity compared to patients with smaller tumors . Our analysis to this point identified modules within each tissue independently . Our focus here is on the relationships between tissues by asking if specific biologies in one tissue are correlated with ( possibly distinct ) biologies in the second tissue . To do this , we constructed a software entitled MIxT ( Matched Interactions across Tissues ) that contains the computational and statistical methods for identifying and exploring associations between modules across tissues ( http://mixt-blood-tumor . bci . mcgill . ca ) . Using MIxT , we first ask if genes that are tightly co-expressed in the primary tumor are also tightly co-expressed in the SR , and vice versa ( Fig 3A , S1 Text ) by investigating the gene overlap between tumor and SR modules ( Fisher’s Exact Test FET , fdr < 0 . 01 ) . Genes that retain strong co-expression across patients regardless of tissue type are likely to be involved in the same biological functions in both tissues as a “system-wide” response to the presence of the disease ( even if patterns of gene expression across tissues might differ ) . Most modules , regardless of tissue , have significant overlap with three to five modules in the other tissue ( Fig 3A ) . In some cases , it appears that a single ( large ) module in one tissue is in large part the union of several smaller modules from the other tissue . For example , the brown tumor module has 2765 genes including many involved in immune-related processes ( T-cell costimulation , the IFN-gamma pathway and inflammation , S2 and S3 Tables ) . All of these genes/processes show very strong co-expression in the tumor however , in the SR , these genes divide into four distinct patterns of co-expression ( Fig 3A ) , captured by four different modules: brown ( inflammation ) , greenyellow ( cytolysis and innate immune response ) , saddlebrown ( B-cell ) and pink ( TNFA inflammatory response ) ( S4 and S5 Tables ) . Of note , MIxT identifies three modules in each tissue ( SR and tumor ) that do not have significant overlap with any module in the other tissue ( Fig 3A ) . For tumors , this includes the purple module enriched for genes involved in estrogen response , the lightcyan module enriched for genes involved in hemidesmosome assembly and cytoarchitecture , and the greenyellow module enriched for genes involved in ECM organization ( Fig 3A , S2 and S3 Tables ) . For the SR , this includes the turquoise module enriched for genes expressed in erythrocytes and involved in hemoglobin production , the purple module enriched for genes in translational termination , and the green module enriched for genes involved in inflammation and specifically expressed in myeloid cells ( Fig 3A , S4 Table ) . This suggests that these processes and responses are either specific to a tissue type ( eg ECM organization specific to tumor , and hemoglobin production specific to blood cells ) or that the co-expression of genes involved in a defined process is unique to a particular tissue ( eg genes specifically co-expressed in peripheral myeloid cells ) . There is only one instance where a single tumor module has significant overlap with only a single SR module: darkturquoise modules of size = 86 and 97 genes in SR and tumor , respectively with 50 common genes , including 20 involved in the type 1 IFN signaling pathway ( S2 and S4 Tables ) . Although these two “mirrored” modules share many genes , their patterns of expression are significantly different between the two matched tissues ( Fig 3B , correlation between ranksums p-value > 0 . 05; S1 Text ) , hinting at a non-concordant expression of the local ( in tumor ) and systemic ( in blood ) IFN-1 mediated signals . Whereas the previous section considers interactions defined by a large number of shared genes between a tumor and a SR module , we also examined more general notions of interactions in MIxT . Here we identify tumor and SR modules that have similar expression patterns ( ie both modules linearly order the patients in very similar manner in both tissues ) but do not necessarily share any genes in common . More specifically , MIxT derives estimates of significance for interactions using a random permutation approach based on the Pearson correlation between ranksums of gene expression in modules across tissues ( S1 Text ) . This type of interaction detects a biological process or response in the primary tumor that is tightly correlated ( or anti-correlated ) with a ( possibly distinct ) biological process or response in the SR , and vice versa . The specific expression pattern in the tissues allows us to then postulate the functional nature of the interaction across tissues . MIxT identified only one tumor module ( of 19 ) that interacts with only a single SR module ( of 23 ) across all patients ( MIxT statistic; p-value < 0 . 005 ) . The paucity of pan-BC interactions across tissues suggest the need to stratify by patient subtype . After stratification for each of the five subtyping schemes ( clinical , PAM50 , hybrid , CIT , and Intclust ) ( Fig 4A ) , we identified 53 interactions involving 15 tumor modules and 19 SR modules ( MIxT statistic; p-value < 0 . 005; Fig 4B , S7 Fig ) . Tumor and SR modules are indicated in columns and rows of Fig 4B , respectively . A non-empty cell corresponds to a significant interaction with color used to indicate in which subtype the association is found , grouping together similar subtypes across schemes ( eg basalL tumors of the pam50 and CIT schemes ) . Nearly all interactions are significant in only a single subtype ( four exceptions indicated by orange arrows , Fig 4B ) . For some subtypes , a single stimulus in the tumor affects several biological processes in the patient SR . For example , within the ER+/HER2- subtype and only within this subtype , the pink tumor module , enriched for genes involved in alternative splicing , is associated with three SR modules , enriched for a diverse range of biological processes ( orange rectangle in Fig 4B ) . The brown tumor module , which is enriched for genes involved in immune processes ( S2 Table ) , has several interactions with SR modules across several subtypes ( orange rectangle in Fig 4B ) . This includes interactions specific to normalL , lumB and IC9 but also several distinct interactions within the ER-/HER2- and basal subtypes . This suggests that immune signals expressed in tumor are associated with changes in expression of different molecular processes in the patient SR for a broad range of subtypes . As alluded to earlier , only a few interactions are significant in two distinct subtypes simultaneously . For example , the brown tumor module is associated with green SR module in both ER-/HER2- and lumB although the directionality of the association differs between the two cases . More specifically , patients with high ranksums in the brown tumor module have low ranksums according to the green SR module , if the patient is of the ER-/HER2- subtype ( Fig 5A , 5C and 5E , MIxT statistic , p-value = 0 . 004 ) . At the same time , patients with high ranksums in the brown tumor module have high ranksum with respect to the green SR module , if the patient is of the lumB subtype ( Fig 5B , 5D and 5F MIxT statistic , p-value < 0 . 004 ) . In this manner the direction of correlation between the biological processes of the brown tumor module and of the green SR module is determined by the subtype of the patient . For the brown tumor module in both subtypes , patients with a high ranksum ( on the left of the ordering in Fig 5B or 5C for both subtypes ) have the strongest immune signals in the tumors . This is because most of the immune-related genes in this brown module ( within the red sidebar in Fig 5B and 5C , S3 Table ) have highest expression in these patients . This includes genes involved in T-cell stimulation ( incl . CD3 , CD4 , CD5 , ICOS , several HLA-DR , -DP , -DQ ) , IFNɣ signaling ( IFNG , IRF1-5 , ICAM1 , IFI30 , HLA-A -B -C ) and inflammation ( incl . several interleukins , chemokines ) . For the green SR module in both subtypes , a high ranksum indicates an inflammatory SR ( patients on the right in Fig 5E for ER-/HER2- , and patients on the left in Fig 5F for lumB ) . This is because almost every inflammation-related genes ( incl . IFNAR1 , IL15 , TLR2 , IL18RAP , RNF144B ) , and B-cell proliferation genes ( incl . BCL6 , IL13RA1 , MIF , IRS2 ) ( within the red sidebar in Fig 5E and 5F , S5 Table ) have highest expression in these patients . Thus , ER-/HER2- patients with low immune activity at the tumor site have a high inflammatory SR ( right side of Fig 5C and 5E ) . In fact , the level of the inflammatory response in these BC patients is higher than healthy controls ( Fig 5I , t-test , p < 0 . 001 ) . However , for the lumB subtype , the relationship between tumor and SR is reversed . Here , it is the patients that have high immune activity at the tumor site that have a high inflammatory SR ( left side Fig 5D and 5F ) . In fact , the CIT subtyping scheme calls these patients on the left side as belonging to the lumC subtype ( Fig 5H ) , the highly immunogenic ER+ subtype . In these lumB patients the inflammatory response is also higher than in healthy controls ( t-test , p-value < 0 . 01; Fig 5J ) . Altogether these results indicate that a high inflammatory SR is observed in both ER-/HER2- and ER+/lumB patients but increase in systemic inflammation is associated with distinct immune activity at the tumor site depending on subtype . Three tumor modules are enriched for genes within amplicons prevalent in BC [48] ( highlighted in orange in Fig 4B , S3 Table ) . Two modules , the darkgrey and turquoise tumor modules , contain 68 genes ( of 110 ) and 48 genes ( on 71 ) located within the 16p11-13 amplicon highly prevalent in luminal tumors [48] , respectively ( S3 Table ) . The darkgrey module interacts with two distinct SR modules for the lumA and ER+/HER2+ subtype , respectively ( S8A and S8B Fig ) . Tumors of both subtypes that over-express genes in the darkgrey module ( left hand side S8C and S8D Fig ) are likely amplified in 16p13 . In these patients , the presence of this amplification is correlated with changes in expression of specific processes within the patient SR and these processes are distinct depending on subtype ( S8E and S8F Fig , p < 0 . 005 in both cases ) . S8G and S8H Fig depicts associations between the presence of this amplification and patient clinico-pathological attributes . For example , in ER+/HER2+ patients ( S8H Fig ) , the presence of 16p13 amplification is correlated with the luminal score of the tumor . In the lumA subtype , patients with the highest expression of the lightyellow SR module are significantly different than healthy controls ( S8I Fig ) , and in the ER+/HER2+ subtype , patients with the lowest expression of the salmon module are significantly different than healthy controls ( S8J Fig ) . The third module enriched for genes involved in BC amplifications is the darkgreen tumor module . This module contains 43 ( of 99 ) genes within the 8q23-24 amplicon prevalent in basal and her2E tumors [48] ( S3 Table ) . Most associations with patient SR modules are specific to the basalL subtype ( Fig 4B ) and again suggest that basalL tumors that harbor this amplification have concomitant changes in expression of specific molecular processes in patient SR . Approximately one-fourth of the interactions identified by MIxT are specific to ER-/HER2- , IC10 and basalL subtypes , indicating that the tumor and SR interact strongly in this family of BCs ( Fig 4B ) . We study two tumor modules in greater depth here: the brown immune-enriched module and the darkgreen 8q-enriched module , and their interactions with SR modules in basalL patients ( Fig 6A–6C ) . Here the brown tumor module interacts with one ( tan ) SR module enriched for genes involved in TOR signaling and cell proliferation ( Fig 6A and 6B ) . BasalL patients with low immune activity at their tumor site ( right side of brown tumor module ) have low expression of the tan SR module , and this expression is significantly lower than healthy controls ( boxplots in Fig 6B , t-test p < 0 . 0005 ) . The darkgreen tumor module interacts with four SR modules in basalL patients ( Fig 6A and 6C ) . High expression of genes in 8q is associated with high expression of the green SR module . This module is enriched for genes involved in inflammation . For the remaining three SR modules associated with the 8q-enriched tumor module , almost all genes in these modules are underexpressed when 8q genes are highly expressed ( ie . the patient orderings are reversed compared to the darkgreen tumor module ) . These SR modules contain genes involved in general cellular processes of blood cells ( RNA/protein processing , cell proliferation; darkgreen module ) , genes involved in cytolysis and lymphoid cell-mediated immunity ( greenyellow module ) , and MYC and CD5 target genes ( darkgrey module ) ( Fig 6A–6C , S5 Table ) . The increase in inflammatory SR and the decrease in the three other molecular processes in the SR of basalL patients whose tumor is amplified on 8q are all significantly different from how these processes are expressed in healthy controls ( boxplots in Fig 6C ) . Overall , we identified one distinct signature in the SR of basalL patients with low immune activity at their tumor site and several immuno-suppressive signals in the SR of basalL patients whose tumor is amplified on 8q .
Molecular profiles of peripheral blood cells and matched tumors were generated and compared for a large cohort of BC patients part of the NOWAC study . The NOWAC consortium provides a highly curated population-based study with extensive gene expression profiling across several tissues from BC patients and controls [35 , 49] . A careful design and our extensive experience in blood-based expression profiles enable a detailed molecular description of the patient SR to the presence of BC where blood molecular profiles represent effectively an “averaging” over the transcriptional programs of the different types of cells in blood . We first asked if the SR could provide accurate univariate markers of tumoral properties such as ER status or subtype . Although thousands of transcripts are differentially expressed in tumors between ER+ and ER- BC [9 , 50] , there is no gene in SR that can reliably predict ER status of the primary tumor . Moreover , the SR does not inform on the intrinsic BC subtype of the tumor such as lumA , lumB or basalL subtype or on IntClust subtypes . Interestingly , univariate markers in the patient SR were only identified for the CIT lumC subtype defined as particularly immunogenic ER+ tumors [8] , suggesting that the SR is informative in cases where the primary tumor exhibits strong immune properties . This is consistent with previous reports that uses blood transcriptomics as a gateway into the patient immune system [51–53] and which is extensively used in the context of autoimmune and infectious diseases [54–56] . This result suggests that it is also applicable in cancer such as BC . To further investigate the molecular changes in the patient SR , we extended our analyses to multivariate approaches where genes are combined into sets or “modules” . In particular , we performed cluster analysis to partition the genes of both tumor and SR profiles into modules with each module representing a distinct pattern of expression across patients . Our user-friendly website ( www . mixt-blood-tumor . bci . mcgill . ca ) provides access to these modules built in each tissue , enables investigation of their expression profiles in each tissue and allow user-defined queries of gene , gene sets , and pathway of interest . Further , our MIxT approach estimates gene module expression in both tissues and find significant associations between modules across tissues in a representative cohort of BC patients . In our dataset , the primary tumor and SR have approximately the same number of modules ( 19 and 23 , respectively ) but their gene composition is qualitatively different . Not surprisingly , many modules in tumors were enriched for genes involved in hallmarks of cancer , while SR modules were enriched for either general cellular processes or specific immune responses . Only one module involved in the IFN-I pathway is highly conserved in both tumor and SR , although the common genes had markedly different expression patterns between the two tissues . This is important as it establishes that genes , whose expression patterns may act as good markers in the primary tumor , are not necessarily expressed in the same manner within blood cells . Our multivariate approach was able to identify modules from the patient SR that could reliably identify not only lumC but also HER2+ and large ( > 2cm ) tumors . These three cases are among the most immunogenic subtypes of BC and are of relatively poor prognosis . For these patients , gene expression in blood cells is mostly decreased compared to other BC and controls . This result also highlights the importance of distinct immune components of the SR for each of these disease groups: B-cells for HER2+ tumors , T-cells for lumC , and aspects of the cellular immune response for large tumors . Interestingly , a previous study showed that her2E tumors have the highest B-cell infiltration and expression of B-cell receptor gene segments , although this was not predictive of improved patient survival [57] . Our study finds an impaired systemic B-cell response specifically in HER2+ patients , consistent with an inefficient anti-tumoral response in these patients , potentially due to a dysfunctional antigen receptor response and cell development . We could also speculate that the dysfunctional thymic T-cell homing signature in lumC patients reflects the well-documented effect of estrogen on thymic T lymphopoiesis [58–61] in patients diagnosed with a highly immunogenic ER+ tumor . These associations would certainly require validation in follow-up studies . Finally , MIxT focuses on molecular associations between tissues and provides a holistic view of molecular changes in BC patients . Although the focus here is towards gene expression of blood and matched tumor , our approach could be extended to multiple tissues ( eg . blood-microenvironment-tumor ) or other levels of molecular data ( eg . DNA level somatic aberrations , gene and miRNA expression , epigenetic profiles ) . Interestingly , associations between BC tumor and patient SR are heavily dependent on subtype . Only one interaction between tumor and patient SR is identified when all BC patients are considered in the analysis but many are identified when we first stratify patients by BC subtype . This is perhaps not surprising given that there is a great deal of molecular heterogeneity between BC subtypes making “one SR fitting all” highly unlikely . We identified molecular stimuli in tumors that change patient SR in multiple ways only for patients within a particular subtype . For example , expression of genes involved in alternative splicing in ER+/HER2- tumors is associated with changes in expression of multiple processes in SR of patients and those associations are observed only within this specific subtype . Of note , immune signals measured at the tumor site are associated with distinct SR across a broad range of subtypes . Immune-related processes are known to be more or less expressed within every subtypes and have prognostic capacity in almost all subtypes [9] . Here we show that a change in immune activity at the tumor site is not associated with equal SR across subtypes . Furthermore , high immune signals in tumor is associated with the patient inflammatory SR in opposite ways depending if the patient is ER-/HER2- or lumB . The high inflammatory SR in ER-/HER2- patients ( with low immune activity at the tumor site ) and in lumB patients ( with high immune activity at the tumor site ) were both significantly different from how systemic inflammation is “normally” expressed in controls . Finally , we identify other examples of interactions between tumor and patient SR that occur in subtype-specific fashions . In particular , three tumor modules were enriched for genes in known large-scale BC amplicons ( 16p11-13 , 8q23-24 ) . The expression of these genes changes in a coordinated manner from high to low , suggesting that these genes measure amplification of the corresponding region in BC tumors . In turn , these patterns of expression were associated with distinct SR depending on subtypes highlighting the significance of each amplicon in defining patient SR for particular BC subtypes ( eg 16p13 in lumA and ER+/HER2+ , and 8q23-24 in basalL and her2E ) . Of note , these patterns of expression also define patients with particular clinico-pathological characteristics . For example , ER+/HER2+ tumors that do not highly express the genes on 16p have a lower luminal score than ER+/HER2+ tumors that highly express the genes on 16p . When we restrict our attention to basalL patients , we observe that both the immune-related module and the presence of a 8q23-24 amplification is associated with the patient SR . In fact , the subset of basal patients with 8q23-24 amplification exhibit high inflammatory SR and underexpress genes involved in general cellular proliferation of blood cells , in immune cytolysis , and in MYC and CD5 targets . Together , our matched profiles offer a detailed map of tumor-permissive SR particularly relevant for basalL tumors amplified on 8q and highlight a signature in the SR of basalL patients with low immune activity at their tumor site . This is especially interesting in the context of BC-immunotherapy combination or for monitoring response to these therapies . Overall , our study set the groundwork for further investigation of promising new ways to tackle and monitor the disease by looking outside the tumor and exploiting the patient SR .
Tumor and blood samples were obtained as part of the NOWAC study [49 , 62] with approval from Regional Committees for Medical and Health Research Ethics in Norway . Between 2006–10 , we collected blood and biopsy samples from BC cases at time of diagnosis , and blood samples from selected age-matched blood controls together with associated lifestyle and clinicopathologic data ( S1 Text ) . In total , and after data preprocessing , profiles include 16 , 792 unique genes expressed in primary tumors and blood from 173 BC patients , and in blood from 290 controls ( S1A Fig ) . We used ER status as measured by IHC and HER2 status measured by FISH or IHC where available . When unavailable , ER and HER2 status was determined using gene expression of the ESR1 gene and 6 gene members of the HER2 amplicon , respectively [9 , 63] ( S1 Text , S1B and S1C Fig ) . In addition , we calculated the HER2 score ( HER2S ) and the luminal score ( LUMS ) as the average expression of the HER2 amplicon gene members and the pam50 luminal genes , respectively . A proliferation score was calculated similarly using 12 mitotic kinases to produce the Mitotic kinase gene expression score ( MKS ) [45] . Samples were labeled according to our subtyping schemes from the literature: PAM50 [5] , hybrid [9] , CIT [8] , IntClust [7 , 39] ( S1 Text ) . Lists of differentially expressed genes in SR according to subtypes were obtained using the R/Bioconductor package Limma [64] . Whenever p-values were adjusted for multiple testing , the false discovery rate [65] was controlled at the reported level ( S1 Text ) . An unsigned weighted co-expression network was constructed independently in each tissue ( SR and tumor ) using the R/Bioconductor package WGCNA [41] ( S1 Text ) . First , a matrix of pairwise correlations between all pairs of genes is constructed across blood and tumor samples , respectively . Next , the adjacency matrix is obtained by raising the co-expression measure to the power β = 6 ( default value ) . Based on the resulting adjacency matrix , we calculate the topological overlap , which is a robust and biologically meaningful measure of network interconnectedness [42] ( that is , the strength of two genes’ co-expression relationship with respect to all other genes in the network ) . Genes with highly similar co-expression relationships are grouped together by performing average linkage hierarchical clustering on the topological overlap . The Dynamic Hybrid Tree Cut algorithm [43] cuts the hierarchal clustering tree , and modules are defined as branches from the tree cutting . Modules in each network were annotated based on Gene Ontology biological processes ( weight01 Fisher test [44] ) , MSigDB [66] and other curated signatures relevant to immune and blood cell responses [33 , 46 , 52] ( S1 Text ) Our approach maps samples to a linear ordering based on expression of genes within a given module or signature of interest ( S1 Text ) . In an univariate fashion , each gene within a given module/signature is used to rank all patients based on their expression . For each patient , the ranks of all k genes from the signature are summed and patients are then linearly ordered from right to left according to this ranksum vector . To identify the left and right boundaries of the low and high regions within the observed linear ordering , we delimit the region of independance ( ROI95 ) for each module . Briefly , we compute ( n = 10K times ) the position of an artificial patient within the observed linear ordering by summing the randomized ranks over all k genes in the module ( S1 Text ) . The ROI95 is defined as the region that contains 95% of the randomly generated samples . The three defined categories of patients correspond to those patients that have high ranskums of the module/signature ( high category ) , low ranksums of the module/signature ( low category ) , and a set of patients where the expression of the genes within the module/signature lose their pattern of pairwise correlation ( mid category ) . Using gene ranksums to capture module expression , we asked how modules are associated with patients’ clinical attributes and how they are associated across tissues . Pearson correlation and Analysis of Variance ( ANOVA ) was used to test association between a given module and continuous patient attributes ( eg . age , weight , MKS , LUMS ) and between a given module and categorical patient attributes ( eg . ER , HER2 , subtypes , lymph node status ) , respectively ( S1 Text ) . For each variable . we computed empirical p-values after permuting clinical labels 1000 times . For each variable , we perform a total of 42 association tests ( 23 blood modules + 19 tumor modules ) and used false discovery rate [65] to correct for multiple testing for each variable independently or for each “family” of tests when dependent variables are very similar ( S1 Text ) . Interactions between modules across tissues are identified using a random permutation approach based on the Pearson correlation between ranksums of gene expression in modules across tissues ( S1 Text ) . ANOVA was used to compare SR module expression between BC patients ( assigned to a given tumor module ROI95 categories ) and controls .
|
We present a novel system ( MIxT ) to identify genes and pathways in the primary tumor that are tightly linked to genes and pathways in the patient systemic response ( SR ) . These results suggest new ways to tackle and monitor the disease by looking outside the tumor and exploiting the patient SR .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"blood",
"cells",
"medicine",
"and",
"health",
"sciences",
"genetic",
"networks",
"body",
"fluids",
"immune",
"physiology",
"pathology",
"and",
"laboratory",
"medicine",
"immune",
"cells",
"immunology",
"signs",
"and",
"symptoms",
"network",
"analysis",
"immune",
"system",
"proteins",
"computer",
"and",
"information",
"sciences",
"white",
"blood",
"cells",
"inflammation",
"animal",
"cells",
"proteins",
"antigens",
"gene",
"expression",
"immune",
"response",
"biochemistry",
"antibody-producing",
"cells",
"diagnostic",
"medicine",
"blood",
"anatomy",
"cell",
"biology",
"b",
"cells",
"physiology",
"gene",
"identification",
"and",
"analysis",
"genetics",
"biology",
"and",
"life",
"sciences",
"cellular",
"types"
] |
2017
|
Interactions between the tumor and the blood systemic response of breast cancer patients
|
Many collective phenomena in Nature emerge from the -partial- synchronisation of the units comprising a system . In the case of the brain , this self-organised process allows groups of neurons to fire in highly intricate partially synchronised patterns and eventually lead to high level cognitive outputs and control over the human body . However , when the synchronisation patterns are altered and hypersynchronisation occurs , undesirable effects can occur . This is particularly striking and well documented in the case of epileptic seizures and tremors in neurodegenerative diseases such as Parkinson’s disease . In this paper , we propose an innovative , minimally invasive , control method that can effectively desynchronise misfiring brain regions and thus mitigate and even eliminate the symptoms of the diseases . The control strategy , grounded in the Hamiltonian control theory , is applied to ensembles of neurons modelled via the Kuramoto or the Stuart-Landau models and allows for heterogeneous coupling among the interacting unities . The theory has been complemented with dedicated numerical simulations performed using the small-world Newman-Watts network and the random Erdős-Rényi network . Finally the method has been compared with the gold-standard Proportional-Differential Feedback control technique . Our method is shown to achieve equivalent levels of desynchronisation using lesser control strength and/or fewer controllers , being thus minimally invasive .
Synchronisation is one of the key mechanisms responsible for self-organisation and emergence in living organisms [1–3] . Regular and periodic activity emerging from the collective behaviour of a set of interacting agents , has been noted to be crucial for the operation of many processes in living organisms [4 , 5] . A prime example are the firing patterns of neuronal populations that form the basis of brain activity [6 , 7] and their coordination among distributed mesoscopic neuronal populations [8] that ultimately controls our behaviour with an impressive and somehow mysterious accuracy [9] . It is therefore not surprising that defects or hypersynchronisation in neural firing patterns can lead to a host of neurological and psychiatric pathologies such as schizophrenia , Alzheimer’s disease , Parkinson’s disease and epilepsy [10–12] . One of the most conspicuous manifestation of neural hypersynchronisation are perturbation in the motor control systems . For example , a lack of dopamine in the basal ganglia is responsible for the uncontrolled and continuous tremors , rigidity and abnormal gait found in Parkinson’s disease ( PD ) [13] . Epilepsy is an even more striking example where strong and violent seizures occur unpredictably [14] and can be caused by an imbalance in neuronal excitation and inhibition [15] . While the exact causes of these diseases have yet to be elucidated [16] , they share a common mechanism: a dysfunction of neuronal firing patterns . Being able to control and restore normal synchronisation patterns could alleviate or even eliminate the symptoms [17] . Long term drug treatments are the reality for most patients suffering from PD or epilepsy , with only partially satisfying results [13 , 18–20] and the potential associated long term and side effects . An alternative to chemical treatments is neurostimulation , which induces a modulation of the neuronal activity in order to desynchronise the phase dynamics of neurons [21–24] . Our methods is to be applied in the framework of standard neurostimulation techniques [25–27] e . g . Deep Brain Stimulation ( DBS ) and is designed to render it as little invasive as possible , both by reducing the number of implanted electrodes and by using weaker applied currents . Although our methods may find clinical application in a number of diseases [28] , we will focus on the control of focal epileptic seizure as an example . Our work is computational in spirit and aims at validating a control strategy using simple but effective computational models already in use in computational neuroscience . The usefullness of such approaches to investigate epilepsy to complement and guide experiments has recently been reviewed [29–31] . In this paper , we will focus on the theoretical description of a novel , minimally invasive , brain neurostimulation method . It is minimally invasive in the sense that with a similar number of electrodes as existing set ups , e . g . the Proportional-Differential feedback method [24] the strength signal needed to control the hypersynchronisation is set at the minimum enough level to desynchronise the neuronal patches ( see Discussion at page 10 and Supplementary Material ( SM ) ) . This work lays the foundations of an applicable desynchronisation technique specifically aimed to control focal seizures , hypersynchronisation events in localised portion of the brain . We will first use the paradigmatic Kuramoto model [32 , 33] ( KM ) to describe synchronisation in networks of small patches neurons that can be targeted by microelectrodes , and then extend our results to the more general Stuart-Landau model ( SLM ) . Neurons are commonly modelled using leak and fire model ( LIF ) , and it has been shown that coherent synchronous behaviour can be obtained from very small patches of neurons [34] . The signal emerging from small neuronal patches can then be considered as phases of non-linearly coupled Kuramoto oscillators and their behaviour is indistinguishable from the more detailed LIF models . Networks of Kuramoto oscillators are particularly adapted to describe synchronisation pattern in neuronal patches: neuronal patches are connected with axonal tracts , forming a network of cells and they can synchronise easily their activation; i . e . , neurons are able to synchronise even when operating in a weakly coupled regime . Indeed the parameter responsible for the interaction among neurons can , without any loss of generality , be considered small . The method presented here simplifies the theoretical control term introduced in [35] to make it operational and show the potential for implementation . With this method , we can reduce the system wide phase synchronisation , or phase-locking , of nonlinearly coupled Kuramoto oscillators . The core mechanism brings the coupling between neurons patches below a certain critical value where partial synchronisation remains , but the system does not hypersynchronise . The magnitude of the control term , even when activated , is much smaller than the interaction among the patches and so minimally affects their activity . Once the coupling among the neurons is strong enough and the system is hypersynchronised , the control term naturally kicks in and induces a desynchronisation of the neuronal dynamics with the consequent suppression of the hypersynchronised behaviour . Keeping the control parameter at its lowest possible value both in the phase-unlocked and in phase-locked regime is important to avoid any side effects such as hallucinations or hypersexuality , commonly observed in other neurostimulation methods due to the stimuli being too strong [36] . For this reason , the proposed procedure for controlling the onset of the symptoms , as in the case of epilepsy , is optimised to get the right balance between managing the seizures and being as little invasive as possible . The basis of the control framework proposed in this work [35] is grounded in the well established Hamiltonian control theory [37–40] , which relies on the Hamiltonian formulation of the synchronisation process proposed in [41] . However , this theoretical control procedure [41] assumes a complete knowledge of the observables of the system: the network topology , phase variable and , more importantly , all the interacting nodes must be directly controlled . This is clearly not directly applicable to the brain where in the best case we can only measure the local dynamics and control only with a very limited number of patches of neurons compared with the whole number of neurons involved . To tackle this problem , we hereby adapt the theoretical control in order to limit the number of necessary microelectrodes to achieve the desired level of control and at the same time reduce the amount of information required on the signal measured from the electrodes . In the following section , we will introduce the mathematical formalism which describes the synchronisation phenomenon and give a short presentation of the Hamiltonian control theory , we invite the interested reader to consult [35] for a detailed discussion . Then , we illustrate our method with neuronal desynchronisation in the framework of the Kuramoto model . Let us observe that the method developed in [35] has been proposed in the framework of unweighted networks , where all the oscillators interact with the same strength , given by the Kuramoto parameter K . However , our method can be straightforwardly extended to the wider class of weighted networks ( see SM ) . We finish by extending our results to the more complex and realistic Stuart-Landau model , which has been used to reproduce brain activity in different settings describing various diseases [42] . Controlling abnormal synchronisation patterns in this model makes a strong point for the applicability of our method in real situations . We then conclude by summing up our results .
Abnormal synchronisation of the neural activity is responsible for the symptoms of many neurological diseases . Despite the very different nature of various systems exhibiting synchronisation , the main features are quite universal and can thus be described using the paradigmatic Kuramoto model [32 , 33 , 43–45] of nonlinearly coupled oscillators . Interestingly , as we show now , the KM is the limit of the more general Stuart-Landau model [24 , 46] . This model is well adapted to describe the normal form of a supercritical Andropov-Hopf bifurcation , which describes the switch from a stationary state to a periodic one—limit cycle— ( and vice versa ) according to a single bifurcation parameter: z ˙ k = ( 1 + ι ω k - | z k | 2 ) z k + Z k , where Z k = K N ∑ j = 1 N A k j z j . ( 1 ) Here the complex variable z k = ρ k e ι ϕ k encodes the information about the amplitude ρk and the phase ϕk of the coupled oscillators and ι = - 1 is the imaginary unity . The others terms are: the natural frequencies of the oscillators ωk , and are drawn from a symmetric , unimodal distribution g ( ω ) , the coupling strength K and the symmetric adjacency matrix Akj encoding the connections among the N oscillator ( Akj = Ajk = 1 if oscillators k and j are coupled and zero otherwise ) . Considering the real part of Eq ( 1 ) and assuming the amplitudes to be almost equal , i . e . ρk ∼ ρj for all k and j ( this statement is true in a weakly coupled regime as the case of neuronal patches ensemble ) , we obtain that the angles ϕk evolve according to the Kuramoto model ϕ ˙ k = ω k + K N ∑ j = 1 N A k j sin ( ϕ j - ϕ k ) . ( 2 ) We remind the reader that the original Kuramoto model corresponds to an all-to-all coupling [32] , Akj = Ajk = 1 , for all k ≠ j , Akk = 0 . The model can be rewritten using the order parameter [33] R e ι Ψ = 1 N ∑ j = 1 N e ι ϕ j , ( 3 ) a macroscopic quantity that measures the strength of the synchronisation; if R ∼ 0 , the oscillators are almost independent each other while if R ∼ 1 they are close to be phase-locked . Substituting the above definition in the original model we get the mean-field equation ϕ ˙ k = ω k + K R sin ( Ψ - ϕ k ) . ( 4 ) Thus the oscillators are no longer directly coupled to each other , but to the mean field oscillator with phase Ψ . The KM is a dissipative system , however an N dimensional Hamiltonian system H ( ϕ , I ) written in angles variables ϕ = ( ϕ1 , … ϕN ) and actions variables I = ( I1 , … , IN ) , has been proposed recently [41] and embeds as particular orbits the ones of the KM H ( ϕ , I ) = ∑ i ω i I i - K N ∑ i , j A i j I i I j ( I j - I i ) sin ( ϕ j - ϕ i ) ≡ H 0 ( I ) + V ( ϕ , I ) , where H0 and V are defined by the rightmost equality . The previous model represents a class of systems able to describe the Lipkin-Meshkov-Glick ( LMG ) model in the thermodynamic limit [47] and of the Bose-Einstein condensate in a tilted optical lattice [48] . The temporal evolution of the angle-action variables is obtained from the Hamilton equations: I ˙ i = - ∂ H ∂ ϕ i = - 2 K N ∑ j = 1 N A i j I i I j ( I j - I i ) cos ( ϕ j - ϕ i ) ( 5 ) ϕ ˙ i = ∂ H ∂ I i = ω i + K N ∑ j = 1 N A i j [ 2 I i I jsin ( ϕ j - ϕ i ) - I j / I i ( I j - I i ) sin ( ϕ j - ϕ i ) ] ( 6 ) for i = 1 , … , N . More precisely , one can define the invariant Kuramoto torus T : = { ( I , ϕ ) ∈ R + N × T N : I i = 1 / 2 ∀ i } and prove , to have a more detailed description of the model and of its properties . that the restriction of time evolution of the angles variables ( ϕ1 , … ϕN ) to this torus coincides with Eq ( 2 ) . We refer the interested reader to [35 , 41] for further details . In [41] , the authors have analytically proved and confirmed numerically that when the Kuramoto oscillators enter in a synchronised state , the dynamics of the actions close to the Kuramoto torus become unstable and exhibit a chaotic behaviour . Based on this result , our aim is to reduce the synchronisation in the KM ( 2 ) by controlling the Hamiltonian system H ( ϕ , I ) by adding a small control term able to increase the stability of the invariant torus T . Based on the previous remark this implies a reduction of the chaotic behaviour close to said torus , and thus impedes the phase-locking of the coupled oscillators . Let us rewrite the Hamiltonian in the form H = H0 + V , where H0 is the integrable part , i . e . the uncoupled harmonic oscillators , and V the non-linear term , namely the KRsin ( Ψ − ϕk ) function in the KM , that can be considered as a perturbation of H0 because of the small magnitude of the parameter K . The main idea of Vittot and coworkers [37 , 39] is to add to H a small control term f ∼ O ( K 2 ) , whose explicit form depends on V , in order to reduce the impact of the perturbation V , i . e . to increase the stability of the invariant torus . The size of f implies that the controlling procedure is much less invasive than other techniques generally used in control theory and is also able to give a rapid response to possible abnormal dynamics and more importantly , without any need for further measurement of the state of the system . Assuming a technical condition on the natural frequencies , namely ω = ( ω1 , … , ωN ) to be not resonant , i . e . for all k ∈ Z \ { 0 } then k ⋅ ω ≠ 0 , one can straightforwardly compute the required control term f ( ϕ , I ) . Let us observe that the theory by Vittot can also handle more general cases where such additional assumption is relaxed . The embedding of the KM in the Hamiltonian system is based on the existence of the invariant torus T , which is no longer invariant for the controlled Hamiltonian H0 + V + f . Nevertheless , it is possible to provide an effective control by truncating the latter to its first term , such that the resulting controlled Hamiltonian system preserves the Kuramoto torus . One can thus transpose this information into the KM and achieve a control strategy: ϕ ˙ k = ω k + K R sin ( Ψ - ϕ k ) + h k ( ϕ 1 , … , ϕ N ) , ( 7 ) where hk ( ϕ1 , … , ϕN ) is the contribution of the control f to the angles dynamics and is explicitly given by: h k ( ϕ 1 , … , ϕ N ) = - K 2 4 N 2 [ ∑ j A k jcos ( ϕ j - ϕ k ) ∑ l A k l ω l - ω kcos ( ϕ l - ϕ k ) + + ∑ j A k j ω j - ω k sin ( ϕ j - ϕ k ) ∑ l A k l sin ( ϕ l - ϕ k ) + - ∑ l ( A k lcos ( ϕ k - ϕ l ) ∑ j A j l ω j - ω l cos ( ϕ j - ϕ l ) + A k l ω k - ω l sin ( ϕ k - ϕ l ) ∑ j A j l sin ( ϕ j - ϕ l ) ) ] , ( 8 ) where , with a slight abuse of notation , we used the same letter to denote the new controlled angular variable . The truncation to the first order of the control term f is justified by the Hamiltonian perturbation theory . Moreover the smaller the perturbation parameter K , being f ∼ O ( K 2 ) , the better the approximation . The details of the derivation of formula ( 8 ) can be found in [35] . To simplify the previous equation , let us introduce a second modified local order parameter that depends on the node index: R ˜ k e ι Ψ k = 1 N ∑ j = 1 N e ι ϕ j ω j - ω k . ( 9 ) Under the hypothesis of an all-to-all coupling and a straightforward computation , the control term can be rewritten as: h k ( ϕ 1 , … , ϕ N ) = - K 2 4 [ R R ˜ k cos ( Ψ - Ψ k ) - B k ] , ( 10 ) where B k is defined by B k = 1 N ∑ l cos ( ϕ k - ϕ l ) cos ( Ψ l - ϕ l ) R ˜ l + ∑ l sin ( ϕ k - ϕ l ) ω k - ω l sin ( Ψ - ϕ l ) R .
Before we enter into the technical details of the proposed method , let us first comment on the analytic result obtained above and discuss its advantage with particular attention to the control of the onset of abnormal synchronisation . As already anticipated earlier in this paper , our principal aim is to develop a novel method to lower the synchronisation level of the neuronal patches situated in regions responsible for causing symptomatic behaviour . However , current neurostimulation techniques achieving this goal are often strongly invasive in terms of its strength . Our aim is to optimise the control strategy by letting the control to act only when necessary and with minimal magnitude . This means that , although the control is always present , it should dynamically “switch on” , i . e . achieve a strength comparable to the one of the signal , when the seizures start and again dynamically “switch off” , i . e . become negligible with respect to the signal ) during the normal neuronal regime . This is exactly what the proposed control term ( 8 ) does; the two main contributions to the control are the prefactor K2 and the denominators containing the differences of natural frequencies ωj − ωk , which are of the order of the width of the frequency distribution g ( ω ) . Because the critical value of the coupling strength Kc ( < 1 ) , the value for which the system exhibits a synchronised state , is of the order ∼ g ( ω ) [33] , the control term becomes of order K in the critical regime , exactly when it is necessary to reduce the synchronisation . On the other hand , during the normal regime the control size is much smaller than the critical one , K2 ≪ Kc , and in consequence the method can be considered to be minimally invasive . Let us now come back to the theoretical control term ( 10 ) and prove that one can realise it as an operational control strategy . The first observation is that the latter requires the control of all neuronal patches , and this is impossible to be achieved in a realistic situation . The second observation is that the control demands the exact knowledge of the connectivity of all the interacting cells . From a practical point of view we a priori know that in order to control the synchronisation we must interfere with the neuronal dynamics , by sending an electrical signal through a microelectrode inserted into a suitable zone of the brain . For this reason the main dilemma inherent with all neurostimulation methods is how to be as little invasive as possible but at the same time as efficient as possible ? To give a possible solution to this issue we will simplify the formula ( 10 ) to fit our goal of having an operational control . We will show that we can obtain a desynchronisation effect using a limited number of controlling microelectrodes , as good as the one involving the control of all the patches . We will work under the hypothesis that the interaction network allows easy global interactions while having a strong local connectivity , i . e . a small-world type of architecture [49] . This assumption is justified by experimental observations [50] which describe mesoscopic brain networks as small-world . Although the control method proposed here operates at a much smaller scale than the ones considered in [50] , there are compelling evidence [51] and models [52] that a form of self-similarity of brain circuitry and function is present , and thus what is observed at a macroscale can be inferred to be similar at a smaller scale . For this reason we believe that the theory previously developed under the assumption of all-to-all coupling , can be applied to a more general network topology , without substantially modifying the resulting dynamics . In this respect , the microelectrodes are supposed to be positioned , for instance as best as possible in the epileptic foci and each of them to directly control a zone which includes a certain number of neurons , optimising the efficiency of the control . The first observation is that the second term in Eq ( 10 ) , B k , is often much smaller than the first one; mathematically this can be understood because this term involves averages of products of oscillatory functions that can thus compensate each other . The second observation is that one can hardly compute the local phase Ψk using a limited number of microelectrodes sampling few neurons , we then decide to replace the latter with the neuron phase ϕk . The last point concerns the term R ˜ k whose computation requires the knowledge of the phases and the natural frequencies of all the neurons . In a real implementation of the control strategy this requirement is too stringent to be achieved , we thus decided to replace it with a term , R ^ k , computed using information obtained only from the neuronal patches where the microelectrodes are implanted in ∀ k = 1 , … , M R ^ k e ι Ψ ^ k = 1 M ∑ j = 1 M e ι ϕ j ω j - ω k . ( 11 ) In the previous formula , we assumed the ordering the neuronal patches j to be such that the first M are the ones upon which the microelectrodes are implanted . We are aware of the impact of these working assumptions , nevertheless the justification of these choices is obtained a posteriori by observing that the effective control performs very well . In conclusion the proposed local control strategy is given by: ∀ k = 1 , … , M h ^ k ( ϕ 1 , … , ϕ M ) = - γ 4 K 2 R R ^ k cos ( Ψ - ϕ k ) , ( 12 ) where we stressed again the dependence of such control term only onto the M neuronal patches upon which the microelectrodes are set in . Let us observe that we added a free parameter γ to take into account the direct action on a small number of nodes , M ≪ N , and the imperfectly known network structure . In particular γ can be set equal to the ratio of the average connectivity with the maximum possible number of links , which is a macroscopic parameter that can be known with good precision in advance . In conclusion let us observe that the local control term is built using a cosine function which is nothing but the coupling term in the KM ( 2 ) delayed by a quarter of its period T . We thus recover the empirical rule proposed by [22] consisting in the re-injection in the microelectrodes used in the DBS of the measured signal delayed by one fourth of its period . The operational control of a given neuronal patches is the following: compute the signal from a given neuron through a microelectrode , delay the signal by T/4 , multiply it by γ K R ^ k / 4 , where R ^ k is computed using a limited number of signals from neurons where the microelectrodes are inserted , and re-inject the new signal in the initial neuron using the same microelectrode . In this way the latter will desynchronise and break away from the whole system acting as a single giant oscillator . We however observe that this is not enough to desynchronise the whole system , but only the controlled nodes where the microelectrode is placed . Because we want to limit the number of implanted electrodes , this strategy will not be able to sufficiently reduce the symptoms . To achieve our goal , it is thus necessary to indirectly influence the behaviour of the other neurons . This can be done be noticing that a microelectrode controlling a given node produces an electromagnetic field potential [53 , 54] . To be more specific , let us denote with S k s t i m the stimulation signal generated on the position of the k-th neuron by the potential produced by the microelectrodes located in all the controlled neuronal patches , mathematically: S k s t i m ( ϕ 1 , … , ϕ M ) = c s ∑ l = 1 M e - 2 r k l h ^ l ( ϕ 1 , … , ϕ M ) ; ( 13 ) where rkl and cs are respectively the distance of node k from the origin of the electromagnetic field l , and the strength of the potential which in our case is taken to be cs = 1 , finally M ≪ N is the number of directly controlled nodes . In conclusion the proposed control strategy will modify the activity of all the N neurons as follows: ϕ ˙ k = ω k + K R sin ( Ψ - ϕ k ) + S k s t i m ( ϕ 1 , … , ϕ M ) , ( 14 ) let us observe that if the k–th neuronal patch has a microelectrode implanted into it , the rightmost term can be rewritten as c s h ^ k ( ϕ 1 , … , ϕ M ) + c s ∑ l ≠ k M e - 2 r k l h ^ l ( ϕ 1 , … , ϕ M ) , namely the direct control term plus the electromagnetic field generated by the remaining M − 1 microelectrodes , while if the k–th neuronal patch doesn’t have any microelectrodes it will feel the resulting electromagnetic field . A schematic illustration of the proposed control method is given in Fig 1 . In Fig 2 , we report the results of a generic simulation for the Kuramoto model; an oscillator is represented by a circle laying on the unit circle whose angular coordinate is given by the oscillator phase . The green circle identifies the Kuramoto order parameter , its angular position being given by Ψ while the distance from the origin , the black segment ( clearly visible on the panel a ) ) , represents R . Let us observe that the longer such segment is , i . e . the larger R , the stronger the synchronisation of the oscillators is . This can be clearly appreciated on panel a ) where most of the circles are very close to the green one . On the other hand ( see panel b ) ) one can observe that in the case of non-synchronisation the oscillators are quite uniformly distributed on the circle , resulting in R ∼ 0 . The network chosen for coupling the 100 oscillators is a Newman-Watts small-world [55] . The Newman-Watts network is a well known and widely used generating model for complex networks , and exhibits the small-world property for a determined set of parameters; it differs from the other widely used model of small-world network , i . e . the Watts-Strogatz [49] , mainly because the resulting network is always connected , and thus does not have isolated nodes . This is extremely important in our case where the neuronal patches by definition form a connected structure . The model contains a single parameter , p ∈ [0 , 1] , which determines the density of the network; indeed the network generation starts from a 1D regular lattice with coordination number 2k , i . e . each node is connected to its first k neighbours counted clockwise and k counterclockwise , then each couple of unconnected nodes is considered and with probability p a link is added . In the limit p → 1 many links can be potentially added and the network can become very dense; in the opposite case the network is sparse and very similar to the 1D regular lattice backbone . In Fig 2 we can clearly observe that in the uncontrolled KM the oscillators tend to synchronise for the chosen coupling parameter . They almost all have the same phase ( see panel a ) ) , as K = 0 . 5 is larger than the critical parameter , here Kc ≈ 0 . 4 . On the other hand , for the same value of the coupling parameter , but applying the effective control using M = 20 oscillators and γ/4 = 4 . 25 , the behaviour is completely different , the oscillators are almost uniformly distributed on the unit circle ( see panel b ) ) , corresponding to a desynchronised system . Let us emphasise that the results reported in Fig 2 are another a posteriori proof of the goodness of the control given by Eq ( 14 ) . Indeed , despite the latter , as well as the theory presented in [35] , has been developed under the assumption of all-to-all coupling , it works perfectly on a different underlying topology such as the Newman-Watts; in the SM we present a complete analysis of the role of the parameter p in the desynchronisation problem . Moreover in the SM we have improved the control strategy and extended it to handle weighted complex networks , and so make a further step towards empirical topologies . The control strategy we proposed depends on two main parameters: the number of controlled microelectrodes M and the strength of the injected signal γ . Intuitively large values are associated to an efficient control for both parameters and thus to a reduction/suppression of the abnormal synchronisation , but with the drawback of being invasive; many microelectrodes have to be implanted and the strength of the signal could induce undesired collateral effects . Let us observe that under the assumption of all-to-all coupling and homogeneous interaction strengths among the oscillators , the results above are indistinguishable and thus the spatial layout of the microelectrodes does not matter . The same result seems to hold in the case of more complex coupling ( see SM ) , for these reasons , we decided to position the microelectrodes uniformly at random among the oscillators . To understand the impact of M and γ and possibly determine an optimal range of values we performed a series of numerical simulations . In Fig 3 we report ( left panel ) the averaged ( over 50 independent repetitions ) Kuramoto order parameters , 〈R〉 , for the controlled system ( 12 ) as a function of the parameters M and γ/4 . One can observe that for large enough M ≳ 33 , the control is able to completely suppress the synchronisation , 〈R〉 ∼ 0 , for all values of γ/4 . For intermediate values , 15 ≲ M ≲ 30 , there exists a non trivial relation γr ( M ) ( see right panel ) such that if γ/4 ≥ γr ( M ) then the control can achieve a partial desynchronisation , 〈R〉 ≤ r ( here r ∈ ( 0 , 1 ) is a parameter defining the amount of partial desynchronisation present in the system ) . Finally for too small values of M ≲ 10 , the proposed strategy is not able to reduce the synchronisation for any tested values of γ . To support the claim that our method is minimally invasive , both in terms of the number of microelectrodes needed and the strength of the signal applied , we compared it with the Proportional-Differential Feedback technique ( PDF ) [24] , whose capability to suppress hypersynchronisation has been already proved . In short , the main idea of the PDF is to split the population of N oscillators into two groups: a first group made by N1 elements whose signal is measured in time; and a second group , containing the remaining N2 = N − N1 oscillators , that will receive the feedback signal which is proportional to the mean-field signal of the N1 first oscillators , with a proportionality parameter P > 0 , and to the derivative of the same signal , with a proportionality parameter D ≥ 0 . We refer the interested reader to [24] and to the SM for a more detailed presentation of the PDF . The PDF , like our method , is thus essentially based on two parameters , the number of controllers N2 and the strength of the feedback signal P + D , a comparison among the two methods is thus straightforward: N2 = M and P + D = γ/4 . We have thus chosen as benchmark the KM composed by N = 100 oscillators coupled with an unweighted all-to-all network . We have numerically computed the asymptotic synchronisation state for several values of the parameters , measured with the Kuramoto order parameter R = | ∑ je ι ϕ j | / N , and averaged the over several independent realisations . The results presented in Fig 4 ( see also Fig . D in S1 Text ) should be compared with the ones of Fig 3 ( the same colour code has been used to help the comparison ) . At first glance both methods exhibit the same behaviour , for a small number of controllers one needs a large control strength , P + D or γ/4 , to remove/reduce the synchronisation , and below a certain values of N2 or M desynchronisation cannot be achieved . However looking at the values of P + D versus γ/4 we realise that the former are 5 times larger . Indeed P + D ranges from 10 to 50 while γ/4 in the interval [2 , 10] . This means that for the same number of controllers our method requires a much weaker signal strength or that for a fixed control strength we can achieve a desynchronisation level with a smaller number of implanted microelectrodes . To mimic the onset of an epileptic seizure in the brain and the action of the proposed control strategy , we realise the following numerical experiment: using N = 100 neurons connected using a Newman-Watts small-network , firstly without control ( reference case ) and then controlled using M = 20 microelectrodes and γ/4 = 4 . 25 . In both cases , during a given period of time , [0 , 5000] , we numerically solve the KM with a small control parameter fluctuating in time to mimic the physiological fluctuations one can observe in neuron; more precisely every Δt = 100 time units , we draw a value for K from a uniform distribution with support [0 . 05 , 0 . 15] , and thus with average 0 . 1 , and we follow the model dynamics during Δt time units . Let us observe that the coupling parameter is smaller than the critical one , Kc ≈ 0 . 4 , and thus the system , for both the reference case and the controlled one remains in a non-synchronised state . This can be appreciated from Fig 5 where we plot the order parameter R as a function of time for the uncontrolled ( blue ) and controlled ( red ) KM . Observe moreover that both systems behave very similarly ( the curves are very close ) , hence the control , even if present , is not changing the dynamics when not needed . Then we assume that the coupling parameter quickly increases and then fluctuates around a large value , mimicking a seizure . Mathematically we assume that during the time interval [5000 , 7500] every Δt = 100 time units , we draw a value for K from a uniform distribution whose average grows linearly in time from 0 . 1 at t = 5000 to reach 0 . 5 for t = 7500 , while in the interval [7500 , 125000] the coupling parameter is drawn from a uniform distribution with support [0 . 55 , 0 . 65] , and thus with average 0 . 6 . The results of the numerical simulations are striking , after a short transient time the uncontrolled system ( blue curve ) almost fully synchronises , R is very close to 1 , while the controlled one remains in the non-synchronised phase , with R very close to 0 . Of course when the coupling parameter starts to decrease to eventually reach again a small average value , the original reference system and the controlled one both exhibit again a non-synchronised behaviour . In the previous section we built an operational control scheme able to effectively reduce the level of phase-locking in the phases of the neurons described by the Kuramoto model , even for large values of the coupling parameter . As previously stated , the KM can be derived from the more general Stuart-Landau model , it is then natural to try to extend the control strategy to act directly on the Stuart-Landau system , self-consistently defined in terms of the control used for the Kuramoto model . Let us consider Eq ( 1 ) under the assumption of all-to-all connection with the additional term Z k c t r l = - ι K 4 R ˜ k Z , where Z = K N ∑ j = 1 N z j , namely z ˙ k = ( 1 + ι ω k - | z k | 2 ) z k + Z + Z k c t r l . ( 15 ) Assuming the amplitudes to be very close each other , one can easily prove that the phase of the complex variable z k = ρ k e ι ϕ k in Eq ( 15 ) evolves according to the controlled Kuramoto system , see Eq ( 7 ) . As before , we performed simulations to mimic the onset of an epileptic seizure to prove the effectiveness of the control strategy applied to the Stuart-Landau model . More precisely we consider a system of N = 100 neurons described by the Stuart-Landau model ( 1 ) and its controlled version ( 15 ) ( M = 20 microelectrodes and γ/4 = 4 . 25 ) using again a Newman-Watts network . Initially the coupling parameter fluctuates around a small value and then becomes larger . In Fig 6 we represent the total ( real part of the ) signal , ∑k ℜzk = ∑k ρk cos ϕk , for the original Stuart-Landau model ( blue curve ) and the controlled model ( red curve ) . In the interval [0 , ∼ 90] , K is small and neither system synchronises , as can be seen in the insets A ( real part of the signal for 10 generic neurons for the original Stuart-Landau model ) and C ( real part of the signal for 10 generic neurons for the controlled Stuart-Landau model ) , the amplitude of the total signal is thus quite small . On the other hand for larger times , [∼ 90 , ∼170] ( roughly corresponding to the shaded central rectangular part of the figure ) , K assumes larger values than in the previous period and the Stuart-Landau system enters in a synchronised state ( see inset B where we plot the real part of the signal for the same 10 generic neuronal patches of inset A ) while the controlled system remains in a non-synchronised state ( see inset D where again we plot the real part of the signal for the same 10 generic neurons of inset C ) . This corresponds to quite a large amplitude for the total signal because now the amplitudes of each single signal add coherently together .
In this paper , we presented a new method to control abnormal synchronisation of neuronal activity based on the Hamiltonian control formalism applied to the paradigmatic Kuramoto model . We focus on the phase dynamics which prepares the foundation for most of the basic functioning of the brain regions . As it is well-known when the coupling strength K exceeds a critical value , the phases of the electrical currents of the neurons of interest get locked and , due to a resonance effect , the neural signal amplifies directly affecting the behaviour . However , sometimes this behaviour is not the desirable and can be associated to neurological diseases , as in the case of epileptic seizures . Often drugs are not sufficient to control , i . e . reduce , the strength of the seizures and invasive brain stimulation becomes necessary . We therefore propose an efficient and minimally invasive control technique aimed to prevent the phase-locking and thus applicable to all cases where over-synchronisation is responsible for undesired negative effects . Starting from a theoretical result [35] , we further develop the control term and adapt it towards potential realistic applications where an abnormal synchronisation state is present , including complex ( weighted ) network topologies . The main idea is to effectively control the interested neuronal patches and brain regions while reducing side effects as much as possible . In terms of control strategy , this amounts to have as few microelectrodes implanted as possible , and which signal injection is directly regulated by the magnitude of the order parameter . The control term then becomes active only when needed . The method is very promising and the desynchronisation level achieved is very good compared with the standard represented by the Proportional-Differential Feedback control . Starting from control scheme developed for the KM we are able to define a control strategy acting directly on the Stuart-Landau model widely used to describe the interaction of coupled neuronal patches [42] and numerically show its effectiveness in suppressing the synchronised state and thus the neuronal disease . The latter result is in our opinion a proof-of-concept that the presented method could be applied to deal with real cases . The control strategy presented in this paper is purely theoretical and need further validation before envisaging a clinical implementation . First we need to carry further in silico investigations on more realistic topologies . The human connectome would be used for the large scale interaction between brain regions , each modelled by smaller network of interacting neural patches , would allow for an extensive investigation of how the control of neural patches within brain regions reverberates to large scale brain dynamics . This goes along the current research lines where brain regions activities are modelled using Stuart-Landau systems , whose bifurcations parameters are used to reproduce the disease we are interested in [42 , 56] . The main difference with respect to the model presented here , based on a 100-nodes networks of coupled Stuart-Landau systems , is the size and the topology of the network and the possibility to have negative bifurcation parameters . However , based on our positive results ( see Fig 6 ) and on the potential robustness of the strategy with respect to changes in the connectivity ( see Fig . B in S1 Text ) , we are confident that this generalisation can be achieved . This first phase will also be used to precisely benchmark the goodness of the synchronisation controllability versus the invasiveness of the strategy , namely the number of used microelectrodes ( M ) and the strength of the signal ( γ ) and thus yield results directly comparable with current implementation . Parallel to these in silico experiments , in vitro experiments could be designed to test our control framework . Indeed Shew and colleagues [57] have performed experiments altering the balance between inhibition and excitation on cortical slices , and this set up could in principle be used to directly test the potential of our control framework to restore the inhibition/excitation balance . Finally , we must point out that our method provides a theoretical framework for empirically determined control strategies proposed in the literature [22–24] , adding credibility to its applicability in real conditions .
|
Synchronisation plays an important role in most of the neuronal activities and in particular in the control of the motor system . However , due to biochemical dysfunction in the brain activity , an abnormal and excessive synchronisation may occur being responsible for severe symptoms of several neurological diseases . For the case of Parkinson’s disease , for instance , an insufficient dopamine production in the basal ganglia causes rigidity or continuous tremors . In the case of epilepsy instead , imbalance between excitation and inhibition causes strong unpredictable seizures . Several neurostimulation techniques have been developed with the aim to control and relieve the symptoms as alternatives to oral medication . In this line of research , we propose a new method which has the property of being as little invasive as possible , in the number of electrodes needed and the strength of the current applied , while still controlling the symptoms . It is based on the consideration that neuronal patches resemble a set of phase-coupled oscillators which dynamics can be described by the celebrated Kuramoto model . The control technique we employ is inspired by a Hamiltonian formulation of the Kuramoto model . To verify the effectiveness of our method , we test it in a more realistic model of coupled neuronal patches described by the Stuart-Landau equations . Numerical simulations validate our approach .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"control",
"theory",
"medicine",
"and",
"health",
"sciences",
"neural",
"networks",
"engineering",
"and",
"technology",
"electronics",
"neurodegenerative",
"diseases",
"signal",
"processing",
"neuroscience",
"epileptic",
"seizures",
"control",
"engineering",
"systems",
"science",
"mathematics",
"epilepsy",
"computer",
"and",
"information",
"sciences",
"animal",
"cells",
"microelectrodes",
"movement",
"disorders",
"cellular",
"neuroscience",
"cell",
"biology",
"neurology",
"neurons",
"parkinson",
"disease",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"physical",
"sciences",
"electrodes"
] |
2018
|
A minimally invasive neurostimulation method for controlling abnormal synchronisation in the neuronal activity
|
Since 2003 , the tropical arthritogenic chikungunya ( CHIK ) virus has become an increasingly medical and economic burden in affected areas as it can often result in long-term disabilities . The clinical spectrum of post-CHIK ( pCHIK ) rheumatic disorders is wide . Evidence-based recommendations are needed to help physicians manage the treatment of afflicted patients . We conducted a 6-year case series retrospective study in Reunion Island of patients referred to a rheumatologist due to continuous rheumatic or musculoskeletal pains that persisted following CHIK infection . These various disorders were documented in terms of their clinical and therapeutic courses . Post-CHIK de novo chronic inflammatory rheumatisms ( CIRs ) were identified according to validated criteria . We reviewed 159 patient medical files . Ninety-four patients ( 59% ) who were free of any articular disorder prior to CHIK met the CIR criteria: rheumatoid arthritis ( n=40 ) , spondyloarthritis ( n=33 ) , undifferentiated polyarthritis ( n=21 ) . Bone lesions detectable by radiography occurred in half of the patients ( median time: 3 . 5 years pCHIK ) . A positive therapeutic response was achieved in 54 out of the 72 patients ( 75% ) who were treated with methotrexate ( MTX ) . Twelve out of the 92 patients ( 13% ) received immunomodulatory biologic agents due to failure of contra-indication of MTX treatment . Other patients mainly presented with mechanical shoulder or knee disorders , bilateral distal polyarthralgia that was frequently associated with oedema at the extremities and tunnel syndromes . These pCHIK musculoskeletal disorders ( MSDs ) were managed with pain-killers , local and/or general anti-inflammatory drugs , and physiotherapy . Rheumatologists in Reunion Island managed CHIK rheumatic disorders in a pragmatic manner following the outbreak in 2006 . This retrospective study describes the common mechanical and inflammatory pCHIK disorders . We provide a diagnostic and therapeutic algorithm to help physicians deal with chronic patients , and to limit both functional and economic impacts . The therapeutic indication of MTX in pCHIK CIR could be approved in future efficacy trials .
Fifty years after its first tropical description in the Newala district of Tanganyika [1 , 2] , chikungunya ( CHIK ) has re-emerged extensively , resulting in 1 . 4 to 6 . 5 million infected individuals between 2004 and 2014 in Africa , as well as regions within the Indian Ocean , Southeast Asia , the Pacific Islands and Europe [3] . Its first autochthonous transmission in the intertropical Americas was identified at the end of 2013 [4] , and six months later this turned into a large outbreak in most of the Caribbean Islands that has now reached northern South America [5] and the United States [6] . Currently , this arboviral disease represents a pressing threat to public health in large areas of the American and European continents that are colonized by the primary disease vector: the Aedes mosquito [4 , 7] . CHIK is usually characterized by an acute febrile and sometimes eruptive polyarthritis , commonly followed by persistent rheumatologic and general disabling symptoms [8] . Historically , post-CHIK ( pCHIK ) rheumatic disorders were first described in South Africa after a local outbreak at the end of the 1970s . In 1979 , Fourie and Morrison first reported a pCHIK rheumatoid arthritic syndrome [9] , and in 1983 Brighton et al . highlighted a high prevalence of chronic polyarthralgia or stiffness occurring three years after disease onset [10] , with one case of destructive pCHIK polyarthritis [11] . After minimal exposure in the literature over the last 20 years , the wide clinical spectrum of pCHIK rheumatic disorders has been rediscovered [12–22] . During its re-emergence in the past 10 years , 92 published articles related to chronic pCHIK status have been indexed in PubMed [keywords: “chikungunya” and “chronic” , last search on October 24th , 2014] , but there are no available evidence-based guidelines with standardized definitions and treatment recommendations . While the overall proportion of patients with chronic symptoms diminishes over time after CHIK onset ( from 100 to 88% during the first 6 weeks , to less than 50% after 3–5 years , with variable results depending on the study ) , the time required to return to pre-CHIK status is still uncertain , as some infected individuals remain symptomatic for 6 to 8 years post-infection [23 , 24] . Simple analgesics and/or non-steroidal anti-inflammatory drugs ( NSAIDs ) provide relief in most patients [25] , but better-targeted drugs are clearly needed to treat inflammatory rheumatic disorders . Hydroxychloroquine and ribavirin were not effective [26–28] , but methotrexate ( MTX ) was of benefit in pCHIK inflammatory polyarthritis [29–31] . MTX efficacy is supported by scientific data that indicates active monocyte/macrophage trafficking into the synovial tissue of chronic patients , possibly maintained by a local viral persistence [32] . Thus , the current challenge for physicians in epidemic areas is to identify and diagnose pCHIK rheumatic disorders and to provide the optimal treatment in order to prevent perpetuation or progression to a potentially destructive disease course . To answer this real-world need , we retrospectively documented the clinical and therapeutic courses of rheumatic and musculoskeletal pCHIK disorders ( pCHIK-RMSD ) that were referred to rheumatologists in Reunion Island over a 6-year period following the 2005–2006 epidemic . We have summarized this practical experience in an algorithm for disease management purposes .
Data were anonymously analyzed and reported . The study was approved by the research ethical committee of Laveran Military Teaching hospital ( registration n° 2014-PRSimon ) . This retrospective study was carried out at two referring rheumatologic practices: University Hospital ( “CHU” ) Félix Guyon and the private office of a rheumatologist in Saint-Denis , the capital of Reunion Island , which was affected by a massive CHIK outbreak in 2005–2006 . All eligible participants were older than 16 years of age and had been referred to the specialist for joint pain that persisted for more than 4 months following typical clinical acute CHIK infection [33] , which was assumed to be related to CHIK-RMSD . Biological confirmation of CHIK infection was obtained for all patients , either by detection of CHIK virus RNA in the blood during the acute stage by reverse-transcriptase polymerase chain reaction ( RT-PCR ) , or the presence of anti-CHIK virus-specific immunoglobulin M and/or G ( depending on the delay relative to the acute CHIK infection ) as detected by enzyme-linked immunosorbent assay ( ELISA ) using CHIK virus antigens produced by the French Referent National Centre for Arbovirus , Marseille , IRBA , France . Data were collected anonymously and retrospectively from January to May of 2012 , based on patient medical files using a structured questionnaire developed for the purposes of this study . Patient details were recorded , including factors considered to be predictors of non-recovery after CHIK according to the literature: age ( >45 year ) , female gender , previous history of an osteoarthritic event , and severe acute CHIK [12 , 13 , 18 , 22 , 29 , 34] . Details on CHIK-RMSD included the date of the first visit to the rheumatologist , the clinical and biological histories , imaging features , and treatments since the acute CHIK infection . We recorded the following clinical data: i ) joint involvement: small joints ( metacarpophalangeal , proximal and distal interphalangeal , metatarsophalangeal , thumb interphalangeal joints and wrists ) , large joints ( shoulders , elbows , hips , knees and ankles ) , vertebral , sacroiliac , temporomandibular and sternocostoclavicular areas; ii ) joint count: polyarticular if more than 4 ( oligoarticular if 4 or less ) ; iii ) articular inflammatory signs or symptoms defining arthritis: synovitis , warmth and/or redness over the joint ( “hot” joint ) , prolonged morning stiffness ( > 30 minutes ) , inflammatory pain ( which improved with exercise , or worsened after rest or during the night ) , or arthritis that was clinically distinguished from oedema without join effusion; iv ) periarticular involvement: enthesitis ( inflammation of the tendon or ligament insertion into bone: Achilles tendonitis , plantar fasciitis ) , tenosynovitis , periostitis ( tibial and ischial tuberosities inflammation ) , tendinopathy , bursitis , myalgia and neural tunnel syndrome . Hyperuricemia and vitamin D deficiency were determined using standard biological assays , as were the elevated erythrocyte sedimentation rate ( ESR ) , and the levels of C-reactive protein ( CRP ) , antinuclear antibodies ( ANA ) , rheumatoid factor ( RF ) , anti-citrullinated peptide autoantibodies 2 ( ACPA2 ) and HLA B27 positive status , when required for scoring . The treatments were classified as follows: painkillers; oral or topical non-steroidal anti-inflammatory drugs ( NSAIDs ) ; oral or intra-articular corticosteroids; conventional disease-modifying anti-rheumatic drugs ( DMARDs ) including MTX ( 7 . 5–25 mg/week ) , hydroxychloroquine ( 200 mg/day ) , leflunomide ( 10–20 mg/day ) , sulfasalazine ( 1 . 5–3 g/day ) ; immune-modulating biologic agents including anti-TNF , a B-lymphocyte depletion agent ( rituximab ) , and an interleukin-6 receptor inhibitor ( tocilizumab ) for indications and at dosages recommended by national guidlines ( former and revised text ) [35 , 36]; physiotherapy; orthopaedic braces; complementary therapeutics such as vitamin D supplementation or gout treatment using colchicine with a urate-lowering diet , such as a xanthine oxidase inhibitor ( allopurinol ) or uricosurics . Patients without previously defined arthritis were classified as pCHIK musculoskeletal disorders ( pCHIK-MSD ) that were either loco-regional or diffuse ( if > 4 painful areas ) . For patients with arthritis , we separated crystalline and non-crystalline rheumatic disorders . Crystalline arthritis was defined as a convincing clinical presentation associated with hyperuricemia that improved after gout treatment . Patients presenting non-crystalline polyarthritis were classified as chronic inflammatory rheumatism ( CIR ) . CIRs were categorized into 3 groups ( Table 1 ) [37 , 38] . Patients who had not previously displayed rheumatic symptoms and developed CIR immediately after CHIK were classified as de novo pCHIK-CIR . We measured the median period from acute CHIK to the first rheumatology consultation for the pCHIK-CIR and pCHIK-MSD patient groups . The severity of all CHIK-RMSDs was evaluated based on the level of joint destruction or deformation ( chondrolysis , joint space reduction , juxta-articular osteopenia , subluxation , bony erosion or destruction , sacroiliitis ) detected in the imaging tests , and based on the need for an internal or external orthopaedic brace . The functional burden was estimated based on a history of being unable to work or requiring adjustment , self-estimated significant reduction of daily activities ( as a percentage of reduction ) [13] and initiation of psychological follow-up or antidepressant medical treatment since 2006 due to CHIK . For each patient , we recorded the therapeutic strategy , including corticotherapy , and the outcome after complete drug withdrawal ( the type of response at 3 and 6 months following therapeutic interruption ) . Rheumatologists treated pCHIK-CIR patients with MTX at the recommended dosage and routes of administration based upon the 2008 ACR Recommendations for Rheumatoid Arthritis Treatments [39] and French national guidelines [35 , 36] . We defined a positive response to MTX if there was no further need for a drug switch or escalation ( treatment in association with other DMARDs ) . Recovery after MTX was defined as the absence of relapse more than 6 months after MTX withdrawal . We documented the DMARDs that were received by patients other than MTX , including hydroxychloroquine and immune-modulating biologic agents . Descriptive results were expressed as the median value and the distribution or the number ( percentage ) of patients according to the category of the variable ( quantitative or qualitative ) .
A total of 159 patients with persistent pCHIK-RMSD were included in this study . The population was predominantly female ( 75% ) and the median age was 51 years-old , ranging from 16 to 80 years-old . Repartition of pCHIK-RMSD is detailed in Fig . 1 . Demographic characteristics and comorbidities are presented in Table 2 , according to the patients’ CHIK-RMSD categories . There was no significant difference between these groups . In the entire cohort , 66% of patients reported a prolonged acute CHIK infection ( fever> 10 days or symptoms> 3 weeks ) . The median time elapsed between CHIK infection and the first consultation with a rheumatologist was 2 years ( median delay: 15 . 0 months for the MSD-group and 38 . 5 months for the CIR group ) ( Fig . 2 ) ; 80% of patients with pCHIK-MSD were referred within the first two years , whereas patients with pCHIK-CIR were regularly referred throughout the 6-year period , independent of the type of CIR . All patients received painkillers as a first-line treatment , and 20% received vitamin D supplementation . The treatment history and medical features according to the different pCHIK-RMSD are summarized in Tables 3 and 4 . The characteristics and specific management of the 43 cases of pCHIK-MSD are shown in Table 5 . Only seven patients ( 15% ) responded to the optimal painkiller treatment ( using a second- or third-step analgesic if necessary , with the additional availability of a painkiller in the case of a painful episode ) . Among the 28 diffuse pCHIK-MSD , 22 consisted of distal polyarthralgia involving the hands ( 18 out of 22 ) and/or the feet ( 17 out of 22 ) , typically associated with bilateral oedema of the extremities ( about 50% of cases ) . This feature was frequently complicated by carpal or tarsal tunnel syndromes and paraesthesia that responded to a short course of corticotherapy ( oral and/or injected into the joint ) and sometimes improved with neuropathic pain medication , when indicated . Five patients with diffuse pCHIK-MSD ( 18% ) received hydroxychloroquine without a self-reported benefit , while 20 out of 28 ( 70% ) were efficiently treated with oral corticosteroids ( 16 out of 20 had a successful withdrawal of systemic corticotherapy without clinical relapse ) . Loco-regional pCHIK-MSD primarily involved proximal joints . In 11 cases , pCHIK-MSD presented as exacerbated pain in previously involved joints: arthrosis ( n = 8 ) and injured areas ( n = 3 ) . Thirteen shoulder capsulitis or tendinitis required physiotherapy to limit stiffness and amyotrophy . Loco-regional CHIK-MSD mainly benefited from joint injections , physiotherapy and/or NSAIDs . Two of the 9 patients presenting with carpal tunnel syndromes developed algodystrophy ( complex regional pain syndrome ) after early surgery . Among the patients with inflammatory arthritis , 4 were diagnosed as early-onset crystalline polyarthritis , and they responded to gout treatment . One hundred twelve patients met the criteria for CIR . Eighteen patients had pre-existing CIR ( diagnosed or suspected ) that was exacerbated immediately after CHIK infection , and the other 94 patients were considered pCHIK-CIR . Twenty-seven percent of patients with pCHIK-CIR reported the inability to work , and 77% reported a significant reduction in their daily activities . Half of the patients with pCHIK-CIR presented radiographical osteoarticular destructions or misalignments , mostly occurring in pCHIK-rheumatoid arthritis ( RA ) ( 83% of the cases ) . The median time from acute CHIK to the diagnosis of radiographical damage was 45 months ( range 3–76 ) . Thirteen patients ( 40% ) with pCHIK-spondyloarthroarthritis ( SA ) had radiographical chondrolysis , 4 of these with sacroiliitis and 3 with bone erosions . Six other patients with SA had non-destructive arthritis with periosteal appositions and reconstructions . Systemic corticotherapy was deemed necessary in 70% ( 13 out of 18 ) of pre-existing CIR , but complete withdrawal was achieved in only half of the cases . MTX was started in 8 patients with pre-existing CIR ( 3 RA , 4 SA and one systemic lupus previously under hydroxychloroquine ) , and it failed in 6 cases ( 3 RA , 3 SA ) , thereby leading to a switch to biologic DMARDs . A therapeutic escalation with immune-modulating biologic agents was required for 3 erosive RA previously controlled with MTX , and this appeared to have a beneficial effect; one patient with RA treated with infliximab was switched to etanercept , but remission occurred only after appropriate urate-lowering therapy , as the final diagnosis for their worsening condition was gout . For the 2 patients with untreated chronic viral hepatitis ( B and C , one case each ) , hydroxychloroquine was ineffective in relieving joint inflammation , and consequently , sequential short courses of oral corticotherapy were given . Lastly , among the 4 patients who started hydroxychloroquine , 3 had to complete their treatment with immune-modulating biologic agents . RA was the most common CIR in this cohort ( 16% pre-existing RA and 36% de novo pCHIK-RA ) and was associated with the highest prevalence of osteoarticular damage , requirement for braces , and functional consequences ( Table 4 ) . Among the 41 SA , we noticed a difference in clinical features: the 8 pre-existing SA presented as mostly ankylosing spondylitis with peripheral involvement ( arthritis and/or enthesitis ) triggered by CHIK , while the 33 de novo pCHIK-SA consisted of 15 cases of psoriatic polyarthritis ( including 6 first occurrences of psoriasis ) and 16 polyenthesitis ( only two cases were HLA B27+ ) . About one-third of patients with de novo pCHIK-CIR ( 14 RA , 11 SA and 3 UP ) presented tenosynovitis of the wrist , hands and/or Achilles tendon , resulting in tunnel syndromes in 16 patients ( 17% ) . Seventy-two pCHIK-CIR ( 77% with 100% of RA , 78% of SA and 30% of UP ) received MTX at a mean weekly dose of 15 mg . Thirty six patients ( 50% ) already had varying degrees of radiographical damage at the start of MTX treatment . With a median follow-up time of 21 months ( a mean of 25 months ) , MTX led to a positive clinical response in 54 out of 72 patients ( 75% with 67% in RA , 80% in SA , 100% in UP ) ; 7 cases developed bone destruction while undergoing MTX treatment , whereas recovery was achieved for 6 patients ( 3 RA and 3 SA ) . Nine patients had to promptly stop taking MTX due to side effects , which included 4 significant increases in transaminases , 2 digestive intolerances , one hair loss , one depression , one rash and cough ) ; only 3 patients fully stopped MTX treatment due to a complete inefficacy . A total of 26 out of 94 patients received other DMARDs as a consequence of MTX contraindication or failure; 12 of these ( 9 RA , 3 SA ) required immune-modulating biologic agents . Finally , algodystrophy after early carpal tunnel surgery was also reported in 3 more cases .
CHIK virus is currently generating an epidemic of chronic rheumatism worldwide . Using a 6-year insight , this rheumatology case series from Reunion Island broadly describes clinical and therapeutic approaches to pCHIK’s long-lasting rheumatic disorders . We highlight the destructive pCHIK-CIR that affects a minority of patients , and we emphasize the importance of reducing the time necessary for the initiation of disease management by a medical specialist . Studies aimed at validating the efficacy of early use of MTX to prevent joint damage and long-term corticotherapy should be conducted in the future . Physiotherapy should also be evaluated in CHIK-MSD . Official guidelines for clinicians are necessary .
|
With a 6-year insight , we extensively and retrospectively describe clinical profiles and specific treatments of mechanical and inflammatory post-chikungunya rheumatic disorders . In the current context of chikungunya’s global spread , we provide the first diagnostic and therapeutic algorithm to guide physicians according to the amount of time that has elapsed since the acute CHIK infection .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[] |
2015
|
Specific Management of Post-Chikungunya Rheumatic Disorders: A Retrospective Study of 159 Cases in Reunion Island from 2006-2012
|
Information sampling is often biased towards seeking evidence that confirms one’s prior beliefs . Despite such biases being a pervasive feature of human behavior , their underlying causes remain unclear . Many accounts of these biases appeal to limitations of human hypothesis testing and cognition , de facto evoking notions of bounded rationality , but neglect more basic aspects of behavioral control . Here , we investigated a potential role for Pavlovian approach in biasing which information humans will choose to sample . We collected a large novel dataset from 32 , 445 human subjects , making over 3 million decisions , who played a gambling task designed to measure the latent causes and extent of information-sampling biases . We identified three novel approach-related biases , formalized by comparing subject behavior to a dynamic programming model of optimal information gathering . These biases reflected the amount of information sampled ( “positive evidence approach” ) , the selection of which information to sample ( “sampling the favorite” ) , and the interaction between information sampling and subsequent choices ( “rejecting unsampled options” ) . The prevalence of all three biases was related to a Pavlovian approach-avoid parameter quantified within an entirely independent economic decision task . Our large dataset also revealed that individual differences in the amount of information gathered are a stable trait across multiple gameplays and can be related to demographic measures , including age and educational attainment . As well as revealing limitations in cognitive processing , our findings suggest information sampling biases reflect the expression of primitive , yet potentially ecologically adaptive , behavioral repertoires . One such behavior is sampling from options that will eventually be chosen , even when other sources of information are more pertinent for guiding future action .
Many spheres of human behavior depend upon gathering and understanding evidence appropriately to inform decision-making . Yet the best way to sample information is a nontrivial problem , necessitating deciding where to sample information [1 , 2] , when to cease information gathering [3 , 4] , and weighing up how such evidence should guide behavior [5 , 6] . Normative approaches can help address these questions [7] , but their computational complexity renders them unlikely candidates for controlling behavior . Instead , these approaches can be better used as a basis for understanding limitations in cognitive processes and why biases emerge in human behavior [8 , 9] . A particularly well-studied bias is that of confirming one’s prior beliefs [10] . Inspired by classic rule discovery and falsification studies of Wason [11 , 12] , explanations of confirmation bias frequently appeal to limits in hypothesis testing as their latent cause . Several alternative accounts have been proposed . The “positive test account” [13] posits that humans form beliefs about a particular hypothesis and subsequently selectively seek and interpret evidence in support of this rule rather than against it . Yet it has been pointed out that this strategy may be normative in situations where possible competing hypotheses to explain the data are sparse [14] . Other accounts suggest that humans are simply limited in the number of hypotheses they can consider at any given time [15] . It is widely acknowledged that humans are also subject to more primitive influences on behavioral control . Whilst these have been overlooked as a potential source of confirmation bias , they are known to impact upon information seeking in other domains . For instance , a primitive behavior present in several species is the observing response [16 , 17] . Here , animals select actions to yield information ( reduce uncertainty ) about the probability of receiving future reward , even when these actions have no bearing upon reward receipt . This can also be related to human preferences for revealing advance information about rewards when that information is immaterial to the task at hand [18] . Critical here is the notion that , in nature , advance information is typically valuable in guiding future action ( unlike in the experimental tasks used to demonstrate these behaviors ) . Preferences for early temporal resolution of uncertainty [19] is thus conserved across humans and other species and persists in influencing behavior even when rendered instrumentally irrelevant . These considerations led us to consider how other primitive behaviors might bias information sampling . A notable characteristic of reward-guided behavior in many species is that of Pavlovian approach . Animals show greater efficacy in learning approach , as opposed to avoidance , actions that will lead to the delivery of reward [20 , 21] . Humans are also subject to similar approach biases [22] . Pavlovian approach effects also spill over into the domain of attentional control , as stimuli previously ascribed a high value capture attention even when they are contextually irrelevant [23] . As the locus of attention is intimately linked to information sampling during choice [24] , this raises the possibility that Pavlovian approach may similarly influence information search . To test this idea , we examined gameplay data from a large-scale smartphone app [25] in which we manipulated several factors of interest whilst probing subjects’ information sampling behavior . In brief , subjects played a card game in which they paid to sample information from different locations prior to deciding which option was most likely to yield reward . A framing manipulation meant that in half of all gameplays , approaching ( choosing ) the “biggest” option would be rewarded , but in the other half , approaching the “smallest” option would be rewarded . Crucially , the information structure of the task was identical across these matched conditions , such that any effects on information sampling could be ascribed to our manipulation as to the option subjects were instructed to approach . We compared observed behavior to predictions derived from a normative dynamic programming model that computes the expected value associated with a perfect model of the task , treated as a Markov decision process ( see Materials and Methods and [26] ) . This enabled us to isolate three distinct biases in subjects’ information search that respectively influenced where information was sought , when information collection terminated , and how information was used to guide eventual choices . Each of these three biases can be considered as a form of “approach” behavior towards locations that are more likely to yield reward . Also relevant here is our recent parameterization of human Pavlovian approach behavior in an approach-avoidance decision model on a separate economic decision task [27] . We demonstrate that the prevalence of all three biases is related to the key parameter from this model .
Subjects played a binary choice game that involved paying escalating costs for information ( by turning over playing cards ) while gambling on which option was best based upon card values that were revealed ( Fig 1A ) . There were six possible conditions that subjects might play ( Fig 1B ) . Across three of these conditions , subjects’ objective was to identify the pair ( row ) of cards with the largest product ( “MULTIPLY BIGGEST” ) , largest sum ( “ADD BIGGEST” ) , or largest single card ( “FIND THE BIGGEST” ) . Across the remaining three conditions , the objective was inverted , such that they now sought the row with the smallest product , sum , or single card . At the beginning of each trial , all cards start face down . Subjects then touch the first card ( randomly located ) to turn it over at no cost . This enters Task Stage 1 ( Fig 1A ) . One of the three remaining cards is made available to be sampled at a cost of 10 points , but subjects can alternatively make a guess ( gamble on which option will be rewarded ) at no cost . If they choose to sample , the value of the second card is revealed and they enter Task Stage 2 . Either of the two remaining cards can then be sampled at a cost of 15 points , or subjects can again choose to make a guess at no cost . If they choose to sample again , they enter Task Stage 3 . The last remaining card can be sampled at a cost of 20 points , or they may again guess at no cost . At any Task Stage , making a guess means that subjects enter the Choice Stage . Here , subjects choose which row they think will be rewarded , and all remaining cards are then turned face up . The subject wins 60 points if the gamble is correct and loses 50 points if incorrect , minus the points paid for information sampling . Card values ranged , with a uniform distribution ( sampled with replacement ) , from 1 to 10 , with “picture cards” removed from the deck . On each gameplay , subjects were randomly assigned to play two short blocks ( 11 trials each ) of two from the six possible conditions . The symmetry between the “approach big” and “approach small” conditions is crucial to our experimental design . Revealing a card of a particular value yields the same information content in both versions of the task ( with the exception of the FIND THE BIGGEST and FIND THE SMALLEST conditions ) . This means that subjects’ information gathering behavior should , normatively , be matched across these conditions . The only behavior that should change is the final gamble made by the subject , which should reverse . By comparing across ADD BIG and ADD SMALL conditions , and across MULTIPLY BIG and MULTIPLY SMALL conditions , we could probe the influence of the approach direction ( i . e . , big/small ) on information sampling behavior and vice versa . The first question we asked pertained to Task Stage 1 ( Fig 1A ) . Here , subjects decided whether to sample or guess based upon two variables: the information seen , i . e . , the card value , and also the location where information was made available for sampling . We label the first row sampled as “row A . ” In some trials , subjects were constrained to sample the next card from row A ( “AA trials” ) , whilst in other trials they were constrained to sample from row B ( “AB trials” ) . As can be seen from the optimal dynamic programming model ( Fig 2A ) , the card value and ( to a lesser extent ) the trial type influences the relative expected value of choosing to guess versus choosing to sample . The U-shaped function of the graph reflects an intuition that high- or low-valued cards are informative about the correct option to approach , making it more valuable to guess early . Mid-valued cards , by contrast , provide less information and make it more valuable to sample more information . The differential influence of AA versus AB trials is because the potential reduction in uncertainty depends upon the information that has already been revealed . Intuitively , on a MULTIPLY TRIAL where a 1 has been revealed , then sampling from row A again yields little information relative to row B , as it is already known that row A will have a low value ( between 1 and 10 ) . On a MULTIPLY TRIAL where a 10 has been revealed , then sampling from row A yields more information than row B as it reduces the range of possible row A values from between 10 and 100 to an exact value . As expected , the dynamic programming model predicts identical behavior irrespective of the subject’s approach goal . As an example , consider revealing a 2 in the MULTIPLY BIG condition on an AA trial . This yields a probability of 0 . 764 that row B will be rewarded , and the expected value of guessing is therefore 0 . 764 * 60 + ( 1–0 . 764 ) * ( -50 ) = 34 . The expected value of sampling again from the A row , calculated using dynamic programming , is 28 . 8 , and so the relative expected value of guessing is 5 . 2 ( Fig 2A ) . Now consider seeing a 2 in the MULTIPLY SMALL condition . This now yields the exact same probability of 0 . 764 that row A will be rewarded . Hence the expected value of guessing remains 34 . The expected value of sampling further information remains 28 . 8 , and so the relative expected value of guessing remains 5 . 2 . In contrast with these model predictions , subjects’ actual behavior showed a systematic difference between MULTIPLY BIG and MULTIPLY SMALL conditions ( compare circles and plus signs in Fig 2B , see also S1 Fig ) . Subjects became more likely to guess if they had seen evidence that supported them approaching row A rather than avoiding it . In MULTIPLY BIG , a high-valued card ( 6 or above ) carries evidence for choosing row A . Subjects become more likely to guess than when seeing the same card in MULTIPLY SMALL . However , a low-valued card in MULTIPLY BIG ( 5 or below ) carries evidence for avoiding row A . Subjects now become more likely to sample than when seeing the same card in MULTIPLY SMALL . This framing effect is seen most clearly when subtracting behavior in MULTIPLY SMALL from MULTIPLY BIG ( Fig 2C ) . The observed bias is one of approaching an option if positive evidence has been provided in support of that option; consequently , we term this “positive evidence approach . ” We observed that positive evidence approach was more pronounced in AB trials than AA trials ( Fig 2C ) . This is again consistent with our hypothesis , as subjects are less inclined to sample further information if it is available on a row they wish to avoid than if it is on a row they wish to approach . To quantify positive evidence approach across our population , we used the following summary statistic: ∑i=610 ( pGuessbig , i−pGuesssmall , i ) +∑i=15 ( pGuesssmall , i−pGuessbig , i ) where pGuessbig , i and pGuesssmall , i denote the average probability of guessing in MULTIPLY BIG and MULTIPLY SMALL , respectively , having revealed card value i . As there should be no difference in the probability of guessing across the two conditions , the expected value of this statistic from the normative model is 0 . By contrast , the value of this statistic across our population was 0 . 42 in AA trials and 0 . 70 in AB trials . To estimate our confidence in this summary statistic , we recomputed it on 1 , 000 bootstrapped samples of 10 , 000 gameplays from our population . This yielded 95% confidence intervals of [0 . 30 , 0 . 54] in AA trials and [0 . 62 , 0 . 79] in AB trials . ( Throughout the paper , we focus on the reporting of effect sizes and 95% confidence intervals rather than p-values , as our large sample size renders p-values less informative [28] . ) Similar results are seen by comparing the ADD BIG and ADD SMALL conditions ( S2 Fig; AA trials: mean 0 . 52 , 95% CIs [0 . 40 , 0 . 64]; AB trials: mean 0 . 79 , 95% CIs [0 . 70 , 0 . 89] ) . See also S3 Fig for FIND THE BIGGEST/FIND THE SMALLEST , which are not directly matched for information content . It is also notable that , overall , subjects’ behavioral choices to sample information were similar to predictions arising from the optimal model ( Fig 2B ) , although not identical ( S1 Fig ) . This alone does not imply that subjects are implementing the optimal model . Instead , it may simply reflect the fact that relatively simple behavioral strategies will often recapitulate many features of more sophisticated strategies [29] . For example , one straightforward strategy would be to compare the value of the presented card to the average value , estimate the current degree of uncertainty in making a choice , and then use these values with a softmax transformation [30] to calculate a probability for selecting row A , selecting row B , or sampling further information . We consider this question further in a latter section of the paper and show that this can approximate the average behavior of subjects in the task without recourse to an optimal model . We next asked how decisions to sample or reject information might influence subsequent choices . If subjects elected to guess at the first stage , they entered the Choice Stage , in which they gambled on which option would be rewarded ( Fig 1A ) . In ADD BIG , the relative expected value of choosing row A over row B increases with the value of the first card ( Fig 3A , blue line ) , while in ADD SMALL , it decreases with the value of the first card ( Fig 3A , purple line ) . This was reflected in subjects’ choices in both sets of trials ( Fig 3B; see S4 Fig and S5 Fig for other conditions ) . However , this decision arises on two different types of trial . The subject will either have just declined the opportunity of sampling information from the A row ( on AA trials ) , or the B row ( on AB trials ) . Our hypothesis was that information sampling depends upon the underlying approach value of an item . A corollary is that declining to sample an item reflects an underlying preference for not approaching it . When we compared choice preferences on AA and AB trials for identical card values on the same condition , we observed that , across all six conditions , subjects showed a systematic shift towards being less likely to choose the option that had just been left unsampled . Hence , subjects presented with the same card value on an AA trial were more likely to choose option B than on an equivalent AB trial ( Fig 3B ) . This effect was most pronounced near the point of subjective equivalence in subjects’ choices and is revealed most clearly by subtracting subjects’ choice behavior in AB from AA trials ( Fig 3C ) . We term this a “rejecting unsampled options” bias . To quantify rejecting unsampled options across our population , we used the following summary statistic: ∑i=110p ( Choice=A ) i , AB−p ( Choice=A ) i , AA where p ( Choice = A ) i , AB denotes the probability of choosing row A having observed card i on an AB trial , and p ( Choice = A ) i , AA denotes the same probability on an equivalent AA trial . There is no difference in the choice that is presented to the subject between AB and AA trials , and the expected value of this statistic is therefore 0 . The mean value of this statistic across the population was 0 . 21 in ADD BIG ( 95% CIs [0 . 10 , 0 . 32] ) and 0 . 30 in ADD SMALL ( 95% CIs [0 . 18 , 0 . 41] ) . We also found the rejecting unsampled options bias to be present across all the other conditions: MULTIPLY BIG ( mean 0 . 24 , 95% CIs [0 . 12 , 0 . 35]; S4 Fig ) , MULTIPLY SMALL ( mean 0 . 27 , 95% CIs [0 . 15 , 0 . 39]; S4 Fig ) , FIND THE BIGGEST ( mean 0 . 22 , 95% CIs [0 . 11 , 0 . 35]; S5 Fig ) , and FIND THE SMALLEST ( mean 0 . 21 , 95% CIs [0 . 09 , 0 . 34]; S5 Fig ) . Our design enabled us to also investigate where subjects chose to sample information . At Task Stage 2 on AB trials , we could determine subjects’ relative preference for sampling from row A versus row B ( Fig 1A ) . Here , different conditions have different predictions for which row is more advantageous to sample . For example , in both the ADD BIG and ADD SMALL conditions , sampling from either row yields exactly the same amount of information about which row might be rewarded . The optimal dynamic programming model predicts no relative advantage for sampling from row A versus row B ( S6 Fig ) . In both MULTIPLY BIG and MULTIPLY SMALL conditions , however , dynamic programming predicts that sampling from the row that currently has the higher-valued card will be more informative . The intuition behind this is that the range of possible outcomes on the row with the higher-valued card is greater , and so sampling further information on this row leads to a greater reduction in uncertainty than sampling the row with the lower-valued card . This is borne out in a heat map of model predictions , showing the difference in relative value from sampling from row A versus row B ( Fig 4A ) . Importantly , these predictions are identical for both MULTIPLY BIG and MULTIPLY SMALL conditions . Somewhat counterintuitively , it is therefore more advantageous to sample from the row with the largest card even in MULTIPLY SMALL . ( Note that this is different from the relative expected value of guessing versus sampling , which is shown in S8 Fig ) In contrast to model predictions , we found that subjects preferred to sample from the option that currently had the higher value in the MULTIPLY BIG condition alone ( Fig 4B , left ) . In the MULTIPLY SMALL condition , they preferred to sample from the option that currently had the lower value ( Fig 4B , right ) . The influence of this bias in subjects’ information sampling is revealed more clearly by subtracting behavior in MULTIPLY SMALL from that of MULTIPLY BIG ( Fig 4C ) . Whereas the optimal model shows no difference between these two conditions ( i . e . , the entire heat map should equal 0 ) , subjects reliably sampled information from the row that they currently sought to approach rather than avoid . We term this bias “sampling the favorite . ” To quantify sampling the favorite , we derived two statistics for trials in which subjects decided to sample a third piece of information . We calculated one “strong evidence” statistic for trials in which the “favorite” ( the item that would eventually be approached ) was clear . We defined this as trials where the difference in card values was 4 or greater in magnitude: ∑i=510∑j=1i−4 ( p ( Sample=A ) i , j , big−p ( Sample=A ) i , j , small ) − ∑j=510∑i=1j−4 ( p ( Sample=A ) i , j , big−p ( Sample=A ) i , j , small ) where P ( Sample = A ) i , j , big refers to the relative probability of choosing to sample from row A over row B , when card i is presented on row A and card j presented on row B , on MULTIPLY BIG trials . The top row of the equation denotes trials where row A has a higher-valued card than row B , favoring approaching A in MULTIPLY BIG but approaching B in MULTIPLY SMALL . The converse is true for the bottom row . We calculated a second “weak evidence” statistic for trials in which the “favorite” was less clear . We defined this as trials in which the difference in card values was between 1 and 3 in magnitude: ∑i=210∑j=max ( 1 , i−3 ) i−1 ( p ( Sample=A ) i , j , big−p ( Sample=A ) i , j , small ) − ∑j=210∑i=max ( 1 , j−3 ) j−1 ( p ( Sample=A ) i , j , big−p ( Sample=A ) i , j , small ) Crucially , because the optimal model predicts identical values for sampling from row A versus row B on MULTIPLY BIG and MULTIPLY SMALL , the expected value for both statistics is always 0 . In contrast , the value of the “strong evidence” statistic across our population was 12 . 35 ( 95% CIs [10 . 55 , 14 . 01] ) , whilst the value of the “weak evidence” statistic was 7 . 21 ( 95% CIs [5 . 85 , 8 . 48] ) . Note this bias was also observed in the ADD BIG/ADD SMALL condition ( S6 Fig ) , where the value of the “strong evidence” statistic was 11 . 93 ( 95% CIs [10 . 39 , 13 . 51] ) , whilst the value of the “weak evidence” statistic was 6 . 86 ( 95% CIs [5 . 61 , 8 . 14] ) . See S7 Fig for FIND THE BIGGEST/FIND THE SMALLEST , which are not directly matched for information content . We consider that subjects are unlikely to be implementing dynamic programming when they perform the task , yet their overall behavior shows a surprising resemblance to model predictions ( e . g . , Fig 2B , Fig 3B ) . We therefore constructed a simpler model that describes subjects’ performance without recourse to dynamic programming . In this model , subjects first compute the value of choosing each option by comparing the presented value to the average value of all possible cards . In ADD BIG trials , at stage 1 , for example , this would be the following: V ( A ) =β1 ( 1stCardValue-<1stCardValue> ) V ( B ) =β1 ( <1stCardValue>-1stCardValue ) where <1stCardValue> is the expected value of the first card ( 5 . 5 ) and β1 is a free parameter . In ADD SMALL trials , we simply inverted the value of the each card , such that 10 became 1 , 9 became 2 , and so on . We also considered an AB trial ( e . g . , Fig 4B ) , where after turning the second card , the values of option A and B become the following: V ( A ) =β1 ( 1stCardValue-2ndCardValue ) V ( B ) =β1 ( 2ndCardValue-1stCardValue ) At both stages , we compute the degree of uncertainty , ω , in choosing either option: ω=− ( 11+eV ( B ) −V ( A ) −0 . 5 ) 2 This is then used to derive the value of sampling information from option A or option B: V ( sampleA ) =β2+β3ω V ( sampleB ) =β2+β3ω The probability of each choice C being option o is finally generated using a softmax choice rule: p ( C=o ) =eV ( o ) τ∑ieV ( i ) τ where i indexes the entire set of available options at a given stage of the task . Hence , at stage 1 of an AA trial , it indexes the set {choose A , choose B , sample A}; at stage 1 of an AB trial , it indexes the set {choose A , choose B , sample B}; at stage 2 of an AB trial , it indexes the set {choose A , choose B , sample A , sample B} . Note that the model therefore assumes a three- or four-way decision at each stage of the game . Although in the structure of the task this was made as two sequential binary decisions ( guess or sample , then select a row ) , it reflects the intuition that when subjects decide to guess , they already have internally committed to choosing a particular row . We fit parameters β1 , β2 , β3 , and τ , using maximum likelihood estimation ( fminsearch in MATLAB ) separately at stage 1 and stage 2 . Fitting was repeated at 50 random seed locations to avoid local minima , and the model was fit to the average behavior of all subjects across each condition using a sum of squared errors cost function . We performed fitting separately for both ADD and MULTIPLY conditions; we do not consider FIND THE BIGGEST/SMALLEST conditions here , as these are not matched for information content . Note that this model does not explicitly feature terms for the costs associated with sampling information; instead , these are implicitly factored into the constant term β2 . This model captures the main features of the behavioral data ( S9 Fig ) . At stage 1 , it displays a U-shaped effect of card value on information sampling ( as in Fig 2B ) caused by the effects of choice uncertainty on the value of sampling information . It also displays a softmax choice curve ( as in Fig 3B ) between options A and B , matching subjects’ real choice probabilities between these two options . At stage 2 , it displays choice probabilities between options A and B that again closely match subjects’ behavior . However , because this model makes symmetric predictions for BIG and SMALL trials , it fails to capture the three biases described above ( S9 Fig ) . We therefore adapted the model with three additional parameters to capture these biases ( Fig 5 ) . At stage 1 , before entering values into the softmax choice equation , we captured the “rejecting unsampled options” bias ( Fig 5D–5F ) by adding an “approach bonus” ( β4 ) to the value of the item that could not be sampled . This makes subjects more likely to choose this option if they do not sample information . V ( A ) =V ( A ) +β41stCardValue[onABtrialsonly]; V ( B ) =V ( B ) +β4<1stCardValue>[onAAtrialsonly] . A natural consequence of this value modulation by β4 is that it also induces a change in the form of the uncertainty ω , which determines the value of sampling information ( see above ) . Whereas previously this would have been symmetric around the average value of the first card , it now becomes asymmetric , but in opposite directions on “add big” versus on “add small” trials . On AB trials , this feature of the model is thus sufficient to also capture the “positive evidence approach” bias ( Fig 5B and 5C ) . On AA trials , however , the “approach bonus” alone predicts the opposite pattern of “positive evidence approach” to that observed in the data . We instead found that we could capture the “positive evidence approach” bias ( Fig 5A–5C ) by modulating the value of sampling option A: V ( sampleA ) =β2+β3ω–β5 ( 1stCardValue ) [onAAtrialsonly] . Notably , we found increasing first card value had a negative influence on the value of sampling A , reflected by the negative sign in front of parameter β5 . We infer that on AA trials ( where option A is available to be sampled ) , subjects are more inclined to choose option A when it is of high value than to sample again from it . This parameter is not needed on AB trials , where “positive evidence approach” is already captured by the β4 parameter alone . It should also be noted that it captures another general feature of the data , which is that subjects are more likely overall to guess on AA trials than on AB trials ( see S1 Fig ) . This is because β5 reduces the value of sampling on AA trials selectively . Finally , at stage 2 , we found that we could capture the “sampling the favorite” bias ( Fig 5G–5I ) by introducing a parameter that affected subjects’ propensity to sample from higher-valued cards: V ( sampleA ) =β2+β3ω+β6 ( 1stCardValue-2ndCardValue ) V ( sampleB ) =β2+β3ω+β6 ( 2ndCardValue-1stCardValue ) Parameter fits for stage 1 and stage 2 for both ADD and MULTIPLY trials are given in S1 Table . To confirm that the additional parameters provided additional complexity to model fits without overfitting , we used 10-fold nested cross-validation . Parameters were fit using 90% of the data ( training set ) , and then the cost function was calculated for the remaining 10% of the data ( test set , not used to train the model ) . This process was iterated ten times using different portions of the data as the test set each time . At both stage 1 and stage 2 , for both ADD and MULTIPLY conditions , the model with additional parameters provided consistently better fits to the test set than the reduced model ( S2 Table ) . The close fit between model predictions and subject behavior reveals that a far simpler framework ( comparing a card value to the average expected value ) can approximate an optimal dynamic programming model . Moreover , subjects’ approach-induced biases in information sampling can be readily parameterized within this framework . We anticipate that further , more refined models will be subsequently tested by downloading the raw behavioral data from Dryad [31] . One potential drawback of the proposed model is that each of the three biases is captured by a separate parameter rather than a single unifying mechanism driven by Pavlovian approach . We therefore tested whether these parameters are related to each other across the entire population of subjects—that is , whether they might show positive covariance with each other . To this end , we adopted an alternative model fitting strategy using a mixed effects analysis to describe population behavior . A mixed effects analysis contains population-level hyperparameters to constrain individual subject model fits ( see reference [32] for details ) . An added benefit is that one can examine the covariance structure of these hyperparameters to explore how β4 , β5 , and β6 are related to each other . Importantly , we found that , after model fitting , the covariance between all three parameters was positive . We normalized by the variance of each parameter to yield correlation coefficients between β4 , β5 , and β6; this yielded a positive correlation between β4 and β5 ( r = 0 . 21 ) , between β4 and β6 ( r = 0 . 06 ) , and between β5 and β6 ( r = 0 . 12; p < 0 . 0001 for all comparisons ) . This analysis provides evidence that subjects who showed stronger expression of one of the biases also tended to show greater expression of the remaining two biases . An advantage of large-scale data collection via a smartphone app is that it allows data to be gathered on a range of cognitive tasks across a large cohort of subjects . Recently , we reported learning and choice behavior on another gambling task contained within the same app platform [27 , 33] . In this simpler gambling task , subjects make binary choices between safe and risky options in three types of trials: gain trials ( a certain gain versus a larger gain/zero gain gamble ) , mixed trials ( certain zero gain versus a mixed gain/loss gamble ) , and loss trials ( a certain loss versus a larger loss/zero loss gamble ) . Notably , subject behavior in this task was best characterized within a Pavlovian approach-avoidance decision model when compared to a range of models that also included a standard Prospect Theory model [27] . This decision model captures the influence on risk-taking behavior of both economic preferences and Pavlovian influences . It describes subjects’ value-independent propensity to approach or avoid gain gambles with a single parameter , βgain , and their value-independent propensity to approach or avoid loss gambles with a second parameter , βloss . Full details of modeling are provided in [27] and Materials and Methods . For each subject who played both games within the app ( n = 21 , 866 users ) , we estimated βgain and βloss and computed the difference between these two parameters . We performed a median split on these values to derive two subpopulations of subjects , one exhibiting a larger bias for approach potential rewards over avoid potential losses and one exhibiting a weaker bias . Next , we calculated the average behavior in our task of the subjects within these two subpopulations . We then fit the model described in the previous section to subjects’ aggregate behavior and compared the fits of β4 ( rejecting unsampled options ) , β5 ( positive evidence approach ) , and β6 ( sampling the favorite ) statistics across the different subpopulations . To estimate our confidence in these statistics , we performed 100 bootstraps using 10 , 000 samples drawn from each subpopulation . All three of our information sampling biases were differentially present in the high approach-avoid versus low approach-avoid groups . Positive evidence approach was greater in the high approach-avoid group in both add and multiply trials ( Fig 6A ) . Rejecting the unsampled option was also greater in the high approach-avoid group in the add condition , although this difference was slightly reversed in the multiply condition ( Fig 6B ) . Sampling the favorite showed a subtler pattern of expression was greater in the high approach-avoid group in both add and multiply trials ( Fig 6C ) . All of the different observed biases are linked by the tendency to sample information from locations that will eventually be approached . The present results show that this is also reflected in the expression of these biases in groups exhibiting differential levels of Pavlovian approach influence on their behavior . An additional advantage of acquiring data via smartphone is that it enables examination of variation in information sampling across a much wider range of subjects than is typically examined in laboratory studies . In an initial exploration of this , we examined variation in a simple measure of information seeking , namely the average number of cards turned relative to the optimal model . Subjects reliably sampled less information than predicted from the optimal model , but there was substantial variation across the population ( Fig 7A ) . It is important to remember , however , that the model is only “optimal” in the sense of maximizing expected points per gameplay . It does not , for example , include additional factors such as the subjective cost of sampling information . Indeed , we found that adding a “subjective sampling cost” of 5 points per turn to the optimal model shifted the distribution in Fig 7A so that it was now centered around zero ( S10 Fig ) . Nonetheless , variability in the extent to which individual subjects sampled information was highly reproducible across repeated gameplays ( Fig 7B and S10 Fig ) , and we also found it to be stable irrespective of which set or ordering of conditions subjects played ( S11 Fig ) . This suggests that it provides a measure that might be related to performance on other cognitive tasks or demographic information about participants . An example of the latter is our finding that the number of cards gathered was positively related to both the highest level of attained education and age group of our participants ( Fig 7C , top panels ) . Importantly , this measure was decoupled from general performance on the task , which was positively related to educational attainment but negatively related to age ( Fig 7C , bottom panels ) . There was a very slight tendency for subjects within the “high approach-avoid group” to gather more evidence versus subjects in the “low approach-avoid group , ” but this difference was negligible relative to the overall variance in information sampling across the population ( mean of 0 . 0066 more cards sampled in high approach-avoid group , 95%CIs [0 . 0037 , 0 . 0094] ) .
Information seeking comprises interlinked decisions that include how much to sample , where to sample from , and , finally , which option to choose based upon sampled information . Whilst the complexity of our task allowed these different features to be indexed simultaneously within a single scenario , the task was sufficiently constrained such that it can be treated as a Markov decision process . As such , an optimal model of the task can be derived using dynamic programming [26] . Dynamic programming has rarely been considered as a normative basis for analysis of information search strategies in human information search [7] . Although computationally expensive , a distinct advantage for our purposes is that it straightforwardly derives a common currency for the expected value of sampling in different locations against the value of choosing a particular option . Subjects rapidly learnt the task , with their performance in terms of points gained becoming relatively stable within ~4 trials; moreover , basic features of subject behavior ( e . g . , Figs 2B and 3B ) matched with the overall pattern of predictions from the normative model . This confirms our previous observations concerning the validity of behavioral data acquired via smartphone [25] . We make the large behavioral dataset freely available for download [31] , providing an empirical testing ground for models of human information seeking . Crucially , three features of subject behavior at different Task Stages showed demonstrable biases in information seeking . Two of these biases , positive evidence approach and sampling the favorite , were elicited as a consequence of our manipulation of which item subjects approached across different conditions . A third bias , rejecting unsampled options , was demonstrated as an effect of rejecting an option on the preference of a subject for choosing that option . All three biases were a consequence of the item that subjects currently sought to approach . Although manifesting as suboptimal biases in our experiment , we contend that these behaviors are present because they are likely to be , and have been , adaptive ecologically [34] . In nature , foraging decisions ( such as whether to stay or depart from a patch , or whether to engage with or reject an item of prey ) are more common than those made between binary mutually exclusive options [35] . In such contexts , we hypothesize that an adaptive strategy is to engage with the most valuable alternative first and then accept or reject this alternative having acquired more information about its value . It would be intriguing to test whether approach-induced information sampling can produce optimal information sampling in more naturalistic foraging paradigms . Further evidence supporting the claim that our biases are related to Pavlovian approach comes from their differential expression in two groups who varied in the degree of a Pavlovian approach-avoid parameter derived from a separate decision task . This provides a tentative suggestion of an underlying dopaminergic mechanism for control of Pavlovian approach on information seeking behaviors , given our recent demonstration that Pavlovian approach is boosted in subjects treated with L-DOPA [27] . We also note that polymorphisms in genes controlling dopamine function have recently been linked to individual differences in confirmation bias [36] . Moreover , recordings from midbrain dopaminergic neurons reveal that they signal information in a manner consistent with the animal’s preference for advance information in the same manner that they encode information about reward [17] . Future studies could easily exploit possibilities of data collection via smartphone to test this and related hypotheses via combined collection of genetic and behavioral data across large populations . It might also be possible to design future versions of our task with a larger number of trials/conditions per subject so as to elicit each of the three observed biases within-subject rather than depending upon examining amalgamated data across a population . It should be noted , however , that this effect was relatively small . This may in part be due to the limited number of trials completed on both tasks , which provides significant challenges for characterization of individual subject behavior . This is particularly the case when multiple conditions/trial types need to be completed to obtain an effect . In the present study , there were only 22 trials per subject , and this is because we explicitly aimed to ensure that the average time to complete each game was less than 5 minutes , as shorter games yield the highest number of gameplays [25] . It is possible that subjects had miscalibrated beliefs about task structure . For example , they may not have realized that there was a card value 1 , which is normally replaced by an “ace” in a regular deck of playing cards; or they may have believed that the average card value is 5 , rather than 5 . 5 . Such beliefs can straightforwardly be factored into the dynamic programming model , as can misunderstandings about the cost structure of the task , or additional opportunity costs for sampling further evidence . We found that such manipulations did indeed influence the relative preference of the model for guessing or sampling at different card values ( S10 Fig ) . Crucially , however , none of these belief-based manipulations predict any of the three biases observed . “Positive evidence approach” and “sampling the favorite” depend upon comparisons of SMALL and BIG conditions: any normative model predicts that subjects’ information sampling should be identical between these conditions , and that they should simply flip their final choice . Similarly , “rejecting unsampled options” depends upon a comparison of final choice behavior in AA and AB trials in situations in which the subject has received identical information in both trial types . It would also be possible to explore alternative versions of the current experiment that might examine the generality of our approach-avoidance account of information seeking biases . For instance , it would be intriguing to manipulate the affective valence of “points” such that they became aversive rather than rewarding . In such an experiment , we would predict that the approach-induced biases in information sampling would reverse . It would also be interesting to parametrically manipulate the costs involved in sampling different cards , as this would allow the experimenter to directly quantify the value of sampling information from different locations . It is also important to bear in mind that , even when information sampling is biased , posterior beliefs can remain unbiased if belief updating is performed normatively [37 , 38] . It would be informative in future experiments to formally dissociate subjects’ apparent biases in information sampling from their biases , if any , in their belief updating . Our findings are closely linked to other evidence from recent studies that relates the value of stimuli to deployment of attention [23 , 39 , 40] . Both these studies and our own suggest that valuable items capture attention and , hence , cause more information to be sampled from the associated location . In contrast with these previous studies , however , we show that the influence of value on information sampling occurs rapidly , can be reshaped depending upon current task goals , and can manifest as several distinct behavioral biases that affect multiple stages of information sampling . Combined , this evidence argues that choice models in which attention and information sampling are determined purely stochastically [6] require revision . Whereas these models convincingly demonstrate an important role for the locus of attention on valuation , the present data imply that the converse is also true . In simple terms , the value subjects ascribe to a location influences how likely they are to sample from it .
Ethical approval for this study was obtained from University College London research ethics committee , application number 4354/001 . Researchers at the Wellcome Trust Centre for Neuroimaging at University College London worked with White Bat Games to develop The Great Brain Experiment [25] , available as a free download on iOS and Android systems ( see http://thegreatbrainexperiment . com ) . Ethical approval for this study was obtained from University College London research ethics committee , application number 4354/001 . On downloading the app , participants filled out a short demographic questionnaire and provided informed consent before proceeding to the games . Each time a participant started a game , a counter recording the number of plays was incremented . At completion of a game , if internet connectivity was available , a dataset was submitted to the server containing fields defining the game's content and the responses given . The first time a participant completed any game the server assigned that device a unique ID number ( UID ) . All further data submissions from that device , as well as the demographic information from the questionnaire , were linked to the UID . No personal identification of users was possible at any time . Users who responded that their age was less than 18 years during the demographic questionnaire ( i . e . , minors ) were excluded from the study: the app allowed these users to play the games , but no data were submitted to the server ( hence the minimum age category on Fig 7C is 18–25 years ) . The information-seeking game was available by clicking on “Am I a risk-taker ? ” , which launched the game . On each gameplay , subjects were randomly assigned to play short blocks ( 11 trials each , as outlined in Fig 1A and main text ) of two different conditions randomly selected from six possibilities ( Fig 1B ) . In two of these , subjects had to select the row that they believe contained the largest sum ( “ADD BIGGEST” ) or largest product ( “MULTIPLY BIGGEST” ) . In a further two conditions , subjects has to reverse their eventual choice and select the row containing the smallest sum ( “ADD SMALLEST” ) or product ( “MULTIPLY SMALLEST” ) . The remaining two conditions required participants to select the row with the largest or smallest individual card ( “FIND THE BIGGEST” and “FIND THE SMALLEST , ” respectively ) . Full instructions for the task can be seen in S1 Text and S1 Movie . The economic gambling task was available by clicking on “What makes me happy ? " Subjects started the game with 500 points and made 30 choices in each play . In each trial , subjects chose between a certain option and a gamble . Chosen gambles , represented as spinners , were resolved after a brief delay . Subjects were presented with the question , “How happy are you at this moment ? ” after every two to three trials . The probabilistic structure of the task means that it is straightforward to derive a normative solution of task performance that maximizes the expected average number of points to be obtained from a given set of moves . This is achieved by applying dynamic programming to the task [30] . At each step , dynamic programming calculates the expected value of every possible action ( seeking more information in a particular location , or making a guess ) . To do so , it takes into account the full probability distribution of currently hidden cards and the possible gain in information that can be obtained from sampling further . Each combination of presented cards is defined as a state s . The best possible action a that a subject can take in a given state is defined as: Qs*=maxaQs , a In a given state , the action value Q of making a particular guess in a particular state s can be calculated as: Qs , guess=60*p ( win ) −50*p ( lose ) +totalcost where p ( win ) is the current probability of winning by making that guess , p ( lose ) is the probability of losing , and totalcost is the incurred costs for sampling information thus far . By contrast , sampling further information has a fixed probability ( 0 . 1 ) of transitioning into one of 10 possible subsequent states ( 10 different card values may be revealed ) . The value of sampling can then be calculated as the best action value in the subsequent state multiplied by the probability of transitioning: Qs , sample=∑i=1100 . 1*Qsi* where si is the state that the subject would transition into if card value i is revealed . To calculate the value of sampling , one works backwards from the terminal state ( all four cards revealed , where Qs , guess = 15 ( = 60–10–15–20 ) ) to calculate Qs* in all previous states . Full details of the approach-avoidance decision model are given in reference [27] . In brief , subjects’ expected utilities for choosing the safe option ( Ucertain ) and risky option ( Ugamble ) were fitted using an established parametric decision model based on Prospect Theory [41] . The probability of choosing to gamble was then modelled by modifying the softmax function: Pgamble=1−β1+e−μ ( Ugamble−Ucertain ) +βifβ≥0 Pgamble=1+β1+e−μ ( Ugamble−Ucertain ) ifβ<0 This causes a value-independent change in the probability of gambling , with mapping choice probabilities to be bounded at ( β , 1 ) if β is greater than zero , and ( 0 , β ) if β is less than zero . β is fit separately for gain trials and loss trials , yielding two parameters , βgain and βloss . Raw data , along with MATLAB scripts for reproducing all figures shown in the paper and code for the dynamic programming model , are freely available for download from http://dx . doi . org/10 . 5061/dryad . nb41c [31] .
|
Human decision-making often appears irrational . A major challenge is to explain why apparently irrational behavior occurs and what potential benefits it might have conferred for our evolutionary ancestors . A well-studied behavior in experimental psychology is “confirmation bias , ” where we sample information that simply confirms what we already believe . In this study , we show that one factor giving rise to such information sampling biases is Pavlovian approach: our natural tendency to approach items that are associated with reward . We demonstrate three novel information sampling biases in a large-scale smartphone experiment with >30 , 000 human subjects . We examine how these three biases are related to Pavlovian approach , as quantified via an entirely independent economic choice task . We also show that , within our population , information sampling is a stable trait of an individual that is related to demographic variables such as age and education . Although irrational in the context of our task , we postulate that approach-induced biases in information sampling may have been adaptive over evolutionary history . They would drive organisms towards gathering information about locations that they will eventually engage with to obtain reward .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"recreation",
"education",
"decision",
"making",
"engineering",
"and",
"technology",
"demography",
"sociology",
"social",
"sciences",
"neuroscience",
"cell",
"phones",
"experimental",
"psychology",
"cognition",
"games",
"behavior",
"educational",
"attainment",
"communication",
"equipment",
"gambling",
"people",
"and",
"places",
"psychology",
"equipment",
"biology",
"and",
"life",
"sciences",
"cognitive",
"science"
] |
2016
|
Approach-Induced Biases in Human Information Sampling
|
The number of child deaths is a potential indicator to assess the health condition of a country , and represents a major health challenge in Bangladesh . Although the country has performed exceptionally well in decreasing the mortality rate among children under five over the last few decades , mortality still remains relatively high . The main objective of this study is to identify the prevalence and determinants of the risk factors of child mortality in Bangladesh . The data were based on a cross-sectional study collected from the Bangladesh Demographic and Health Survey ( BDHS ) , 2011 . The women participants numbered 16 , 025 from seven divisions of Bangladesh – Rajshahi , Dhaka , Chittagong , Barisal , Khulna , Rangpur and Sylhet . The 𝟀2 test and logistic regression model were applied to determine the prevalence and factors associated with child deaths in Bangladesh . In 2011 , the prevalence of child deaths in Bangladesh for boys and girls was 13 . 0% and 11 . 6% , respectively . The results showed that birth interval and birth order were the most important factors associated with child death risks; mothers’ education and socioeconomic status were also significant ( males and females ) . The results also indicated that a higher birth order ( 7 & more ) of child ( OR=21 . 421 & 95%CI=16 . 879-27 . 186 ) with a short birth interval ≤ 2 years was more risky for child mortality , and lower birth order with longer birth interval >2 were significantly associated with child deaths . Other risk factors that affected child deaths in Bangladesh included young mothers of less than 25 years ( mothers’ median age ( 26-36 years ) : OR=0 . 670 , 95%CI=0 . 551-0 . 815 ) , women without education compared to those with secondary and higher education ( OR =0 . 711 & . 628 , 95%CI=0 . 606-0 . 833 & 0 . 437-0 . 903 ) , mothers who perceived their child body size to be larger than average and small size ( OR= 1 . 525 & 1 . 068 , 95%CI=1 . 221-1 . 905 & 0 . 913-1 . 249 ) , and mothers who delivered their child by non-caesarean ( OR= 1 . 687 , 95%CI=1 . 253-2 . 272 ) . Community-based educational programs or awareness programs are required to reduce the child death in Bangladesh , especially for younger women should be increase the birth interval and decrease the birth order . The government should apply the strategies to enhance the socioeconomic conditions , especially in rural areas , increase the awareness program through media and expand schooling , particularly for girls .
Child deaths , which are one of the most important health indicators for a country , represent the socio-economic development of a population and have received considerable attention from national and international agencies in the last few decades because of their inclusion in the Millennium Development Goals ( MDGs ) [1 , 2 , 3] . A significant number of projects and programs have been conducted worldwide to reduce the under-five child deaths , especially in resource-limited countries , by two-thirds in 2015 . In 2013 , WHO reported that among the total of 6 . 3 million deaths of children aged under-five , 74% of them ( 4 . 6 million ) died within the infancy period . However , 45% died within the first few months after birth . Almost 83% of this child mortality was due to neonatal , infectious or nutritional conditions . Although the number of under-five deaths has declined worldwide in recent decades ( from 12 . 7 million in 1990 to 6 . 3 million in 2013 ) , the number is still alarmingly high [4] . In addition , the high burden of child mortality still exists in South Asia ( one child dies out of 15 before reaching 5 years of age ) as well as Sub-Saharan Africa ( one child dies out of 8 ) [2 , 5] . About 33% of global child mortality occurs in South Asian countries , compared to less than 1% in high-income countries [6] . In Bangladesh , the mortality of children has followed a declining trend , having reduced from 133 deaths per 1000 in 1993 to 53 deaths per 1000 in 2011 , which confirms that Bangladesh is more likely to reach target 4 of the MDG ( 48 deaths per 1000 children under 18 years of age ) by 2015 [7] . Child deaths , as reported in previous studies , is associated with various socio-demographic , and health related characteristics , e . g . , lower parental education , lower socioeconomic status , and higher order of birth are associated with increasing risk of child mortality [8 , 9 , 10] . Whereas , large birth spacing , lower birth order , urban dwelling , and high socioeconomic status are associated with a lower risk of child mortality [10 , 11 , 12] . Moreover , another study in Bangladesh identified parent’s education , parent’s occupation , delivery and size of child as significant determinants of child mortality [13 , 14] . A few studies also reported a significant difference in child deaths between urban and rural areas [15 , 16] . In addition , a multi-country study confirmed that higher national income is associated with a lower rate of child deaths [17] . In developing countries like Bangladesh , women are neglected in almost all aspects of life . In addition , such negligence starts from the childhood , as households , especially from rural areas and among the uneducated have a desire for a male child . Even the family is unhappy for the birth of a female baby . Moreover , social disparity and gender differences for health care exist; for example , girls experience a delay in the intervention for their illness compared to boys [18] . Therefore , the same factor may affect child deaths in a different fashion ( severity ) for male and female children . Although some studies have been conducted in the field of child mortality in Bangladesh [10 , 13 , 14 , 19] , they did not study child deaths separately for male and female children . Therefore , the objective of this study is to determine the risk factors that influence child deaths in Bangladesh among males and females separately .
This is a cross-sectional study using data from the BDHS , 2011 . There are seven administrative divisions in Bangladesh—Dhaka , Rajshahi , Rangpur , Chittagong , Khulna , Barisal and Sylhet . One division is subdivided into districts ( zilas ) , and each district is divided into administrative units ( upazilas ) , which are further divided into urban and rural areas . An urban area is divided into wards and city corporation units ( mohallas ) within a ward , while a rural area is divided into union parishes ( UP ) and villages ( mouzas ) within a union parish . The 2011 population and housing census , together with the Bangladesh Bureau of Statistics ( BBS ) , were used as the sampling frame for the list of enumeration areas ( EAs ) in this survey [7] . A two-stage stratified sample of households was the basis for this survey . In the first stage , 207 clusters in urban areas and 393 in rural areas were selected totaling 600 enumeration areas with proportional probability . In the second stage , on average , 30 households were selected based on the demographic and health survey variables , for both the urban and rural areas in seven divisions by systematic sample . The survey selected 17 , 842 residential households within this study design [7] . The main objective of the study is to determine the prevalence and risk factors of child deaths in Bangladesh . The statistical analysis of the results was measured using the IBM statistics Version 21 . The χ2 test was used to determine the significant associations with the child deaths ( boys and girls ) and respondents age , place of residence , educational level , socioeconomic status , total number of children ever born , birth interval , mode of delivery and size of child at last birth . Logistic regression analysis was conducted to determine the risk factors with child deaths ( boys and girls ) and respondents age , place of residence , educational level , socioeconomic status , total number of children ever born , birth interval , mode of delivery and size of child at last birth ( Table 1 ) . A P value of 0 . 05 was considered significant at the 95% CI ( Confidence Interval ) level . The dependent variable used in the model was the dichotomous binary variable: Y = 1 if have child deaths ( boys and girls ) and Y = 0 , otherwise . Respondents age , place of residence , educational level , socioeconomic status , total number of children ever born , birth interval , mode of delivery and size of child at last birth were considered as predictor variables in this model . Perform preliminary analyses using univariate tests such as chi-square test . Such initial analyses provide a better grasp of what is happening in the data and may point the potential important variable ( s ) to be used for the multivariate analysis . As a general rule , perform univariate analyses for each independent variable , if they are significant at p-value of 0 . 25 then select them to the multivariate analysis . If the χ2 test for a single independent variable is not significant there is no need to include it in the logistic regression analysis . The rural population refers to the people living in rural areas and urban refers to the people living in urban or city areas . There is a difference between urban and rural in that urban people have access to medical facilities , health care centers , good communication facilities , living standards and schooling compared to rural areas . Poor people refers to those with a monthly income of less than 3000 Bangladeshi Taka , middle class people refers to those with a monthly income of 3000–15 , 000 and rich people refers to those with a monthly income of more than 15 , 000 Bangladeshi Taka . If the baby size is 2700–4000 grams ( 6–9 pounds ) that means average baby size , less than 2700 grams is small and more than 4000 grams is large [7] . Ethical approval was obtained from the Ministry of Health and Family Welfare , Dhaka , Bangladesh . International Credit Finance ( ICF ) provided financial and technical assistance for the survey through USAID/Bangladesh . The BDHS is part of the worldwide Demographic and Health Surveys program , which is designed to collect data on child health [7] . Informed written consent was obtained from all women prior to the study .
The total women participants included in the study numbered 16 , 025 . The results of the associations of child mortality for several socio-demographic characteristics of women are presented in Table 2 . The findings from the present study indicated that the prevalence of male children dying was 13 . 0% and 11 . 6% for female children . For both sexes , child deaths was significantly associated with mothers’ age , place of residence , mothers’ educational level , total number of children ever born , socio-economic status of household , preceding birth interval , mode of delivery , and size of child at birth . In observing the reported child mortality from several socio-demographic characteristics , an increasing percentage of child mortality was observed in the case of the mothers’ age . Child mortality was significantly higher in rural areas ( 14 . 2% for male child and 12 . 8% for female child ) compared to those in urban areas ( 10 . 7% for male child and 9 . 4% for female child ) . The mothers’ educational level was significantly associated with child deaths . The child mortality rate for males was higher among women with no formal education ( 22 . 0% ) followed by primary educated ( 14 . 5% ) , secondary educated ( 6 . 3% ) and higher educated women ( 3 . 4% ) . A similar pattern of mortality was also observed in the case of female children; however , for each level of mothers’ education , the percentage was comparatively lower than the mortality for males . For both sexes , the reported child mortality was high among poor households ( 17 . 4% for males and 14 . 9% for females ) and the percentage decreased with the increasing socio-economic status . Women with more than six children reported a higher percentage of child mortality for both sexes , whereas it was lower among women with 1–2 children . However , the percentage of child mortality was higher among those children whose preceding birth interval was less than 24 months and lower among those children whose preceding birth interval was more than 48 months ( 12 . 5% for males and 11 . 8% for females ) . The percentage of child deaths was higher among those children who were delivered normally ( 13 . 5% for males and 12 . 2% for females ) than for children who were delivered by caesarean section ( 6 . 0% for males and 4 . 7% for females ) . The percentage of mortality was also higher among those children whose size at birth was small ( 16 . 2% for males and 14 . 6% for females ) ; however , the percentage was lower among those whose size at birth was large ( 9 . 0% for males and 7 . 4% for females ) . The results in table 3 indicated a positive high relationship with the place of residence and socioeconomic status , a negative high correlation with educational level and socioeconomic status , a positive correlation with the total number of children ever born and the mode of delivery and the other variables are negatively or positively correlated with each other . From table 4 , the results showed an inverse moderate relationship with respondents’ age and total number of children ever born , a positive high relationship with place of residence and socioeconomic status , a positive moderate relationship with the total number of children ever born , and positive moderate relationship with the mode of delivery and total number of children ever born . Table 5 represents the risk factors associated with child deaths . Two separate binary logistic regression models were fitted to identify the socio-economic correlates of child mortality for males and females . The selected socio-demographic variables considered in the two models are mothers’ age , place of residence , mothers’ educational level , total number of children ever born , socio-economic status of household , preceding birth interval , mode of delivery , and size of child at birth . According to the fitted models , all of the selected variables except place of residence and child size at last birth appeared to be statistically significantly correlated to child mortality in the case of males; whereas all of the selected variables except place of residence appeared to be significant correlates of child mortality in the case of females . The findings from this study revealed that women who are aged 26–36 years have 33% and 26% lower risk of child mortality for their male children ( OR = 0 . 67 , 95%CI = 0 . 55–0 . 81 ) and female children ( OR = 0 . 74 , 95%CI = 0 . 59–0 . 91 ) , respectively , compared to women aged 25 years and below . There was a negative association between higher education and child mortality . The risk was significantly lower among women who were higher educated followed by secondary and primary educated compared to women with no formal education . For example , in the case of female mortality , the risk was about 50% lower among higher educated women ( OR = 0 . 51 , 95% CI = 0 . 35–0 . 77 ) compared to women with no formal education . Household socio-economic status was an important indicator that was correlated with child deaths between both males and females . Middle class and upper households were 0 . 81 times ( OR = 0 . 81 , 95%CI = 0 . 69–0 . 93 ) and 0 . 71 times ( OR = 0 . 71 , 95%CI = 0 . 61–0 . 81 ) less likely to have sons die , respectively , than poor households; whereas , rich households had 0 . 81 times less ( OR = 0 . 81 , 95% CI = 0 . 70–0 . 93 ) chance of having daughters die . In addition , women whose total number of children ever born was 3–6 and more than 6 were 5 . 33 times and 21 . 42 times more likely of having sons die compared to women whose total number of child ever born was one to two . An almost similar risk was seen in the case of girl child’s mortality . The preceding birth interval also appeared to be another important correlate for both models . The risk of death was 0 . 71 ( OR = 0 . 71 , 95% CI = 0 . 62–0 . 81 ) times less for those whose preceding birth interval was 25–48 months compared to those whose preceding birth interval was less than 25 months . Sons who were delivered by normal section had 1 . 68 times ( OR = 1 . 68 , 95% CI = 1 . 25–2 . 27 ) more risk of mortality compared to sons delivered by caesarean . However , the risk was 1 . 47 times ( OR = 1 . 47 , 95% CI = 1 . 06–2 . 04 ) more in the case of daughters . Daughters whose size at birth was average were 1 . 53 times ( OR = 1 . 53 , 95% CI = 1 . 22–1 . 91 ) more likely to face mortality compared to daughters whose size at birth was large , which indicates that large size at birth is a protective factor for child deaths . For the fitted model , in respect of the Cox and Snell R2 and Nagelkerke R2 , 12% and 20% , respectively , of the variance for boys , and 12% and 21% , respectively , of the variance for girls can be predicted from the linear relationship of the eight independents variables . In respect of the overall percentage , 88% and 87% women were predicted correctly . The overall model was significant when all eight independent variables were entered .
The child death rate reflects a country’s level of socioeconomic improvement and quality of life . It depends on monitoring and evaluating the population and health programs and policies . These rates are also useful in identifying promising directions for the health and nutrition programs in Bangladesh [7] . The results of this study indicate that for the death of children less than 18 years of age by gender difference , a child’s death depends on mothers’ age , mothers’ education , socioeconomic status , geographic difference and mass media awareness program . Births to younger women can generally be identified as high-risk births [2 , 20] . This study confirms that children born ( especially boy ) to women of more than 25 years of age are at lower risk of child deaths . In Bangladesh , majority younger women are not developed economically , emotionally , physically , and are less mature , which , together with other aspects of their life , have a detrimental impact on their children . Women’s education has an inverse relation with children’s deaths . Educational level is highly association with lower mortality risks because women’s education provides the information about better pregnancy and child health care [7 , 21] . The results show that the child deaths for boys is 0 . 711 times lower risk for those whose mothers have completed higher education than for those whose mothers are illiterate ( 6 . 3% to 22 . 0% deaths ) , and for the child deaths of girls it is 0 . 588 times lower for those whose mothers have completed higher education than for those whose mothers are illiterate ( 5 . 2% to 21 . 0% deaths ) . Mothers with higher education reduce the risk of child deaths by about half . Educated women have better income , higher health literacy and power to make healthier decisions for their health and that of their children [3] . Child deaths in urban and rural areas in Bangladesh have a statistically significant risk difference . Some studies have identified that child mortality in rural areas is at higher risk than in urban areas [3 , 10–14 , 15 , 22] . This study shows that there is a significant association with urban-rural and gender difference , such as the deaths of urban boys and girls ( 10 . 7% to 9 . 4% ) and that of rural boys and girls ( 14 . 2% to 12 . 8% ) . This is because urban areas have good access to basic medical facilities and health care compared to rural areas . The wealth index of women is the main way to determine the health status in a country . Some studies have shown a positive statistical association between low income and child deaths but an opposite relationship between high income and child deaths [3 , 8 , 17] . Similarly , in this study the lower income women are at greater risk for child deaths ( boys and girls ) . Because lower income women have access to fewer medical facilities in rural areas , such as hospitals , doctors , paramedics or community based health workers . Birth order is an essential measure of child deaths . Although some previous studies have shown that higher ranked birth order present a higher risk of child deaths , a few other studies have indicated that lower ranked birth order experience an increased risk of child deaths . The infants of first birth have a higher risk of neonatal mortality than fourth or higher-ranked births in India [23 , 24 , 25] . A study conducted in Taiwan [26] showed that children with first and fifth-ranked births are at higher risk of early child deaths , with same in Nigeria [27] . In this study , women who have children ever born 3–6 is 5 . 33 times more risk of boys child death , and children ever born more than 7 is 21 . 42 times more risk of sons child death as compared to women who had at least one child ever born . An almost similar risk is seen in the case of girl child deaths . Short birth interval and child deaths have a significant association in Bangladesh [3 , 28] . Another study also found a similar risk association [29] . The risk of death is 0 . 71 times less for those women having a birth interval of 25–48 months compared to those women having a birth interval of less than 25 months . One of the limitations in this study is that the information was self-reported .
The causes of child deaths is important for health sector planning , especially in determining the program needs , and monitoring for improving intervention and reassessment of health priorities . Increasing mothers’ education and making them productive to improve their income are important aspects for reducing child deaths . Hence , it is having needed to increase rural based community education programs about child deaths . Being a younger mother at birth , short birth interval and birth order have been identified as risk factors for increased child deaths in Bangladesh . Interventions targeted at empowering women and much effort in emerging rural areas is necessary . Reducing young motherhood and increasing satisfactory birth spacing are also necessary to reduce child deaths in Bangladesh .
|
Children are a significant asset of a country . Child deaths are an important way to determine the health sector development . The effectiveness of the interventions is required to prevent child deaths . The purpose of this study is to identify the prevalence and risk factors of child deaths in Bangladesh . Data were collected from the Bangladesh Demographic and Health Survey , 2011 . The results indicate that in Bangladesh there is an association with child deaths and mothers’ age , mothers’ education , social-economic status , birth interval , birth order , baby size and place delivered . For Bangladesh , this study recommends expanding female education to increase mothers’ knowledge , an awareness program about birth order ( take one child ) and an increase in the birth interval .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion",
"Conclusion"
] |
[] |
2015
|
Prevalence and Determinants of the Gender Differentials Risk Factors of Child Deaths in Bangladesh: Evidence from the Bangladesh Demographic and Health Survey, 2011
|
Here we present the first global functional analysis of cellular responses to pore-forming toxins ( PFTs ) . PFTs are uniquely important bacterial virulence factors , comprising the single largest class of bacterial protein toxins and being important for the pathogenesis in humans of many Gram positive and Gram negative bacteria . Their mode of action is deceptively simple , poking holes in the plasma membrane of cells . The scattered studies to date of PFT-host cell interactions indicate a handful of genes are involved in cellular defenses to PFTs . How many genes are involved in cellular defenses against PFTs and how cellular defenses are coordinated are unknown . To address these questions , we performed the first genome-wide RNA interference ( RNAi ) screen for genes that , when knocked down , result in hypersensitivity to a PFT . This screen identifies 106 genes ( ∼0 . 5% of genome ) in seven functional groups that protect Caenorhabditis elegans from PFT attack . Interactome analyses of these 106 genes suggest that two previously identified mitogen-activated protein kinase ( MAPK ) pathways , one ( p38 ) studied in detail and the other ( JNK ) not , form a core PFT defense network . Additional microarray , real-time PCR , and functional studies reveal that the JNK MAPK pathway , but not the p38 MAPK pathway , is a key central regulator of PFT-induced transcriptional and functional responses . We find C . elegans activator protein 1 ( AP-1; c-jun , c-fos ) is a downstream target of the JNK-mediated PFT protection pathway , protects C . elegans against both small-pore and large-pore PFTs and protects human cells against a large-pore PFT . This in vivo RNAi genomic study of PFT responses proves that cellular commitment to PFT defenses is enormous , demonstrates the JNK MAPK pathway as a key regulator of transcriptionally-induced PFT defenses , and identifies AP-1 as the first cellular component broadly important for defense against large- and small-pore PFTs .
Pore-forming toxins ( PFTs ) are proteinaceous virulence factors that play a major role in bacterial pathogenesis [1] , [2] , [3] . PFTs constitute the single largest class of bacterial virulence factors , comprising ∼25–30% of all bacterial protein toxins [3] , [4] . PFTs have been demonstrated to be important in vivo virulence factors for Staphylococcus aureus , Group A and B streptococci , Streptococcus pneumoniae , Enterococcus faecalis , uropathogenic Escherichia coli , Clostridium septicum , and Vibrio cholerae [2] , [5] , [6] , [7] , [8] . There are various ways of grouping PFTs , based on the structure of the pore , based on their pore size as determined by structural or functional data , or even based on the organisms that produce them [9] , [10] . With regards to pore size , they generally can be grouped into two categories , those that form small ( 1–2 nm ) diameter pores and those that form large ( ≥30 nm ) diameter pores [9] , [10] . Regardless of these groupings , PFTs as a class play a singularly important role in bacterial pathogenesis in mammals . The mode of PFT action is simple yet elegant—they poke holes in the plasma membrane , breaching cellular integrity and disrupting ion balances and membrane potential [10] . As a consequence , cells die or malfunction , which significantly aids bacterial pathogenesis . Although unregulated pores at the cell surface might be expected to be catastrophic , cells have apparently evolved some mechanisms to protect against low-moderate doses of PFTs . First shown in Caenorhabditis elegans and then demonstrated in mammalian cells , the p38 mitogen-activated protein kinase ( MAPK ) pathway was the first intracellular pathway demonstrated to protect cells against PFTs [11] , [12] , [13] , [14] . C . elegans animals or mammalian cells lacking p38 MAPK are more susceptible to killing by PFTs . Three different downstream targets of the p38 PFT defense pathway were identified in C . elegans—two toxin-regulated targets of MAPK called ttm-1 and ttm-2 , and the IRE-1 – XBP-1 unfolded protein response ( UPR ) pathway [11] , [14] . The ttm genes and the UPR are all required for PFT defenses , are all induced by crystal toxin PFT in C . elegans , and all require the p38 pathway for their induction . Conserved induction of the UPR in mammalian cells by a PFT that makes similar in size to crystal toxin was demonstrated [14] . A few other scattered studies have identified hypoxia inducible factor ( hif-1 ) , wwp-1 ( part of the insulin pathway ) , and sterol regulatory element binding protein ( SREBP ) as involved in cellular defenses against PFTs [15] , [16] , [17] . These studies raise the question as to how extensive cellular defenses to PFT attack are . In a broader sense , since PFTs likely act similar to membrane damage that occurs in daily the life of cell [3] , [10] , these studies raise the question as to how cells deal with unregulated holes at their membranes . How many genes are involved ? Are PFT defenses relatively limited or are they extensive ? Is there a coordinated pathway for defensive responses or are multiple parallel pathways involved ? Little work has been done in this area since it was assumed that unregulated pores at the membrane are catastrophic , likely leading to osmotic lysis . In essence , PFT attack was assumed to be too simple for detailed scientific research . To address the extent to which cells respond to PFT attack , we report here on the first high-level systematic study of PFT responses in cells . Namely , we perform a C . elegans RNAi screen to characterize on a genome-wide scale the genes involved in PFT defenses . Follow up of this data led us to investigate the relative importance of two MAPK pathways in regulating PFT defenses . The combination of these data with other functional and molecular data using both small- and large-pore PFTs in C . elegans and mammalian cells reveal important insights into the extent of their genome that cells employ to neutralize proteinaceous membrane pores and how cellular responses to PFTs are regulated .
C . elegans is susceptible to crystal ( Cry ) protein PFTs , such as Cry5B , made by the soil bacterium Bacillus thuringiensis ( Bt ) [18] , [19] . Based on homology modeling using several available Cry toxin structures as templates [20] , Cry5B is a member of the three-domain Bt Cry PFTs that generally form small 1–2 nm diameter pores similar in size to those of α-toxin from S . aureus and cytolysin from V . cholerae [1] , [3] , [21] . To directly demonstrate that Cry5B is a PFT , we examined the ability of purified , activated Cry5B to permeabilize artificial phospholipid membranes . Current transitions between open and closed states demonstrate clearly that Cry5B forms ion channels in planar lipid bilayers ( Figure 1A ) , and is therefore a functional PFT . Furthermore , the conductance of the pores is 125pS or less , consistent with that of other Cry toxins tested under identical experimental conditions [22] , and whose pore size has been determined to be in the 1–1 . 3 nm radius [23] . Thus , Cry5B intoxication of C . elegans is a valid model of a small-pore PFT attack upon eukaryotic cells in vivo . To ascertain globally how extensive cellular commitment to protection against PFTs is , we performed a genome-wide RNAi screen using the Ahringer bacterial RNAi feeding library , which targets 16000+ C . elegans genes or about 85% of the C . elegans genome [24] ( Figure 1B ) . Each of the 16000+ individual gene knock-downs were fed Cry5B PFT and assayed in duplicate wells for the Hpo ( hypersensitive to pore-forming toxin ) phenotype ( Figure 1C ) . Hpo animals display a significantly higher level of intoxication compared to wild-type animals when exposed to a sublethal PFT dose . Intoxicated animals are dead or pale , shrunken , and less motile . Several hundred initial hits were retested for the Hpo phenotype in a second round of semi-quantitative knock-down assays ( see Materials and Methods ) . At this round , each gene knock-down was also counter assayed for health in the absence of PFT attack; RNAi clones that caused obvious ill health in the absence of PFT were eliminated from future consideration . RNAi treatments that replicated the Hpo phenotype in the second round were then subjected to a third round of testing , namely quantitative mortality assays in the presence or absence of a fixed dose of Cry5B PFT ( 10 µg/mL , triplicate wells ) . We kept genes for which knock down resulted in ≤60% viability on 10 µg/mL PFT relative to empty vector controls and ≥83% viability in the absence of PFT relative to empty vector controls . These criteria were selected to maintain a balance between gene knock-downs with solid Hpo phenotypes and relatively good health in the absence of toxin . One hundred and six gene knock-downs passed the three rounds of testing ( Table 1 ) . Relative to no knock-down controls , 1 ) for 92/106 genes , there is ≤10% lethality in the absence of toxin , 2 ) for 91/106 genes there is ≥50% lethality in the presence of toxin , and 3 ) there is ≥4X more lethality in the presence of toxin than in the absence of toxin for all but two genes . To demonstrate the robustness of the list , we selected 12 of these hpo genes that span the range of PFT hypersusceptibility and , in a fourth round of testing , repeated PFT lethality assays with knock-downs in four independent trials . All twelve knock-downs were statistically confirmed to be Hpo relative to no knock-down controls and healthy in the absence of PFT ( Table S1 ) . Interestingly , of the eight known hpo genes identified in previous screens and present in the RNAi library , we found all three of the genes that mutate to a strong Hpo phenotype and none of the four genes that mutate to a weak to moderate Hpo phenotype ( see Table 1 legend; knock down of the eighth gene is sickly in the absence of toxin and would have been screened out ) . Thus , we have likely identified many of the genes that mutate to a strong Hpo phenotype and likely missed an equally large number of genes that mutate to a weak-moderate Hpo phenotype . The 106 strong hpo genes constitute >0 . 5% of the C . elegans genome ( ∼20 , 000 genes ) . Based on the Wormbase annotation , we manually classified them into the following nine categories: G-protein-coupled receptor ( GPCR ) , ligand and cell surface related , signal transduction , transcription factor and gene expression , transporter , vesicular trafficking , metabolism , miscellaneous , and uncharacterized ( Figure S1 ) . Thirteen of these genes have been previously implicated in , or are homologues of genes involved in innate immunity , indicating that much of the defense against PFTs is uncharacterized relative to established immune pathways . The most highly represented group is that containing genes associated with metabolic activities , suggesting cells may alter metabolism to mount an appropriate defense and repair in response to PFTs . Among the 106 hpo genes , we also identified many genes associated with vesicle trafficking and membrane transporters , both of which have been previously implicated , but not demonstrated to be involved , in PFT defenses [25] , [26] . Other significant groupings include GPCR , signal transduction , and transcription factor and gene expression . Such genes might include key components for the sensing of upstream signals , and downstream effector genes for PFT defenses [11] , [13] . We found that the majority of these 106 genes have orthologs or homologs in fly ( 71% ) , mouse ( 71% ) , and humans ( 75% ) ( Table 1 ) . Taken together , these findings support the idea that metazoans employ an extensive and conserved response to counteract PFTs and breaches of the plasma membrane . To understand how functional responses against PFTs compare to responses to other stressors and conditions , we compared our list of 106 hpo genes with published genome-wide RNAi screens in C . elegans for hypertonic stress sensitivity [27] , osmo-regulation [28] , life-span regulation [29] , [30] , fat content regulation [31] , protein aggregation regulation [32] and irradiation sensitivity [33] ( no other screens involving a challenge related to bacterial pathogenesis have been published to our knowledge ) ( Figure 1D ) . The comparison revealed a statistically significant ( P = 0 . 000051 ) overlap of the hpo genes with only one set of genes , those that upon inactivation lead to reduced life-span [30] . The lack of overlap with most other biological processes suggests that protection against small pores involves a specialized protective response and/or integrates many different responses , none of which are singularly dominant . To make sense of these 106 genes , we mapped potential interactions among them ( along with a few others from our previous publications ) using the C . elegans interactome database that contains protein-protein interactions and genetic interactions based on yeast-2-hybrid data and literature [34] . Using this unbiased approach , a major interconnected network containing one-third ( 38 ) of all the hpo genes emerges . This network contains at its center two MAPK pathways , the p38 MAPK pathway and the c-Jun N-terminal kinase ( JNK ) MAPK pathway ( Figure 1E ) . Thus , based on interactions from our genome-wide RNAi screen , we hypothesized that MAPK pathways play a central role in cellular PFT responses . The identification of p38 MAPK as important in PFT defenses in C . elegans , insect , and mammalian cells has been noted above and characterized in detail [11] , [12] , [13] , [14] . The conservation of the p38 pathway in PFT defenses suggested it might be the central regulator of these defenses . On the other hand , the JNK-like MAPK pathway , in C . elegans represented by the JNK-like MAPK KGB-1 , was briefly noted as playing a role in Cry5B defenses [11] but not functionally studied in detail in any system . Whether it might be central to PFT defenses or a peripheral player was unknown . The quantitative data from our genome-wide RNAi screen suggested kgb-1 might be important since knock-down results in a quantitatively strong hypersensitivity to PFT phenotype ( Table 1 , Table S1 ) . Knock-down in the MAPKK gene , mek-1 , known to function upstream of KGB-1 JNK-like MAPK [35] , also results in hypersensitivity to PFT ( Figure S2 ) . To determine the relative importance of the JNK and p38 pathways in PFT defenses , we turned to microarray analyses . Since inductive transcriptional responses are an important part of MAPK-mediated responses , we hypothesized that , if important for PFT defenses , KGB-1 might play an important role in PFT-induced gene transcription . We previously used microarray analyses to determine to what extent the p38 MAPK pathway controlled Cry5B PFT-induced transcriptional responses and to identify two downstream targets of the p38 pathway , ttm-1 and ttm-2 , both of which phenotypically play a minor role in Cry5B PFT defenses [11] . Here , we carried out a similar microarray study with the JNK MAPK pathway in order to determine how many gene transcripts that are normally induced by Cry5B PFT are dependent upon the JNK pathway for their induction . We exposed wild-type and kgb-1 ( um3 ) JNK-like MAPK mutant animals either to Cry5B PFT-expressing bacteria or to bacteria carrying the empty vector ( no-toxin control ) , each for three hours . The kgb-1 ( um3 ) loss-of-function mutant lacks 1 . 2 kb of genomic DNA and deletes most of the kinase domain [36] , [37] . That kgb-1 ( um3 ) represents a null mutant is supported by our Western blotting experiments ( see below ) . RNA was isolated from these animals in three independent repeats and hybridized to Affymetrix-based C . elegans whole genome microarrays . We then compared these data with those found in the p38 MAPK microarray experiment [11] . From the total of six wild-type expression profiling microarrays ( three repeats for the sek-1 set and three for the kgb-1 set , both with and without toxin ) , we found that at a 2-fold cut-off , 572 transcripts were induced in wild-type C . elegans upon exposure to Cry5B PFT; that number increases to 1117 transcripts if the cut-off is set to 1 . 5-fold ( P<0 . 01 for both ) . Of repressed transcripts , we found that 707 transcripts repressed at the 2-fold cut-off and 1428 repressed at the 1 . 5-fold cut-off ( P<0 . 01 for both ) ( Figure 2A; see Supporting Information S1 for a complete list of differentially expressed transcripts ) . We cross-checked these transcripts with the 106 hpo genes identified above , and found that 15 hpo genes were induced ≥1 . 5 fold and 8 were induced ≥2 fold . Statistically , there is a significant enrichment of induced genes on the hpo list ( P = 0 . 00041 and P = 0 . 0148 respectively ) . In contrast , the overlap between repressed genes and hpo genes is not significant ( P = 0 . 345 with 1 . 5 fold cutoff , seven gene overlap , and P = 0 . 652 with 2 fold cut-off , four gene overlap ) , indicating that the appearance of repressed genes on the list of hpo genes is likely from random chance . We then classified the dependence of all the transcripts ( ≥2-fold cut-off ) into one of four categories: 1 ) transcripts dependent upon only the p38 MAPK pathway for their induction or repression by Cry5B PFT , 2 ) transcripts dependent upon only the JNK MAPK pathway for their induction or repression by Cry5B PFT , 3 ) transcripts dependent upon both p38 and JNK MAPK pathways for their induction or repression , and 4 ) transcripts whose induction or repression by Cry5B PFT are independent of either MAPK pathway ( Figure 2B ) . The results of these analyses were stunning: whereas the p38 MAPK pathway was responsible for regulating 8% ( 48/572 ) and 2% ( 14/707 ) of the PFT-induced and -repressed transcripts respectively , the JNK MAPK pathway controlled 51% ( 290/572 ) and 24% ( 172/707 ) of the PFT-induced and -repressed transcripts respectively . Thus , the JNK pathway controls more than half of the induced transcripts and more than six times the total number of transcripts controlled by the p38 pathway . Furthermore , the JNK pathway controls 85% ( 41/48 ) of the p38 MAPK-dependent induced transcripts and 43% ( 6/14 ) of these p38-dependent repressed transcripts . To validate the microarray data , we chose one up- and one down-regulated gene from each of the four categories , isolated RNA from three independent experiments ( minus or plus Cry5B PFT ) , and performed quantitative real-time ( qRT ) PCR analyses ( Figure 2C ) . The qRT-PCR results validate the microarray results . Since the JNK MAPK pathway is important for the PFT-induced transcripts , we predicted the pathway itself would be activated upon exposure to PFT . We first showed that there is no detectable KGB-1 protein in kgb-1 ( um3 ) mutant animals and further confirmed that the JNK-like MAPK pathway is activated in C . elegans treated with Cry5B PFT ( Figure 3A ) , as has been seen with the mammalian cell responses to PFTs [38] , [39] . Interestingly , in kgb-1 ( um3 ) mutant animals , we detected the phosphorylated form of p38 MAPK ( Figure 3A ) induced by Cry5B PFT indicating that the p38 MAPK activation is not directly regulated by KGB-1 . In addition to the fact that JNK MAPK is important for induction of most of the PFT-induced transcripts , including those regulated by p38 , we wanted to know if JNK MAPK is also important for induction of PFT-induced transcripts that are functionally important ( e . g . , knock-down to Hpo phenotype ) . We therefore tested whether or not KGB-1 regulates p38-controlled transcripts/pathway that play a role in PFT defenses , namely ttm-1 and ttm-2 and the UPR [11] , [14] . We find that the induction of both ttm-1 ( Figure 2C ) and ttm-2 ( Figure 3B ) in response to PFT is dependent upon the JNK MAPK pathway . Cry5B PFT-induced activation of the UPR PFT defense pathway is also dependent upon JNK MAPK , as ascertained by induction of the downstream transcriptional target hsp-4 transcript levels and by induction of xbp-1 splicing ( Figure 3B ) . As with p38 MAPK-dependent activation of the UPR by PFT [14] , JNK-dependent UPR activation occurs in the intestine in response to Cry5B PFT ( the tissue directly targeted by the PFT; Figure 3C , upper panels ) but not in response to an unrelated stressor like heat ( Figure 3C , lower panels ) . The UPR-activation data suggest that the JNK pathway functions cell autonomously in the intestine for PFT defenses . To confirm this , we performed intestine-specific RNAi of kgb-1 , and found that animals lacking KGB-1 in the intestine are hypersensitive to Cry5B PFT ( Figure 3D ) . Thus , the JNK MAPK pathway regulates all currently known p38-dependent PFT protection genes in the target tissue of the PFT . We next set out to determine , apart from known p38-dependent PFT defense genes , what other defense genes JNK MAPK might regulate . For this information , we turned back to our genome-wide RNAi list of hpo genes , examined all eight of the genes that are induced by PFT ( 2X cut-off ) , and asked , based on our microarray data , if their induction is dependent upon either the p38 or JNK MAPK pathways . We found four of eight ( 50% ) induced hpo genes ( namely kin-18 , Y54E5A . 1 , F42C5 . 10 and kgb-1 itself ) require JNK MAPK for their induction whereas none of the eight genes require the p38 MAPK pathway for their induction . The dependence of these four genes upon the KGB-1 and not the p38 MAPK pathway for their induction was confirmed via qRT-PCR ( Figure 3E ) . The same conclusion is reached if one examines genes required for PFT defenses induced at the 1 . 5-fold cutoff: 7/15 or ∼50% of these genes are dependent upon JNK; none are dependent upon SEK-1 . Although the p38 MAPK pathway has been studied in more detail , these data taken together indicate that the JNK MAPK pathway plays a more central role in coordinating induced PFT defenses . Since KGB-1 is a master regulator of ∼50% of Cry5B PFT-responsive genes , including those that are functionally relevant , we were interested to see how susceptible kgb-1 mutant animals are to Cry5B PFT attack . Using a dose-dependent lethality assay , we determined the LC50 ( lethal concentrations at which 50% of the animals die ) of the wild-type , sek-1 ( km4 ) p38 MAPKK-minus , and kgb-1 ( um3 ) JNK-like MAPK-minus animals treated with Cry5B PFT for 8 days . We find that both kgb-1 ( um3 ) and sek-1 ( km4 ) mutant animals have similar LC50 values and are both significantly hypersensitive to PFT relative to wild type animals ( Table 2 ) . These results were unexpected since , based on the fact that KGB-1 regulates more PFT-induced genes functionally required for PFT protection than the p38 pathway , we predicted that kgb-1 mutant animals would be more hypersensitive to PFT than sek-1 mutant animals . There are several explanations for this finding . First , it might suggest that KGB-1 , but not SEK-1 , transcriptionally regulates many genes required for induced PFT defenses and also some genes that inhibit PFT defenses . Such genes could be temporally segregated—e . g . , KGB-1 JNK-like MAPK could initially up-regulate transcriptional responses that are protective ( our microarray data is taken at 3 hr post-PFT treatment ) and later down-regulate responses that inhibit protection ( our LC50 data is taken at 8 days ) . Eventual down-regulation of protection is an important part of induced immune responses and serves to protect cells from the adverse effects of a prolonged or overwhelming immune response . Such positive and negative regulation of a cellular response by the JNK pathway is with precedent [40] , [41] . Second , it might also suggest that whereas KGB-1 JNK-like MAPK is a master regulator of transcriptionally-induced PFT defenses , the p38 pathway may play an important role in other ( e . g . , constitutive ) PFT defenses , hence resulting in a more severe sek-1 phenotype than predicted . Given its importance in the hierarchy of coordinating transcriptionally-induced defenses against a small-pore PFT and its activation in both C . elegans treated with small-pore PFT ( our data; see above ) and mammalian cells treated with large-pore PFT ( referenced above ) , we hypothesized that JNK MAPK might broadly coordinate defenses against PFTs—i . e . , it might be required for defense against both small- and large-pore PFTs . To date , no gene has been shown to play a protective role against both large- and small-pore PFTs . We therefore tested whether a large-pore PFT could intoxicate C . elegans using streptolysin O ( SLO ) . SLO binds to cholesterol in cell membranes , forms large 30 nm diameter pores ( versus 1–2 nm pores for Cry PFTs ) , and is an important virulence factor of human-pathogenic streptococci [42] . The repair of the pores formed by SLO in mammalian cells membranes is calcium dependent [25] , and a lytic-mutant N402 SLO has been shown to be defective in generating pores in membranes [43] . Whether SLO or any other cholesterol-dependent cytolysin affects C . elegans has not been previously reported . We find that wild-type C . elegans are intoxicated by wild-type SLO at 50 µg/mL with 58% of animals killed in calcium-free medium and 11% killed in calcium-containing medium in 6 days ( Figure 4A ) . The inhibitory effect of calcium in SLO-mediated killing of C . elegans is consistent with known effects of SLO on mammalian cells [25] . Furthermore , the killing of C . elegans by SLO is completely dependent upon pore formation since wild-type worms are not killed by the non-lytic SLO mutant N402 ( Figure 4B ) . Thus , the interaction of SLO with C . elegans parallels that of the toxin with mammalian cells and the intoxication of C . elegans by SLO is a valid model of large-pore PFT intoxication . We then tested whether or not JNK is required for defense against SLO in C . elegans . Wild-type animals and kgb-1 ( um3 ) mutant animals were exposed to SLO at a dose of 50 µg/mL . Significantly more kgb-1 ( um3 ) mutant animals are killed by SLO than wild-type animals ( Figure 4C ) , demonstrating that the JNK MAPK pathway is also required for defense against a large-pore PFT . C . elegans fos-1 and jun-1 were found in our genome-wide RNAi screen as genes important for defense against Cry5B PFT ( Table 1 ) . Together , these proteins make up the heterodimeric transcription factor known as activating protein 1 ( AP-1 ) . AP-1 is a well-defined regulator of innate immunity and stress responses in mammalian cells and a known downstream target of JNK MAKP pathway in these responses [44] . To date , neither jun-1 nor fos-1 has been demonstrated to be involved in protective responses in C . elegans nor in PFT protection in any system . Both C . elegans AP-1 homologs , fos-1 and jun-1 were found to be up-regulated by Cry5B PFT in our microarray study ( P<0 . 001 ) . To confirm their induction , we performed qRT-PCR analysis of C . elegans jun-1 and fos-1 expression upon treatment with PFT at 1 , 2 , 4 , and 8 hr . Both are robustly induced by Cry5B , with maximum induction at the earliest time point tested ( Figure 5A ) . jun-1 induction by Cry5B PFT is KGB-1-dependent and SEK-1 independent whereas fos-1 induction is independent of both ( Figure 5B ) . These inductions are consistent with our microarray data ( 1 . 5-fold cut-off ) and demonstrate that jun-1 is a downstream target of KGB-1 during PFT protective responses . To quantitate the protective function of AP-1 during small-pore PFT attack , we performed Cry5B PFT mortality assays using the available jun-1 mutant , jun-1 ( gk551 ) ( as opposed to jun-1 RNAi; the jun-1 ( gk551 ) allele is likely a null as it deletes 1 . 4 kb of DNA from the jun-1 locus , including the DNA binding domain ) . We found that jun-1 ( gk551 ) animals are significantly more sensitive than wild-type animals at all doses of Cry5B tested and overall are ∼10 fold more sensitive to PFT ( Figure 5C ) . This level of jun-1 ( gk551 ) hypersusceptibility to PFTs is greater than that of mutants in other non-MAPK hpo genes quantitated to date ( namely , xbp-1 , hif-1 , ttm-1 , ttm-2 , wwp-1 ) . We also exposed jun-1 ( gk551 ) mutant animals to heat stress , and find that , compared to wild-type animals , they are actually slightly resistant to heat stress ( Figure 5D; Table S2; Figure S3 ) . When tested against a second form of pathogenic attack , namely Pseudomonas aeruginosa PA14 , we find that jun-1 ( gk551 ) mutant animals are resistant relative to wild-type animals ( Figure 5E , Table S2 , Figure S4 ) . Thus , the sensitivity of jun-1 mutant animals to PFT is not due to generally compromised health and there is specificity in its role as protecting against PFTs . To determine whether C . elegans jun , like JNK MAPK , is protective against large-pore PFTs , we treated jun-1 ( gk551 ) animals with 50 µg/mL SLO and compared the percentage killed to wild-type animals exposed to the same dose . Animals lacking jun-1 are hypersensitive to SLO PFT in the absence or presence of calcium , with 82% and 43% killed respectively ( Figure 5F; for wild-type the numbers killed are 58% and 11% ) . jun-1 ( gk551 ) animals are not killed by SLO N402 , indicating that killing by SLO in jun-1 mutant animals is dependent upon pore formation ( Figure 5F ) . Thus , jun-1 is required for C . elegans protection against both large- and small-pore PFTs . Since KGB-1 , as the master regulator of PFT defenses , controls the induction of p38-dependent sub-pathways ( ttm-1 , ttm-2 , hsp-4 , and xbp-1 ) and p38-independent sub-pathways ( kin-18 , Y54E5A . 1 , and F42C5 . 10 ) , we further examined whether JUN-1 being the downstream of KGB-1 contributes to these gene regulation . The qRT-PCR analysis in wild-type versus jun-1 ( gk551 ) mutant animals indicated that jun-1 is required for the Cry5B-triggered ttm-2 and hsp-4 but not ttm-1 , kin-18 , Y54E5A . 1 , and F42C5 . 10 induction ( Figure 5G ) . Consistently , we found that increased splicing ( activation ) of xbp-1 in response to Cry5B does not occur in jun-1 ( gk551 ) mutant animals ( Figure 5G ) . These results indicate that KGB-1-regulated genes could be further parsed to be either JUN-1 dependent or independent . We then considered whether our finding that AP-1 is important for PFT responses in C . elegans extends to mammalian cells . Using commercially available antibodies , we found that 125 ng/mL of lytic , wild-type SLO induces activation ( phosphorylation ) of c-JUN in HaCaT cells and that the non-lytic SLO mutant does not ( Figure 6A ) . To test whether AP-1 is functionally important for survival of SLO-treated mammalian cells , we employed a transcription factor decoy approach [45] . HaCaT cells were simultaneously exposed to SLO and biotinylated double-stranded deoxyoligonucleotides comprising a consensus AP-1 binding site ( decoy oligonucleotide ) , which competes with AP-1 sites of genomic promoters for binding of the transcription factor . A mismatched oligonucleotide , which does not bind AP-1 , served as a control . The AP-1 decoy oligonucleotide specifically increased the proportion of pyknotic nuclei ( Figure 6B ) . The combination of activation and decoy oligonucleotide data supports the notion that AP-1 is protective against PFT attack in mammalian cells , as it is in C . elegans .
We present the first genome-wide study of functional cellular responses to a PFT . We identify 106 hpo genes important for cellular protection against an attack by a small-pore PFT , Cry5B . Interactome network analysis of these genes emphasizes the importance of two MAPK pathways , p38 and JNK , in coordinating this protection . The stringency of our experimental design led us to all previously known genes that mutate to a strong Hpo phenotype and none that mutate to a weak to moderate Hpo phenotype . All of the genes on this list except two are previously uncharacterized with regards to PFT defenses . A comparison of the genes induced by PFT with these 106 important for protection against PFT revealed a statistically significant enrichment of transcriptionally-induced , but not transcriptionally-repressed , genes on the list of PFT-protective genes . These data indicate that transcriptional induction is important for PFT protection . Our findings demonstrate that >0 . 5% of genes in the C . elegans genome play a major role in protection against pore-forming toxins . This number is large by any standard for a single cellular process . Given the likely parallels between cellular responses to pore-forming toxins and membrane ruptures that occur to cells during their normal existence , these data suggest that cells devote such a large portion of their genome to protecting against membrane holes and that this list of genes represents a rich starting point for understanding how cells cope with PFT attack and membrane damage . It is interesting to note that plasma membrane pore-formation has been suggested to play a significant role in neurological diseases of ageing , such as Alzheimer's or Parkinson's disease [10] . The significant overlap of the 106 hpo PFT protection genes with genes involved in promoting long life-span , suggest indeed that small ruptures of the plasma membrane play a significant role in the aging process . The fact that JNK and AP-1 are important for protection against both large-pore and small-pore PFTs indicates that at least some of the pathways cells use to deal with pores of various sizes are conserved , in contrast to previous suggestions [10] , [13] . Based on interactome analyses from our genome-wide RNAi screen , we examined the requirement of the JNK MAPK pathway for PFT defenses . JNK MAPK has not been studied in any detail for its role in PFT defenses . Indeed , to date the only signal transduction pathway studied in detail has been the p38 MAPK pathway . Our study of the JNK MAPK pathway identifies it , and not p38 , as the first master regulator of PFT transcriptionally-induced cellular protection . This conclusion is based on the following observations and results: 1 ) the JNK MAPK pathway is induced in mammalian and C . elegans cells by small-pore and large-pore PFTs ( this study and others [38] , [39]; 2 ) more than 50% of all genes transcriptionally induced by PFT depend upon the JNK pathway for their induction ( this study ) ; 3 ) 85% of the p38-dependent PFT-induced transcriptional response is controlled by JNK MAPK ( this study ) ; 4 ) all three of the known p38-dependent PFT-induced functional responses are dependent upon JNK ( this study ) ; 5 ) 50% of the hpo genes that mutate to a strong Hpo phenotype and that are induced by PFT are dependent upon JNK; and 6 ) JNK protects against both large and small-pore PFTs in C . elegans , the first such gene so identified . Our analyses also identify the first known downstream targets of the JNK MAPK-regulated PFT protection pathway , namely , kin-18 , jun-1 , Y54E5A . 1 , and F42C5 . 10 . While JNK MAPK is the master regulator of induced responses , our Western blotting data suggest that JNK is not hierarchically upstream of p38 MAPK activation in response to PFT . Based on our findings we propose a model whereby the JNK pathway converges in parallel on p38-regulated PFT responsive genes to regulate the transcriptionally-induced defense response to Cry5B ( Figure 6C ) . Our genome-wide analyses also led to the discovery of AP-1 , a key regulator of mammalian innate immunity downstream of JNK and the toll-like receptor ( TLR ) pathway , as important for defense against small- and large-pore PFTs in C . elegans and against a large-pore PFT in mammalian cells . This study is the first to identify AP-1 as involved in immunity in C . elegans . We also demonstrated that the UPR and ttm-2 Cry5B-induced defense sub-pathways , but not the ttm-1 , kin-18 , Y54E5A . 1 , and F42C5 . 10 Cry5B-induced defense sub-pathway , are transcriptionally regulated via JUN-1 . Thus , AP-1 mediates some , but not all , of JNK-regulated PFT defenses ( Figure 6C ) . As indicated by our data and reflected in our model , regulation of transcriptionally-induced cellular defenses against PFTs involves an intricately connected network of sub-pathways , with the JNK pathway hierarchically at the top of known regulators . AP-1 is the first transcription factor to be shown to be broadly required for metazoan cells to defend against PFTs . Since the TLR does not appear to be important for C . elegans immunity [46] , our data suggest that the role of the JNK-AP-1 pathway in protection against PFTs is one of its most ancestral functions , predating that of its role in TLR-mediated immunity . In summary , we report the first large-scale functional study of genes involved in protecting against a small-pore PFTs , place JNK as the master regulator of known cellular defenses against small- and large-pore PFTs , and identify AP-1 as an ancient factor conserved from worms to humans involved in PFT defenses . This study demonstrates the power of integrated genome-scale approaches in vivo to provide vital insights and significant breakthroughs into what has been to date a daunting scientific challenge , namely understanding defenses against breaches of plasma membrane integrity .
Strains were maintained at 20°C as described [47] , except strains containing glp-4 ( bn2 ) , which were maintained at 15°C . Images for the genome-wide RNAi screen were acquired using an Olympus SZ60 dissecting microscope . Strain VP303 rde-1 ( ne219 ) ; kbEx200 [rol-6 ( su1006 ) ;pnhx-2::RDE-1] was provided by Kevin Strange ( Vanderbilt University ) [48] . Phsp-4::GFP and Phsp-4::GFP;kgb-1 ( um3 ) ) animals were placed on 2% agarose pads containing 0 . 1% sodium azide and were imaged on an Olympus BX60 microscope with a 10x objective . Purified Cry5B was prepared as described [49] and suspended either in water for the genome-wide RNAi screen , or dissolved in 20 mM HEPES ( pH 8 . 0 ) for liquid mortality assays . E . coli-expressed Cry5B was prepared as described previously [50] , except the overnight culture was 5 times diluted before IPTG induction in OP50 expression system . For the microarray experiment , conditions were carried out as described [11] . Solubilized Cry5B protoxin was incubated at 30°C for two weeks , allowing co-purifying native Bt protease ( s ) to activate it into a 59 kDa fragment . The 59 kDa Cry5B was purified by gel filtration ( Superdex 75 16/60 ) . Ionic currents passing through activated Cry5B proteins ( 5-10 µg/mL ) reconstituted into planar lipid bilayers were recorded at various holding voltages applied to phospholipid membranes ( phosphatidylethanolamine∶phosphatidylcholine∶cholesterol ( 7∶2∶1 w/w ) ) painted across a 250-µm hole drilled in a Teflon wall separating two 600-µl chambers filled with 150 mM KCl , 10 mM Tris , pH 9 . 0 . Data were recorded with an Axopatch-1D amplifier ( Axon Instruments , Foster City , CA ) , filtered at 600Hz , digitized and analyzed using pClamp6 software ( Axon Instruments ) [51] . RNAi feeding for the genome-wide screen was performed in a 24-well format modified from a published protocol [52] . Bacteria containing each RNAi clone were cultured in 100 µL Luria Broth media containing 50 µg/mL carbenicillin in a 96-well format for 16–18 hrs . 30 µl of each culture was seeded in one well of a 24-well plate containing NGM agar , 1 mM IPTG and 50 µg/ml carbenicillin ( each plate was set up in duplicate ) . For each batch of RNAi clones tested , empty vector ( L4440 ) and tir-1 RNAi clones were included as negative ( no knock-down ) and positive ( Hpo ) controls respectively . After overnight incubation , ∼20 synchronized RNAi-sensitive rrf-3 ( pk1426 ) L1 worms were placed in each well and allowed to develop to L4 stage at 20°C . 30 µL of total 1 µg purified Cry5B resuspended in water was added to RNAi-fed worms in each well ( rrf-3 mutants have a normal response to Cry5B [11] ) . The plates were covered with porous Rayon films ( VWR , West Chester , PA ) to allow air exchange . Hypersensitivity to PFT relative to no knock-down ( empty vector ) controls was assessed after 48 hours by light phase microscopy . The initial Hpo hits were individually retested in duplicate under the same conditions ( second round ) . In addition , for each RNAi clone , a single well was set up and treated with water alone ( no toxin ) to assay for ill health in the absence of toxin . To be considered a hpo gene , we looked for lack of ill health and developmental problems in the no-toxin well and either >50% moderately Hpo animals in both wells or >33% severely Hpo animals in both wells . Two hundred and sixty nine such genes were found in total . For the quantitative validation test ( third round ) , bacteria from these RNAi hits were cultured overnight , induced with 1mM IPTG for 1 hour at 37°C , and resuspended in S-media . We then added ∼20 glp-4 ( bn2 ) ;rrf-3 ( pk1426 ) L1 animals to each well of a 48-well plate with RNAi bacteria ( triplicate wells for each clone ) . The plates were gently rocked at 25°C for ∼30 hours until the animals had developed to the L4 stage . Then we added 10 µg/mL Cry5B or 20 mM HEPES pH 8 . 0 control into each well and quantitated the percentage of worms that were killed after six days at 25°C . The worm viability was determined based on movement and worms failed to respond to several gentle touches with a platinum pick were scored as dead . The survival on each RNAi treatment was normalized to L4440 treatment in the same batch , to help account for differences in toxicity from batch to batch of the experiment . We sequenced the 106 hpo gene clones to confirm their identities . The quadruplicate repeat triple-well verification was performed with the similar procedures except that no plate rocking was performed and data were analyzed via analysis of variance ( ANOVA ) and Dunnett's post-test . Statistical data analyses here and elsewhere were performed using Prism 5 . 0 ( GraphPad , San Diego , CA ) . For intestinal-specific liquid RNAi , bacteria for kgb-1 RNAi or L4440 empty vector were first cultured overnight , induced with 1mM IPTG for 1 hour at 37°C , and resuspended in S-media . Approximately 20 synchronized L1 stage VP303 worms were fed with kgb-1 or L4440 RNAi in 48 wells plates , grown to the L4 stage in ∼40 hours at 25°C and FUdR was then added to a final concentration of 200 µM . Then Cry5B to a final concentration of 10 µg/mL or 20 mM HEPES pH 8 . 0 control were added into each well and the percentage of worms that were killed after six days at 25°C was scored . The comparison of survivals between kgb-1 RNAi and empty control was done with one-tailed Student's t-test . The RNA samples were collected from three independent experiments as described [11] . glp-4 ( bn2 ) animals lack a functional germline and have otherwise normal response to Cry5B [11] . Use of these animals removes a major tissue from the animals , allowing for intestinal mRNAs to represent a larger portion of the total RNA population . Raw microarray datasets were normalized using Bioconductor's [53] Robust Multi-array Analysis ( RMA ) [54] , [55] in R language . Linear Models for Microarray Data ( LIMMA ) [56] was used to determine a set differentially expressed genes . The cutoff p-value used was 0 . 01 with minimum 1 . 5 or 2 fold change . The data were clustered to reveal prominent groups of transcripts with similar changes in expression pattern using a κ-means ( κ = 4 ) algorithm . The gene enrichment analysis for microarray was done with DAVID PANTHER annotation tool [57] . Only enriched categories with three or more counts and P<0 . 05 were kept . The statistical significance of the observed overlap between different gene lists ( e . g . , genome-wide RNAi screens , hpo vs . induced genes ) was calculated by a modified Fisher's exact test [58] . The PFT immunity interaction network was generated by combining data from our genome-wide RNAi screen and a modified C . elegans integrated functional network ( WI8∪Literature∪Interolog∪Genetic ) database with several manually curated interactions [34] . Interactions were only maintained if at least one of the nodes was hpo . The resulting networks were visualized using Cytoscape 2 . 6 . 3 [59] . Total RNA was isolated from approximately 5 , 000 L4 worms and cDNA was made from 2 µg of total RNA with MoMLV-reverse transcriptase ( Promega , Madison , WI ) and oligo-dT primers for 90 minutes at 42°C in 20 µl reaction mixtures . qRT-PCR ( triplicate independent experiments , normalization to eft-2 ) was then carried out as described [14] . Primers for qRT-PCR are listed in Table S3 except for eft-2 , ttm-2 , hsp-4 , and xbp-1 spliced form , which have been described before [14] , [15] . Transcript levels among mutants and wild-type animals for a given gene were compared using one-way ANOVA with Tukey's HSD post hoc test or one-tailed Student's t-test ( Figure 3A and 5G only ) . Approximately 750 L1 stage worms were grown in a single well of a 48 well plate , treated with indicated concentrations of Cry5B and prepared as described before for phospho p38 MAPK and α-tubulin [14] . The same PVDF membranes were stripped with Restore Western Blot Stripping Buffer ( Thermo Fisher Scientific Inc . , Waltham , MA ) and reprobed as needed . The probing order is phospho-KGB-1 ( 1∶1000 ) first , phospho-p38 MAPK and α-tubulin second , and total KGB-1 antiserum ( 1∶5000 ) third . Cry5B mortality assays with wild-type animals and kgb-1 and sek-1 mutants were carried out for 8 days at 20°C in three independent experiments as per standard protocol[50] . LC50 values and confidence intervals were calculated as described [60] . For SLO and N402 assays , synchronized L4 worms were prepared in complete or calcium-free S-medium with 50 µg/mL of toxin dissolved in phosphate buffered saline ( PBS; PBS alone was used in negative controls ) . The survival of animals was scored after six days at 25°C and differences analyzed by analyzed with one-tailed Student's t-test . The P . aeruginosa PA14 survival assay was performed on slow-killing plates as described [14] . For the heat stress analysis , synchronized N2 and jun-1 ( gk551 ) animals were shifted to 35°C as two-day-old adults and survival was scored every two hours after the treatment as described [61] . Three lifespan assays per worm strain were done , and all show the same trends . The data presented in the paper are from a representative experiment . Data are plotted as a Kaplan–Meier survival plot and analyzed for significance using the Log Rank test . Recombinant streptolysin O and non-lytic mutant N402 were prepared as described previously [43] , [62] . Growth of the human keratinocyte cell line HaCaT has been previously described [63] . Cells were seeded at a density of 0 . 5x106 cells in 12-well plates , cultured overnight and treated with 125 ng/ml SLO or non-lytic mutant N402 for indicated times . After washing with PBS , cells were lysed by addition of 50 µl 2xSDS-loading buffer . Proteins were separated by SDS-PAGE and electroblotted onto nitrocellulose membrane , incubated with rabbit anti-phospho Ser63-c-jun ( 1∶1000; Cell Signaling Technology ) over night at 4°C , washed , and incubated with goat anti-rabbit IgG HRP-conjugate ( 1∶2000; New England Biolabs ) . After three washes , bound antibody was detected by ECL . The same membrane was stripped with TBS containing 2% SDS and 100 mM β-mercaptoethanol for 50 min at 50°C and reprobed with rabbit anti-c-jun ( 1∶1000; Cell Signaling Technology ) . For the AP-1 decoy experiment , HaCaT cells were treated in Ca2+-free medium with 500 ng/mL SLO and 3 µM biotinylated AP-1 decoy oligodeoxynucleotide ( ODN ) , or mismatched AP-1 decoy ODN for 30 min , ( Genedetect , Auckland , New Zealand ) . Subsequently , cells were washed twice with PBS and incubated in full culture medium . After incubation for 4 hours , cells were fixed with 4% paraformaldehyde , permeabilized with 0 . 1% triton-X100 . Biotinylated ODN were stained with AlexaFluor594-conjugated streptavidin; nuclei were visualized with DAPI stain ( 10 µg/mL ) . After washing , cells were examined by fluorescence microscopy using an Axiovert 200M microscope ( Zeiss , Germany ) . The relative intensity of DAPI staining was quantified for ca . 800 cells per treatment and experiment , applying the analysis tool of Photoshop to digital images in RGB-mode . Nuclei were considered pyknotic if they were condensed to a diameter<50% of average and showed more intense DAPI uptake ( >2-fold of average intensity ) [64] .
|
The plasma membrane surrounds cells and protects their interior from the environment and from attack by disease-causing agents like bacteria and viruses . Bacteria that cause disease have discovered that an effective way to attack cells is to secrete proteins ( pore-forming toxins ) that breach , i . e . , form holes in , the plasma membrane . How cells deal with and survive this kind of attack is poorly understood . Here , we report on the first large-scale study of the genes and mRNA transcripts that respond to pore-forming toxin attack in cells . We find that a remarkable portion , >0 . 5% , of the cell's genome protects it against pore-forming toxins . These data led us to look more closely at mitogen-activated protein kinase pathways as regulators of pore-forming toxin defenses . We find that half of the PFT-induced protective response is controlled by a single , conserved signaling pathway in cells , which controls a complex array of downstream targets and which protects against both large pore and small pore toxins . Our results indicate that defense against pore-forming toxins is a very ancient defense that utilizes a much more complex and extensive response in cells than previously demonstrated .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"infectious",
"diseases",
"genetics",
"and",
"genomics/functional",
"genomics",
"immunology/innate",
"immunity",
"immunology",
"infectious",
"diseases/bacterial",
"infections",
"genetics",
"and",
"genomics"
] |
2011
|
Global Functional Analyses of Cellular Responses to Pore-Forming Toxins
|
The acquisition and analysis of datasets including multi-level omics and physiology from non-model species , sampled from field populations , is a formidable challenge , which so far has prevented the application of systems biology approaches . If successful , these could contribute enormously to improving our understanding of how populations of living organisms adapt to environmental stressors relating to , for example , pollution and climate . Here we describe the first application of a network inference approach integrating transcriptional , metabolic and phenotypic information representative of wild populations of the European flounder fish , sampled at seven estuarine locations in northern Europe with different degrees and profiles of chemical contaminants . We identified network modules , whose activity was predictive of environmental exposure and represented a link between molecular and morphometric indices . These sub-networks represented both known and candidate novel adverse outcome pathways representative of several aspects of human liver pathophysiology such as liver hyperplasia , fibrosis , and hepatocellular carcinoma . At the molecular level these pathways were linked to TNF alpha , TGF beta , PDGF , AGT and VEGF signalling . More generally , this pioneering study has important implications as it can be applied to model molecular mechanisms of compensatory adaptation to a wide range of scenarios in wild populations .
Modelling the responses and compensatory adaptations of living organisms to a changing environment is extremely important both in terms of scientific understanding and for its potential impact on global health . Although computational modelling of ecological systems has been utilised in ecotoxicology , the application of systems biology approaches to non-model organisms in general presents formidable difficulties , partly due to limited sequence information for environmentally relevant sentinel species . Moreover , the number of samples and the depth of information available are often limited and there may be a lack of truly relevant physiological endpoints . Thus , omics have proven effective in finding responses of aquatic organisms to model toxicants in laboratory-based experiments [1] but the environment poses a greater challenge as anthropogenic contaminants are present as complex mixtures and responses will additionally be dependent upon natural life history traits and other environmental factors . Relatively few omics studies have focussed upon the ecotoxicology of environmentally sampled fish [2]–[7] . Although we have previously shown [8] , [9] that expression of stress response genes could be used to distinguish fish from environmental sampling sites with different underlying contaminant burdens , this gave little insight to the health outcomes of these molecular differences . In this context , identifying molecular mechanisms of compensatory and toxic responses from observational data ( reverse engineering ) , an approach that has been so successful in clinical studies and in laboratory model organisms , is highly challenging in field studies . We addressed this challenge by developing a novel network inference strategy based on the integration of multi-level measurements of populations of fish exposed to a diverse spectrum of environmental pollutants . This provides a useful model for a network biology approach generally applicable to non-model species and represents a breakthrough in the way we study the mechanisms whereby organisms respond to chemical exposure in the environment . We directed our efforts towards modelling molecular networks representative of populations of the flatfish European flounder ( Platichthys flesus ) sampled from marine environments of North Western Europe , including locations significantly impacted by anthropogenic chemical contaminants . The study integrated measurements representing a broad spectrum of samples characterized using transcriptomics , metabolomics , conventional biomarkers and analysis of chemicals in sediments from the sampling sites . Previous studies have shown both anthropogenic contamination and higher prevalence of pre-neoplastic and neoplastic lesions in flounder from the Elbe estuary [10] and from the Mersey and Tyne [11] , together with elevated levels of hepatic DNA adducts at these sites [12] . Data integration was achieved by implementing a systems biology framework for network reconstruction , starting from cross-species mapping of sequence information to the integration of multi-level datasets within a framework for network inference [13] and culminating in the identification of network modules predictive of physiological responses to chemical exposure , valuable for marine monitoring [14] . The networks we identified demonstrate a remarkable parallel between human liver carcinogenesis and environmental effects on fish liver as well as revealing potentially novel adaptation mechanisms . The broader application of network biology approaches to other non-model species sampled from the environment is therefore likely to profoundly change our understanding of how living systems are likely to adapt to complex environments .
An important assumption in many eco-toxicology studies is that the molecular states of organisms reflect their biological responses to complex chemical mixtures present within that environment . Indirect evidence suggests that this hypothesis may be correct . For example , consistent with previous studies [9] , we have identified genes and metabolites that were differentially expressed between environmental sites ( the results obtained are shown in detail in Table 1 and Text S1 , Tables S1 and S2 ) . Many of these were either known to be associated with stress responses or were previously shown to respond to anthropogenic chemical contaminants in fish . Although these results were encouraging they did not provide a direct link between molecular status and response to specific chemicals . Since sediment chemistry data was available , we assessed whether chemical contaminant profiles could be inferred from gene expression data and whether these would at least partially match the known sediment composition . Our analysis was performed by linking genes differentially expressed between each sampling site and the reference site , with chemical-gene relationships within the Comparative Toxicology Database ( CTD ) [15] . The Alde estuary was chosen as the reference site due to its low concentrations of major anthropogenic chemical contaminants ( Table 1 ) , both in sediment and in flounder livers . These significant associations may be regarded as predictive of the most important classes of chemicals exerting their biological effects upon flounder gene expression amongst the highly complex chemical mixtures within the sediments at these sites . Results were consistent with the initial hypothesis , as where contaminants were highlighted both by chemistry and CTD analysis; sediment concentrations all exceeded the lower OSPAR ecotoxicological assessment criteria , except for PAHs at the Morecambe site . At Brunsbuttel , elevated chromium and polychlorinated biphenyls ( PCBs ) ; at Cuxhaven , chromium , nickel , lead , zinc , polycyclic aromatic hydrocarbons ( PAHs ) and PCBs; at Helgoland nickel , zinc , manganese and PCBs , at Mersey PCBs; at Morecambe Bay arsenic , nickel and PAHs; and at Tyne arsenic and PCBs were all predicted by the CTD analyses and confirmed by chemistry data ( Table 2 and Table S3 ) . PAHs were predicted at Morecambe and Cuxhaven , with the AhR-inducer beta-naphthoflavone predicted at Brunsbuttel , Helgoland and Tyne , consistent with our finding of CYP1A transcriptional induction at all sites in comparison with the Alde . Additionally , Ingenuity Pathway Analysis ( IPA ) of all genes identified as significantly differentially expressed between sites showed significant associations with a number of toxicologically important processes and outcomes ( Table 3 ) . As there were clear relationships between geographical location , chemical exposure and molecular profiles of flounder livers , we proceeded to reconstruct a network model representing the relationships between transcriptomic and metabolomic data , morphological measurements , protein biomarkers and microsatellite markers ( Figure 1 ) . This network was constructed from all data , not limited to molecules that differed between sites . Inspection of the resulting network ( Figure 2 ) showed that transcriptional and metabolic networks separated into two different areas of the network layout . Interestingly , modules whose hubs were fish morphometric measurements occurred exactly at the interface between these two areas and these modules contained metabolite ( 46% ) and transcript ( 50% ) measurements as well as fish morphometric measurements ( 4% ) . Different areas of the inferred network ( Figure 2 ) were characterised with different functional profiles . The modules close to the interface with metabolism ( A ) showed enrichment ( FDR<0 . 05 ) for the annotation terms mitochondrion ( GO:0005739 ) , oxidoreductase activity ( GO:0016491 ) , endoplasmic reticulum ( GO:0005783 ) , protein folding ( GO:0006457 ) and antioxidant activity ( GO:0016209 ) , and the two KEGG pathways hsa03050:proteasome and hsa00480:glutathione metabolism . The second sub-network ( B ) was enriched for immune response ( GO:0006955 ) and response to stress ( GO:0006950 ) . The third sub-network ( C ) was enriched for proteolysis ( GO:0006508 ) and digestion ( GO:0007586 ) . Each individual network module was tested for its ability to predict geographic sampling sites ( Figure 3 ) , the presence of parasites ( Figure 4A ) and the presence of any of the liver histo-pathological abnormalities shown in Table 1H ( Figure S2 ) . Modules that were predictive of environmental sampling site were concentrated in two sub-networks . The larger ( Figure 3 , area A ) was centred on the interface between metabolic and transcriptional networks and consequently included 14 modules consisting of morphometric indices as well as metabolites and transcripts ( 1% morphometric indices , 36% metabolite bins , 63% transcripts ) . Modules that were predictive of parasitic copepod infection by Acanthochondria sp . and Lepeophtheirus sp . were similarly distributed in the network , but with additional modules localized in the sub-network B that were enriched in annotation to the immune response . Modules that were predictive of infection by Anisakid nematodes ( Figure 4A ) displayed a different profile , being more concentrated within group B . Hierarchical clustering of the profile of modules that were predictive of parasite infections showed that responses to copepod infections by Lepeophtheirus and Acanthochondria clustered together and were distinguished from responses to infection by the Anisakid nematodes . We have previously shown that there is a strong link between laboratory exposure to individual chemicals and flounder hepatic gene expression [8] . It was therefore reasonable to hypothesize that genes differentially expressed in laboratory exposures may map onto modules predictive of sampling location . Fisher's Exact Test was therefore used to identify modules where genes differentially regulated as a result of single chemical laboratory exposures were over-represented ( these were determined by ANOVA FDR<0 . 05 over 16 day time-courses post-intraperitoneal injection ) . Responses to lindane are shown as an example in Figure 4B . This highlighted the temporal change in responses to toxicants , with the majority of overlapping modules occurring in both sub-networks A and B at early timepoints , followed by a shift towards sub-networks B and C at later timepoints . We have previously shown [8] that this temporal change is associated with an early induction of transcripts for chaperones , phase I and II metabolic enzymes , oxidative stress and protein synthesis that diminishes by the later timepoints and is replaced by induction of protein degradation , immune-function and inflammation-related transcripts . The results for all treatments are illustrated graphically in Figure S2 E to L . All treatments showed overlap with modules in group A , at the metabolite/transcript interface , and this was clearest for cadmium , that only affected this area , apart from one module in group C . All other treatments showed overlap between responsive genes and group B modules to varying extents and all except estradiol and cadmium overlapped with at least two modules within group C . These results are supported by our previous study [9] in which we found that employing transcripts altered during laboratory exposures to a range of individual toxicants improved predictivity of environmental sampling sites . Having defined network modules predictive of geographical location , Ingenuity Pathway Analysis was used to elucidate the detailed structure of molecular pathways and their potential association with specific signatures of liver pathology . We performed these analyses under the hypothesis that the underlying response to chemical exposure would be consistent with what is known of human liver molecular pathophysiology . It was therefore expected that significant associations between the modules defined by our analysis and networks stored in the Ingenuity database would be informative of the underlying molecular mechanisms . We indeed observed a remarkable overlap between modules predictive of geographical location and modules containing genes whose transcriptional profile has been previously associated with liver fibrosis , cirrhosis and hepatocellular carcinoma in mammals . Modules whose component genes related to hepatotoxicity are shown in Figure 5 . The major group of site-predictive modules shows significant overlap with modules relating to liver cholestasis and hepatocellular carcinoma , whereas the secondary group overlaps with liver fibrosis . The annotation gained from Ingenuity , with key regulators inferred from networks based on interaction information , was combined and clustered in the TMEV software package using 5 different algorithms . These show ( Table 4 ) that genes and metabolites a ) involved in bile acid synthesis , transport and amino acid metabolism b ) predictive of parasite infection c ) linked to hepatocellular carcinoma , reproductive disorders and liver cirrhosis d ) responding to oxidative stressors tert-butylhydroperoxide ( tBHP ) and cadmium , the hormone estradiol and rodent peroxisome proliferator perfluoro-octanoic acid ( PFOA ) are closely linked to differences between environmental sites . Additional relationships with inflammation , immune response , energy , fatty acids and nucleic acid metabolism , response to other toxicants and regulation by insulin , huntingtin , MYC and hepatocyte nuclear factor HNF4A were also highlighted . Functional analysis of the modules that were both site-predictive and associated with hepatocellular carcinoma showed significant overlap with mitochondrion , proteasome , tricarboxylic acid cycle , melanosome , protein dimerization activity , membrane-enclosed lumen , glutathione metabolism , coenzyme binding , microsome , translation , protein transport and carbohydrate catabolism ( enrichment score >2 , FDR<0 . 05 ) . The models we have developed are a high level representation of the molecular network's underlying response to environmental exposure . In order to generate specific hypotheses on the molecular pathways modulated during compensatory adaptation and toxicity further in-depth analyses of the specific interactions between genes and metabolites were performed . In this context , we combined the genes and metabolites represented in each group of predictive modules ( Groups A and B in Figure 3 ) and input these to IPA software . The most statistically significant networks derived from each group are shown in Figure 6 and Figure S3 , coloured by expression represented as a ratio between a highly polluted site and the reference site ( Brunsbuttel versus Alde ) . The component genes and metabolites were clustered and the resulting expression profiles are shown in Figure S1 . The Ingenuity networks are further described in Text S1 and are discussed below .
The chemical exposures predicted from CTD interactions were partly confirmed by chemical data ( Table 2 ) despite the complexity of the environment , potentially including mixture effects , bioaccumulation and non-chemical stressors . Additional stressors were indicated that had not been chemically measured . Taking the Brunsbuttel site as an example , ethinyl-estradiol was a predicted contaminant and serum VTG protein , a canonical marker of endocrine disruption , was induced relative to the Alde ( Table 1 ) . Perfluorooctane sulfonic acid [16] and other persistent organic pollutants including PCBs , dieldrin and endosulfan [17] have been detected at elevated concentrations in the Elbe estuary and floodplain , and were all identified by our approach . Additional chemicals highlighted included systhane and vinclozolin fungicides , the halogenated aromatic hydrocarbon pesticide lindane , chlorine and tetradecanoylphorboyl acetate ( TPA ) . It is uncertain whether these compounds are in fact present at this site as , for example , the presence of TPA appears unlikely . However TPA is a well-known tumour promoter [18] , so detection of its associated gene expression changes might be viewed as a biomarker of effect , not necessarily of a specific exposure . At a number of sites flavonoids and flavonols , such as epicatechin gallate , were predicted , potentially indicating plant-derived exposures not of anthropogenic origin . At Morecambe Bay and the Tyne the prediction of paraquat perhaps reflected an oxidative stress response rather than the presence of this particular compound . These results support the use of a knowledge-based approach to infer chemical exposure profiles from molecular responses and validate the underlying assumptions in the study . Predictions from interrogation of the CTD database ( Table 2; Table S3 ) differed between sites suggesting that the approach can be sufficiently sensitive to specific differences in the exposure profiles . However , we do not propose that these associations necessarily indicate the presence of each specific contaminant at each site , for example ‘tobacco smoke pollution’ in the Mersey , we instead hypothesise that these represent the effects of related stressors , for example , AhR inducers at the Mersey site . The development of a modular network , representing the integration between molecular and physiological readouts , provided us with an interpretive framework to analyse the complex molecular signatures linked to exposure . One of the most interesting findings is that the modules that predict environmental exposure with greatest accuracy represent the interface between metabolite and transcriptional networks and link to higher level indicators of fish health , such as condition factor and hepatosomatic index ( Figure 2 ) . Consistent with this observation , network modules at the interface between metabolite and transcriptional networks were also differentially regulated in response to single chemical laboratory exposures . It should be borne in mind that the environmentally sampled fish have been chronically exposed to pollutants , and that chronic exposure can result in different responses than acute exposure [19] , [20] . In addition bioavailability , mixture effects , metabolism and bioaccumulation affect compound-specific responses within the livers of these fish . This is illustrated by the modules containing genes that responded to 16-day treatments of flounder with individual toxicants ( Figure 4 B , Figure S2 ) . While all toxicants induced changes in the metabolic-interface genes , they also affected the secondary area of the network that related more to acute stress and immune response ( Figure 2 , area B ) , in contrast to the differences between environmental sites , where only one module ( 40 ) in this area was affected . The characterisation of transcripts and metabolites that differed between sites was undertaken to provide insights into the molecular mechanisms that they describe , and to inform on the potential health outcomes for the fish . Canonical pathways that contributed to these differences included those relevant to metabolism of toxicants; AhR signalling , metabolism of xenobiotics by cytochromes P450 , the NRF2-mediated oxidative stress response , glutathione metabolism and bile acid bioysnthesis ( Table 3 ) . Together these describe phase I and II metabolism of xenobiotics , such as aromatic hydrocarbons , and their excretion via the bile . Additional endobiotic metabolic pathways were affected . Changes in glycolysis , pyruvate metabolism , the citric acid cycle and oxidative phosphorylation implied disturbances to the energy pathways of the liver that could reflect the energetic requirements of xenobiotic metabolism and lead to further metabolic disruption . Changes in amino acid synthesis and proteasomal protein degradation also indicated reorganisation of metabolism . This change in metabolic state and gene expression could be viewed as a successful compensatory response to toxicants and thus of little concern for the health of individual fish and these fish populations . Further examination of the annotation of transcripts and metabolites differing between sites implied that this hypothesis was false . As illustrated in Figure 5 , and shown in Tables 3 and 4 , there is a remarkable overlap between site-predictive modules and modules associated with hepatocellular carcinoma ( HCC ) . Additionally , liver cholestasis -annotated modules overlapped with HCC and site predictive modules and this area of the network was highly associated with bile acid biosynthesis . Apart from this metabolic interface group only one other module ( module 40 ) was predictive of site . This was also associated with hepatocellular carcinoma , and additionally with liver fibrosis , indicative of chronic liver damage , and occurred in an area of the network associated with inflammation . Therefore flounders inhabiting differentially contaminated sites show transcript and metabolite changes that have been associated with liver carcinogenesis in mammals . A question remains as to whether this simply represents the detection of HCC in the liver samples , as histopathology data were unavailable for the fish sampled off Germany . By comparison with studies of tumours from the closely related flatfish dab ( Limanda limanda ) this does not appear to be the case . In dab tumours the metabolites choline , phospocholine and glycine were reduced in concentration and lactate increased , an indication of the switch to anaerobic metabolism in the bulk tumours [21] . In , for example , the Brunsbuttel samples compared with non-tumour bearing Alde fish , choline , phosphocholine and glycine increased , and lactate decreased . Additionally , transcripts for ribosomal proteins showed co-ordinated induction in bulk tumours from dab , indicative of proliferation [22] , but no such induction was apparent from the present samples . The changes in gene expression and metabolites detected in this study do not recapitulate those found in bulk tumours , and may be viewed as indicating either an earlier stage of tumourigenesis or a permissive micro-environment in which hyperplastic tissue may form and lead to tumour formation . Ingenuity networks , based on mammalian interaction data , permitted more detailed biological characterisation of the site-associated modules . Complete pathways were not recapitulated by these analyses , as only a minority of the transcripts and metabolites from flounder liver were examined . Nevertheless , the analyses highlighted important processes and inferred key regulators . Here the most significant network derived from site-predictive modules is discussed in detail and additional networks are discussed in terms of their key inferred regulators . The most striking finding from the Ingenuity analyses was the co-ordinated repression of proteasomal subunit genes at the Brunsbuttel site ( Figure 6A; Figure S3A1 ) . This was not so marked at other sites ( Figure S1 ) , indeed at Morecambe Bay these genes were induced in comparison with the Alde fish . Proteasome maturation protein ( POMP ) has been found to be a critical regulator of proteasomal activity [23] and has been shown to be repressed by the halogenated aromatic hydrocarbon 2 , 3 , 7 , 8-tetrachlorodibenzodioxin ( TCDD ) in an AhR-independent manner [24] . Although TCDD concentration was not measured , the mean expression of proteasomal genes was inversely correlated ( r = −0 . 79 ) with fish liver PCB concentrations but did not correlate well with sediment PAH or PCB concentrations . Tyne fish , for example , displayed relatively high proteasomal gene expression and had low liver PCB but high PAH concentrations ( Table 1 , Figure S1 ) . Therefore the repression of proteasomal genes may represent a halogenated aromatic hydrocarbon-related response ( Figure 7 ) . In trout Oncorhynchus mykiss , a proteasome inhibitor reduced PAH-dependent CYP1A induction [25] , in contrast to mammalian studies [26] . This difference may contribute to the lower inducibility of CYP1A in flounder in comparison with many mammals . Ingenuity analysis also predicted an interaction between the proteasome and NF kappa B , a key regulator of mammalian hepatocarcinogensis [27] . The proteasome represses NK kappa B activation , and potentially disruption of proteasomal activity could have extensive additional effects on intracellular protein levels due to its role in the degradation of numerous proteins . We found no significant changes in NF kappa B gene expression between sites , and the consequences of putative activation at the Brunsbuttel site , and repression at the Morecambe site , due to changes in the proteasome , are difficult to predict , as in the early stages of carcinogenesis NF kappa B can have a protective effect , whereas in later stages it can promote tumourigenesis [27] . From the Ingenuity networks a number of key regulatory molecules were inferred . These included insulin ( Figure 6B , Figure S3A2 , S3A4 , S3B1 ) , estrone , luteinizing hormone ( LH ) and follicle stimulating hormone ( FSH ) ( Figure S3A3 and S3A6 ) , platelet derived growth factor beta ( PDGFBB ) ( Figure 6B , Figure S3A2 , S3B1 ) , transforming growth factor beta ( TGF-beta ) ( Figure S3A8 ) , vascular endothelial growth factor ( VEGF ) ( Figure 6B , Figure S3A7 , S3B1 ) , tumour necrosis factor ( TNF ) ( Figure S3A5 ) , and angiotensinogen ( Figure 6B , Figure S3B1 ) . Insulin , in fish as in mammals , is a key hormonal regulator of energy , glucose and lipid metabolism , all pathways that were identified as affected by sampling site . By the Ingenuity networks it was linked to protein kinases , metabolites ( including glucose and lactate ) and the glucose transporter SLC2A4 . The most obvious explanation for changes in insulin and related parameters would be differences in diet between fish from different sites . Amino-acid levels are more important regulators of insulin in carnivorous fish such as the flounder than sugars [28] . Dietary parameters would be expected to be highly variable depending upon recent feeding history of the fish , which was unknown for these individuals . However , insulin can also be modulated by exposure to toxicants including organophosphates [29] that was suggested to lead to an increase in lipogenesis , in agreement with our observations of phospholipidosis in fish from polluted sites ( Table 1 ) . Mild estrogenic endocrine disruption was suggested by VTG induction in Brunsbuttel fish , and networks shown in Figures S3A6 and S3A3 inferred that estrogen receptor alpha ( ESR1 ) , FSH and LH target genes were modulators of the different responses between sampling sites . ESR1 and HNF alpha were linked in Figure S3A6 and both are involved in hepatic cholestasis , indeed EE2-induced hepatotoxicity has been linked to alterations in bile acid biosynthesis in mice [30] . PDGFBB is the dimeric form of platelet derived growth factor beta ( PDGF-B ) . Notably , PDGF-B over-expressing mice spontaneously developed liver fibrosis [31] , and PDGF-BB was inferred as part of the network deriving from the liver fibrosis-annotated module 40 in our analysis . Additionally PDGF-B over-expressing mice developed hepatocellular carcinoma in response to phenobarbital and diethylnitrosamine treatment and induced TGF-beta and VEGF expression . TGF-beta was inferred to be an important regulator in site-specific responses ( Figure S3A8 ) and is a well-known mediator of cancer initiation , progression and metastasis , via interaction with the inflammatory response [32] . Furthermore , the pro-inflammatory cytokine TNF-alpha , an initiating signal for the innate immune response in fish as well as mammals [33] , was also identified by Ingenuity analysis ( Figure S3A5 ) . Release of TNF alpha from Kupffer cells leads to hepatocyte cell death , regeneration and fibrosis that can lead to hepatocellular carcinoma [34] . VEGF , best known as a stimulator of angiogenesis , was also highlighted in both the fibrosis-related and carcinoma-related sections of the network , and was linked with cell cycle , oncogenes and tumour suppressor genes ( CDKN1A , TP53 , MYC ) . Angiogenesis is a key requirement for the transition from fibrosis to hepatocellular carcinoma [35] . Angiotensinogen ( AGT ) is the precursor of angiotensin and was found to be repressed at all sites in comparison to the Alde reference site ( Table S1 ) . Angiotensin is a signal for vasoconstriction in mammals and in fish its expression is related to osmoregulation [7] with repression in liver in response to higher salinity . As the sampling sites differed in salinity , alteration of AGT transcription was not a surprising finding . As shown in Figure 6B , AGT was a member of the fibrosis-related module 40 and was predicted to form part of a complex network with VEGF , PDGF and intracellular kinases . Angiotensin has indeed been linked to stimulation of inflammatory liver fibrosis [36] , via fibroblast proliferation and production of inflammatory cytokines and growth factors , including TGF-beta . Inhibition of the angiotensin system by antagonism of its receptor [37] or inhibition of angiotensin-converting enzyme [38] has been shown to reduce hepatic fibrosis . VEGF , TGF beta , TNF alpha , PDGF and AGT are all intimately related to the progression of fibrosis to cirrhosis and hepatocellular carcinoma in mammals . These molecules were all highlighted as important regulators of the differences between molecular profiles of flounder livers from different sampling sites using an unbiased approach combining network inference and predictive algorithms . A combination of omics , multiple biomarkers and bioinformatics were used to identify and characterise hepatic molecular changes between fish sampled from several environmental sites . Based on these data , parasite infection , fish morphology and genetics do contribute to the differences between sites , but do not explain the majority of changes seen . For example , within-site tests showed that morphometric parameters and parasite infections could be significantly associated only with a small proportion ( <3% ) of the gene expression differences between sites ( Table S1 , Table S4 ) . Taken as a whole with our previous studies [8] , [9] , we find that anthropogenic chemical contamination of the marine environment is a major factor in explaining the molecular differences between fish sampled from these sites . The different methodologies employed displayed different strengths and weaknesses . Histopathology was a good guide to broad levels of pollution effect , but provided little information upon the nature of the contaminant profile . Protein biomarkers and enzyme activities were useful for categorising sites by major classes of toxicant , but gave little information on the potential health outcomes . 1H NMR metabolomics showed low technical variability , and metabolite profiles alone were more predictive of sampling site than gene expression profiles alone , however the annotation of metabolites is not yet well advanced , limiting the functional information currently available . Transcriptomics exhibited higher variability than metabolomics , but was more informative due to better annotation . Overall the methodologies were highly complementary , allowing analyses that would be impossible if one were limited to a single technique . The gene expression signatures associated with fish from each sampling site were used to predict the presence of chemical contaminants using the CTD gene expression-chemical interaction database . Mixture effects , other environmental influences and the similarity of certain stressors , such as the metals , might be expected to confound this approach . Additionally the incomplete nature of the flounder microarray and the CTD database and the limited numbers of samples for certain sites , which is a common issue in field studies , reduce the potential of this analysis . Therefore we did not expect to predict all environmental contaminants by this method . While this approach was useful with the current dataset , it may be expected to improve in future as both the CTD database and transcriptomic data become more comprehensive . Data integration and network analyses were essential; both to predicting health outcomes and to identifying and examining affected biological pathways . They allowed visualisation of the highly complex dataset and facilitated comparison of the effects of different stimuli upon the model system . Modules associated with specific parameters could then be examined in detail , utilising interaction databases ( Ingenuity ) for further characterisation . Detailed examination of these networks illustrated the changes detected by broader classification of modules by annotation terms . In addition to potential interactions with diet and salinity , the majority of networks contained key regulators of inflammation , hepatic fibrosis and hepatocellular carcinoma . Therefore we propose that network biology approaches can lead to the identification of health impacts of environmental pollutants upon non-model organisms . The molecular differences between reference and contaminated sampling sites were associated with carcinogenesis , and this outcome is supported by previous histopathology [10] , [39] . Flatfish hepatic histopathology has long been associated with chemical contamination [39] and our results demonstrate the linkages between toxicants and histopathology via alterations in molecular signalling pathways and metabolism .
The sampling sites employed in this study were: In UK waters; on the Irish Sea , the Mersey estuary , at Eastham Sands , Liverpool ( lat 53°19N , long 2°55W ) and Morecambe Bay ( lat 54°10N , long 2°58W ) ; on the North Sea , the Alde estuary , Suffolk ( lat 52°95N , long 01°33E ) and the Tyne estuary at Howdon , Tyne and Wear ( lat 54°57N , long 1°38W ) : In North Sea waters off Schleswig-Holstein , Germany; the Elbe estuary at Cuxhaven ( lat 53°53N , long 08°15–19E ) and Brunsbuttel ( lat53°52N , long 09°09–10E ) and off Helgoland ( lat 54°06N , long07°15–08°00E ) . Adult European flounders ( Platichthys flesus ) were caught during statutory monitoring programs carried out by the Centre for Environment , Fisheries and Aquaculture Science ( Cefas ) at UK sites in April 2006 and by AWI Bremerhaven at FRG sites in October 2004 and April 2005 . Fish were caught using beam trawls and held in tanks of flowing sea water onboard ship and were dissected either onboard ship or on return to shore . Livers were immediately removed and 100 mg samples for microarrays and metabolomics and 200 mg samples for biomarker assays were flash frozen in liquid nitrogen , with liver slices taken for histopathology . Blood was extracted and stored at 4°C overnight before plasma preparation for vitellogenin ( VTG ) analysis . Fin clips ( 1 cm2 ) were preserved in 70% ethanol at 4°C for genotyping . After sexing , livers from males that were 12 to 34 cm long were used for further analyses , this sample set included n = 20 for Alde , n = 16 for Tyne , n = 22 for Mersey , n = 23 for Morecambe Bay , n = 22 for Helgoland , n = 24 for Cuxhaven and n = 48 for Brunsbuttel . Fish lengths and weight , condition factor ( K; body wt/length3×100 ) , liver weight and hepatosomatic index ( HSI; liver wt/body wt×100 ) were determined for all samples , gonad weight and gonadosomatic index ( GSI; gonad wt/body wt×100 ) for FRG fish only . Chemical determinations were carried out on sediment samples and independent sets of flounder liver samples from the same samplings by Cefas and Deutsches Ozeanographiches Datenzentrum , Germany and submitted to the International Council for the Exploration of the Sea ( ICES ) , Copenhagen , Denmark as part of the national marine monitoring programmes . UK data was analysed from that collected as part of the Clean Safe Seas Environmental Monitoring Programme ( CSEMP ) and archived in the UK's Marine Environment Monitoring and Assessment National database ( MERMAN ) . For sediment; metal concentrations ( Al , As , Cd , Cr , Cu , Fe , Hg , Li , Mn , Ni , Pb and Zn ) ; polycyclic aromatic hydrocarbons ( PAHs ) ( anthracene , benzo[a]anthracene , benzo[a]pyrene , benzo[ghi]perylene , chrysene/triphenylene , fluoroanthrene , indo[123-c]pyrene , naphthalene , phenanthrene and pyrene ) ; total organic carbon and polychlorinated biphenyls ( PCBs ) ( congeners CB28 , 52 , 101 , 118 , 138 , 153 and 180 ) were determined and the sum of PAHs and the sum of ICES 7 priority PCBs calculated for all sites . For flounder livers; metals ( As , Cd , Cr , Cu , Fe , Hg , Ni , Pb , Se and Zn ) and PAHs ( acenaphthylene , acenaphthene , benzo[a]anthracene , C1- , C2- and C3- naphthalene , C1-phenanthrene/anthracene , chrysene , fluoroanthrene , fluorene , naphthalene , phenanthrene and sum of PAHs ) were determined for Alde , Tyne and Mersey fish , with partial metal concentration data for Morecambe Bay , Helgoland , Cuxhaven and Brunsbuttel samples . Polychlorinated biphenyls ( PCBs ) ( congeners CB28 , 52 , 101 , 118 , 138 , 153 , 180 and sum of ICES 7 PCBs ) were determined for liver samples from all sites . Data are available from the Merman database ( http://www . bodc . ac . uk/projects/uk/merman/ ) . UK flounders were examined for external lesions , liver gross appearance and parasite infection . Liver pathology was assessed according to the criteria of Feist et al . [40] . Sections of liver tissue were removed , placed into individual histological cassettes , transferred to 10% neutral buffered formalin and processed for histopathology as described previously [11] . The presence of toxicopathic lesions , foci of cellular alteration , benign neoplasia , malignant neoplasia and non-specific inflammatory lesions was determined . Plasma vitellogenin ( VTG ) concentrations ( mg/ml ) were determined by the method described by Kirby et al [41] . Hepatic metallothionein ( MT ) concentration ( µg per mg ) and glutathione reductase ( GR ) ( nmol/mg ) , glutathione-S-transferase ( GST ) ( µmol/mg ) and ethoxyresorufin-o-deethylase ( EROD ) ( pmol/mg ) activities were determined by the methods of George and Young [42] . These assays were carried out for all except Cuxhaven and Helgoland fish . Flounder fin-clip samples ( n = 50 ) from all sites were surveyed for six neutral microsatellite markers ( all polymorphic ) and 13 detoxification gene-associated size variants within introns of flounder cytochrome P450 1A ( CYP1A ) [43] , GST-A [44] and peroxisome proliferator activated receptors ( PPAR ) alpha , beta and gamma [45] . Following targeted PCR spanning each polymorphic site; DNA fragments were detected and sized by fluorescent capillary electrophoresis ( Beckman CEQ8800 sequencer ) . Chromatogram files were individually inspected , and alleles were identified/scored manually . Four standard flounder DNA samples were analysed in each genotyping run ( 96 sample plate ) to maintain scoring consistency . Standard genetic analyses for both single and multi-locus conformance to Hardy-Weinberg expectations within samples and examination of potential allelic differentiation among sites were undertaken using GENEPOP [46] . PHYLIPv3 . 5 software [47] was then employed to compute and compare four different measures of genetic distance ( Nei's standard and Da distances; Cavalli-Sforza chord distance; Reynolds distance ) and to construct unrooted neighbour-joined dendrograms ( branch points being bootstrap-supported ) . The GENIPOL flounder cDNA microarray has been described previously [48] , [49] . The methods and design were similar to those employed in earlier experiments , with minor modifications [8] , [9] . Briefly , liver tissue from individual flounders was homogenised in a methanol/water mixture [50] and aliquots were taken for both metabolomics and transcriptomics . Liver homogenates were used to prepare total RNA ( Qiagen , Crawley , UK ) , reverse-transcribed to cDNA and labeled with Cy5-dCTP fluorophore ( GE Healthcare , Amersham , UK ) . Labeled cDNAs were individually statically hybridised overnight to the microarray versus a common Cy3-labeled synthetic reference , before stringent washing and scanning ( Axon 4000B; Molecular Devices , Wokingham , UK ) . Data were captured using Genepix software ( Molecular Devices ) , and each slide was checked in detail , with spots showing poor morphology or arrays showing gross experimental artefacts being discarded . The data consisted of local background-subtracted median 635 nm intensities . MIAME-compliant gene expression data are available from ArrayExpress under accession E-MTAB-396 . As the microarray is redundant , CAP3 clustering [51] had been used to identify contiguous sequences [48] . For metabolomics , liver homogenate aliquots were further extracted individually using methanol/chloroform/water ( 2∶2∶1 . 8 final volumes ) [50] , [52] . One-dimensional 1H NMR spectroscopy was performed upon the hydrophilic fraction as previously described [53] . Briefly , NMR spectra were measured at 500 . 11 MHz using an Avance DRX-500 spectrometer and cryogenic probe ( Bruker , Coventry , UK ) , with 200 transients collected into 32k data points . NMR data sets were zero-filled to 64k points , exponential line-broadenings of 0 . 5 Hz were applied before Fourier transformation , and spectra were phase and baseline corrected , then calibrated ( TMSP , 0 . 0 ppm ) using TopSpin software ( version 1 . 3; Bruker ) . The subsequent processing and statistical analyses of the NMR data have been described in detail in a previous study [53] . Briefly , taurocholic acid , an abundant bile acid with highly variable concentration in the liver extracts was subtracted from each spectrum using Chenomx NMR metabolomics software ( version 4 . 6; Chenomx , Edmonton , Canada ) . Next , residual water was removed , each spectrum was segmented into 0 . 005 ppm bins , and the total area of each binned spectrum was normalized to unity so as to facilitate comparison between the samples . Subsequently to statistical analyses , significantly changing metabolite ‘bins’ were identified as particular metabolites by comparison with spectral libraries of reference compounds and were annotated with PubChem CID accessions ( NCBI ) . Microarray data were filtered to remove spots where 20% or more of the data were undetectable over all samples and background-subtracted intensity values of 0 or below were set to 0 . 5 . Data were log2 transformed , quantile normalised and de-noised by a ) removing data where SD/mean was more than 0 . 9 and b ) removing data where maximum–minimum was less than 1 . 5 . Missing data were estimated using MetaGeneAlyse probabilistic principal components analysis ( PCA ) algorithm [54] . Array slide batch effects were resolved using an empirical Bayes correction [55] . A representative clone with greatest average expression across all samples was chosen for each contiguous sequence cluster where the Pearson correlation score was greater than 0 . 6 to other members of the cluster . Where the correlation failed to pass this cut-off , data were discarded . The noise level for each metabolomics NMR spectrum was estimated by dividing the spectrum into 32 regions and calculating the smallest bin SD for each region and multiplying this by 3 . These results were used to de-noise the data [56] . Data from metabolomics , transcriptomics and fish measurements ( K , length , weight , liver weight , HSI ) were then combined where all were available . The final data set therefore consisted of n = 15 for Alde , n = 9 for Tyne , n = 9 for Mersey , n = 13 for Morecambe Bay , n = 21 for Helgoland , n = 23 for Cuxhaven and n = 36 for Brunsbuttel . An additional dataset was generated for the omics samples that also possessed genetic data . Normalised combined microarray and metabolomic data were input to Genespring GX 7 . 3 . 1 ( Agilent Technologies , Santa Clara , CA , USA ) . Statistically significantly changing genes were found by 1-way ANOVA with a multiple testing correction [57] for a false discovery rate ( FDR ) <0 . 05 , and with Welch T-tests employing the same FDR . Fold change cutoffs of 1 . 5-fold were additionally applied . A classification algorithm was used to compare previous data [9] with the current data; this employed the Support Vector Machines algorithm within Genespring with the Kernel Function Polynomial Dot Product ( Order 3 ) , a Diagonal Scaling Factor of 0 , for all genes passing QC cut-offs in both experiments . Gene ontology ( GO ) analyses were carried out within Blast2GO [58] , [59] employing the GOSSIP package [60] . As flounder is a non-model species , genes were annotated with gene symbols of their putative human orthologs , found by employing a Conditional Stepped Reciprocal Best Hit approach between flounder and zebrafish ( Danio rerio ) and human transcriptome databases , similar to Herbert et al . [61] , with additional manual curation . Chemical-gene expression interactions were downloaded from the Comparative Toxicology Database ( CTD ) [15] , for all annotated genes . These represent a database of the previous literature on chemical-gene expression interactions . The chemical-gene pairs from this list were segregated into inducers and repressors , duplicates were removed , and the two lists uploaded into TMEV [62] , thereby annotating each gene with its ‘chemical inducers’ and ‘chemical repressors’ . Lists of genes ( ANOVA , FDR<0 . 05 , fold change versus Alde>1 . 5; illustrated in Table S1 ) were interrogated for enrichment of chemical associations using EASE ( Expression Analysis Systematic Explorer ) within TMEV , and FDR calculated . Where associations were found between an inducing chemical and induced genes and also between the same chemical acting as a repressor and repressed genes , FDRs were multiplied to produce a final FDR value . Lists of genes and metabolites were additionally interrogated by Ingenuity Pathway Analysis ( Ingenuity IPA 8 . 5; Ingenuity Systems , Redwood City , CA , USA ) , employing Human Gene Organisation ( HUGO ) gene identifiers and PubChem CID compound identifiers , with statistical tests using Benjamini and Hochberg multiple testing corrections . The overall approach taken for network construction and analysis is shown in Figure 1 . It is conceptually sub-divided into: 1 ) Selection of network hubs: 2 ) Construction of a fully connected network: 3 ) Identification of network modules representing the neighbourhood of the hubs: 4 ) Assembly of the final modules and graphical representation: 5 ) In-depth analysis of gene interactions using Ingenuity Pathway Analysis ( IPA ) software . The network was constructed from all measured variables , including transcript , metabolite , morphometric , protein biomarker and genetic data . Within the network each individual variable is described as a node . We selected 99 ‘hub’ nodes representing transcripts with known toxicological and regulatory relevance in order to identify the molecular network representing the interactions between these hubs and all the other nodes in the multi-level dataset . In addition , morphometric indices and metabolite peaks were also included in the list of hubs to represent the complexity of the metabolic networks , which , we reasoned are likely to closely influence liver physiology . The network inference methodology ARACNE [13] was used to create the network . Statistically significant interactions were selected on the basis of mutual information between the nodes at cut-off of P<1e-6 . We defined 99 modules derived from each selected hub and its neighbourhood . Many nodes were present in multiple modules . The overlap index was calculated between each pair of modules by dividing the number of overlapping nodes by the total number of nodes in the smaller module . The final network was then constructed as the union of all network modules and visualized using a force driven layout available in the software application Cytoscape [63] . In the final network , the edge distances between the modules are relative to the overlap index and the node sizes are relative to the size of the module . We also compared this strategy to develop mutual information-based modules from hub variables with a more complex method [64] , shown in Text S1 , and discovered that they both gave similar results . Subsequently the multivariate selection algorithm GALGO [65] was applied to each module to determine its predictivity for parameters including fish sampling site , parasite infection , and the presence or absence of liver pathologies . The cut-off employed for identification of predictive modules was >70% specificity and >70% sensitivity . Genes were annotated with HUGO identifiers for their putative human orthologs . DAVID v 2008 and v6 . 7 [66] , [67] was used to classify module genes and groups of modules inferred from the topology of the module graph , by their shared Gene Ontology ( GO ) and other annotation terms . Flounder laboratory treatment data was employed to relate gene expression changes seen in the environmental samples to those elicited by model toxicant treatments . These treatments consisted of a single intraperitoneal injection with cadmium chloride ( Cd , 50 µg/kg ) , 3-methylcholanthrene ( 3-MC , 25 mg/kg ) , aroclor 1254 ( 50 mg/kg ) , tert-butyl-hydroperoxide ( tBHP , 5 mg/kg ) , lindane ( 25 mg/kg ) , perfluoro-octanoic acid ( PFOA , 100 mg/kg ) , estradiol ( l0 mg/kg ) and furunculosis vaccine ( killed Aeromonas salmonicia , Aquavac Furovac 5; 1 ml/kg ) with gene expression monitored over a 16-day timecourse versus appropriate controls . Full details are shown in Williams et al . [48] , Williams et al . [8] and Diab et al . [49] , with data available from ArrayExpress under accessions E-MAXD-32 and E-MAXD-38 . Overlap between module genes and genes differentially expressed by toxicant treatments over the 16 day timecourse , ANOVA , FDR<0 . 05 was determined by Fisher's Exact Test with a cut-off of P<0 . 01 . Modules and groups of modules were interrogated by Ingenuity Pathway Analysis . Key regulatory molecules were inferred from networks generated within Ingenuity , that also output functional enrichment within the lists of nodes ( P<0 . 05 ) . Modules were annotated with associated diseases , functions , canonical pathways , toxicity and hepatotoxicity terms within Ingenuity and with inferred key regulators , as well as with environmental and parasitological predictivity and overlap with laboratory treatment data to produce a binary matrix . This was clustered within TMEV [62] using hierarchical clustering , SOTA self organising tree , K-means , QT and SOM self organising map algorithms . Grouped modules that were predictive of sampling site were subjected to Ingenuity analyses and were overlaid with Brunsbuttel expression data relative to Alde , the reference site . These genes and metabolites were subjected to K-means clustering within TMEV and the clusters functionally annotated using DAVID .
|
Understanding how living organisms adapt to changes in their natural habitats is of paramount importance particularly in respect to environmental stressors , such as pollution or climate . Computational models integrating the multi-level molecular responses with organism physiology are likely to be indispensable tools to address this challenge . However , because of the difficulties in acquiring and integrating data from non-model species and because of the intrinsic complexity of field studies , such an approach has not yet been attempted . Here we describe the first example of a global network reconstruction linking transcriptional and metabolic responses to physiology in the flatfish , European flounder , a species currently used to monitor coastal waters around Northern Europe . The model we developed has revealed a remarkable similarity between network modules predictive of chemical exposure in the environment and pathways involved in relevant aspects of human pathophysiology . Generally , the approach we have pioneered has important implications as it can be applied to model molecular mechanisms of compensatory adaptation to a wide range of scenarios in wild populations .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"genome",
"expression",
"analysis",
"functional",
"genomics",
"environmental",
"sciences",
"marine",
"monitoring",
"marine",
"biology",
"toxicology",
"environmental",
"protection",
"marine",
"and",
"aquatic",
"sciences",
"regulatory",
"networks",
"biology",
"microarrays",
"systems",
"biology",
"predictive",
"toxicology",
"biochemistry",
"ecology",
"earth",
"sciences",
"genomics",
"computational",
"biology",
"genetics",
"and",
"genomics"
] |
2011
|
Towards a System Level Understanding of Non-Model Organisms Sampled from the Environment: A Network Biology Approach
|
We present a computational model of spatial navigation comprising different learning mechanisms in mammals , i . e . , associative , cognitive mapping and parallel systems . This model is able to reproduce a large number of experimental results in different variants of the Morris water maze task , including standard associative phenomena ( spatial generalization gradient and blocking ) , as well as navigation based on cognitive mapping . Furthermore , we show that competitive and cooperative patterns between different navigation strategies in the model allow to explain previous apparently contradictory results supporting either associative or cognitive mechanisms for spatial learning . The key computational mechanism to reconcile experimental results showing different influences of distal and proximal cues on the behavior , different learning times , and different abilities of individuals to alternatively perform spatial and response strategies , relies in the dynamic coordination of navigation strategies , whose performance is evaluated online with a common currency through a modular approach . We provide a set of concrete experimental predictions to further test the computational model . Overall , this computational work sheds new light on inter-individual differences in navigation learning , and provides a formal and mechanistic approach to test various theories of spatial cognition in mammals .
Neurobehavioral evidence supports a prominent role for interactions between multiple anatomically distinct memory systems in the mammalian brain underlying the coordination of different behavioral strategies during learning ( e . g . , [1] ) : A cognitive memory system , relying on a network comprising the hippocampus , prefrontal cortex and associative parts of the basal ganglia ( i . e . , the dorso-medial striatum ) , would mediate goal-oriented planning strategies; While a stimulus-response/habitual memory system , relying on sensorimotor parts of the cortex and basal ganglia ( i . e . , the dorso-lateral striatum ) , would in parallel mediate the progressive acquisition of routine strategies that would take over with overtraining [2–8] . Recently , a growing computational effort has been put forward to model the coordination of such behavioral strategies , with more and more computational models employing such a dual learning systems framework to account for changes in animals’ behavioral strategies between different stages of learning during the task [4 , 9–12] , as well as between different subparts of the action sequence or movement trajectory during the trial [9 , 13 , 14] . In particular , when dealing with instrumental conditioning experimental data , these dual systems models well explain animals’ tendency to alternate between initial flexible goal-oriented strategies , where the animal is hypothesized to use an internal model to plan and infer future consequences of action ( model-based ) , and more automatic and habitual strategies at late stages of learning , where behavior is supposed not to rely on an internal model but rather on stimulus-response associations ( model-free ) ( see e . g . , [5 , 15] for reviews ) . In the case of navigation paradigms , the model-based / model-free dichotomy has been found to better account for the diversity of navigation behaviors than the old classical distinctions between place and response strategies , or between allocentric / egocentric strategies [8] . Moreover , such a distinction provides a possible explanation of the distinct roles of the hippocampus and different subparts of the striatum during navigation [8 , 15] . Nevertheless , how learning systems dynamically interact during navigation is still little understood . In particular , it is not clear how a unified coordination principle or mechanism can explain both cases of strategy competition ( when a lesion impairing one strategy but leaving another one spared can produce an improvement in the animal’s behavioral performance , e . g . , [16] ) and cases of strategy cooperation ( when two strategies together produce a better performance than one strategy alone , e . g . , [17] ) . Existing computational models have proposed various criteria to coordinate multiple learning systems , but each criterion has been evaluated on specific experimental paradigms . For instance , system coordination has been proposed to depend on the uncertainty in the model-free system alone [11] , relying on the strong assumption that the model-based system always has perfect information; Alternatively , some models have released the assumption of perfect information [18] , but they still bias the coordination towards a default model-free control , which cannot explain why some actions always remain under model-based control even after training [4] . Other models propose a coordination that can also depend on uncertainty in the model-based system [4 , 19] , an approach that does not scale up to tasks involving a large number of states [20]; some authors have used a fixed coordination weight per individual [12 , 21] , which cannot account for dynamic changes in coordination strength along training; system coordination has also been proposed to depend on working-memory load [18 , 19 , 21] in human experiments where it is considered that accessing working-memory has a cost , and more recently in similar experiments in monkeys [22] . Few multiple-systems models have addressed rodent navigation data . The model proposed by [10] could solve a variety of navigation tasks in a simulated robot but employed fixed pre-learned behavior in the model-free system and did not perform formal comparisons between their simulated robot and experimental results in rats . The model proposed by [9] combined place-based and cue-guided learning systems , coordinating them by choosing at each timestep the system with the smallest reward prediction errors and the largest reward expectations . However , since the two learning systems are model-free , the model cannot account for flexible strategies enabled by model-based learning . In general , most previous computational models of rodent navigation have employed what is called a Locale strategy to account for place-based behavior [9 , 13 , 17 , 23 , 24] , which learns place-action associations through model-free learning and can thus not account for model-based behavior . We previously proposed a computational model of navigation where a gating-network coordinates model-free and model-based systems with a commun currency: their measured instantaneous performance [14] . In the model the gating-network is an associative module which learns through model-free reinforcement learning which system is the most efficient in each location of the spatial and perceptual spaces , which implements a certain degree of hierarchy in learning [25] . This enables to gate the appropriate system at the right moment during performance , depending on input from place cells’ and visual cells’ activity ( Fig 1 ) . The fact that the gating-network learns to select which navigation strategy to follow based on model-free reinforcement learning is consistent with the hypothesis that the same selection mechanisms learned through dopamine reinforcement signals are employed in different striatal territories for movement selection , action selection , and strategy selection [5 , 26–32] . Previous studies reported how the model could reproduce rodent experimental data in two specific navigation tasks . Nevertheless , the previous version of the model employed a hand-tuned mixture of Gaussians to model artificial hippocampal place cells with a fixed distance between place fields , and it is thus not clear how general these previous results were . Here , we extend this model by integrating a more realistic hippocampus model [33] and show how the mechanisms of the gating network within this new model can explain a wide range of experimental data during navigation paradigms involving spatial memory as well as associative phenomena such as generalization and blocking . Specifically , we reproduce the classical reference memory experiment in the hidden water maze [34]; a delayed matching to place task [35]; cases of competition between strategies previously classified as cue-guided and place-based in a water maze [16]; a gradual competition between distal and proximal cues [36]; generalization gradient [37] and blocking [38] . In particular , we show that phenomena such as generalization gradient and blocking observed within the navigation paradigm , which to our knowledge have never been accounted for by computational models before , cannot be explained by each learning system alone , but rather by their interaction through the proposed gating network . While the debate is still vivid in psychology and experimental neuroscience between the cognitive map theory and the associative theory of mammal navigation [39–44] , our work highlights together with recent previous computational models that a variety of learning behaviors can result from a single coordination mechanism for the interaction between these two types of strategies . Moreover , while most previous computational models focus on mechanisms for the competition between learning systems , our work shows that a set of rodent navigation behaviors can be explained in terms of cooperation between systems . Finally , by proposing a common currency for learning systems coordination , our model can generalize to the coordination of N systems whose individual learning mechanisms may be of different nature . This could help predict behavior in paradigms involving more than two navigation strategies , which has so far rarely been experimentally studied .
One of the best known experimental paradigms , the Morris water maze show that intact rats are able to learn the location of a hidden , stable platform [34] ( Fig 2a ) . In contrast , hippocampal-lesioned animals are impaired in such tasks as shown in Fig 2b . While previous computational models have already reproduced these classical results ( e . g . , [17] ) , we present here new simulations with our model to show that it can also reproduce them ( Fig 2c ) , but also to analyze which variants of the model fail to do so . The simulated hippocampus-lesioned model , where only the Direction ( D ) strategy is operational , shows significantly higher latencies to reach the platform during the first 10 trials than the full model , where both strategies ( DP ) are operational ( Mann-Whitney test for non-matched paired samples , p < 0 . 001 ) . In the simulations , when P and D strategies are available simultaneously , the gating mechanism learns to privilege the former ( S3a Fig ) which uses the configuration of distal cues to estimate the allocentric position of the platform and to plan a sequence of movements towards it . In contrast , the performance of an associative model with a D strategy only , where distal cues compete against each other , was impaired as is the case with hippocampal lesioned animals . Interestingly , our simulations predict that if the experiment is performed for a sufficient number of trials , animals with impairments in hippocampal processing ( i . e . , D strategy alone ) should eventually reach the platform with performance that is not statistically different than control animals ( i . e . , DP strategies together ) . This is consistent with more recent experimental results showing that the blocking of hippocampal sharp-wave ripples oscillations , known to be important for memory consolidation , impairs performance in a spatial memory task but still spares a slow improvement in performance in the tested animals [45] . Moreover , our simulations predict that Striatum-lesioned animals ( i . e . , P strategy alone ) should not be impaired in this task ( Fig 2c ) . This is again consistent with more recent experimental results in the water maze where striatum-lesioned animals had non-different espace latencies than controls [16] . So far , these results are not novel compared to the large body of computational simulations of this experiments that have been previously done [17 , 23 , 24] . Nevertheless , it is interesting to note that many such models have reproduced these results using a model-free allocentric Locale ( L ) strategy in contrast to the model-based one used here . Simulation of a variant of our model where the Planning strategy is replaced by a Locale can also reproduce the experimental results ( S3b Fig ) . Nevertheless , the simulation results strikingly lead to different predictions: that the performance of Hippocampus-lesioned animals should never reach that of the control animals even after a large number of simulated trials; and that the performance of Striatum-lesioned animals ( i . e . , L strategy alone in S3b Fig ) should also be impaired ( but less ) compared to control animals . Hence in this variant of the model , the two strategies together produce a better performance than each strategy alone , which reveals a potential collaborative interaction between strategies that will be discussed later . These predictions constitute possible ways to disentangle the two alternative models . However , here we argue that the ability of a model without model-based strategy to reproduce these results is mainly due to the stationarity of the task: the platform always remains at the same location , which can be easily learned by a model-free strategy . The next simulated experiments will show that in non-stationary cases , a model-based strategy is necessary to reproduce rats’ ability to adapt in a few trials to each change in the platform location . In an extension to the previous paradigm , the hidden platform was moved every session made of four trials ( Fig 2d ) , thus allowing the animal to remember its position for a few trials once it has been found , but nevertheless requiring an adaptation to frequent goal location changes [35] . Experimental results show that escape latencies of intact animals increased on the first trial after the platfom is moved , but decreased quickly in the following ones ( Fig 2e ) . Results of model simulations showed that this quick adaptation of behavior can be reproduced when a model-based Planning strategy is available ( Fig 2f ) , but not with a model-free Locale one nor an associative Direction one ( S4a and S4b Fig ) . The Planning strategy permitted the quick within-session adaptation observed in rats , while both Locale and Direction strategies were much slower at learning the platform location and hence did not display much within-session reduction in escape latency . Analysis of the evolution of the contribution of strategies within the full model shows that while the Direction strategy contributed to the model decisions of movement during the first simulated session , the model quickly learned to avoid using it during next sessions ( S4c Fig ) . This explains why the performance of the full model shows smaller within-session reduction in escape latencies during the first session than during later sessions . At that stage , the model automatically learned that solving this task can be achieved through a combination of Planning and Exploration strategies . When the contribution of the Exploration strategy increased , such as during session #5 , the model started the session with a lower escape latency because the model relied less on the Planning strategy at the first trial of the session and hence spent less time searching around the previous platform location . This suggests that an ideal combination of strategies in this type of tasks , once the structure of the task is learned by the model , would be to rely most of the time on the Planning strategy—to enable the quick within-session improvement in performance—while keeping a certain level of exploration to prevent the model from being stuck at the previous platform location . As we will see in Experiment IV , the gating network of the model can achieve this sort of cooperation between strategies—hence suggesting a way in which rats may do it—when the presence of an intra-maze cue enables the Direction strategy to be efficient at the first trial of each session . Before that , the next simulated experiment will illustrate a case where the use of an intramaze cue enables the Direction strategy to reach a good performance and hence to enter competition with the Planning strategy . In this experiment , animals learned to reach a cued and stable platform , also identified by surrounding distal cues . During some trials , the cue was hidden , forcing the animals to learn its location also by distal cues ( thus discarding the possibility of overshadowing—i . e . , neglecting—distal cues because of the presence of the proximal one ) [16] . In the last trial block ( 4 trials ) , the cued platform was moved at the opposite place , testing whether rats reach it following its spatial location or the cue ( Fig 3a ) . In these trials , hippocampal-lesioned animals went directly towards the new cued platform position , as did half of the control animals—named cue-responders . In contrast , the remaining control animals—named place-responders—first swam towards the previous platform location ( presumably following distal cues ) and then went directly to the cued goal . This suggests a competition between both strategies taking place at these trials ( Fig 3b ) . Simulation results in this task with a previous version of our model have already been reported in [14] . That study focused on whether output actions in the model should have an egocentric or an allocentric frame . Here , using simulations relying on an allocentric frame for output actions , we address two questions: how the gating-network can manage inter-strategy competition in order to solve the task; what are the different experimental predictions raised when the model-free Direction strategy competes with a model-based Planning strategy versus a model-free Locale one . As previously reported [14] , the model can reproduce the experimental results both in the control case ( when strategies P and D are available ) and in the hippocampal lesion case ( when only the D strategy is operational ) ( Fig 3c ) . Simulations reproduce the fact that control and lesion groups perform comparably in the trials where the intra-maze cue is visible ( trials #1 , 2 , 4 , 5 , 7 , 8 ) and during the competition trial #10 , as well as the significantly larger escape latencies of the lesion group in trials where this cue is hidden ( trials # 3 , 6 , 9 ) . An interesting new prediction from the model is that lesions to the striatum ( putatively impairing the Direction strategy while sparing the Planning strategy ) would produce an intermediate performance ( P group in Fig 3c ) . More precisely , the performance should not be impaired in the hidden cue case because the Planning strategy can still rely on distal cues to locate the platform . Nevertheless , the performance in the visible case should not be as good as the full model , indicating that the full model solves this case through a cooperation between strategies rather than by the Planning strategy alone . Such a cooperation is illustrated in S5d–S5h Fig where the trajectory produced by the agent during a given trial expresses a D strategy during the initial part of the trajectory and a P strategy later on . This enables to spend less time far from the new platform location by preventing the P strategy from driving the agent towards the previous platform , as is the case with the P model alone during the competition trial #10 . This contributes to a better performance of the full model also in that case . Such a cooperation is nevertheless characterized by a strong dominance of the Planning strategy in the behavior of the simulated agents ( S5a Fig ) . Separating selection rates by trial types clearly shows that the model manages to increase the contribution of the Planning strategy when the intra-maze cue is hidden ( hence when the Direction strategy is inefficient ) and to decrease it during the competition trial in order to reduce the time spent at the previous location of the platform ( S5b Fig ) . A second line of simulation results can be illustrated when repeated simulations with the same parameter-set enable the full model to exhibit behavior alike to the two distinct populations of experimentally observed rats: cue-responders and place-responders ( S5c Fig ) . The most important prediction of the model in this case is that the behavior of both populations should at the same time reflect a dominance of each individual’s preferred strategy ( Planning for place-responders and Direction for cue-responders ) , but in neither group this behavior results from the complete absence of the other strategy ( S5c Fig ) . In our simulations , the Direction strategy still contributed to 20% of the choices made by the simulated place-responders . Conversely , the Planning strategy contributed to nearly 15% of the choices made by the simulated cue-responders . An important consequence of this feature is that individual simulated trajectories within the competition trial #10 reflect an alternation between movements guided by the three different strategies ( Exploration included; S5d Fig ) . This illustrates a cooperation between strategies , the Planning strategy being the one which attracts the simulated agent towards the previous platform location in this case ( obviously more strongly for place-responders than for cue-responders ) . These results suggest that even experimental situations of apparent competition between navigation strategies can be solved through different degrees of cooperation , the respective contribution of each strategy being dynamically adjusted by the model to achieve the task properly . We performed additional simulations to analyze the case where the model-based Planning strategy is replaced by a model-free Locale strategy , as in previous computational models [9 , 17 , 24] . As before , the model shows an increased contribution of the spatial strategy ( here the Locale instead of the Planning ) during trials where the intra-maze cue is hidden , and a strong decrease in its contribution during the competition trial to avoid losing time at the previous platform location ( S5e Fig ) . Nevertheless , these results reveal an overall increase in the contribution of the Direction strategy ( S5f Fig ) . This can be explained as a mechanism of compensation for the lower flexibility of the model-free Locale strategy compared to the model-based Planning one . Interestingly , this version of the model is still able to reproduce the experimental results , both in the control and lesion cases ( S5g Fig ) . Like in Experiment I , we argue that this is made possible because the platform location is stable during all trials except test trial #10 . A different prediction from this version of the model is that , while place-responders should still perform mixed strategies relying on a cooperation between strategies , the involvement of the Locale strategy should be much weaker in cue-responders , resulting in frequent homogeneous trajectories only controlled by the Direction strategy ( S5h Fig ) . As in Experiment II , the platform was moved every four trials ( a session ) , but in this paradigm a proximal cue indicating the position of the platform was held at a constant distance and direction from it [36] ( Fig 3d ) . This was meant as a way to enable the Direction strategy to also solve the task on its own , and hence to trigger a fair competition between Planning and Direction strategies . Central questions addressed by the authors of the original study are whether hippocampal lesions would specifically impair a particular strategy and how this would bias the competition . Interestingly , they observed that both control and hippocampus-lesioned animals were able to ( at least partially ) learn the task since they both show a gradual improvement in performance , as illustrated by the decrease in escape latencies across sessions ( Fig 3e ) . This session-by-session progressive improvement in performance converged to a point where both groups reached similar espace latencies in the last sessions . The important observation is that the two groups showed different performance characteristics within each session . Control animals were able to display a fast adaptation ( i . e . , within 4 trials ) to the new position of the platform at each session . In contrast , hippocampus-lesioned animals did not show a significantly different performance between the first and the fourth trial of each session ( Fig 3e ) . Strikingly , hippocampus-lesioned animals were nevertheless better than control animals at the first trial of the session . Further analyses reveal that this is explained by the tendency of hippocampus-intact animals to spend time at the previous platform location [36] . This experiment thus reveals a set of intrincate phenomena which support a dual learning systems approach: the hippocampus appears as necessary to enable fast adaptation to new platform location; Nevertheless , lesion of the hippocampus led to a reduction in the time spent around the previous platform location; In consequence , both groups were eventually able to learn the task . One important computational question is how the model should balance the competition/cooperation between the strategies to produce such a performance ? As for Experiment III , our previously reported results in this paradigm [14] focused on whether output actions in the model should have an egocentric or an allocentric frame . Here we show new simulations to ( i ) further analyze the balance between cooperation and competition mediated by the gating network , and ( ii ) to see whether a model-free Locale strategy could solve the task similarly to a model-based Planning strategy in the model . The combination of Planning and Direction strategies in the full model can reproduce the behavior of intact animals in this experiment ( Fig 3f ) . Emulation of hippocampal lesions in the model—leaving only the Direction and Exploration strategies spared—can reproduce the behavioral performance of the lesioned animals in this task: a better performance than the control group at the first trial of each session; and a lack of fast adaptation between the first and the fourth trial of each session; hence an impaired performance compared to controls at the fourth trial of each session . Interestingly , as observed in our previous work [14] , an artificial lesion to the striatum in the model—leaving only the Planning and Exploration strategies spared—predicts an impaired performance in this task ( S6a Fig ) : while the fast adaptation between the first and the fourth trial of each session is preserved , the striatum-lesioned model shows larger escape latencies than controls and hippocampus-lesioned models at the first trial of each session . This directly results from the tendency of the Planning strategy to be attracted by the previous platform location at the first trial of each session . We found this tendency to be stronger in the behavior of the simulated agent in the striatum-lesioned model ( P ) than in the full model ( DP ) ( S6b Fig ) . Another interesting property of the striatum-lesioned model is that it does not show the progressive improvement of performance across sessions seen in the two other models , and which could be the signature of a slow model-free learning process ( S6a Fig ) . This constitutes a strong prediction of the model which could be tested experimentally . Importantly , simulation results of each strategy alone enable us to well decompose the overall behavior of the control group into two clearly distinct components: a model-based learning component responsible for the fast within-session adaptation and a model-free learning component responsible for the slow across-session adaptation . Importantly , these two components are clearly visible in the performance of the full model ( Fig 3f ) . It is thus interesting that the model , which has been mainly designed to regulate the competition between navigation strategies—since it gives full control of the movement to a single strategy at each timestep – , learns to achieve some degree of cooperation between strategies so as to benefit from the advantages of each of them . Plotting the rate by which each strategy is selected by the gating network during the first and the fourth trial of each session reveals how the model learned to operate this cooperation ( S6c Fig ) . The Exploration strategy was more selected during the first trial of the two first sessions until the gating network learned to decrease its contribution to the movement . In parallel , the gating network learned to decrease the contribution of the model-free Direction strategy which is not yet efficient at the beginning of the experiment , and to increase the contribution of the model-based Planning strategy which can lead to fast adaptation . Very interestingly , from the second session onwards , the gating network learned to progressively reduce the contribution of the Planning in parallel to the improvement of the Direction strategy with learning . This resulted in the simulated agent spending less and less time at the previous platform location ( S6b Fig ) . After the eighth session , the Direction strategy is selected more often than the Planning strategy , because it is now sufficient to successfully solve the task . This is the explanation that the model offers relative to the hippocampus-lesioned group in the experimental data which eventually reached the same performance as the control group in the last sessions ( Fig 3e ) . Finally , it is worthy of note that the selection rate of the Planning strategy not only decreases during the first trial of each session , but also during the fourth one ( S6c Fig ) . This is because the input that the gating network receives only provides it with information about visual cues and activity in the planning graph ( Fig 1 ) . The gating network is thus not able to discriminate between the different types of trials . A prediction from this is that any learning occurring in one type of trial will affect behavior in the other type of trials , which contributed here in making the Direction strategy more prevalent in the behavior of the simulated agents in the late sessions . We further evaluate the predictions of this approach when the model-based Planning strategy is replaced by a model-free Locale strategy . Interestingly , the learning of the Locale strategy is too slow to learn the new platform position within only 4 trials , making the performance of this version of the model at the fourth trial not better than the Direction group ( S6d Fig ) . Hence , as it was the case in the previous experiment , the important message is that fast adaptations within a few trials experimentally observed in animals are more likely to be well accounted for by a model-based learning strategy than by a model-free one . The last two experiments presented here highlight the role that associative learning processes can play in navigation paradigms . In particular , both experiments were originally designed as attempts to experimentally contradict the cognitive mapping theory—relying on localization processes based on a constellation of distal cues—by showing that individual cues could induce associative phenomena previously observed in non-navigation learning paradigms to support the associative learning theory . These associative phenomena are generalization gradient and blocking effects , which we will define hereafter while showing at the same time that only a competitive interaction of the associative Direction strategy with others ( e . g . , Exploration , cognitive-mapping-based Planning strategy ) can reproduce these effects , not an associative Direction strategy alone . The spatial generalization gradient effect was studied by [37] in a navigation task involving a hidden platform under opaque water but marked by a proximal cue B , where a gradient of occupancy of the zone near cue B was recorded as this cue was progressively moved away from a distal cue F ( Fig 4a ) . The authors expected a gradual loss of response to the proximal cue proportional to the distance increase . This decrease was supposedly due to the competition between cues—leading to a specific decrease of the proximal cue’s associative strength—rather than due to a competition between strategies . The experimental protocol was composed of two training stages followed by one test trial . During Stage 1 , a training of four sessions of eight trials was performed with two cues present , the proximal cue B ( for Beacon ) being initially close to the distal cue F ( for Frame of reference ) . Stage 2 was composed of 10 sessions of nine trials each . In all sessions , eight of these trials were performed as in the previous stage ( hereby termed escape trials ) . In the 9th trial of sessions 2 , 4 , 6 , 8 , and 10 ( gradient trials ) , the platform was removed and the proximal cue B was rotated 0° , 45° , 90° or 135° from its original position ( Fig 4a ) . This rotation was done either clockwise or counterclockwise , but the direction was kept constant for each animal . In the remaining sessions ( 1 , 3 , 5 , 7 , 9 ) , the 9th trial was conducted with the F cue only , without the cue B nor the platform ( extinction trials ) . These extinction trials were performed to reduce overshadowing of B by F , assumed to bias the generalization gradient . The main experimental result shows that the greater the angle of the proximal cue B rotation during gradient trials , the less time the animal spent in the vicinity of this cue ( i . e . , a generalization gradient ) ( Fig 4b ) . In contrast , the occupancy of the area near the distal cue F did not exceed chance level . Thus , the proximal cue may have been overshadowing the distal one , and the obtention of the gradient suggests that the strength of the proximal cue was learned in an associative way . Nevertheless , during extinction trials rats occupied the octant F above chance level , hence revealing that behavior could still be under the control of the distal cue in the absence of the proximal cue . Our model simulations suggest that only the competitive interaction between associative and cognitive mapping strategies could produce such effects . We found that Direction or Planning alone cannot reproduce the experimental results ( Fig 5a ) . However , the modular approach allowing the selection among these two strategies in the full model was able to do so ( Fig 4c ) . Analysis of the session-by-session evolution of the selection rate of each strategy reveals that the model could achieve this performance by progressively learning during Stage 1 that the Direction strategy is more efficient and accurate than the Planning strategy in this task and should thus be progressively more selected ( Fig 5b ) . Indeed , plotting the escape latencies for the simulations with the Planning strategy alone shows a progressive improvement during Stage 1 followed by degradation of performance during Stage 2 ( Fig 5d ) , which the model tried to compensate by selecting more and more the Exploration strategy instead of the Planning one ( S7a Fig ) . This degradation of performance with the Planning strategy alone was not observed in the experimental data ( Fig 5c ) and only the simulations with the Direction strategy alone or with the full model ( i . e . , DP ) could reproduce the performance during Stage 2 ( Fig 5d ) . This suggests that the contribution of the Direction strategy was required to reproduce the characteristics of the learning process , and that within the Direction strategy the associative strength of cue B overwhelmed that of cue F ( S7b Fig ) . Nevertheless , simulations with the Direction strategy alone cannot reproduce the occupancy rates above chance level in octant F during extinction trials ( Fig 5e ) . Chance levels were here obtained by simulating a Chance group , consisting of only one Exploration strategy , in the same conditions as the other groups . Only the full model and the Planning strategy alone could reproduce the animal behavior during extinction trials . Overall , only the full DP model could reproduce the ensemble of observed results in this experiment . The selection rates of the strategies in the full model can give a further clue about the cooperation between Planning and Direction strategies which was employed to solve the task ( Fig 5b ) . During Stage 1 , selection rates indicate that both Direction and Planning strategies contributed to locate the platform . At the end of this stage , the gating network gave an advantage to the Direction strategy , but the Planning strategy remained selected at a rate above chance ( 36 . 4% ) . In the simulations without the Planning strategy , the Direction strategy was mainly helped by the Exploration strategy ( averaged selection rate of 31% ) . At the beginning of this Stage , its performance was lower than in the full model—suggesting that the performance of the full model during these first trials was due to the cooperation between Planning and Direction strategies . This suggests that even if the Planning strategy was not the most efficient in this task nor sufficient to explain the experimental data alone , reproduction of rats’ performance by the full model still relies on the cooperation of the Planning strategy with the other strategies . Importantly , the spatial generalization gradient effect in the full DP model was mainly due to the associative rules underlying the interactions between strategies rather than the associative rules within the Direction strategy itself ( as the original authors hypothesized ) . S7c and S7d Fig detail strategy selections during gradient trials in groups DP ( left ) and D ( right ) by distinguishing the moments before the simulated agents had reached the octant B for the first time , and the moments after , when they occupied this octant and the other ones . Strikingly , as can be seen in the first column ( “Before B” ) , the generalization gradient was not expressed until the simulated agents reached the octant B for the first time ( light grey dashed line; no significant difference between test trials for 0° , 45° , 90° and 135° ) . The association between the proximal cue and the response leading to the goal was however well learned , as the simulated agents were able to reach the zone of the displaced proximal cue without any gradient . Yet the gradient itself was generated after , by the recruitment of other strategies when searching for the absent platform , i . e . , without getting a reward ( S7c and S7d Fig , right column “After B” , dark grey dashed line ) . This is also depicted by typical trajectories during gradient trials 45° and 135° ( S7e and S7f Fig ) : octant B was rapidly reached with the Direction strategy and the Planning and Exploration strategies gave their contribution after , the former attracting the simulated agents towards octant F . These results contrast with the original authors’ hypothesis considering that the gradient resulted from a gradual loss of the associative strength of B during its learning . Unfortunately , the original experiment did not analyze the octant occupancy within gradient trials , but this prediction of the model would be easily verified . This experiment proposed another associative task [38] to investigate the expression of spatial blocking . This effect was proposed to depend on the amount of training with both distal and proximal cues , and on the change of the physical characteristics of the proximal cue . The hypothesis was that the blocking of the distal cues by a proximal cue would be due both to the presence of the same proximal cue during the experiment ( which could be tested by the replacement by a different proximal cue in a different group of animals ) and to the weak reliability of distal cues during training ( which could be tested by changing the number of trials available to learn the position of a hidden platform based on distal cues ) . In this experiment [38] , four groups of animals are defined: Session-Same , Trial-Same , Session-Diff , Trial-Diff . The experimental protocol is decomposed into three different experimental stages ( Fig 4d ) . In Stage 1 , animals learned to find a cued platform ( with a proximal cue A ) in the presence of surrounding distal cues . For Session animals , the platform was moved every session ( a session being composed of four trials ) . For Trial animals , the platform was moved every trial , so that rats did not anchor their learning process on the distal cues , contrary to groups Session . In Stage 2 , the platform remained at the same location and was signaled either by the same proximal cue A ( Same animals ) or a different proximal cue B ( Diff animals ) . Lastly , in Stage 3 , the platform and its attached proximal cue were removed and the time spent near the previous platform location was recorded . The original experimental results showed the following main phenomenon: Only Trial-Same animals exhibited blocking ( i . e . , the time spent near the previous platform location lasted no more than chance level ) , whereas in other groups , animals spent more time near the previous platform location , demonstrating their learning of the platform location with the help of distal cues [38] ( Fig 4e ) . The proposed explanations are that the two Session groups could not express blocking since , in Stage 1 , distal cues were relevant to locate the platform . In contrast , distal cues were irrelevant for the two Trial groups during Stage 1 , thus susceptible of being blocked . However , the change of proximal cue in Stage 2 in Group Trial-Diff prevented blocking , leaving Group Trial-Same as the only one to express a blocking phenomenon . A second important experimental result in [38] relates to the escape latencies of the animals . In the original article , the observed escape latencies were reported only for Stage 2 . Groups Session-Same and Trial-Same showed no learning improvement ( with respect to Stage 1 ) , whereas groups Session-Diff and Trial-Diff expressed a re-learning of the association between the new proximal cue and the platform , resulting in larger escape latencies than groups Same during the first session of Stage 2 ( i . e . , Session 13 ) ( Fig 6a ) . Simulation of the full model ( containing a Direction strategy D , a Planning strategy P and an Exploration strategy E ) can reproduce these experimental results ( Fig 4f ) : In the Trial-Same group , the selection module learned that the Planning was inefficient due to its poor performance in Stage 1 and was thus no more selected in the later stages ( Fig 6g ) . In Diff groups , the change of proximal cue discarded the selection of the Direction strategies , thus leaving room for the Planning strategy to take place in Stage 3 . And the Planning strategy was not discarded in Session conditions , because of its satisfactory performance in Stage 1 . The full model could also reproduce the significantly different escape latencies during Stage 2 in the case of a different proximal cue B ( Fig 6d ) . The model predicts that these escape latencies for the Diff conditions in the first session of Stage 2 should not be as high as those in the first session of Stage 1 , thanks to the cooperation of strategies P and E which enabled some generalization between these two situations with a different proximal cue . This prediction can however not be confronted with the original article which does not show such comparison between Stage 1 and Stage 2 . We have also tested three other versions of the model in this task: A DL version where the model-based Planning strategy is replaced with a model-free Locale strategy; a D version where the Planning strategy is removed , leaving only the Direction and Exploration strategies; and a P version where the Direction strategy is removed . The latter completely fails to reproduce the experimental results because: ( i ) the time spent in the quadrant containing the previous platform location is significantly above chance in all conditions , unlike experimental results; ( ii ) the escape latencies during Stage 2 do not show an improvement and are significantly different between conditions Trial and Session , unlike experimental results ( S8 Fig ) . A standard cognitive mapping approach is thus not appropriate to explain blocking in this context . Interestingly , while the two other versions ( models DL and D ) can reproduce the escape latencies profile in Stage 2 ( Fig 6e and 6f ) , as model DP also does , they lead to different predictions . The DL model predicts smaller escape latencies in the first session of Stage 2 compared to the first session of Stage 1 , for the same reasons as the DP model . In contrast , the D model predicts even larger escape latencies since the Direction strategy starts Stage 2 with a performance lower than chance because of its synaptic weights resulting from learning during Stage 1 with a different proximal cue ( S9a Fig ) . In addition , the D model predicts a quicker learning across sessions during Stage 1 than the two other models ( Fig 6d–6f ) because it is not polluted by the presence of an inefficient P or L strategy anchored on distal cues . Conversely , model P , but not model D , demonstrates a difference of performance between Session and Trial conditions ( S8b Fig ) , and thus confirms that only the Planning strategy is able to learn within-sessions rather than across-sessions . This complementarity between the Direction and the Planning strategies is confirmed in model DP by the drastic improvement of performance between the first and the fourth trial of each session of Stage 1 in the Session-same condition compared to the Trial-Same condition ( S9b Fig ) . Most importantly , neither the DL model nor the D model can produce a blocking effect only in the Trial-Same condition ( Fig 6b and 6c ) . Model D also showed a blocking effect in the Session-Same condition ( unlike animals ) because of the absence of a Planning strategy to rely on distal cues for the localization of the platform in this condition . S9c Fig illustrates how the DP model could avoid to express blocking in this condition by learning large weights and thus high confidence in distal cues used by the Planning graph ( PG ) when learning with the distal cues was possible for several trials ( Session conditions ) . This results in the prediction that animals will hippocampal lesions should also show a blocking effect in the Session-Same condition , similar to model D . Analysis of the behavior of the DP model during the test trial ( Stage 3 ) provides further insights on the strategy coordination dynamics that may explain animal behavior in this task . We recorded the details of strategy selection in the model before reaching the quadrant of the platform and after , when the simulated agents occupied this quadrant and the others . Because in model DP the Direction strategy was preferred just before Stage 3 , reaching the goal quadrant during the test trial mostly relied on this strategy ( S9d Fig , 1st column of each condition ) , even if the corresponding proximal cue was absent , thus at random . The lowest selection rates of the Planning strategy ( 0 . 57% ) were obtained during the Trial-Same condition . However , the Planning strategy was recruited after , when searching for the absent platform ( S9d Fig , 2nd and 3rd columns ) . According to these results , in all conditions—and not only Trial-Same—did the model avoid to mainly rely on distal cues for reaching the goal quadrant in the absence of the proximal cue . Moreover , in all conditions—Trial-Same included—could the model quickly re-use distal cues after , which constitutes an interesting prediction of the model . Thus , in our simulations , the low occupancy rate of the goal quadrant in the Trial-Same condition could not be assumed to be due to a total blocking of the learning of distal cues—since distal cues were eventually used during the test trial – , but to a decrease in the confidence in these cues , acquired early in the experiment ( as attested by S9c Fig , right ) . In model D , the blocking phenomenon was also expressed in the Session-Same condition , as expected by the authors but not observed in animals . A competition between strategies happened very early in the experiment , giving to the proximal cue a too high relevance during Stage 2 ( S9a Fig ) . This reinforced the use of the Direction strategy in the Same conditions during Stage 3 , leading to a longer time spent out of the goal quadrant ( S9e Fig , 3rd column , Same compared to Diff conditions ) . An important remaining question is whether alternative models employing a model-free Locale mechanism for the place-based strategy instead of the model-based Planning mechanism used here can also reproduce these results . Strikingly , unlike animals , model DL showed no blocking effect at all . This is because this task again involves a platform with a constant location , which gives an advantage to the Locale strategy over the Planning one by enabling the former to learn with more precision . As a consequence , while the Direction strategy is also the most selected in the DL model , the Locale strategy still contributes substantially to the behavior in the Trial-Same condition , hence preventing the blocking effect . Altogether , these results highlight that the complex mechanisms underlying the blocking effect in some conditions but not others can here not be reproduced by a purely associative model containing only a Direction strategy , but can instead be reproduced only by a modular approach which coordinates an associative strategy ( here Direction ) with a cognitive mapping strategy ( here Planning ) . Only such a modular approach was in our simulations capable , like animals , of expressing both blocking and its absence .
In this work , we have presented a computational model for navigation paradigms combining a model-based Planning strategy , a model-free Direction strategy and a random Exploration strategy . The three strategies are coordinated by a gating network which learns in a model-free associative manner which strategy is the most efficient in each situation ( i . e . , depending on visual input and planning graph activity ) . The model could reproduce a set of behavioral and lesion data observed in navigating rodents in six different experiments ( the main results are summarized in Table 1 ) . The model can account for these data by achieving both competition and cooperation between strategies , which results in non-trivial behavior both within and across-trials . It is a striking feature of the model to be able , with a single coordination mechanism through a gating network , to produce both cases of competition between strategies , where lesion of one strategy leads to an improvement of performance , and cases of cooperation where two strategies together produce a better performance than each strategy alone . The model moreover permits a precise quantification of this cooperation/competition trade-off by plotting the evolution of weights assigned by the gating network to different strategies at different times across learning . This permits concrete predictions that could be tested experimentally . Importantly , these behavioral properties result from dynamic activation of different strategies . These dynamics were different from those obtained by a model composed only of model-free strategies , or of only a Direction strategy . The fact that these different variants of the model could not reproduce all the aimed experimental results highlights that a combination of navigation strategies of different nature is key to account for these experimental data . The simulations moreover yield a series of predictions which could be tested in future experiments to further assess the model ( Table 2 ) . We also summarize predictions raised by the alternative model DL , which only relies on model-free strategies , in order to guide future experiments that aim at further comparing the two models ( Table 3 ) . Several previous models have used a Locale strategy [9 , 23 , 24 , 46 , 47] , which associates places to movements without really building a cognitive map ( no topological graph; see detailed comparisons in [8 , 48] ) . Here we have shown that the Planning strategy better explains behavioral results observed in protocols involving frequent changes of goal location , because a model-based strategy is more flexible than a model-free one [4] . We also found that the Locale works well when the goal location is stable , suggesting a possible co-existence of the two strategies in a modular architecture . Such a co-existence has been previously discussed in [8] , arguing that model-free learning processes involved in the Locale strategy could take place in the dorsolateral striatum , while model-based learning processes involved in the Planning strategy could take place in the hippocampus and prefrontal cortex . The possible involvement of the dorsolateral in a model-free place-based strategy is consistent with electrophysiological recordings showing that activity in the dorsolateral striatum correlates with place when the task requires knowledge of spatial relationships [49] . Inactivations of these different regions could be a way to test the different predictions raised in this manuscript relative to Planning versus Locale strategies . Several previous models have already proposed a coordination of model-based and model-free reinforcement learning mechanisms to account for various rodent behavioral data [4 , 11 , 12 , 18] and could thus be considered as possible candidates to model the experiments addressed here . Nevertheless , the Lesaint model [12] proposes a fixed coordination of MB and MF through time: each individual has a specific weight attributed to each system determining its contribution in decision-making . The models proposed by Daw [4] , Keramati [11] and Pezzulo [18] do incorporate a dynamic coordination of MB and MF , based on uncertainty . Nevertheless , these models were designed to account for the sequential shift from initial goal-directed behavior to habitual behavior after overtraining , explaining the insensitivity in the latter case to outcome devaluation , which is a specific case of the questions addressed here . In Experiment IV studied in the present work , animals progressively learn to reduce their use of the cognitive mapping strategy , which we explain in the model by the fact that the gating network learned to use less and less this strategy at the first trial of each session to avoid being attracted by the previous platform condition . The Daw and Keramati models should in principle not be able to explain this because the uncertainty associated to the MB system should be lower and lower sessions after sessions , while uncertainty in the MF system should remain high because the platform changes location every four trials . Besides , the Pezzulo model biases its system coordination towards a default model-free control , which cannot explain why some actions remain under model-based control even after training , as argued in [4] and as observed in several experiments considered here ( e . g . , S3a , S4c , S5a , S7c–S7e , S9d Figs ) . Moreover , the fact that the gating-network of our model learns to coordinate strategies ( which is not the case for these three other models ) also enables the model to learn to increase the contribution of the Exploration system when necessary ( S5b Fig ) , which corresponds to a dynamic exploration rate which is absent from these other models . Moreover , the generalization gradient in Experiment V is produced by the model at the level of the associative rules within the gating network ( thus at the level of strategy coordination ) rather than at the level of associative rules within the model-free Direction strategy itself . The Daw , Keramati and Pezzulo models proposed a coordination criterion which depends on instantaneously measured signals ( i . e . , uncertainty ) rather than on learned signals ( i . e . , their models cannot learn that strategy X is efficient in a particular part of the environment while strategy Y is efficient in another part ) , hence they cannot reproduce this effect . Nevertheless , these models account for a variety of other experiments involving outcome devaluation , contingency degradation as well as hippocampal off-line replays , which our model does not address . Thus it would be particularly interesting in future work to study if combining mechanisms from all these models can account for a wider array of experimental data . Most animal experiments have aimed at distinguishing between only two strategies ( place-based versus associative ) , without subtly distinguishing subtypes of these two categories . Our model enabled to show that different subtypes of place-based strategies ( i . e . , planning versus locale ) are more efficient/relevant depending on the protocol . Similarly , we have previously illustrated how different subtypes of response strategies ( taxon , direction ) which differ in the frame of reference for actions ( resp . egocentric and allocentric ) can also display complementary behavioral properties [14] . Together these computational results predict that new elaborated protocols should permit to isolate more than two concurrent strategies ( for instance , planning+locale+direction or planning+direction+taxon ) . The common currency proposed here enables in principle to coordinate any number of strategies of any different nature , because the model just needs to be able to evaluate their current performance in different states of the task . Moreover , these various subtypes of strategies should engage different parallel memory systems ( for instance subterritories of various cortico-striatal loops , depending on the input-output of these territories and their respective learning mechanisms ) . This predicts that specific lesions of these subterritories should affect only particular subtypes of strategies . The present computational results have important implications relative to the debate between the cognitive mapping theory and the associative theory of spatial cognition in mammals [50 , 51] . These two theories propose alternative mechanisms to explain spatial learning . According to the associative theory , spatial learning is dependent of a single type of mechanism—abundantly studied within the framework of classical and operant conditioning—by which a new response is incrementally acquired by the association of a stimulus and a reward [52–55] . In the associative paradigm , stimuli or group of stimuli available in the environment are assumed to compete to control animal navigation . Those which are not favored by this competition are not going to contribute to the achievement of the task . The cognitive mapping paradigm rather attests the existence of non-associative spatial rules ( i . e . , not incremental , independent from reward ) , in which all cues participate to develop a spatial representation [56] . This theory has received a strong support from the discovery of hippocampal place cells [57] , which enables the animal to quickly build a reliable spatial representation of their environment [58] , independently from the reward ( latent learning ) . The debate between the two theories is still vivid in that the cognitive mapping paradigm is not able to explain blocking or overshadowing effects , and since the actual existence of such “cognitive map” enabling animal and humans to plan shortest paths or shortcuts aroused and still arouses controversies [39 , 40] . On the other hand , opponents to the associative theory highlight a number of experiments failing to display overshadowing between proximal and distal cues [41 , 42] or revealing potentiation between cues during attempts to look for spatial blocking and overshadowing ( for a review , see [43] ) . We have tried to show here that these theories could however be reconciled by a modular paradigm which proposes that both kinds of mechanisms may cohabit in distinct neural systems and may be learned in parallel [44] . Indeed , a large amount of studies have shown that inactivation of specific neural zones in rodents selectively impair only part of their navigational capacities [2 , 44 , 59–69] . The modular approach is also strengthened by several experimental procedures that have shown animals shifting from one type of spatial strategy to another one , either within a navigation trial , or as learning takes place across sessions [36 , 44 , 70–78] . This suggests the existence of mechanisms ruling the selection among navigation strategies in distinct neural structures from those which learn each strategy . In support of this view , lesions of prelimbic and infralimbic areas of the medial prefrontal cortex prevent the shift of a place-based strategy towards a cue-guided one but does not prevent the strategies themselves to be learned or displayed [79] . Similarly , lesion and electrophysiological studies of the ventral striatum suggest an evaluative role of the structure , important for initial learning and flexibility , but not necessarily a substrate for learning a specific navigation strategy ( e . g . , [62 , 80]; see more thorough discussions in [7 , 8 , 81 , 82] ) . The computational model proposed here constitute a refutable proposition concerning the mechanisms that may underly such a modular organization combining associative and cognitive mapping memory systems . Several criticisms of the cognitive mapping theory have argued against the assumption that a global topographical representation ( i . e . , a cognitive map ) exists and this information is available at all times during training . Whether this is a valid assumption and whether real rats benefit from such a representation is open to debate . However , it is important to emphasize that our computational model does not assume that a global map is learned . The mapping mechanism that we used rather focuses on the representation of areas that have been extensively visited [83] , and it leads to local , partial and sometimes approximative maps that can produce suboptimal planning behavior in embedded , noisy tests [84] . Such a mechanism is supported by the observation that successful “planning” of trips does not necessarily depend upon a global representation ( see , e . g . , [85] ) . Moreover , a number of studies over the past 20 years have provided empirical evidence of local , non-global maps ( e . g . , [86–88] , the first one providing clear evidence that non-global ( at the very least ) representations are involved in rodent spatial navigation in the water task ) . Related to this , recent work , both behavioral and physiological , has emphasized the important distinction between local boundary and distal landmark control [89–92] . These results also have important implications for the understanding of the coordination of learning and decision-making systems in humans , beyond spatial navigation . While we focused here on the modeling of experimental data in rodents for consistency , the coordination of model-based and model-free learning principles has also been highlighted in humans during instrumental learning tasks [93] . Moreover , cognitive mapping models have implications beyond spatial navigation , including roles in information contextualization [94] , navigation between conceptual relationships in a manner similar to that of space [95] , mapping of social relationships [96] , and more generally in the integration of memories to guide future decisions [97] . Within this framework , an important question relates to the nature of the interaction between brain networks that underlies these cognitive functions . As mentioned above , previous contributions have emphasized the role of different parts of the striatum in different types of learning [5 , 7 , 8] , the hippocampus being in a position to provide transition information between places for the building of model-based information in the medial prefrontal cortex and more ventromedial parts of the striatum [8] . Interestingly , studies in humans have demonstrated the recruitment of the striatum during learning with immediate feedback in a probabilistic learning task , and increased activation of the hippocampus with delayed feedback [98 , 99] . Strikingly , in these tasks human subjects with Parkinson’s disease—whose striatum is known to be degraded—were impaired in learning from immediate but not delayed feedback . Such results appear consistent with the separation within the model between dorsolateral striatum-dependent model-free learning and hippocampus-prefrontal cortex-dependent model-based learning . Nevertheless , the precise role of different subparts of the prefrontal cortex in these learning processes is probably more difficult to disentangle . One currently attractive theory proposes that the orbitofrontal cortex participates to the learning of relationships between states within the model-based system , which in humans can also be useful to learn cognitive maps of non-spatial tasks [100] . In contrast , hippocampal projections to regions homologous to the dorsolateral and anterior cingulate prefrontal cortex are thought to play an important role in performance monitoring , with increased between-regions coherence upon task learning [101] . Such a process could relate to the performance monitoring mechanisms that underlie systems coordination within our gating-network . Nevertheless , more investigations would be required to further test the hypothesized roles of different prefrontal cortex subregions with respect to the different computations in the model . The proposed coordination of learning systems also offers an opportunity to discuss about the possible role ( s ) of dopamine in mediating memory formation . Here , to be conservative , one could argue that the only role of dopamine on which to postulate relates to the production of phasic model-free reinforcement signals to update action values [102] . Following previous work on the combination of model-based and model-free learning in Pavlovian conditioning [12] , we could further predict that dopamine blockade would only impair model-free navigation strategies , but not model-based ones , thus predicting similar behavior to the one shown through simulations of the Planning system alone . Such a prediction , specific to the navigation domain , could be interesting to experimentally test in order to further assess the model . Nevertheless , dopamine is known to play a role beyond the learning of action values based on reinforcement: For instance , it has been shown that dopamine contributes to the successful binding between experiences that are separated in time [103] , which have been interpreted in terms of inference-based processes at the time of generalization . While dopamine reinforcement signals hypothesized to subserve model-free learning in our model could in principle slowly produce some binding between delayed events , notably through the association of reward values to stimuli and places that precede it , true off-line inference in the model relies on model-based processes ( which enable action planning through a tree-search process [104] ) . Hypothesizing that dopamine plays no role in model-based learning [12] would at first glance fail to explain the coupled changes in learning-phase activity between the hippocampus and the dopaminergic system during information binding [103] . Nevertheless , the possibility to include in the model some off-line replay mechanisms—which permit another form of systems cooperation through the transfer of knowledge from model-based to model-free [105]—could be a promising extension of the model to explain off-line hippocampus drives over model-free dopaminergic learning signals [106] without using these signals for model-based learning per se . Finally , some simplifications and limitations of the present model should be stressed in order to highlight possible ways to improve it . A first criticism that can be raised against the model presented here is the important number of parameters needed . Some of them need to be tuned differently according to the experiment ( S2 Table ) . As a consequence , this can weaken the explanatory power of the model , that could be seen as an unnecessarily complex mixture of experts [107] , where each strategy is considered as an expert whose selection becomes then irrelevant . In order to tackle this issue , we limited ourselves to two free parameters only , and changed their values within constrained boundaries . These parameters are the model-free learning rates of , respectively , the Gating Network and the Direction strategy—thus 2 parameters among a total of 18 . The neurobiological meaning of such parameters ( inherent to any RL model ) has been investigated [108] , and could account for motivational levels like , for instance , a stress induced by the experiment [109] . Moreover , it is not unreasonable to consider that animals may have changed their learning rates between task conditions [110] . While adding mechanisms to dynamically adapt learning rates based on some measures of the statistics of each task ( such as reward volatility as done in [110] ) would have added unnecessary complexity with respects to the phenomena investigated here—making the interpretation of our results more difficult – , one particularly interesting continuation of this work could consist in modeling neural systems responsible for such task monitoring and motivational effects and their influence on learning rates . A second important limitation of the model is that it does not address the question of which precise model-free learning mechanisms should be employed . Instead , it rather focuses on the comparison at the global level between learning properties of model-based and model-free families of reinforcement learning algorithms [104] . Thus here we have not tested different types of model-free ( MF ) learning algorithms ( e . g . , Q-learning , Actor-Critic , SARSA ) . Comparing these different MF algorithms is particularly important when examining the precise profile of neural activity in different brain regions , as done by Hagai Bergman’s group and Geoff Schoenbaum’s group [111 , 112] ( see [113 , 114] for extensive discussions ) who investigated which of these different algorithms could best explain dopamine neurons’ phasic activity in instrumental learning tasks . Such an analysis goes beyond the present work and extensions of the model would be required to account for this . Nevertheless , in previous work , we have shown that these precise MF learning algorithms do not make very different predictions in terms of behavioral adaptation [115 , 116] , the behavior of animals in such tasks instead appearing to also rely on a more flexible MB learning algorithm . This is why the present study focuses on the comparison between learning algorithms of different natures ( MB , MF , random exploration ) to account for animal behavior . In summary , we presented a new computational model of navigation that successfully reproduced a set of different experiments involving cognitive mapping and associative phenomena during spatial learning . The fact that these experimental results have for a long time been considered contradictory while they could here be accounted for by a unified modular principle for strategy coordination opens a promising line of research to systematically assess computational predictions of this type of modular computational models of navigation . This type of model can also be used to design new experimental protocols and assess new hypotheses about complex behavior arising from the interaction of different navigation strategies . In parallel , such models could contribute in translating important inspiration from animals’ behavioral flexibility to autonomous agents having to display fast adaptation to rapid changes in the environment from a small amount of data , a paradigm which has been called micro-data learning [117] , by opposition to big data learning where the data perimeter is already known in advance . The computational work presented in this manuscript thus highlights the importance of cross-talk between disciplines interested in biological and artificial cognition to contribute to a better understanding of brain and behavior .
The model is sketched in Fig 1 and described in more details in S1 Text .
|
We present a computational model of navigation that successfully reproduces a set of different experiments involving cognitive mapping and associative phenomena during spatial learning . The key ingredients of the model that are responsible for this achievement are ( i ) the coordination of different navigation strategies modeled with different types of learning , namely model-based and model-free reinforcement learning , and ( ii ) the fact that this coordination is adaptive in the sense that the model autonomously finds in each experimental context a suitable way to dynamically activate one strategy after the other in order to best capture experimentally observed animal behavior . We show that the model can reproduce animal performance in a series of classical tasks such as the Morris water maze , both with and without proximal cues , which support the cognitive mapping theory . Moreover , we show that associative phenomena such as generalization gradient and blocking observed within the navigation paradigm cannot be explained by each learning system alone , but rather by their interaction through the proposed coordination mechanism . The fact that these experimental results have for a long time been considered contradictory while they could here be accounted for by a unified modular principle for strategy coordination opens a promising line of research . We also derive model predictions that could be used to design new experimental protocols and assess new hypotheses about complex behavior arising from the interaction of different navigation strategies .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"learning",
"medicine",
"and",
"health",
"sciences",
"brain",
"social",
"sciences",
"neuroscience",
"learning",
"and",
"memory",
"simulation",
"and",
"modeling",
"cognitive",
"psychology",
"animal",
"behavior",
"animal",
"management",
"zoology",
"animal",
"performance",
"research",
"and",
"analysis",
"methods",
"behavior",
"neostriatum",
"agriculture",
"psychology",
"animal",
"migration",
"hippocampus",
"anatomy",
"animal",
"navigation",
"biology",
"and",
"life",
"sciences",
"cognitive",
"science"
] |
2018
|
Interactions of spatial strategies producing generalization gradient and blocking: A computational approach
|
The Drosophila protocadherins Dachsous and Fat regulate growth and tissue polarity by modulating the levels , membrane localization and polarity of the atypical myosin Dachs . Localization to the apical junctional membrane is critical for Dachs function , and the adapter protein Vamana/Dlish and palmitoyl transferase Approximated are required for Dachs membrane localization . However , how Dachs levels are regulated is poorly understood . Here we identify the early girl gene as playing an essential role in Fat signaling by limiting the levels of Dachs protein . early girl mutants display overgrowth of the wings and reduced cross vein spacing , hallmark features of mutations affecting Fat signaling . Genetic experiments reveal that it functions in parallel with Fat to regulate Dachs . early girl encodes an E3 ubiquitin ligase , physically interacts with Dachs , and regulates its protein stability . Concomitant loss of early girl and approximated results in accumulation of Dachs and Vamana in cytoplasmic punctae , suggesting that it also regulates their trafficking to the apical membrane . Our findings establish a crucial role for early girl in Fat signaling , involving regulation of Dachs and Vamana , two key downstream effectors of this pathway .
Precise coordination of growth and morphogenesis during development is critical for formation of organs of correct proportions and optimal function . This is achieved through the cumulative effect of biochemical signaling through morphogens and biomechanical signals mediated through the actomyosin network . The protocadherins Dachsous ( Ds ) and Fat initiate a signaling cascade ( Fat signaling ) , which functions to restrict growth by activating the Hippo signaling pathway [1 , 2] and influences morphogenesis by modulating planar cell polarity ( PCP ) [3–6] . Fat signaling regulates the Hippo pathway and influences PCP by modulating membrane localization of the atypical myosin Dachs . Multiple studies have provided insight both into the mechanisms by which Dachs influences the Hippo pathway and PCP , and into how Fat regulates Dachs [7–18] . However , the mechanisms that control Dachs levels and membrane localization are still not completely understood . Ds and Fat are protocadherins with large extracellular domains ( ECD ) and small intracellular domains ( ICD ) and localize to the apico-lateral membrane just apical to the adherens junctions . The ECDs of Ds and Fat interact with each other across the cell-cell junctions in a heterophilic manner and the Golgi resident kinase , Four-jointed ( Fj ) , modulates this interaction by phosphorylating their ECDs [19–22] . In most developing tissues Ds and Fj are expressed in opposite gradients under the influence of morphogens [2] . The heterophilic interaction between Ds and Fat and the graded expression of Ds and Fj contributes to planar polarized localization of Ds and Fat within cells , where in the developing Drosophila wing Fat is preferentially enriched on the proximal side while Ds is enriched on the distal side [23–25] . Ds and Fat then regulate the levels and polarity of Dachs at cell membranes to influence Hippo signaling and PCP . In absence of Fat , increased amounts of Dachs accumulate around the entire circumference of cells [8 , 24] . Conversely , overexpression of Fat or just the Fat ICD displaces Dachs from the membrane into the cytoplasm [8 , 16] . The Hippo signaling pathway plays a central role in growth regulation and includes Hippo and its cofactor Salvador and Warts ( Wts ) and the Warts cofactor Mats [26] . Hippo phosphorylates and activates Wts , which in turn phosphorylates the transcriptional co-activator Yorkie ( Yki ) . Phosphorylated Yki is sequestered in the cytoplasm . When the pathway is inactive , unphosphorylated Yki translocates into the nucleus , where it associates with Scalloped and induces the transcription of downstream target genes such as expanded ( ex ) , bantam ( ban ) , and cyclin E , which modulate Hippo pathway activity , stimulate growth , and regulate cell fate . Multiple upstream regulators impinge on Wts to regulate Hippo signaling , including Fat signaling . Fat affects the membrane localization of Expanded , the levels of Wts and the interaction of Wts with Mats [7 , 11 , 27–30] . Dachs is required for each of these effects on Hippo signaling . Similarly , Fat signaling influences PCP signaling partly by regulating the Spiny legs isoform of the prickle locus [12 , 13] and partly through an influence on junctional tension [31] , and Dachs is involved in both of these processes . Dachs localizes to the apico-lateral plasma membrane in a planar polarized manner [8] . In the developing wing discs Dachs localizes to the distal side of the cell [8 , 23 , 24 , 32] . Several proteins that influence membrane localization of Dachs have been identified as components of the Fat signaling pathway . vamana ( vam/Dlish ) encodes an adapter protein with three SH3 domains that physically connects Dachs with the Fat and Ds ICDs , interacting with Dachs through its second SH3 domain and interacting with the Fat and Ds ICD through its first and third SH3 domains [14 , 15] . Vam physically interacts with the C-terminal region of Dachs . Like Dachs , Vam localizes to the apical plasma membrane with preferential enrichment on the distal side of wing cells . In absence of Vam , Dachs fails to localize to the membrane and accumulates in the cytoplasm . Approximated encodes a palmitoyl transferase and is also required for Dachs membrane localization [33] . Although , the exact mechanism is still not clear , it can physically interact with both Dachs and Vam and can palmitoylate Vam and the Fat ICD when overexpressed [15 , 34] . Despite progress in our understanding of the Fat signaling pathway , how Dachs localizes to the apical membrane and how its levels at the membrane are regulated is still not completely understood . Here we report the isolation and characterization of the early girl ( elgi ) gene as a key regulator of Fat-Hippo signaling . Loss of elgi function promotes growth by up-regulating Yki activity . This arises from accumulation of high levels of Dachs and Vam at the apical plasma membrane . Elgi physically interacts with Dachs and regulates Dachs and Vam by controlling their protein levels . Further , we show that concomitant loss of Elgi and App prevents Dachs and Vam localization to the subapical membrane and results in their accumulation in cytoplasmic punctae , suggesting that Elgi and App coordinately regulate trafficking and membrane localization of Dachs and Vam . Under this condition , Dachs is primarily affected and Vam gets mislocalized due to its interaction with Dachs . Our observations establish Elgi as an important component of the Fat signaling pathway .
To identify contributions of protein stability to wing growth and patterning , we conducted a RNAi-based genetic screen , in which we depleted the RING domain E3 ubiquitin ligases encoded by the Drosophila genome within the wing pouch region of the developing wing imaginal disc using UAS RNAi lines and a nub-Gal4 driver . From this screen , we identified elgi , based on an increase in size of the adult wing , and a reduced spacing between the anterior and posterior cross veins when it is knocked down by either of two independent RNAi lines ( Fig 1A–1E ) . These phenotypes are hallmark features of mutations in components of the Fat signaling pathway [8 , 14 , 15 , 33 , 35–39] , suggesting that elgi could participate in Fat signaling . elgi was previously characterized for a role in oogenesis , and named for a mutant phenotype including premature meiotic maturation [40] . Loss of function alleles that were isolated based on this oogenesis phenotype , elgi1 and elgi2 , also increase the size of adult wings and reduce spacing between the anterior and posterior cross veins when homozygous , or when hemizygous with a chromosomal deficiency that deletes the elgi locus ( Fig 1F–1K ) . elgi1 mutant wing discs are also larger than control wing discs ( Fig 1L and 1M ) . These observations confirm that elgi regulates wing growth and patterning . Fat signaling regulates growth by controlling Hippo signaling . To examine the possibility that elgi regulates wing size through Hippo signaling , we depleted elgi in the posterior compartment of wing imaginal discs by expressing UAS-RNAi elgi under the control of en-gal4 , and then examined the effect of this Elgi depletion on the levels of the Yki target genes ex and ban . Using ex-lacZ and ban-lacZ reporter genes , we observed that loss of elgi increased Yki activity ( Fig 1N and 1O ) , consistent with the inference that the increased wing size observed is due to impairment of Fat-Hippo signaling . To further examine the possibility that Elgi could be involved in Ds-Fat signaling , and to begin to investigate how it might influence this pathway , we examined the effect of loss of elgi on the key Fat signaling components , Ds , Fat , Dachs and Vam . Depletion of elgi by RNAi did not result in any visible effect on the localization or levels of Ds ( S2 Fig ) . Loss of elgi leads to a slight increase in Fat levels ( Fig 2D ) , but this cannot explain the increased Yki activity and wing growth observed , because over-expression of Fat decreases Yki activity and wing growth [7 , 41] . In contrast , in cells lacking elgi , both Dachs and Vam , examined using genomic GFP-tagged transgenes , accumulate at very high levels ( Fig 2B and 2C , S4A Fig ) . Since over-expression of Dachs or Vam can increase wing size [8 , 14 , 15] , this could potentially account for the increased wing growth associated with loss of elgi . Absence of fat also leads to increased Dachs levels . To compare the increases in Dachs generated by loss of elgi versus loss of fat , elgi or fat were knocked down by RNAi in the posterior compartment of wing discs expressing Dachs:GFP , which were examined using the same imaging settings for both genotypes . This revealed that apical Dachs levels appear higher in the absence of elgi than in the absence of fat ( Fig 2A and 2B ) . However , in the absence of fat Dachs localizes evenly around the cell circumference at the sub-apical membrane , whereas in the absence of elgi Dachs localizes to the membrane in a punctate pattern , similar to its localization in wild-type cells . To examine whether the increase in the levels of Dachs is transcriptional or post-transcriptional , we examined the influence of elgi-RNAi and elgi1 mutant clones on a transgene expressing Dachs:GFP under the control of the Act5c promoter . Act5c-Dachs:GFP accumulated at high levels in the apical cortex in the absence of elgi ( Fig 2E , S4C and S4E Fig ) , just like Dachs:GFP expressed from its endogenous promoter , which indicates that elgi regulates Dachs at a post-transcriptional level . Consistent with this , western blotting of wing disc lysates revealed that loss of elgi leads to a significant increase in the levels of V5-tagged Dachs expressed ubiquitously under the control of a tub-Gal4 driver ( Fig 2F ) . Mutation of upstream components of the Fat signaling pathway , including Fat , Ds , and Dco , increase Yki activity and wing growth by increasing levels of Dachs at the apical membrane [7 , 8 , 24] . To confirm that the increased Yki activity observed with elgi loss-of-function is due to increased Dachs , we investigated whether the elgi phenotype genetically depends upon dachs . Indeed , depletion of dachs by RNAi completely suppressed elevation of ex-lacZ expression by elgi RNAi , and wing disc cells with reduction of both proteins instead exhibit the loss of ex-lacZ expression characteristic of dachs RNAi ( Fig 3A–3C ) . Overexpression of Fat can displace Dachs from the membrane into the cytoplasm [8 , 16] . To examine the genetic relationship between elgi and fat , we investigated the influence of elgi on this regulation of Dachs by Fat . Fat was overexpressed in the absence or in the presence of elgi RNAi in the posterior compartment of wing discs expressing Dachs:GFP . Overexpression of Fat caused a displacement of Dachs from the plasma membrane even in the presence of elgi RNAi ( Fig 3E ) . Thus , Fat over-expression is epistatic to elgi loss of function . Together with observations that elgi RNAi increases Fat levels , and that loss of Elgi or loss of Fat result in different patterns of increased Dachs , this suggests that fat and elgi function in parallel to regulate Dachs . Normally Dachs localizes in a planar polarized manner , with a preferential localization to the distal side of wing disc cells [8] . To examine whether elgi affects Dachs polarity , we expressed Dachs:GFP in small clones in wing discs expressing elgi RNAi in the posterior compartment . While loss of elgi leads to an increase in the levels of Dachs:GFP , it is still mostly polarized to the distal side ( Fig 3F ) . Fat signaling regulates PCP as well as Hippo signaling , and loss of Fat results in Dachs-dependent disruption of hair polarity in the proximal wing [13 , 42] . However , wings from elgi1 mutant flies do not exhibit any defect in hair polarity ( S1 Fig ) . These observations are consistent with the idea that the levels of membrane localized Dachs influence Hippo signaling , whereas the polarity of membrane Dachs influences PCP , and also further support the conclusion that Elgi and Fat act in parallel to regulate Dachs . To further investigate how Elgi regulates Dachs , we examined the localization of Elgi protein . We created a GFP-tagged genomic copy of Elgi , but the extremely weak signal from this construct was insufficient to clearly establish its localization . Therefore we expressed a HA-tagged Elgi in the wing pouch under nub-Gal4 control . Overexpression of Elgi-Myc-HA under UAS-Gal4 control results in a slightly smaller wing size ( Fig 4D and 4F ) , opposite to the effect of elgi loss-of-function , while also causing a very mild decrease in cross vein spacing ( Fig 4C–4G ) . This UAS-elgi-Myc-HA construct could rescue the reduced cross-vein spacing of an elgi mutant ( S3 Fig ) . These observations indicate that it has Elgi activity . Anti-HA antibody staining revealed a dispersed cytoplasmic localization for Elgi ( Fig 4A ) . elgi encodes for a predicted RING domain E3 ubiquitin ligase , homologous to vertebrate RNF41 protein . Since we found that Elgi regulates the levels of Dachs and Vam , we examined whether Elgi physically interacts with them . To test this , we co-expressed Myc-HA-epitope-tagged Elgi along with V5-epitope-tagged full length Vam or Dachs , or N-terminal , middle or C-terminal fragments of Dachs , in S2 cells and conducted co-immunoprecipitation experiments . While Elgi does not interact with Vam , it can co-precipitate both full length Dachs , and the middle region of Dachs ( Fig 4B ) . Since Dachs can influence levels of Vam [14 , 15] , this suggests that elgi could primarily affect Dachs , with the increase in Vam levels arising indirectly from the increase in Dachs levels . The middle region of Dachs is homologous to regions of myosin family proteins . Interestingly , a pulldown experiment identified Myo10A as an interactor of Elgi [43] , suggesting that Elgi interacts with a subset of myosin domains . Since Elgi physically interacts with Dachs , influences Dachs levels , and encodes a predicted ubiquitin ligase , we examined whether it can ubiquitinate Dachs . We used a recently described method that employs biotinylated-Ubiquitin , as it is highly sensitive [44] . Briefly , when avi-tagged Ubiquitin fused to E . coli biotin ligase BirA ( Bio-Ub ) is expressed in cells , biotinylated Ubiquitin is produced , which is then incorporated by the ubiquitination machinery ( Fig 5A ) . We expressed Dachs:V5 in cultured Drosophila S2 cells along with Bio-Ub with or without Elgi , immunoprecipitated Dachs:V5 using a mouse anti-V5 antibody , and then immunoblotted with rabbit anti-V5 to detect Dachs:V5 and with fluorescently-conjugated streptavidin to detect biotinylated-Ubiquitin . As a positive control , we examined Elgi auto-ubiquitination , which was readily detected ( Figs 5B and 6H ) . When Dachs was expressed along with Bio-Ub we detected a smear of biotinylated bands , both at higher molecular weight ( presumably corresponding to poly-ubiquitinated Dachs ) and at lower molecular weight ( presumably corresponding to ubiquitinylated degradation products ) . When Elgi was co-expressed along with Dachs:V5 and Bio-Ub , this signal was not altered , aside from the detection of the co-immunoprecipitated Elgi band . These results suggest that Dachs is ubiquitinated by an E3 ligase present in S2 cells . However , because this signal does not change when Elgi is over-expressed , it appears that Elgi does not ubiquitinate Dachs . We also examined if Dachs ubiquitination in S2 cells depends upon Elgi . Using dsRNA , Elgi could be efficiently depleted in S2 cells ( Fig 5D ) . However , elgi RNAi did not decrease Dachs ubiquitination ( Fig 5C ) . To evaluate the potential contribution of Elgi catalytic activity to Dachs regulation , we created a mutant version of elgi , elgiCS , in which the two cysteines ( C18 and C21 ) in the highly conserved RING domain were mutated to serines . In other E3 ubiquitin ligases , this type of mutation has been reported to abolish catalytic activity [45 , 46] . Consistent with this , while wild type Elgi can auto-ubiquitinate , ElgiCS lacks this activity ( Fig 6H ) . Over-expression of elgiCS results in a significant decrease in cross vein spacing in the adult wing , reminiscent of loss of elgi function ( Fig 4E and 4G ) . However , unlike loss of elgi , it does not lead to an increase in wing size ( Fig 4E and 4F ) . At the cellular level , expression of elgiCS resulted in a dramatic increase in the levels of Dachs ( Fig 6B ) , reminiscent of elgi RNAi , and opposite to the decrease in levels of Dachs observed when wild-type Elgi is over-expressed ( Fig 6A ) . These observations suggest that ElgiCS interferes with endogenous Elgi function . This could result from ElgiCS binding to its substrate and preventing access to wild type Elgi . Alternatively , Elgi could function as a homodimer or heterodimer with another E3 ligase , as many RING-domain ligases are known to function as dimers [47] . To test if Elgi forms homodimers , we coexpressed 3xFlag-tagged Elgi with either V5-tagged Vam or V5-tagged Elgi in S2 cells and immunoprecipitated 3x-Flag tagged Elgi and examined for their interaction . While Elgi:3xFlag did not interact with Vam:V5 it interacted with Elgi:V5 ( Fig 6I ) , indicating that Elgi forms homodimers , which could explain why ElgiCS interferes with endogenous Elgi function . ElgiCS also causes an additional effect on Dachs , as when it is expressed , Dachs localization to the subapical membrane is diminished , and instead it accumulates in bright punctae that are dispersed throughout the cytoplasm ( Fig 6B , S4G and S4H Fig ) . This could explain why unlike elgi RNAi or mutants , animals expressing elgiCS do not display wing overgrowth ( Fig 4E and 4F ) , as Dachs function depends upon its membrane localization . Expression of elgiCS also does not upregulate the Yki target gene , ex-lacZ ( S5A Fig ) . The formation of cytoplasmic puncta is a Dachs-specific effect , as several other proteins including Fat , Ds , Crumbs , Armadillo and Jub that normally localize to the apical cortex are not affected by elgiCS ( S5B–S5F Fig ) . Vam also gets mis-localized with Dachs when elgiCS is expressed ( S5G Fig , Fig 8A ) , but this could be because Vam forms an obligate heterodimer with Dachs [14 , 15] . Since membrane localization of Dachs and Vam is promoted by App [14 , 15 , 33] , the loss of Dachs and Vam from the apical membrane when elgiCS is expressed suggested that it might interfere with App function . We also note that expression of elgiCS causes rounder wings with a mild hair polarity defect in the proximal wing , similar to app loss-of-function ( Fig 4E , S5I Fig ) [33] . If ElgiCS interferes with App , concomitant depletion of elgi and app should cause similar localization defects of Dachs and Vam . To examine this , we compared loss of elgi or app alone with simultaneous knockdown of both genes . Loss of elgi and app together , but not individually , leads to accumulation of Dachs and Vam in cytoplasmic punctae similar to those observed when elgiCS is expressed , with Dachs and Vam co-localized ( Fig 6B–6E ) . If elgiCS interferes with app function , we would also predict that overexpression of App could rescue the mis-localization of Dachs induced by expression of elgiCS . Indeed , over-expression of App restored Dachs membrane localization in cells expressing elgiCS ( Fig 6G ) , indicating that elgiCS interferes with App function . The mechanism by which ElgiCS affects App is unclear . Elgi can bind to App and one of its identified substrates , the Fat-ICD ( Fig 6J ) . However , Elgi does not detectably promote their ubiquitination ( Fig 5B and 5C ) . We were unable to determine the nature of the cytoplasmic accumulations of Dachs and Vam observed when elgi and app are knocked down or ElgiCS is expressed . They fail to colocalize with known markers of subcellular compartments including fast recycling endosomes ( Rab4 ) , early endosomes ( Rab5 ) , late endosomes ( Rab7 ) , recycling endosomes ( Rab11 ) , lysosomes ( LAMP1 ) , multi vesicular bodies ( Hrs ) , mitochondria ( Mito-GFP ) and autophagosomes ( Atg8a-GFP ) ( S6 Fig ) . It is possible that they correspond to unidentified vesicular compartments . Alternatively , they could represent a transient vesicular trafficking structure that accumulates in the absence of Elgi and App . They could arise from coacervation . It is also possible that in absence of App and Elgi , Dachs accumulates in misfolded aggregates . Interestingly , palmitoylation is known to be required for the proper function of a number of chaperones [48 , 49] , but it remains to be examined if App affects any chaperones . Expression of elgiCS or simultaneous depletion of elgi and app led to accumulation of both Dachs and Vam in cytoplasmic punctae ( Figs 6B , 6D , 6E and 8A ) . App can palmitoylate Vam in S2 cells [15] . If App acts through Vam in vivo to regulate Dachs localization , then we would expect simultaneous depletion of elgi and vam to result in a similar mislocalization of Dachs . However , Dachs does not mislocalize to cytoplasmic punctae when elgi and vam are both knocked down ( Fig 7B ) , suggesting that Vam may not be a bona fide substrate of App in vivo . The mislocalization of both Dachs and Vam in the absence of elgi and app could be explained either by Elgi and App directly affecting both Dachs and Vam , or alternatively by only directly affecting one of them , with the other mislocalized due to the physical interaction between Dachs and Vam . To distinguish between these possibilities , we examined whether Vam is required for Dachs mislocalization in absence of elgi and app . Depletion of Vam in cells lacking elgi and app did not affect the mislocalization of Dachs ( Fig 7C and 7D ) , indicating that Vam is not required for Dachs mislocalization . To investigate the potential requirement for Dachs in Vam mislocalization , we took advantage of the observation that the second SH3 domain of Vam is specifically required for its interaction with Dachs [14 , 15] . We examined how deletion of the different SH3 domains of Vam affected its localization in presence of elgiCS . Full length Vam-RFP as well as Vam-RFP lacking either the first ( Vam-ΔSH3-1-RFP ) or third ( Vam-ΔSH3-3-RFP ) SH3 domains were mislocalized along with Dachs:GFP ( Fig 8A , 8B and 8D ) , but Vam-RFP lacking the second SH3 domain ( Vam-ΔSH3-2-RFP ) , which fails to interact with Dachs , did not mislocalize with Dachs:GFP in presence of elgiCS ( Fig 8C ) . Together , these experiments indicate that elgi and app primarily affect Dachs , and that in absence of both elgi and app , Vam gets mislocalized along with Dachs because it physically interacts with Dachs . In addition , they suggest that Vam does not seem to be the substrate of App that regulates Dachs localization . If Vam is affected by Elgi solely through its interaction with Dachs , then Vam that cannot interact with Dachs could be more stable than wild-type Vam . To test this , we expressed either full length Vam-RFP or Vam-RFP lacking one of the SH3 domains , using a ubiquitously expressed tub-Gal4 and UAS-vam transgenes inserted at the same genomic locus . Western Blotting of wing disc lysates revealed that Vam-ΔSH3-2:RFP , which cannot interact with Dachs , is expressed at a higher level than the other Vam constructs ( S7 Fig ) , suggesting that association with Dachs regulates Vam protein stability .
Our results identify the E3 ubiquitin ligase Elgi as playing a crucial role in modulating Fat-Hippo signaling . Elgi regulates the stability of Dachs and , together with App , its membrane localization . Through its effects on Dachs , Elgi also indirectly regulates the levels and localization of Vam . Our observations suggest a model ( Fig 8E ) where , under normal conditions , Elgi maintains limiting levels of Dachs and Vam by facilitating their degradation , and controls their localization in coordination with App . In the absence of Elgi , Dachs and Vam levels increase , but App is still sufficient for their membrane localization . However , when elgiCS is expressed or when elgi and app are simultaneously depleted , the increased levels of Dachs and Vam are unable to localize to the membrane and instead accumulate in punctate structures in the cytoplasm . Due to the impact that membrane levels of Dachs have on Warts activity , Elgi influences the activity of Yki , the key transcription factor of the Hippo pathway . Our observations indicate that Elgi regulates Dachs in parallel to Fat , and the distinct ways in which they regulate Dachs provide insight into how Dachs regulates Hippo signaling and PCP . The increase in Dachs at membranes appears greater when Elgi is knocked down than when Fat is knocked down but the increase in wing growth is greater when Fat is knocked down than when Elgi is knocked down . This might be explained by an influence of elgi on other proteins , which counteract the growth promoted by Dachs . Alternatively , it could be explained by the distinct patterns of Dachs membrane localization in fat versus elgi loss-of-function . In the absence of Fat , Dachs localizes to the membrane around the entire circumference of the cell , whereas in the absence of Elgi , Dachs still remains polarized . Investigations of Hippo signaling have revealed that Warts is activated in discrete complexes at the sub-apical membrane , where multiple upstream components co-localize [50 , 51] . We suggest that when Dachs is membrane localized around the entire cell circumference , it is able to broadly disrupt Warts activation , whereas when Dachs is polarized , there could be regions of the sub-apical membrane where Warts activation can occur normally , resulting in a lower overall level of Yki activity than can be achieved with uniform Dachs membrane localization . The normal polarization of Dachs , together with the normal hair polarity in elgi mutants , is consistent with the conclusion that Fat signaling regulates PCP through the polarization of Dachs localization . Thus Elgi effectively separates the two distinct downstream branches of Fat signaling–it disrupts Hippo signaling , but leaves PCP unaffected . Elgi is a homolog of mammalian RNF41/NRDP1 , which regulates ErbB3 and ErbB4 receptors , BRUCE , BIRC6 and Parkin [52–54] . RNF41 also regulates trafficking of certain JAK2-associated type1 receptors [55] . Moreover , it can regulate PCP by ubiquitinating Dishevelled ( DVL ) [56] . Interestingly , RNF41 physically interacts with VANGL2 but does not ubiquitinate it [56] . Rather it ubiquitinates DVL , which is associated with VANGL2 . RNF41 also stabilizes AP2S1 in an E3 ligase activity-independent manner [57] . Although Elgi can physically interact with Dachs , App and Fat-ICD , we were unable to detect direct ubiquitination of these proteins by Elgi . Thus Elgi might regulate Dachs stability through an unknown protein . Alternatively , it is possible that Elgi might function as an adapter to facilitate proteosomal or endolysosomal degradation of Dachs . For example , Elgi might directly link Dachs ( or a Dachs-Vam complex ) to the proteasome . Some E3 ligases , such as Stuxnet , have been reported to regulate protein degradation without ubiquitination by acting in this way [58] . If Elgi plays a role in vesicular trafficking , it might not only promote trafficking of Dachs ( or a Dachs-Vam complex ) to the lysosome , it could also affect the localization and levels of other proteins . The increase in Fat levels in elgi mutant cells is consistent with this possibility . Dachs is also regulated by FbxL7 [16 , 17] , which like Elgi is predicted to be a ubiquitin ligase . It is also not clear if FbxL7 regulates Dachs by directly ubiquitinating it or instead by affecting vesicular trafficking . However , FbxL7 differs from Elgi in that FbxL7 regulates both the levels and the polarity of Dachs , and in that FbxL7 itself exhibits polarized membrane localization . The presence of multiple mechanisms to limit amounts of Dachs emphasize the importance of strict control over Dachs levels . The palmitoyltransferase App is critical for membrane localization of Dachs [33] . Although App can palmitoylate Vam and the Fat ICD when overexpressed [15 , 34] , the mechanism by which it regulates Dachs localization in vivo is still not clear . Our studies revealed that concomitant loss of Elgi and App leads to accumulation of Dachs and Vam in cytoplasmic punctae . The detection of these puncta might reflect a role for App in trafficking of Dachs and Vam . In the presence of wild-type Elgi , the low levels of Dachs could preclude detection of this trafficking defect , but the dramatic increase in Dachs levels in the absence of elgi could exacerbate accumulation of Dachs/Vam in vesicles and make them easier to detect . Consistent with this possibility , basal puncta of Dachs were detected when Elgi was depleted and Dachs was expressed using the act5C promoter . Alternatively , Elgi and App might impair distinct activities that are redundantly required for trafficking of Dachs and Vam . Palmitoylation is well known to play a key role in vesicular trafficking [59] . Further characterization of the nature of these cytoplasmic accumulations of Dachs could help understand the exact mechanism by which App regulates Dachs localization . Our genetic experiments also suggest that Vam may not be a relevant substrate of App in vivo , as concomitant loss of vam and elgi does not result in the accumulations of Dachs observed in the absence of elgi and app . This is consistent with the finding that mutating a predicted palmitoylation site within Vam does not abrogate its membrane localization [14] . Our experiments also revealed that elgi and app primarily affect Dachs , and only indirectly affect Vam through its physical association with Dachs . This is consistent with the observation that Elgi associates with Dachs but not with Vam . Interestingly , Vam lacking the second SH3 domain is expressed at a higher level compared to the full length Vam or Vam lacking the first or third SH3 domains . This suggests that Elgi-mediated degradation normally maintains Vam at low levels by acting on the Dachs-Vam heterodimer .
The following previously described alleles and transgenes were used , elgi1 , elgi2 [40] , Df ( 3L ) BSC575 ( BL27587 ) , en-Gal4 ( BL30564 ) , nub-Gal4 ( BL25754 ) , tub-Gal4 , UAS-dcr2 , ex-LacZ ( BL44248 ) , ban-lacZ ( BL10154 ) , UAS-mCD8:RFP ( gift of G . Morata , Universidad Autónoma de Madrid , Madrid ) , Dachs:GFP [31] , Ds:GFP [21] , UAS-Dachs:V5 [8] , UAS-Fat [41] , UAS-vam-RNAi ( BL38263 ) , UAS-fat-RNAi ( VDRC 9396 ) , UAS-d-RNAi ( VDRC 12555 ) , UAS-app , UAS-app-RNAi [33] , Vam:GFP , UAS-vam-3x-FLAG-RFP , UAS-vam-ΔSH3-1-3x-FLAG-RFP , UAS-vam-ΔSH3-2-3x-FLAG-RFP , UAS-vam-ΔSH3-3-3x-FLAG-RFP and ACT>Stop>Vam-3x-FLAG-GFP [14] , Jub:GFP [60] , UAS-D:RFP , UAS-elgi RNAi ( VDRC109617 , NIG 17033R-3 ) UAS-Mito-GFP ( BL8442 ) , UAS-atg8a-GFP ( 52005 ) , LAMP1-YFP ( DGGR 115517 ) , Rab4-YFP ( BL62542 ) , Rab5-YFP ( BL62543 ) , Rab7-YFP ( BL62545 ) , Rab11-YFP ( BL62549 ) . The transgenes UAS-elgi-Myc-HA and UAS-elgiCS-Myc-HA were created in this study . Dissected wing discs were fixed in 4% paraformaldehyde for 10 minutes at room temperature and stained with primary antibodies , rat anti-E-cad ( 1:400 , DSHB DCAD2 ) , rat-anti-Fat ( 1:4000 ) ( Feng and Irvine , 2009 ) , rat-anti-Ds ( 1: 5000 ) ( Ma et al . , 2003 ) , mouse-anti-Crb ( 1:200 , DSHB ) , mouse-anti-Arm ( 1:200 , DSHB ) mouse anti-β-gal ( 1:400 , DSHB JIE7-c ) , Guinea pig anti-Hrs ( gift from Hugo Bellen ) ( 1:200 ) ; and secondary antibodies , donkey anti-rat-647 ( 1:100 , Jackson , 712-605-150 ) and donkey anti-rat-Cy3 ( 3:400 , Jackson , 712-165-150 ) . GFP and RFP were detected by autofluorescence . Confocal images were captured using a Leica SP8 confocal microscope . To create pUAST-elgi-Myc-HA , elgi cDNA was PCR amplified with elgi-forward 5’TGAATAGGGAATTGGGAATTCATGGGCTACGATGTGAATCGCTT3’ and elgi-MYC-HA-rev CGCAAGATCTGTTAACGAATTCCTAAGCGTAATCTGGAACATCGTATGGGTACAGATCCTCTTCTGAGATGAGTTTTTGTTCCTCGATGCCATGGGCGAAG3’ primers and cloned into pUAST plasmid digested with EcoRI by Gibson assembly . To create pUAST-elgiCS-Myc-HA , elgi cDNA was PCR amplified with elgiCS forward primer , 5’TGAATAGGGAATTGGGAATTCATGGGCTACGATGTGAATCGCTTTCAGGGGGAGGTGGACGAGGAGCTCACCTCTCCCATCTCCTCCGGAGTGCTT 5’ and elgi-MYC-HA-rev primer and cloned into pUAST plasmid digested with EcoRI by Gibson assembly . All plasmids were sequence verified . pUAST-elgi-V5 and pUAST-elgi-3xFLAG were similarly created using the elgi-forward primer and elgi-V5 reverse 5’GCAAGATCTGTTAACGAATTCCTAGGTGCTGTCCAGGCCCAGCAGGGGGTTGGGGATGGGCTTGCCCTCGATGCCATGGGCGAAG 3’ and elgi-FLAG-reverse 5’ GCAAGATCTGTTAACAATTCTTACTTGTCATCGTCATCCTTGTAATCGATGTCATGATCTTTATAATCACCGTCATGGTCTTTGTAGTCCTCGATGCCATGGGCGAAG primers respectively . pACU2-app:V5 was created by amplifying the app CDNA using forward 5’gttcaattacagctcgaattcATGAATCTGCTGTGCTGCTGTTG 3’ and reverse primer 5’ cgcagatctgttaacgaattcTTAGGTGCTGTCCAGGCCCAGCAGGGGGTTGGGGATGGGCTTGCCGACTATGGCCACGTTTGTGGTC and and cloned into pACU2 plasmid digested with EcoRI , by Gibson assembly . To create pUAST-Vam-V5 , Vam cDNA was amplified with Vam-for 5’TGAATAGGGAATTGGGAATTCATGGCATTTCTTTGCCCCGT3’ and Vam-V5-Rev 5’GCAAGATCTGTTAACGAATTCTTACGTAGAATCGAGACCGAGGAGAGGGTTAGGGATAGGCTTACCAAGGCTGGTCATCGCGGGTGGT3’primers and cloned into pUAST plasmid digested with EcoRI , by Gibson assembly . To create pUAST-D-RFP , Dachs cDNA was amplified with D-for 5’TGAATAGGGAATTGGGAATTCATGTTGACTACGACGATCTGGACAG 3’ and D-Rev 5’ GCTCTTCGCCCTTAGACACCATTTTACTGAGCGTCATGAACTGGAAGG 3’ primers . Tag-RFP was amplified with RFP-For 5’ ATGGTGTCTAAGGGCGAAGAG and RFP-Rev 5’ TCCTTCACAAAGATCCTCTAGATCAATTAAGTTTGTGCCCCAGTTTGC3 and cloned into pUAST plasmid digested with EcoRI and XbaI by Gibson assembly . All plasmids were verified by sequencing . Small fragments of elgi or EGFP coding sequences were selected using SnapDragon tool to avoid any off-target effect . For generating the template for dsRNA synthesis for elgi , elgi T7 forward 5’TAATACGACTCACTATAGGGGGATGCATTAACGAGTGGCTAACC 3’ and T7 Reverse 5’ TAATACGACTCACTATAGGGATAGTCAGCTGCTGGTCGGTG 3’ primers were used . For dsRNA synthesis for EGFP , a fragment was amplified with the EGFP-T7 forward 5’TAATACGACTCACTATAGGGACGTAAACGGCCACAAGTTC 3’ and EGFP-T7 reverse 5’ TAATACGACTCACTATAGGGTGTTCTGCTGGTAGTGGTCG 3’ primers . dsRNA synthesis was carried out using MEGAscript T7 transcription Kit ( Invitrogen ) , following manufacturer’s instruction . After synthesis the dsRNA was purified using RNAeasy mini Kit ( Quiagen ) . S2 cells were grown in Schneider’s media and transfected with plasmids aw-GAL4 , pUAST-D-:V5 [8] pUAST-D-N:V5 , pUAST-D-M:V5 , pUAST-D-C:V5 [10] , pUAST-Vam-V5 [14] along with pUAST-elgi-Myc-HA , using effectene transfection reagent ( Quiagen ) following the manufacturer’s instruction . Cells were lysed in RIPA buffer supplemented with protease inhibitor cocktail and phosphatase inhibitor cocktail ( Calbiochem ) . Cell lysates were incubated with affinity resin conjugated with anti HA antibody ( Sigma ) for two hours at 4 °C , following which they were washed four times . Proteins samples were boiled with Laemlii buffer and subjected to SDS-PAGE using 4–15% gradient gels ( Bio Rad ) . Western transfer was carried out using Trans-Blot Turbo transfer system ( Bio Rad ) and immunoblotting was performed using the primary antibodies , rabbit-anti-V5 ( Bethyl Laboratories , 1:5000 ) , mouse anti-V5 ( Invitrogen , 1:5000 ) , mouse-anti-GAPDH ( Covance , 1:1000 ) , mouse anti α-Tubulin ( 1:20 , 000 ) , mouse anti FLAG ( Sigma , 1:5000 ) Rabbit anti FLAG ( Sigma , 1:5000 ) and Guinea Pig anti App ( gift from Seth Blair , 1:5000 ) ; and as secondary antibodies goat anti-mouse 680 ( Li-Cor , 926–68 , 020 ) , goat anti-rabbit-680 ( Li-Cor , 926–68 , 021 ) and Donkey anti-Gunea Pig-780 ( Li-Cor , 926–32411 ) , all at a dilution of 1:10 , 000 . Blots were scanned using an Odyssey Imaging System ( Li-Cor ) . Co-immunoprecipitation experiments in Fig 6I and 6J were performed following similar protocols where S2 cells were transfected with aw-Gal4 , pUAST-Elgi-3x-FLAG together with either pUAST-Vam-V5 or pUAST-Elgi-V5 ( Fig 6I ) or pACU2-App-V5 or pUAST-Fat-ICD-V5 ( Fig 6J ) . Elgi-3x-Flag was immunoprecipitated using EZ view Red FLAG M2 affinity Gel and immunoblotted with Mouse anti-FLAG and Rabbit-anti-V5 primary antibodies and donkey anti-mouse 680 and donkey anti-rabbit 780 secondary antibodies . Ubiquitination assays were carried out as previously described [44] . Briefly , S2 cells were transfected with either pUAST-Vam:V5 , pACU2-App:V5 , pUAST-Fat-ICD:V5 or pUAST-D:V5 with or without pUAST-BioUb-BirA and pUAST-elgi-3x-FLAG along with pmt-Gal4 . After incubation for 72 hours , protein expression was induced by adding 0 . 5mM CuSO4 . In all ubiquitination experiments , to prevent degradation of the ubiquitinated products , cells were treated with the proteasome inhibitor MG132 ( 10uM ) at the same time as CuSO4 was added . After overnight incubation , cells were lysed in RIPA buffer supplemented with protease inhibitor cocktail ( Roche ) and phosphatase inhibitor cocktail ( Calbiochem ) and 10mM N-ethyl Maleimide ( deubiquitinase inhibitor ) . Cell lysates were incubated with affinity resin conjugated with mouse anti-V5 antibody ( Sigma ) for two hours at 4 °C , following which they were washed four times . After washing , samples were boiled in Laemmli buffer and subjected to SDS-PAGE followed by Western transfer . Immunoblotting was performed using rabbit anti-FLAG and rabbit anti-V5 primary antibodies and goat anti-rabbit-800 ( Licor ) secondary antibody . IRDye-680RD-Streptavidin ( Licor 925–68079 ) was used at 1:10000 dilution to detect biotin , and the blots were scanned using an Odyssey imaging system ( LiCor ) . Elgi autoubiquitination was performed following a similar protocol , where S2 cells were transfected with pUAST-Elgi-3x-Flag or pUAST-ElgiCS-3x-Flag along with pUAST-BirA or pUAST-BioUb-BirA and Elgi-3x-Flag or ElgiCS-3x-Flag was immunoprecipitated using EZ view Red FLAG M2 affinity Gel ( Sigma ) and immunoblotted with Rabbit-anti-FLAG ( Sigma ) and above described secondary antibody and streptavidin . For RNAi , cells were resuspended in serum free Schneider’s medium at a concentration of 1-4x105 cells/ml and to 1ml of of cell suspension , 30μg of either EGFP-shRNA or elgi-shRNA was added and incubated for 30 minutes , following which 1ml of fresh medium containing 20% serum was added . Immediately , they were transfected with either pUAST-Vam:V5 , pACU2-App:V5 , pUAST-Dachs:V5 , or pUAST-Fat-ICD:V5 plasmid along with the pMT-Gal4 and pUAST-BirA-bioUB plasmids , using Effectene transfection reagent ( Quiagen ) following manufacturer’s protocol . After incubation for 72 hours , protein expression was induced by adding 0 . 5mM CuSO4 , in the presence of 10uM MG132 . After overnight inductions the cells were collected and the proteins of interest were immunoprecipitated and processed as described above .
|
During development , organs grow to achieve a consistent final size . The evolutionarily conserved Hippo signaling network plays a central role in organ size control , and when dysregulated can be associated with cancer and other diseases . Fat signaling is one of several upstream pathways that impinge on Hippo signaling to regulate organ growth . We describe here identification of the Drosophila early girl gene as a new component of the Fat signaling pathway . We show that Early girl controls Fat signaling by regulating the levels of the Dachs protein . However Early girl differs from other Fat signaling regulators in that it doesn’t influence planar cell polarity or control the polarity of Dachs localization . early girl encodes a conserved protein that is predicted to influence protein stability , and it can physically associate with Dachs . We also discovered that Early girl acts together with another protein , called Approximated , to regulate the sub-cellular localization of Dachs and a Dachs-interacting protein called Vamana . Altogether , our observations establish Early girl as an essential component of Fat signaling that acts to regulate the levels and localization of Dachs and Vamana .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"rna",
"interference",
"developmental",
"biology",
"immunoprecipitation",
"epigenetics",
"co-immunoprecipitation",
"morphogenesis",
"cellular",
"structures",
"and",
"organelles",
"research",
"and",
"analysis",
"methods",
"lipids",
"genetic",
"interference",
"proteins",
"fats",
"gene",
"expression",
"ubiquitination",
"imaginal",
"discs",
"cell",
"membranes",
"precipitation",
"techniques",
"cytoplasm",
"biochemistry",
"rna",
"cell",
"biology",
"nucleic",
"acids",
"post-translational",
"modification",
"genetics",
"biology",
"and",
"life",
"sciences"
] |
2019
|
Early girl is a novel component of the Fat signaling pathway
|
RecQ helicases , including Saccharomyces cerevisiae Sgs1p and the human Werner syndrome protein , are important for telomere maintenance in cells lacking telomerase activity . How maintenance is accomplished is only partly understood , although there is evidence that RecQ helicases function in telomere replication and recombination . Here we use two-dimensional gel electrophoresis ( 2DGE ) and telomere sequence analysis to explore why cells lacking telomerase and Sgs1p ( tlc1 sgs1 mutants ) senesce more rapidly than tlc1 mutants with functional Sgs1p . We find that apparent X-shaped structures accumulate at telomeres in senescing tlc1 sgs1 mutants in a RAD52- and RAD53-dependent fashion . The X-structures are neither Holliday junctions nor convergent replication forks , but instead may be recombination intermediates related to hemicatenanes . Direct sequencing of examples of telomere I-L in senescing cells reveals a reduced recombination frequency in tlc1 sgs1 compared with tlc1 mutants , indicating that Sgs1p is needed for tlc1 mutants to complete telomere recombination . The reduction in recombinants is most prominent at longer telomeres , consistent with a requirement for Sgs1p to generate viable progeny following telomere recombination . We therefore suggest that Sgs1p may be required for efficient resolution of telomere recombination intermediates , and that resolution failure contributes to the premature senescence of tlc1 sgs1 mutants .
Telomeres are critical for genome stability and normal cell physiology because they cap the ends of chromosomes; if uncapped , telomeres behave as DNA breaks and thus elicit damage responses and are subject to nucleolytic degradation and recombination [1 , 2] . Capping depends on telomere architecture , which is mediated by chromatin factors , and on telomere length . The enzyme telomerase can counteract the shortening of telomeres that accompanies DNA replication or DNA damage , but dividing cells lacking sufficient telomerase can develop critically short , uncapped telomeres that signal cell cycle arrest ( cell senescence ) or death . Some cells bypass these barriers by up-regulating telomerase expression and thus elongating telomeres . In other cases , bypass involves the use of recombination to maintain telomere length . Examples of the latter case are so-called “survivors” of telomerase deletion in Saccharomyces cerevisiae and alternative lengthening of telomeres ( ALT ) cells in mammals [3 , 4] . A growing number of proteins are recognized as participating in telomere maintenance [2] . Among these are members of the RecQ family of DNA helicases [5] , including the human Werner syndrome ( WS ) and Bloom syndrome proteins ( WRN and BLM , respectively ) and S . cerevisiae Sgs1p . Deficiencies in these helicases lead to genome instability caused by defects in recombinational repair of DNA damage , replication fork stability , and checkpoint signaling , and can lead to the premature onset of cancer and age-related pathologies [5 , 6] . The precise mechanisms by which RecQ helicases help maintain telomeres are not yet clear , but there is evidence that they are important for telomere replication , repair , and recombination [7–18] . A well-characterized function of RecQ helicases throughout the genome is the regulation of homologous recombination , by which they facilitate resolution of recombination intermediates and perhaps avoid the initiation of inappropriate recombination events [5] . Yeast survivors of telomerase deletion and mammalian ALT cells are two settings in which RecQ helicases are important in recombination-dependent telomere maintenance . For example , Sgs1p is required for emergence of type II survivors , which depend on recombination among telomere repeat sequences [15–17]; the Schizosaccharomyces pombe RecQ homolog SPAC212 . 11 similarly facilitates survivor emergence [7] , and WRN regulates the generation of ALT cells from murine telomerase knockout cells [19] . In addition to their roles in survivors and in ALT cells , RecQ helicases function in telomere maintenance in primary cells that have little or no telomerase activity . For example , human WS fibroblasts suffer occasional complete loss of a telomere , which occurs preferentially at the guanine-rich telomere strand , which is replicated by lagging-strand synthesis [11 , 20] . These loss events presumably contribute to the premature senescence of cultured WS cells and their arrest at longer mean telomere lengths than control cells [21]; even though the shortening of most telomeres may be normal in WS cells , the increased frequency of occasional and critically shortened telomeres could signal senescence . Further , mutations in Wrn or Blm synergize with short telomeres in telomerase knockout mice to cause several degenerative pathologies , indicating that the helicases play important roles in telomere maintenance [10 , 12] . And in yeast , although sgs1 mutants maintain telomeres of normal length in the presence of telomerase , tlc1 sgs1 mutants senesce faster than tlc1 mutants [15 , 17] . The rapid senescence of tlc1 sgs1 mutants is due to an increased propensity of cells lacking Sgs1p to suffer G2/M arrest at a given average extent of telomere shortening; this suggests a role for Sgs1p in the repair of rare , critically shortened telomeres that would otherwise be repairable by telomerase if it were active . We recently described evidence that Sgs1p uses recombination functions to maintain telomeres during senescence , similar to its role in survivors of senescence . In particular , Sgs1p was shown to function in a RAD52-dependent pathway during senescence , and to use known recombination functions , including helicase activity and cooperation with topoisomerase III [22] . Here we use more direct methods , including 2DGE and sequence analysis of telomeres from a single chromosome end to investigate telomere maintenance in tlc1 and tlc1 sgs1 mutants . Our findings indicate that Sgs1p is required for efficient resolution of recombining telomeres during senescence . Moreover , telomere sequence analysis suggests that cells that have entered into telomere recombination require Sgs1p to give rise to viable progeny . These findings emphasize the importance of RecQ helicase recombination functions in telomere maintenance during senescence , and suggest that failure of recombination functions contributes to rapid senescence in cells with mutations in RecQ helicases .
Nondenaturing 2DGE and Southern analysis were used to examine possible changes in telomere replication or recombination caused by sgs1 and tlc1 mutations . Each S . cerevisiae telomere contains from zero to four tandem copies of a subtelomeric element called Y′ [23] . Thus Y′ elements may be “terminal” and followed only by telomere repeat sequences , or “internal” and followed by at least one Y′ element prior to the telomere terminus . There are two Y′ element size classes ( Y′ long [Y′L] and Y′ short [Y′S] ) , with the longer , Y′L , containing a 1 . 5-kilobase ( Kb ) element absent from Y′S ( Figure 1A ) . To simplify analysis , and because in our yeast strain Y′L is the predominant class and most Y′L elements occupy a terminal position , a Y′L-specific probe was used to visualize telomeres . Replication of Y′ and telomere repeat DNA was observed predominantly in a Y-arc pattern indicating that most replication of the terminal Y′ elements derives from origins centromeric to the terminal Y′ elements ( Figure 1B ) . No change in the replication pattern was observed in the sgs1 mutant , consistent with a previous report [24] , nor in senescing tlc1 or tlc1 sgs1 mutants ( Figure 1B ) . In particular , there was no accumulation of replication forks in the telomere repeat DNA near the end of the Y arc ( 2N spot ) , indicating that , at least under typical circumstances , SGS1 is not required for fork progression through telomere repeat sequences . A “spike” extending upward from the 2N spot was visible ( Figure 1B ) ; species running in this position have been attributed to X-shaped structures , including convergent replication forks and various recombination intermediates ( see below ) [25–27] . The intensity of the spike ( normalized to the intensity of the replication arc ) did not increase significantly in sgs1 compared to wild-type cells ( Figure 1B and 1D ) . However , the spike intensity increased significantly as telomeres shortened in the tlc1 mutant ( Figure 1B and 1C ) . This is consistent with observations in Kluyveromyces lactis that telomere recombination rates increase at short telomeres [28] . Moreover , the spike intensity was 50% higher in tlc1 sgs1 compared to tlc1 cells at equivalent population doublings ( PD ) after loss of telomerase ( p < 0 . 025; Figure 1D ) . Similar results were also obtained with a probe that detected all Y′ elements ( Supplementary Figure 1 in Protocol S1 ) . To address the specificity of the increased X-structures at telomeres , X-structure levels were examined in the ribosomal DNA ( rDNA ) . Although rDNA X-structures were elevated by sgs1 mutation , consistent with elevated rDNA recombination and higher X-structure levels in sgs1 mutants in earlier reports [29 , 30] , telomere shortening caused by tlc1 mutation caused no increase in levels ( Figure 2 ) , indicating specific accumulation at telomeres with senescence . X-structures at several genomic loci have been found to accumulate in late S or G2/M phases of the cell cycle . Because senescent telomerase mutants similarly accumulate in the G2/M phase of the cell cycle [15 , 31 , 32] , this raised the possibility that elevated telomere X-structures in senescent cells might simply reflect a cell cycle effect . However , the lack of increased rDNA X-structures with senescence indicates that this is not the case . Similarly , the lack of increased telomere X-structures in sgs1 mutants indicates that perturbations in cell cycle progression reported by some investigators in these mutants [29 , 33–35] do not explain the increases that occur during senescence . Rather , elevated telomere X-structures are somehow related specifically to telomere shortening , and Sgs1p plays a role in attenuating the accumulation of telomere X-structures in this context . Such a role of Sgs1p might be to prevent the formation of telomere recombination intermediates or to facilitate their resolution . We considered several possible identities for the X-structures , including Holliday junctions ( HJs; and structurally similar regressed replication fork or “chicken-foot” structures ) , convergent replication forks , hemicatenanes ( HC ) , and so-called rec-X structures ( Figure 3A ) . HC have an interlink between single strands from two duplexes , and such a link forms in a RAD52-independent fashion between sister chromatids behind advancing replication forks , whereas rec-Xs are thought to form in a RAD52-dependent fashion at stalled forks when one nascent strand switches its template transiently for the other nascent strand and then returns to the original template ( see also Figure 7A ) [27 , 36–38] . The different types of X-structures can be distinguished based on their susceptibility to cleavage by HJ resolvases such as RuvC , their ability to branch migrate , and their dependence on RAD52 . To characterize the X-structures , we first tested whether they could be cleaved by the HJ-specific resolvase RuvC [39] . RuvC ( 350 ng ) did not cleave telomere X-structures from tlc1 sgs1 cells , despite fully cleaving a synthetic X-junction added to the same reactions containing yeast DNA ( Figure 3B , top left and right , and bottom right panels , and unpublished data ) . Five-fold higher levels of RuvC ( 1 . 75 μg ) partially degraded both the replication arc and X-structures , but had no preferential effect on the X-structures ( Figure 3B , bottom left panel ) . Therefore the X-structures are apparently not HJs . Incubation of gel slices from the first dimension in branch migration buffer at 65 °C prior to running the second dimension caused selective loss of the spike and retention of the replication arc in tlc1 and tlc1 sgs1 mutants ( Figure 3C ) . The X-structures therefore cannot be convergent forks , because replication forks do not branch migrate [25] , thus leaving HC or rec-Xs as possibilities because they do branch migrate [27 , 36] . Further , the X-structures branch migrated in the presence of Mg+2 , consistent with a HC or rec-X , but not HJ , identity ( unpublished data ) [27] . To test RAD52 dependence of the X-structures , we first re-examined the effect of rad52 mutation on senescence of tlc1 and tlc1 sgs1 mutants . This was done because our previous studies used a rad52 disruption allele beginning after amino acid 167 [22 , 40] , potentially allowing for expression of the N-terminus . Because N-terminal fragments of Rad52p have been shown to be sufficient for DNA binding and strand-annealing activities in vitro [41 , 42] , this raised the possibility that the disruption allele might have residual activity . However , a full-deletion rad52 allele yielded the same results as the disruption allele: rad52 deletion sped senescence of tlc1 mutants , and additional sgs1 mutation had no further effect on the rate of senescence ( Figure 4A ) , confirming that RAD52 is epistatic to SGS1 during senescence . As reported previously [21 , 43] , RAD52 was required for survivor formation in tlc1 mutants , because they depend on recombination . Curiously , at later times than those shown , tlc1 sgs1 Δrad52 cells were actually able to form poorly growing survivors , as will be described elsewhere ( J . Y . Lee and F . B . Johnson , unpublished data ) . Comparison on 2DGE of PD-matched cultures of tlc1 sgs1 with tlc1 sgs1 Δrad52 cells showed a significant 47% reduction , but not elimination of telomere X-structures caused by the rad52 deletion ( Figure 4B and unpublished data ) . RAD52 deletion also had no effect on X-structures in TLC1+ and early generation tlc1 mutants prior to significant rise in X-structure levels ( Supplementary Figure 2 in Protocol S1 , and unpublished data ) . We conclude that the elevated level , but not the basal level , of X-structures in tlc1 sgs1 mutants is dependent on RAD52 . Rec-X formation is RAD52-dependent , whereas HC formation is not [27 , 36] , and so , as detailed in the Discussion , a reasonable interpretation of the data is that the basal X-structures are HC and the elevated levels represent rec-Xs formed secondary to stalled replication at telomeres . To further test the idea that the elevated level of X structures might correspond to rec-Xs , we examined the dependence of the telomere X-structures on RAD53 . At stalled replication forks , Rad53p functions to stabilize the replisome , signal the intra-S phase checkpoint , and facilitate the resumption of replication [6 , 44] . The kinase-defective rad53K227A allele has been shown to lead to the loss of rec-X structures at stalled forks , including suppression of elevated rec-X levels in sgs1 mutants treated with methane methyl sulfonate [27 , 36] . tlc1 rad53K227A mutants senesced faster than tlc1 mutants ( Figure 5A ) , consistent with a role for Rad53p in preventing replication fork collapse at telomeres and thus premature senescence in the absence of telomerase . Further , tlc1 sgs1 rad53K227A mutants did not senescence significantly faster than tlc1 sgs1 mutants , as expected if Sgs1p and Rad53p function in the same pathway to prevent rapid senescence . tlc1 sgs1 rad53K227A mutants showed a significant 2-fold reduction in X-structure levels compared with PD-matched tlc1 sgs1 mutants ( Figure 5B and 5C ) , supporting the hypothesis that rec-Xs account for the elevated X-structure levels . The greater accumulation of X-structures in tlc1 sgs1 mutants could reflect either increased formation or reduced resolution of recombination intermediates in the absence of Sgs1p . To address these alternatives , and as a second test for telomere recombination during senescence , we used telomere PCR [45] to clone and sequence examples of telomeres from a single chromosome end in the populations of senescing cells . The yeast telomere repeat is imperfect , with an approximate consensus of ( TG1–3 ) n , and individual telomeres thus differ in their precise sequence . Telomerase adds new and variable sequences to telomere ends , but in tlc1 mutants , the sequence is fixed and typically does not change as a telomere shortens . However , occasional recombination events append new sequences at telomere ends during senescence [46] . If Sgs1p resolves telomere recombination intermediates and thus allows continued cell division , telomere recombinants should be reduced in tlc1 sgs1 mutants ( because these cells would arrest and yield no progeny ) , whereas if it inhibits the formation of telomere recombination intermediates , an increase in recombinants would be predicted ( because recombination events would be more frequent and recombinant cells would yield progeny ) . Genomic DNA was isolated from tlc1 and tlc1 sgs1 mutants at 44 PD after loss of telomerase , far in advance of survivor formation . DNA ends were polyC tailed , and then amplified by PCR using a polyG primer specific to the tail and one specific to sequences internal to the telomere on the left arm of Chromosome I ( telomere I-L ) . Sequence analysis of 434 tlc1 and 439 tlc1 sgs1 cloned products showed that , as expected , the majority of the G-rich strands from telomere I-L had a distribution of lengths , which ranged from 8–178 nucleotides ( nt ) , but were identical in sequence over their shared lengths ( Figure 6A ) . The different lengths in the population of senescing cells reflect the stochastic natures of senescence and telomere shortening , as observed previously [46] . Analysis of telomere sequences that diverged from the original showed that 5 . 1% ( 22/434 ) of sequences in tlc1 mutants were recombinant , similar to the 6 . 6% reported previously [46] . There was a trend toward fewer recombinants in tlc1 sgs1 mutants , with recombinants accounting for only 3 . 9% ( 17/439 ) of telomeres ( p = 0 . 196 ) . Recombinants included apparent cases of inter-telomeric recombination as well as intra-telomere recombination resulting in duplication of telomere sequences ( Figure 6B ) . Importantly , there was a significant difference in the distribution of recombinants as a function of the length of the nonrecombined telomere repeat tract , with recombinant telomeres occurring at a significantly reduced frequency at longer telomeres in the tlc1 sgs1 mutants ( Figure 6A ) . Comparing telomeres with nonrecombinant tract lengths of greater than 85 nt , there was a 2 . 6-fold reduction in the frequency of recombinant telomeres in the double mutants ( 14/283 vs . 5/264; p = 0 . 026 ) . As detailed in the Discussion , these results can be explained if Sgs1p facilitates the resolution of telomere recombination intermediates and if unresolved intermediates in tlc1 sgs1 mutants lead to cell cycle arrest , thus inhibiting the accumulation of progeny bearing recombinant telomeres .
Previous studies have indicated that RecQ helicases , including the human WRN and BLM and S . cerevisiae Sgs1p , are important in the maintenance of telomeres . Here , we have used physical and genetic methods to investigate how Sgs1p slows the rate of senescence in yeast tlc1 mutants . Our new findings indicate a role for Sgs1p in the resolution of telomere recombination intermediates and lend mechanistic insight to earlier observations of the function of Sgs1p during senescence [15 , 17 , 22] . Further , they may help explain telomere defects caused by deficiencies in other RecQ helicases . We observed an accumulation of X-shaped structures at telomeres in tlc1 sgs1 mutants . Because RuvC cannot selectively cleave these X-structures , they do not appear to be HJs . This is supported by our recent report that the C-terminal 200 amino acids of Sgs1p are dispensable for slowing senescence [22]; this C-terminus contains the HRDC domain , which is important for HJ binding and double-HJ dissolution [47–49] , arguing that HJ-targeted functions of Sgs1p are not involved in slowing senescence . These X-structures can branch migrate , indicating that they cannot be convergent replication forks , but this is consistent with them being HC or rec-X structures . In S . cerevisiae , HC have been observed to form behind replication forks in a RAD52-independent fashion [36 , 38] . Liberi et al . [27] suggested that at a stalled fork , resumption of replication might use a HC to facilitate template switching whereby one nascent strand leaves the parental template and copies the other nascent strand ( to bypass the cause of the stall ) before returning to the original template in a RAD52-dependent step , thus forming a rec-X structure . Our finding that the elevated level , but not the basal level , of X-structures in tlc1 sgs1 mutants is RAD52- and RAD53-dependent is consistent with the elevated level structures being rec-Xs and the basal level structures being HC . According to this model ( Figure 7A ) , the stalling of replication forks that occurs naturally in the telomere repeats [24 , 50] might be somehow enhanced by changes related to telomere shortening ( see below ) and thus might lead to rec-X formation and eventual resolution by Sgs1p; in the absence of Sgs1p , rec-X structures would accumulate . This is analogous to the recently reported accumulation of rec-X structures at non-telomeric loci in sgs1 mutants after the stalling of replication forks by methyl methane sulfonate [27] . Sgs1p would be expected to function in tandem with its Top3p cofactor to effect strand transfer reactions that would enable resolution of the rec-X structure [27 , 51 , 52] , consistent with our finding that such cooperation is required to prevent rapid senescence [22] . If unresolved , rec-Xs might lead directly to cell cycle arrest; if resolved by other means ( e . g . , nucleases ) , the shortened or aberrantly structured telomere ends might hasten the onset of senescence . In TLC1+ cells , telomerase could repair such ends , thus explaining the normal telomere length in sgs1 mutants and the synergy of sgs1 mutation with tlc1 mutation to accelerate senescence . Previously , we observed that senescent tlc1 sgs1 mutants appear unable to segregate nuclei between mother and daughter cells [15] , and a possible explanation is that unresolved recombination intermediates interfere with chromosome segregation . We note also that suppression by Sgs1p of X-structure accumulation at stalled replication forks was recently shown to cooperate with a parallel pathway that is dependent on SUMOylation [18] . This might explain our recent observation that , like Sgs1p , Slx5p and Slx8p are required to prevent rapid senescence of tlc1 mutants , because Slx5p and Slx8p function in parallel with Sgs1p for cell viability and also show genetic interaction with SUMO pathway factors [22 , 53] . It is not yet clear why telomere shortening should increase X-structure levels at telomeres , although one possibility is that changes in chromatin might contribute to increased replication fork stalling and thus rec-X formation . For example , decreased Rap1p binding at shortened telomeres might allow telomere repeats to adopt DNA structures involving hydrogen bonding between guanines , e . g . , G-quadruplexes , that could impede replication . The recent demonstration of impaired telomere replication in S . pombe cells lacking the Taz1 telomere repeat binding protein supports this idea , and led the authors to also propose that Rap1p might serve a similar function to facilitate telomere replication in S . cerevisiae [54] . Of note , however , a mutant Rap1p protein lacking the C-terminus does not impact telomere fork stalling [24] , although this mutant possesses the N-terminal DNA binding domain and so might have retained the function proposed to facilitate replication . We further note that we did not observe in senescing tlc1 or tlc1 sgs1 cells an increase in apparent stalling near the 2N spot , corresponding to the telomere repeats , although this might be explained by efficient conversion of stalled forks into rec-X structures . A second possibility is that a shortened and less heterochromatic telomere might be more accessible to recombination factors like Rad52p , thus facilitating rec-X formation . The elevated recombination rates at shortened telomeres in K . lactis and telomerase knockout mice are consistent with these possibilities [55 , 56] . During senescence , Sgs1p may inhibit the formation of telomere recombination intermediates , or facilitate their resolution . Using telomere PCR and sequencing , we observed a significantly decreased frequency of recombinants occurring at longer telomeres in tlc1 sgs1 mutants , supporting the model that Sgs1p helps resolve telomere recombination intermediates into mature products . For this reason , we suggest that decreased resolution , rather than increased formation of X-structures explains their higher levels in tlc1 sgs1 mutants . Two aspects of the PCR assay used to detect the recombinant telomeres must be understood to explain why recombinants are less frequent at longer telomeres in the absence of Sgs1p . First , the number of cell divisions after loss of telomerase at which a recombinant arises will influence the apparent frequency of that event: productive recombination events that occur early and give rise to progeny will remain at a frequency approximately equal to that at the time of their occurrence , whereas events that occur late will appear at a lower frequency that reflects the larger size of the pool of cells at that later time point ( Figure 7B , left vs . right ) . Although only one time point was examined for each strain , the stochastic natures of senescence and telomere shortening caused some cells to be closer to senescence than others , and so information spanning a large range of telomere lengths was obtained . The second aspect of the telomere PCR assay that must be appreciated is that it will detect both resolved products and unresolved intermediates , and therefore , any stalled recombination intermediates in cells lacking Sgs1p will still be observed; this fact minimizes the measured difference in recombinants between tlc1 and tlc1 sgs1 mutants when telomeres of all lengths are examined ( because recombinants forming at short telomeres are expected to suffer little from the sgs1 defect; see below ) . Telomere recombination appears to increase as telomeres shorten ( Figure 1D and [28] ) , yet the frequency of recombinants measured by telomere PCR in tlc1 cells was not greater at shorter than at longer telomeres . A reasonable explanation is that the recombination events at short telomeres were more likely to have occurred in cells that were closer to senescence and so gave rise to fewer progeny than the cells experiencing recombination at long telomeres . These competing effects of more frequent recombination at short telomeres but fewer recombinant progeny arising from cells with short telomeres could balance each other in the tlc1 mutants so that the distribution of recombinants is similar among telomeres of all sizes . In contrast , if Sgs1p is required for efficient resolution of recombination intermediates and if unresolved intermediates cause cell cycle arrest , then cells with long telomeres , and thus high replicative potential , would be most affected by stalled recombination events; tlc1 sgs1 mutants arrested by stalled recombination intermediates at long telomeres will become diluted by the other dividing cells ( Figure 7B , middle ) . Cells with short telomeres are unlikely to divide much further regardless of the outcome of a telomere recombination event , and thus absence of Sgs1p would have relatively little effect on the measured frequency of recombinants at short telomeres ( Figure 7B , right ) . This explains why the decrease in recombinants in the tlc1 sgs1 mutants occurs preferentially at longer telomeres . As an interesting aside , by this view , the distribution of recombinants among telomeres of different lengths in tlc1 sgs1 mutants most accurately reflects the propensity of short telomeres to engage in recombination because this distribution is not skewed , as it is in tlc1 mutants , by the opposing effect of recombination at longer telomeres tending to occur in cells with greater remaining replicative potential and thus giving rise to more progeny . Consistent with the interpretation that stalled recombination events in tlc1 sgs1 mutants lead to permanent cell cycle arrest , WRN is required in cultured human cells for the resolution of recombination intermediates that enable cells to generate viable progeny [57] . If tlc1 mutants can complete telomere recombination and give rise to viable progeny , then repeat examples of the same recombination event should be detectable . Indeed , four independent examples were obtained in the tlc1 cells ( indicated by the number sign [#] , Figure 6A ) . No such repeat events were observed in the tlc1 sgs1 cells , consistent with telomere recombination often being a terminal event in the absence of Sgs1p . Furthermore , no such events were observed in the shortest ( <85 nt ) telomeres of tlc1 mutants , consistent with the recombination events at short telomeres occurring in cells that are near the end of their lifespan . The action of Sgs1p during senescence need not reflect any telomere-specific function , but rather may be one manifestation of a general role in the restart of replication forks stalled for various reasons , for example , hydroxyurea treatment , DNA alkylation by methane methyl sulfonate , or as proposed here , chromatin changes at shortened telomeres . We note , however , that stalled forks in the terminal telomere repeats would be particularly problematic because there is no replication origin distal to the stall to generate a rescuing fork , thus perhaps contributing to the dependence of telomeres on Sgs1p-dependent restart during senescence . Sgs1p helps activate the checkpoint response to DNA damage in S phase , and also helps to stabilize DNA polymerases alpha and epsilon at stalled forks [35 , 58] . The former function , but not the latter , is thought to occur in collaboration with Rad53p [58] . Nonetheless , Rad53p appears to help stabilize stalled replication forks [27 , 36 , 59 , 60] , although the extent to which this reflects stabilization of DNA polymerases [35 , 61] , the MCM helicase [62] , or other functions , and the degree to which Sgs1p is required for these functions , are not resolved at present . The reduction in X-structure levels caused by the rad53K227A allele and modest acceleration of senescence in tlc1 rad53K227A mutants is consistent with the model that rec-X–dependent fork restart contributes to optimal telomere replication during senescence ( Figure 7A ) . The larger effect of sgs1 mutation on senescence may reflect a hypomorphic effect of the rad53K277A allele with respect to HC- and rec-X–mediated fork rescue , such that some stalled forks are still routed through this pathway and thus depend on Sgs1p function . Alternatively , the capacity of Sgs1p to stabilize stalled forks may be greater than that of Rad53p . Our findings leave open the possibility that replisome stabilization by Sgs1p may contribute to slowing senescence , in addition to the proposed role in rec-X resolution . Given the increased loss of telomeres replicated by lagging-strand synthesis in WS cells [11] , it is interesting that fork stalling was not increased in sgs1 mutants . Therefore , Sgs1p is not required for telomere replication in most instances . However , the WS defect affects only about 2% of telomeres [11] , and it is possible that similarly infrequent replication defects that are below the limit of detection of the 2DGE assay do occur in sgs1 mutants . Another possibility is that WRN has a function in telomere replication that is different from Sgs1p . However , we note that the helicase domain of WRN , which is conserved among all RecQ family helicases , is critical for its telomere maintenance function [11] , and further , that human and mouse BLM [10 , 13] and a S . pombe RecQ homolog [7] also appear to have roles in telomere maintenance , and so this likely represents a conserved function of several RecQ proteins . Perhaps recombination defects like those observed here in tlc1 sgs1 mutants contribute to the replication-related telomere defects of WS cells . If so , our model does not address why the defect in Werner cells should selectively affect the telomere strand copied by lagging-strand synthesis . Given the propensity of RecQ proteins to unwind G-quadruplexes [48] , one possible explanation is that , in the absence of a RecQ helicase , persistence of a G-quadruplex on the unpaired G-rich strand of a rec-X intermediate might lead to cleavage by a G-quadruplex–specific nuclease ( e . g . , Mre11 [63] ) and thus , selective loss of this strand . Alternatively , differences in the structure of the termini at telomeres generated by lagging- versus leading-strand synthesis may affect the propensity for recombination , since the product of lagging-strand synthesis has a 3′ overhang , whereas the initial leading strand product would have a less recombinogenic blunt end . Further investigation of these possibilities , and of the interface between replication , recombination , and telomere maintenance , should improve understanding of the mechanisms underlying the cancer and age-related diseases caused by deficiencies in RecQ helicases .
All strains were isogenic derivatives of YBJ133 ( Mata/α Δho Δhml::ADE1 Δhmr::ADE1 ade1 ura3–52 , leu2–3 , 112 , lys5 , TLC1/Δtlc1::kanMX , and SGS1/Δsgs1::hisG-URA3 ) [15] . RAD52 was deleted by PCR-mediated open reading frame ( ORF ) replacement with LEU2 in one allele of YBJ133 to generate YJL4 . The rad53K227A allele was introduced using plasmid pCH8 as described [64] . into a YBJ133 derivative that was TLC1/Δtlc1::HygB , rather than TLC1/Δtlc1::kanMX , to generate YBJ436 . Senescence experiments were as described [15] , using haploid spore products derived from diploids that were heterozygous for mutations in TLC1 , SGS1 , and in some cases , RAD52 or RAD53 , and cells were cultured at 30 °C in standard YPD medium . DNA was purified from log-phase cells using hexamine cobalt ( III ) to limit branch migration as described [65] . Telomere length was measured by Southern analysis of XhoI-digested DNA using a Y′ probe as described [15] . 2DGE for telomere analysis was as described using ClaI-digested DNA [50] , except the first dimension was run at 4 °C at 0 . 6 V/cm for 68–70 h , and the second dimension was at 3 V/cm for 20–22 h . The Y′L-specific probes were generated by amplification of genomic DNA with primers 5′-ggcgttgcaatgtggaaatg and 5′-gaccggcaaaagcgagtagc . The 2DGE for rDNA analysis was performed as for telomeres except that DNA was digested with SnaBI , and the first dimension was for 19 h at 1 V/cm , and the second dimension was for 4 h at 4 . 8 V/cm . The rDNA probe ( to the 18S ribosomal RNA region ) was generated using PCR amplification of genomic DNA with primers 5′-CTGGTGAGTTTCCCCGTGTTGAG and 5′-CCTTGTGCTGGCGATGGTTC . 32P-probed blots were washed and visualized as described [15] using a Molecular Dynamics Phosphoimager ( Molecular Dynamics/GE Healthcare , http://www . gehealthcare . com/usen/index . html ) , and levels of X-structures , replication forks , and 1N and 2N spots were determined with ImageQuant software . For branch migration , the first-dimension gel slice was incubated in 10 mM Tris-HCl ( pH 8 . 0 ) , 0 . 1 mM EDTA , and 100 mM NaCl ( and in additional tests , with 10 mM MgCl2 ) at 4 °C for 2 h and then at 65 °C for 4 h . For RuvC treatment , ClaI-digested DNA was ethanol precipitated and dissolved in 45 ml of RuvC buffer ( 12 mM Tris-Cl [pH 8 . 0] , 10 mM MgCl2 , 1 mM DTT , and 100 μg/ml BSA ) and treated with 350 ng of one of two different preparations of purified RuvC ( Gifts of the M . Whitby and R . Lloyd laboratories; similar results were obtained with both preparations ) at 37 °C for 3 h . To confirm RuvC activity , the synthetic X-junction J11 was constructed as described [66] , 5 ng was added to control digests that were otherwise identical to experimental digests , reactions were separated on 10% native PAGE gels , and the blotted products probed with 32P end-labeled oligonucleotide J11–2 [66] . Telomere PCR was performed as described [45] with modifications . DNA was isolated using a Qiagen genomic DNA kit ( http://www1 . qiagen . com ) , denatured , C-tailed , and the telomere ends of Chromosome I-L were amplified using o286S MluI: 5′-CACGCGTGGTTGGCCAGGGTTAGATTAGGGCTG-3′ and an equimolar mixture of G18A BamHI: 5′-CGGGATCCG18A-3′ and G18C BamHI: 5′-CGGGATCCG18C-3′ . Unexpectedly , one of the Chromosome I copies in YBJ133 possessed an uncharacterized polymorphism that prevented amplification; this was discovered because there was a 2:2 segregation in the ability of DNA from spore products to be amplified; conveniently , this ensured that all amplifiable spore products inherited the same telomere . PCR amplification included 1 . 25 M betaine because its addition to control reactions with a 262-base pair ( bp ) cloned telomere repeat sequence yielded a sharp band of the expected size that was more distinct and at higher yield than standard conditions . PCR products were separated on 2% agarose gel , 50–1 , 000-bp fragments were excised and cloned into the pCR4-TOPO vector ( Invitrogen , http://www . invitrogen . com ) , transformed into DH5α Escherichia coli cells , and individual clones were sequenced by Certigen ( http://www . certigen . com ) or by the University of Pennsylvania DNA Sequencing Facility . Telomere sequences were aligned using MegAlign software and compared using the National Center for Biotechnology Information BLAST . Rules used to classify telomere sequences are detailed in Supplementary Figure 3 in Protocol S1 . Two-tailed unpaired t-tests were performed for all comparisons except the frequency of recombinants , for which a one-tailed chi-square test was performed .
|
Because telomeres are situated at the ends of chromosomes , they are both essential for chromosome integrity and particularly susceptible to processes that lead to loss of their own DNA sequences . The enzyme telomerase can counter these losses , but there are also other means of telomere maintenance , some of which depend on DNA recombination . The RecQ family of DNA helicases process DNA recombination intermediates and also help ensure telomere integrity , but the relationship between these activities is poorly understood . Family members include yeast Sgs1p and human WRN and BLM , which are deficient in the Werner premature aging syndrome and the Bloom cancer predisposition syndrome , respectively . We have found that the telomeres of yeast cells lacking both telomerase and Sgs1p accumulate structures that resemble recombination intermediates . Further , we provide evidence that the inability of cells lacking Sgs1p to process these telomere recombination intermediates leads to the premature arrest of cell division . We predict that similar defects in the processing of recombination intermediates may contribute to telomere defects in human Werner and Bloom syndrome cells .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"oncology",
"cell",
"biology",
"yeast",
"and",
"fungi",
"geriatrics",
"genetics",
"and",
"genomics",
"saccharomyces"
] |
2007
|
Evidence That a RecQ Helicase Slows Senescence by Resolving Recombining Telomeres
|
Estimates suggest that more than 25 , 000 to 125 , 000 people die annually from snakebite envenomation worldwide . In contrast to this major disease burden , thorough bibliometric studies do not exist so far that illustrate the overall research activity over a long time span . Therefore , the NewQIS-platform conducted an analysis on snakebite envenoming using the Thomson Reuters database Web of Science . To determine and assess changes regarding the scientific activities and to specifically address the more recent situation we analyzed two time intervals ( t ) . During the first time interval from 1900 to 2007 ( t1 ) 13 , 015 publications ( p ) were identified . In the following period ( 2008–2016 = t2 ) 4 , 982 publications were identified by the same search strategy . They originate from 114 ( t1 ) respectively 121 countries ( t2 ) , with the USA ( p = 3518 ) , Brazil ( p = 1100 ) and Japan ( p = 961 ) being most productive in the first period , and the USA ( p = 1087 ) , Brazil ( p = 991 ) and China ( p = 378 ) in the second period , respectively . Setting the publication numbers in relation to GDP/capita , Brazil leads with 92 publications per 10 , 000 Int$GDP/capita , followed by India with 79 publications per 10000 Int$GDP/capita ( t1 ) . Comparing the country’s publication activity with the Human Development Index level indicates that the majority of the publications is published by highly developed countries . When calculating the average citation rates ( citations per published item = CR ) mainly European countries show the highest ranks: From 1900–2007 Sweden ranks first with a CR = 27 , followed by the Netherlands ( CR = 24 . 8 ) , Switzerland ( CR = 23 ) , Spain , Austria and the USA ( CR = 22 ) . From 2008 to 2016 the highest rate achieves Switzerland with a value of 24 . 6 , followed by Belgium ( CR = 18 . 1 ) , Spain ( CR = 16 . 7 ) , Costa Rica ( CR = 14 . 9 ) and Netherlands ( CR = 14 ) . Compared with this , the USA was placed at rank 13 ( CR = 9 , 5 ) . In summary , the present study represents the first density-equalizing map projection and in-depth scientometric analysis of the global research output on snakebites and its venoms . So it draws a sketch of the worldwide publication architecture and indicates that countries with a high incidence of snakebites and a low economical level still need to be empowered in carrying out research in this area .
Snakebites envenomation is a neglected tropical disease and lead to an enormous burden of disease in many parts of the world [1–3] . Precise epidemiological data is missing but estimates suggest that 25 , 000–125 , 000 deaths and about 400 , 000 permanent disabilities are caused by snakebites annually [4–7] . The highest incidences were reported for ( up to 500/100 , 000 inhabitants per year ) Papua New Guinea , West Africa and Guinea [6 , 8 , 9] . Snakebites may also cause psychological morbidity [10] . However , these facts were previously largely neglected . India plays a prominent role in snakebite epidemiology due to the vast size of the country and the number of inhabitants . However , potentially biased hospital-based statistics has led to widely ranging estimates of total annual snakebite mortality ( 1 , 300 to 50 , 000 ) . Therefore , Mohapatra et al . calculated estimates of direct snakebite mortality from a national mortality survey using data from 123 , 000 deaths from 6 , 671 randomly selected areas between the years 2001 and 2003 [11] . They reported a number of 562 deaths ( 0 . 47% of total deaths ) , which was assigned to snakebites occurring mostly in rural areas ( 97% ) . This proportion was suggested to represent about 45 , 900 annual snakebite deaths nationally ( 99% CI 40 , 900 to 50 , 900 ) with higher rates in rural areas ( 5 . 4/100 , 000; 99% CI 4 . 8–6 . 0 ) [11] . The most reported annual snakebite deaths occur in the states of Uttar Pradesh ( 8 , 700 snakebite deaths ) , Andhra Pradesh ( 5 , 200 snakebite deaths ) , and Bihar ( 4 , 500 snakebite deaths ) [11] . In view of this underestimation , snakebites need to be considered a neglected problem in twenty-first century India and South Asia in general [2 , 12] . Despite this substantial global burden , snakebite injuries have received only little attention from governmental and non-governmental institutions all over the world , the pharmaceutical industry , and public health authorities and advocacy groups . Also , research funding and resources for health programs are on a low level in comparison to other diseases . In this respect , stakeholders and decision-makers have a huge requirement for scientifically validated recommendations that are still not sufficiently available . The inclusion of snakebites in the WHO list of Neglected Tropical Diseases ( NTDs ) as well as the development of project and initiatives by the WHO may help to improve global awareness . The fight against snakebite-related diseases is also crucially dependent on research funding . Usually , the funding allocation processes get improved by adding scientometric features to the election processes . Unfortunately , there is a lack of basic information about snakebite research . Therefore , we carried out a combined density equalizing mapping and scientometric analysis . The here presented patterns of the research efforts and the publication output empower scientific institutions , planners , and stakeholders to allocate research funding better , and support the most affected regions .
The present study is part of the NewQiS-platform , which uses novel visualizing techniques in combination with new and established scientometric analysis tools [13 , 14] . The principle algorithm of density-equalizing map projections ( DEMP ) was reported by Gastner and Newman [15] and incorporated in NewQIS-studies [16] . In brief , software applying these algorithms is used to determine correlations and differences between countries that are publishing on snakebite-research . The method resizes countries proportionally according to a predefined variable . In this process , the nation with the highest value of the analyzed parameter is depicted largest on the associated map , whereas regions without or with a very low value are proportionally scaled down . All analyzed data was retrieved from the Thomson Reuters online-database ‘Web of Science’ ( WoS ) . The evaluation period was divided into two intervals . The first time frame was limited to the period between 1900 and 2007 ( t1 ) in order to assess a closed and defined interval to reach a historical bibliometric evaluation . This time period starts after the epochal invention , and introduction into clinical medicine , of antivenom treatment for snakebite envenoming by Albert Calmette [17] and Vital Brazil at the end of the 19th century [18] , and ends with the publication of the key conceptual snakebite advocacy paper by Gutiérrez et al . in 2006 [19] . Publications from the subsequent years up to the time of analysis ( 2008–2016 ) were addressed separately ( t2 ) because this timespan has been marked by a fresh interest in snakebite injuries and related advocacy initiatives which have led to substantial changes and new hope for this neglected field of research . These included , for example , events like WHO regional and bi-regional workshops leading to the publication of ‘WHO Guidelines on the production and use of snake antivenom immunoglobulins’ [20] with an extensive associated internet database of venomous snakes and antivenoms [21] , a revised edition of ‘WHO-SEARO Guidelines for the management of snakebite in South and Southeast Asia’ [22] , ‘WHO Guidelines for the management of snakebite in Africa’ [23] , the ‘Global Issues in Clinical Toxicology’ conference in 2008 with the formation of the ‘Global Snakebite Initiative ( GSI ) ’ [24] , several regional conferences in Africa with the foundation of the ‘Société Africaine de Venimologie / African Society of Toxicology’ [25] , and a series of influential high-profile articles [26–29] that have helped generate awareness and a re-emerging interest for snakebite envenoming in professional societies as well as funding agencies . The methodology of data categorization was performed as reported also in previous NewQIS-publications [30–32] . Using this method , the retrieved bibliometric data has been categorized according to various parameters ( e . g . publication countries , publication year , publishing authors , published document type , and assigned subject categories ) . Afterwards it has been transformed into a database format . Data adjustments led to more meaningful findings by the correction of author’s data . This has proved necessary because the software-supported assignment of bibliometric data is faulty and cannot be evaluated accurately when e . g . the country or origin or the author’s specifications are erroneous or outdated . The following composed query term has been applied via the WoS Topic-Search: ( ( snake* ) AND ( venom* OR envenomation* OR poison* OR toxin* OR antidote* OR antiserum* OR antivenom* ) ) OR ( ( bit* OR venom* OR envenomation* OR poison* OR toxin* OR antidote* OR antiserum* OR antivenom* ) AND ( viper* OR elapid* OR colubrid* OR atractaspid* OR naja OR cobra* OR crotalus OR rattlesnake* OR bothrops OR lancehead* OR agkistrodon OR moccasin* OR bungarus OR krait* OR echis OR saw-scaled viper* OR dendroaspis OR mamba* OR trimeresurus OR asian pit viper* OR bitis OR puff adder* OR notechis OR tiger snake* OR oxyuranus OR taipan* OR lachesis OR bushmaster* OR cerastes OR horned viper* OR dispholidus OR boomslang* OR hydrophii ? ae OR sea snake* ) ) . The Boolean operator connects the individual terms as disjunction . The asterisk is used as a placeholder for several characters , while the question mark stands for one letter only . The information about the address of the author’s institutions was analyzed in order to determine the country of origin . The interpretation of international publications is based on added up values . Herewith every national contribution has been counted separately , so that the overall publication sum of the country-specific analysis is much above the actual amount of the retrieved articles . Countries that are no longer existing ( e . g . USSR , Yugoslavia ) , or that have been formed later ( e . g . Germany from the former FRG and GDR ) have been compared with an updated list of countries and self-governed regions , and corrected respectively . Additionally , chronological analyses have been performed . By this procedure , it was possible to assess the total number of publications and citations per publication year until the date of analysis . Also , the average citation per year was computed for years with at least 30 published items . This threshold was implemented for all citation rate analyses to reduce the impact and the consequent bias of both years in which very few articles have been published or of countries with a very number of publications . The Human Development Index ( HDI ) of the Human Development Report ( HDR ) of the UN Development Project ( UNDP ) was used to relate the research activity to social and economic data [33] . Also , the GDP/capita in 10 thousand Int$ was used . Due to the exclusion of Taiwan from the World Bank , the data was retrieved from the World Factbook that includes Taiwan [34 , 35] . Epidemiologic indices were used from the estimates of Kasturiratne et al . [4] . An analysis of international cooperations was carried out to assess research networks [36][35] . In brief , a bilateral cooperation between two countries was defined when at least one author originates from one country and at least one other author from a second country . A matrix with all identified countries was set up and filled with the corresponding values for the cooperation for each pair of countries . A second software module was used to translate the matrix and to transform the figures into vectors .
In t1 a total number of 13 , 015 publications were identified , and the second evaluation period ( t2 ) delivered 4 , 982 publications . A strong increase between the years 1990 and 1991 is due to the inclusion of abstracts and keywords in the search mode of WoS . In the first evaluation period ( t1 ) 87% of the publications were attributable to one or more countries , viz . 11 , 323 publications , whereas in t2 only 1 , 4% was not attributable , so that 4 , 913 publications were assigned to one or more countries . In total , 114 ( t1 ) , respectively 106 countries ( t2 ) contributed to the overall publication output . Ninety-four percent ( t1 ) and alternatively 97% in t2 of all publications were written in English . The most common non-English articles were published in French ( 2% ) , Russian ( 0 . 97% ) , and German ( 0 . 96% ) in timespan t1 , while in t2 French ( 0 , 64% ) , Spanish ( 0 , 62% ) and Portuguese ( 0 , 54% ) were the most frequent non-English languages . Despite the already high level of English-written publications in t1 , this proportion has even more increased in t2 . The USA was the leading country concerning research output with a total of 3 , 518 published items representing 31% of the attributable publications ( t1 ) , respectively 1 , 087 publications ( 22% ) in t2 . On second position in both periods , Brazil was listed with 1 , 100 publications ( 10% ) in t1 and 991 publications ( 20% ) in t2 , respectively . In t1 , these two countries were followed by Japan with p = 961 ( 9% ) , the UK with p = 862 ( 8% ) , France with p = 700 ( 6% ) , Taiwan with p = 617 ( 5% ) , Australia with p = 506 ( 4% ) , Germany with p = 496 ( 4% ) , China with p = 454 ( 4% ) , and India with 317 publications ( 3% ) . Costa Rica ( p = 284; 3% ) , Italy ( p = 236; 2% ) , South Africa ( p = 230; 2% ) , Israel ( p = 215; 2% ) , and Russia ( p = 209; 2% ) . Sweden ( p = 194; 2% ) , Switzerland ( p = 187; 2% ) , Singapore ( p = 161; 1% ) , the Netherlands ( p = 145; 1% ) , Thailand ( p = 132; 1% ) , Canada ( p = 130; 1% ) and Spain ( p = 115; 1% ) have more than 100 published items each . The other 92 countries , with less than 100 publications each , accounted for 1594 publications in total ( Fig 1A and Fig 2A ) . In the second evaluation period ( t2 ) the third most active country was China with 378 items ( 7 . 7% ) . The next countries were Australia with p = 373 ( 7 . 6% ) , India with p = 335 ( 6 . 8% ) , UK with p = 303 ( 6 . 2% ) , Germany with p = 224 ( 4 . 5% ) , France with p = 210 ( 4 . 3% ) , Costa Rica with p = 202 ( 4 . 1% ) , Japan with p = 185 ( 3 . 8% ) , Spain with p = 178 ( 3 . 6% ) , Taiwan with p = 140 ( 2 . 8% ) , and Italy with p = 113 ( 2 . 3% ) . Less than 100 publications were published by another 108 countries ( Fig 1B and Fig 2B ) . When comparing the research activity to incidence of snakebites and economic data of the publishing countries , a two-fold result was obtained ( Tab . 1 ) : Due to the low incidence , the country with the highest research output ( USA ) was increasing its distance to the second ranked country , publications per snakebites per 100 , 000 inhabitants in the US versus 356 for Japan . Other countries with low incidences such as the UK , or France and Germany also raised in this ranking . By contrast , when relating the output data to GDP/capita , Brazil with 92 publications per 10 , 000 Int$ GDP/capita was ranked number one . Second was India with 79 publications per 10 , 000 Int$ GDP/capita . The adjustment of the research activity data to the six WHO regions ( WHO African Region , WHO Region of the Americas , WHO South-East Asia Region , WHO European Region , WHO Eastern Mediterranean Region and WHO Western Pacific Region ) led to the following ranking: The WHO Region of the Americas was leading with p = 5 , 343 ( 47% ) attributable publications , followed by the WHO European Region with p = 4 , 015 ( 36% ) and the WHO Western Pacific Region ( Fig 3A ) . By contrast , the snakebite incidence was higher in the WHO South-East Asia Region . When adjusting the output data to the HDI level according to the Human Development Report , in the majority of the publications ( p = 11 , 004 , 97% ) a highly developed country was at least collaborating ( Fig 3B ) . By contrast , only in 1 , 650 ( 14 . 6% ) of the publications , a country with a medium HDI level is participating and only in 69 publications , a low level country is participating . The number of international collaborations between two or more countries increased steadily since 1972 ( Fig 4A ) to the end of period t1 in 2007 . Also , between 2008 and 2016 ( t2 ) the trend towards an increasing number of collaborations has been remaining–with exception of 2016 , due to the missing values of this incomplete evaluation year ( Fig 4B ) . The maximal levels were reached with 132 ( t1 = 2005 ) and 193 collaborations ( t2 = 2012 ) . In total , 1 , 739 cooperation articles ( CA ) were identified from 1972 to 2007 ( t1 ) and 1 , 322 until 2016 ( t2 ) . Out of them , CA = 1 , 480 in t1 ( t2: CA = 997 ) are a result of a bilateral cooperation; t1: CA = 211 , t2: CA = 238 publications are trilateral . Four countries collaborated in CA = 38 ( t1 ) and CA = 57 ( t2 ) . The USA participated in CA = 843 until 2007 ( t1 ) and in CA = 633 in t2 , the UK in CA = 456 ( t1 ) and CA = 358 ( t2 ) , Brazil in CA = 365 ( t1 ) and CA = 342 ( t2 ) , France in CA = 352 ( t1 ) and CA = 226 ( t2 ) , and Germany in CA = 303 ( t1 ) and CA = 278 ( t2 ) . Forty cooperation country pairs were found with USA and Brazil cooperations being highest in number ( CA = 92 ) until 2007 ( Fig 5A ) . In t2 ( 2008–2016 ) the bilateral works of the USA and Brazil ranked only third with an amount of CA = 52 . Here , the highest number has been reached by the collaboration of Australia and the USA with CA = 63 common publications ( Fig 5B ) . In respect to the number of citations ( c ) , the USA led in both time intervals with t1: c = 75 , 614 and t2: c = 10 , 420 . Until 2007 ( t1 ) the UK is ranked 2nd with c = 17 , 908 , followed by Japan ( c = 17 , 606 ) , France ( c = 14 , 612 ) and Brazil ( c = 12 , 028 ) ( Fig 6A ) . In contrast , the second evaluation period ( 2008–2016 ) showed another continuing order . The second most cited country in t2 was Brazil with c = 6519 , followed by Australia ( c = 4 , 209 ) , the UK ( c = 3 , 868 ) and Costa Rica with c = 2 , 998 ( Fig 6B ) . When calculating the citation rate per published item of each country in a DEMP with a threshold of at least 30 publications per country , a different global research landscape has been obtained with European countries taking the lead position . In the first time span ( t1 ) , Sweden ranked number 1 with a citation rate ( CR ) of 27 , followed by the Netherlands ( CR = 24 . 8 ) , Switzerland ( CR = 23 ) , Spain and Austria ( CR = 22 ) . At position six , the USA came up with a citation rate of 22 , followed by other European countries ( France CR = 21 , the UK CR = 21 , Denmark CR = 21 and Germany CR = 20 ) . At position 11 , non-European countries followed with Costa Rica ( CR = 19 ) , Japan ( CR = 18 ) , Canada ( CR = 17 ) and Taiwan ( CR = 16 ) leading the field . Brazil has a citation rate of 11 citations per publication ( Fig 7A ) . The analysis of the citation rate of the second time period ( t2 ) revealed a different picture . Here , Switzerland had the leading position with CR = 24 . 6 , followed by Belgium ( CR = 18 ) . Placed next were Spain ( CR = 16 . 7 ) , Costa Rica ( 14 . 8 ) , the Netherlands ( CR = 14 ) , UK ( CR = 12 . 8 ) , France ( CR = 12 . 2 ) , Australia ( CR = 11 . 3 ) , Italy ( CR = 11 . 2 ) and Colombia ( CR = 10 . 9 ) . The USA was ranked only 13th with a rate even under 10 ( CR = 9 . 6 ) ( Fig 7B ) . The 13 , 014 identified publications of t1 were published by a total of 3 , 922 institutions ( i ) . When analyzing the number of institutions per country , the USA were leading with i = 896 institutions , followed by Japan with i = 284 , Brazil with i = 283 , France with i = 259 and the UK with i = 217 ( Fig 8A ) . The analysis of the institutions participating in t1 showed a range of 1 to 279 publications per single institution . The National Taiwan University was ranked number one with 279 publications , followed by the Brazilian Instituto Butantan with p = 277 and the Universidad de Costa Rica ( Instituto Clodomiro Picado ) with p = 275 . The highest modified country-specific h-index within the analyzed set of p = 11 , 302 was found for the Universidad de Costa Rica with a value of 38 ( i . e . 38 publications that were at least cited 38 times ) ( Fig 9A ) ; all of these from the university’s Instituto Clodomiro Picado . In t2 there were 705 institutions in the USA publishing on snakebites , followed by Brazil with i = 540 , China ( i = 291 ) , India ( i = 284 ) , and France ( i = 158 ) ( Fig 8B ) . The range of publications per institution varied from 1 to 284 snakebite related publications ( University of Sao Paolo , Fig 9 ) . As in t1 , the second most active institution was the Brazilian Instituto Butantan with p = 267 from 2008 to 2016 ( t2 ) . Another Brazilian institution followed with p = 126 , the Universidad Estadual Campinas . The highest snakebite-specific h-index in t2 was achieved by the Spanish National Research Council CSIC ( ‘Consejo Superior de Investigaciones Científicas’ Fig 9B ) .
PLOS Neglected Tropical Diseases has published in the past years numerous studies related to the field of snakebites [37 , 38] . They range , for example , from neurotoxicological aspects [39] , immune responses [37] , phylogeny , venom composition of diverse species [38] , the analysis of geographical information [40] to basic science issues [41–43] and public health [44] . Despite the burden of disease caused by snakebite envenoming and the need for increased transnational and national funding , there has been no in-depth scientometric analysis of this topic so far . Therefore , we used the NewQIS-platform methodology to conduct a combined DEMP and scientometric study on publications related to snakebites until 2016 . Bibliometric tools are commonly used to dissect research profiles also for neglected tropical diseases [45–47] and using such approaches , a number of methodological issues need to be discussed: First , it has to be considered that the analysis of snakebite related articles in the present study cannot be regarded as completely representative of global snakebite research activity , since the data was retrieved from only one database ( Web of Science ) , denoting a potential bias . However , we used the WoS since this database enabled us to assess also qualitative aspects ( citations ) . Whereas the WoS is among the largest global biomedical databases , there are of course still publications , which cannot be traced by the use of this system . Nonetheless , it can be hypothesized that the present findings represent common trends in the research on snakebite and snake venoms and toxins . Second , the employed indicators that refer to the significance of the publications in the scientific community ( number of citations , citation rate ) need to be regarded critically and therefore , the data should not be over interpreted concerning research quality , as indicated by numerous previous articles [26–28] . In fact , assessing the quality of research is only possible by advanced meta-analysis using , for example , Cochrane approaches [48] . Insofar , the findings of the citation-related analyses refer to the resonance and the attention that the publications gained in the scientific community . This is certainly not always connected with the publications’ quality , but it indicates the interest with which scientific peers take them into account . Third , the research output data was related to epidemiological estimates [4] which do not necessarily display the actual case numbers in different countries exactly [4 , 11 , 49] . Fourth , there is a language bias present within the biomedical databases . Publications written in English have a higher chance of being included [50] . Also , established journals are listed more frequently than novel journals although the latter may have the same quality standards [51] . Therefore , the Matthew effect needs to be considered [52]: The communication systems in science are directed towards a reward for highly productive and renowned journals , scientists and institutions which leads , for example , to a pyramidal citation scheme . Over two hundred years ago , the Scottish surgeon Patrick Russell published a work entitled "An Account of Indian Serpents Collected on the Coast of Coromandel" ( 1796 ) which can be regarded as one of the first scientific publications that point to the need of differentiating between venomous and non-venomous snakes and the necessity to develop therapeutic options for snakebite envenoming [53] . About one hundred years later , Léon Charles Albert Calmette developed the first snake antivenom at the Institute Pasteur de Saigon [54 , 55] . We therefore decided to start the analysis at the beginning of WoS entries in 1900 . Thenceforward , we analyzed two different time intervals . The first time frame was closed in 2007 due to the time that is needed for each publication to show their highest impact measured as Cited Half-Life . This value has been introduced by the originator of the JCR ( Journal Citation Report ) Eugene Garfield , meaning the time span that is necessary to get at least 50% of the citations [56] . Regarding the biomedical literature the mean Cited Half-Life can be considered up to eight years with a growing annual trend [57] . Some journals publishing on snakebites reaches even a Cited Half-Life of more than ten years ( e . g . Archives of Ophthalmology ) . The second evaluation time frame was set from 2008 to 2016 to include also the more current research outputs up to now . The analysis shows that from 1900 to 2016 , the yearly published amount of snakebite related publications increases from 5 ( 1900 ) to 677 ( 2012 ) . This exemplifies the increase in scientific activity in this scientific area but is also related to the general bibliometric principle that research articles usually double within a time frame of 10 to 20 years [58] . The years after the maximum in 2012 show a slight decrease to 659 publications in 2014 . The even lower publication numbers in 2015 ( p = 586 ) and 2016 ( p = 82 ) are due to the fact that on the one hand not all accepted publications are already listed ( 2015 ) and on the other hand the year 2016 is not terminated at the time of analysis , so that only a part of the real publication output of these years is taken into account . Maxima of publications numbers in single years are often related to important scientific findings and usually occur 2–5 years after the publication of the important finding . In this respect , the observed maximum in 1978 ( 198 publications ) is most likely related to the development of Captopril from the venom of the South American Jararaca ( Bothrops jararaca ) . Also , the maximum in 1985 ( 281 publications ) is most likely related to the finding of dendrotoxin in 1980 , which was isolated from Mamba venom . The strong increase in 1991 is not due to a scientific but to a methodological reason caused by the implementation of the Topic-search tool of WoS . When analyzing the contribution of single countries , neither the estimated incidence ( 1 . 7 snakebite injuries per 100 , 000 inhabitants ) nor the mortality ( 0 . 01 per 100 , 000 inhabitants ) but the general standing of the USA as the world leading country concerning research activity points to the large gap in overall snakebite-research activity . This indicates that countries with a high incidence of snakebites and a low economical level need to be empowered to carry out research . In striking contrast to other scientometric studies that addresses diseases such as gout [59] , silicosis [31] , or infectious diseases including influenza [35] , or hepatitis B [60] , the global ranking of snakebite research activity is different to the usual picture with the USA being followed by the UK , Germany or Japan: This usual pattern has also been found in a study analyzing over 5 . 5 million publications on the global publication activity within the following 21 organ systems: Brain , heart , artery , vein , lung , muscle , eye , nose , ear , throat , neck , skin , breast , stomach , intestine , pancreas , kidney , genital , hormone , arm , feet . Here , an almost uniform pattern was present for every single organ . The USA was ranked first in every of the 21 organ systems . Number two and three concerning research output were either the UK , Japan , Germany or France . Interestingly , a dichotomy was present between Western countries such as the USA , UK or Germany and Asian countries such as Japan , China or South Korea concerning research focuses . Western countries each had the following ranking concerning research activity: the most frequently focused organ was the heart . By contrast , the Asian countries had the liver as number one organ of interest . In contrast , the current analysis shows that Brazil , a country with a relatively high incidence and mortality and with a rather low GDP contributed with the second highest research activity . The Brazilian publication performance corresponds with the findings of another neglected disease study on yellow fever [61] . Here , a recent NewQIS-study identified a total of 5 , 053 yellow fever-associated publications , which were published by 79 countries . The difference of the overall numbers of publications between yellow fever and snakebite related publications is most probably caused by the inclusion ( search term ) of studies that use venom ingredients to address basic mechanisms of human physiology and to develop novel pharmaceutics . Until now basic research on snakebites and venoms is more or less privileged to high-income countries , firstly , because they have the means for high tech research and secondly , there are no or only very little problems with snake envenomation . This is obviously the opposite in middle-income and low-income countries . The lack of resources limits the research on snakebite envenoming in these countries . Nevertheless , the emerging nation Brazil established–even enhanced–an important role in snakebite research . Also , as shown here , other countries from South and Middle America , like Costa Rica and Colombia , have appeared in the recent decade on the global landscape of snakebite research . Hence , it should be requested that other directly affected low and middle nations are also integrated to the worldwide research networks .
The present study represents the first density-equalizing mapping and scientometric analysis of the worldwide research activities on the subject of snakebites and draws a sketch of its overall global research architecture . The ten and more years old calls for global snake-bite control and procurement funding needs to be re-emphasized [62] .
|
Snakebite injury is a neglected tropical disease and lead to an enormous burden of disease in many parts of the world with about 25 , 000–125 , 000 estimated deaths . Therefore , research on this area of medicine is crucial since new diagnostic and therapeutic pathways may help to diminish disease burden . This study provides the first detailed landscape of the global snakebite research , which can be used by funding agencies and politicians to plan new programs . From 1900 to 2007 we found over 13 , 000 publications related to snakebites originating from over 110 countries , with the USA , Brazil and Japan being most active . The second period from 2008 until 2016 was characterized by 4 , 982 publications out of 121 countries . Here , the rank order was USA , Brazil and China . When efforts are analyzed from a socioeconomic perspective applying the GDP per Capita , Brazil and India take a lead position . The Human Development Index indicates that the vast majority of research is performed by highly developed countries . Closer investigations demonstrated that a large part of the research deals with scientific studies that used venom ingredients to unravel basic mechanisms of human physiology or to develop new pharmaceutical compounds . In summary , we here draw the first sketch of the overall global research architecture concerning snakebite envenoming . We found that countries with a high incidence of snakebites and a low economical level need to be empowered to carry out research in this area .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion",
"Conclusions"
] |
[
"medicine",
"and",
"health",
"sciences",
"toxins",
"pathology",
"and",
"laboratory",
"medicine",
"japan",
"bibliometrics",
"tropical",
"diseases",
"geographical",
"locations",
"vertebrates",
"animals",
"toxic",
"agents",
"toxicology",
"reptiles",
"scientometrics",
"neglected",
"tropical",
"diseases",
"snakebite",
"research",
"and",
"analysis",
"methods",
"venoms",
"south",
"america",
"brazil",
"research",
"assessment",
"snakes",
"people",
"and",
"places",
"asia",
"squamates",
"biology",
"and",
"life",
"sciences",
"amniotes",
"organisms"
] |
2016
|
Snakebite Envenoming – A Combined Density Equalizing Mapping and Scientometric Analysis of the Publication History
|
The development of high-throughput biotechnologies allows the collection of omics data to study the biological mechanisms underlying complex diseases at different levels , such as genomics , epigenomics , and transcriptomics . However , each technology is designed to collect a specific type of omics data . Thus , the association between a disease and one type of omics data is usually tested individually , but this strategy is suboptimal . To better articulate biological processes and increase the consistency of variant identification , omics data from various platforms need to be integrated . In this report , we introduce an approach that uses a modified Fisher’s method ( denoted as Omnibus-Fisher ) to combine separate p-values of association testing for a trait and SNPs , DNA methylation markers , and RNA sequencing , calculated by kernel machine regression into an overall gene-level p-value to account for correlation between omics data . To consider all possible disease models , we extend Omnibus-Fisher to an optimal test by using perturbations . In our simulations , a usual Fisher’s method has inflated type I error rates when directly applied to correlated omics data . In contrast , Omnibus-Fisher preserves the expected type I error rates . Moreover , Omnibus-Fisher has increased power compared to its optimal version when the true disease model involves all types of omics data . On the other hand , the optimal Omnibus-Fisher is more powerful than its regular version when only one type of data is causal . Finally , we illustrate our proposed method by analyzing whole-genome genotyping , DNA methylation data , and RNA sequencing data from a study of childhood asthma in Puerto Ricans .
Because of major advances in high-throughput biotechnologies , large amounts of omics data have been collected to study the biological mechanisms underlying complex diseases at different levels , such as genomics , epigenomics , and transcriptomics . Such different types of omics data can help us understand a disease from several perspectives . However , each of the arrays or sequencing technologies is designed to collect a specific type of omics data , such as SNPs , DNA methylation markers , and RNA sequencing . Thus , the association between a complex disease and one type of omics data is usually tested individually , but this strategy is suboptimal and has some disadvantages . Researchers often find that only a small proportion of disease variation can be explained by one type of omics ( e . g . , genetic ) data , leading to “missing heritability” [1] . Moreover , molecular variants identified by different studies usually suffer from poor reproducibility [2 , 3] . Most importantly , only partial information is used for each individual analysis . Therefore , in order to better characterize biological processes and increase the consistency of variant identification , omics data from separate platforms need to be integrated and analyzed . Integrating information from different biological datasets has the potential to yield better insight into causal mechanisms of complex diseases than that from individual omics datasets . Although integrative analysis of omics data is clearly needed , the complexity of disease mechanisms , the large number of collected molecular variables , and relatively small datasets can make such analysis quite challenging . Bersanelli et al . [4] summarized a list of existing statistical approaches for integrative analysis . When developing integrative analysis methods , two inevitable issues are: 1 . handling a large number of variables and 2 . dealing with data from relatively small studies . In genetic studies , to handle a large number of genetic variants in a gene , gene-based approaches [5–15] have been developed to evaluate the joint effects of genetic variants in the same gene on the disease of interest . Of the existing methods , the sequence kernel-machine-based associations test ( SKAT ) [16 , 17] is a powerful , flexible , and computationally efficient test . In this kernel machine ( KM ) approach , the test statistic follows a mixture of chi-square distributions , and thus p-values can be computed analytically and quickly without using resampling techniques . Although gene-based tests were originally developed for genetic studies , the same concept can be applied to studies of multi-omics data . Another issue is small sample size , especially for epigenomic or transcriptomic data . For example , large-scale genome wide association studies ( GWASs ) have been widely conducted for genetic studies for many years , so researchers usually have hundreds or thousands of genotyped samples . However , genome- wide methylation studies are more recent , and thus researchers often have a small number of samples . Moreover , incomplete samples may be wasted when using methods requiring complete samples ( e . g . , methods incorporating multi-omics data variables into one regression model ) . In this scenario , methods combining multiple p-values can be applied to make full use of data . For example , the p-values for association testing of a disease and SNPs , DNA methylation markers , and RNA sequencing data are be calculated separately , and then these separate p-values can be appropriately combined into one final p-value . In order to test for overall gene-level significance , we here present an approach to use a modified Fisher’s method ( denoted as Omnibus-Fisher ) to combine separate p-values for association testing of a disease or trait and SNPs , methylation markers , and RNA sequencing data calculated by KM regression into an overall gene-level p-value accounting for correlation between omics data . This method can be applied to either samples with all three types of omics data or samples with one or two types . To account for all possible disease models , we further extend the modified Fisher’s method to an optimal test by using perturbations . In our simulation studies , we show that a usual Fisher’s method has inflated type I error rates when directly applied to correlated omics data . In contrast , our Omnibus-Fisher test preserves the expected type I error rates when employed in correlated omics data . Moreover , the Omnibus-Fisher method has increased power compared to its optimal version when the true disease model involves all of SNPs , methylation markers , and RNA sequencing data . On the other hand , the optimal Omnibus-Fisher method is more powerful than its regular version when only one type of data is causal . Finally , we illustrate our proposed methodology by analyzing whole-genome genotyping , DNA methylation , and RNA sequencing data from a study of childhood asthma in Puerto Ricans .
When applied to samples with independent SNPs ( G ) , methylation markers ( M ) , and RNA sequencing data ( E ) , all of the methods used ( i . e . , the Fisher’s methods with and without considering p-value covariance [Omnibus-Fisher and usual Fisher] , and the optimal test with p-values from Omnibus-Fisher as inputs [optimal Omnibus-Fisher] ) had empirical Type I error rates close to the nominal level ( Fig 1A and Table 1 ) . When the usual Fisher’s method without considering covariance was applied to G and E correlated data , the Type I error rate was inflated ( Fig 1B and Table 1 ) . In contrast , the optimal and regular Omnibus-Fisher methods with considering covariance retained the desired Type I error rates as evidenced by the patterns observed in the QQ plots shown in Fig 1B and Table 1 . Similar results were observed when extending to 100 , 000 datasets for evaluation ( S1 Table ) . When we compared the power of the statistics on the samples with independent G , M and E ( Fig 2 ) , the power of optimal Omnibus-Fisher was consistent higher than that of the regular Omnibus-Fisher method when G was the only causal factor , but when G , M and E were all causal factors , the optimal methods had lower power . This was expected because the optimal methods automatically searched for the appropriate disease model; in contrast , the regular Omnibus-Fisher assumed that G , M and E were all causal factors . Thus , when the simulation matched the assumption of the regular version method , they performed better than the optimal version and vice versa . However , when G and M were causal factors , no methods were consistently better than another . Furthermore , similar patterns were observed , when evaluated using the samples with G and E correlated ( Fig 3 ) . Since the causal SNPs in G were correlated with E , GM causal was equivalent to G , M and E causal . Note that the usual Fisher’s method is not included in Fig 3 because of its inflated Type I error rate with correlated data . We used the proposed optimal Omnibus-Fisher statistic and its regular version to analyze the Puerto Rican childhood asthma data from WBCs for associations between asthma status and 14 , 808 genes with all SNPs , DNA methylation markers , and gene expression , with adjustment for age , gender and first two principal components calculated based genotypes . In addition , batch effect and cell type composition were also adjusted for DNA methylation and RNA sequencing data . We found that ZPBP2 was the most significant gene from both optimal ( P = 1 . 40×10−5 ) and regular ( P = 3 . 39×10−5 ) Omnibus-Fisher tests , although it didn’t reach a Bonferroni corrected significance level ( P = 3 . 38×10−6 ) ( Fig 4 ) . The ZPBP2 region from chromosome 17q21 has been consistently replicated as an asthma-susceptibility locus across diverse ethnic groups [18–28] including Puerto Ricans [29] and this region regulates its gene expression in Puerto Ricans [30] . In a meta-analysis of GWAS in Puerto Ricans [29] , the only region associated with asthma was the ZPBP2 locus and the current genotypic dataset was analyzed as a part of the data . This gene could be served as a positive control in asthma genetic studies . In the optimal Omnibus-Fisher test , the significance of ZPBP2 as well as GSDMB was mainly driven by their genetic effect ( P = 2 . 89×10−6 for ZPBP2 and 2 . 36×10−6 for GSDMB ) . Moreover , five additional genes ( KAT2A , HIST1H1C , NFRKB , C14orf178 and ZNF213-AS1 ) were suggestively associated with asthma ( P < 0 . 0001 [Table 2] ) from the regular Omnibus-Fisher test . Of these genes , KAT2A had moderate effects for SNPs , DNA methylation , and RNA expression separately , which could be overlooked by a single type of data analysis . The results also indicate that the optimal Omnibus-Fisher test was more powerful than its regular version when the significance was driven by one type of data . Conversely , when statistical significance was driven by two or three types of data , the regular Omnibus-Fisher had overall better power than the optimal version . These observations were generally consistent with the simulation results . Here , the optimal Omnibus-Fisher test does not outperform its regular version that assumes all types of omics data are in the disease model . Since only three types of omics data were analyzed in this study , it was still fine to assume they all were in the disease model . However , when more omics data are analyzed , the optimal test could be more useful than simply assuming all types of data are causal . We additionally output the p-value correlation for each gene across the whole genome ( S1 Fig ) : 1 . between SNPs and DNA methylation markers , 2 . between SNPs and expression genes , and 3 . between DNA methylation markers and expression genes . Analysis of the WBC genome-wide data with 1 , 116 samples and 14 , 808 genes took ~108 . 8 hours on a single computing node with a 3 GHz CPU and 4 GB memory . Using a computer cluster with multiple nodes , we anticipate that genome-wide data analysis should be finished within hours using our proposed methods .
In this work , we developed an Omnibus-Fisher statistic using a kernel machine ( KM ) regression framework , which can be employed to test overall gene-level significance by combining separate p-values of association testing for a disease and SNPs , methylation markers , and expression genes , accounting for correlation between omics data . The separate p-values are calculated by gene-based KM regression . The gene-based analysis methods can improve power by testing a set of variants jointly and by reducing the multiple testing penalty . In addition , the method using a gene as the unit can easily combine different types of omics data that are mapped to the same gene and thus easily interpret the results . Since we do not know the exact disease model in reality , the extended optimal Omnibus-Fisher test can account for all possible disease models . Moreover , our proposed tests can be applied to either samples with all three types of omics data or with one or two types . In other words , samples with incomplete data can still contribute to the test statistic . The information about whether the different types of omics data are from the sample can also be accounted . In the simulation studies , we showed that using a usual Fisher’s method on correlated omics data results in an inflated Type I error rate , while the modified Fisher’s method , Omnibus-Fisher , had the correct Type I error rate because it considered the omics data correlation in the model . The Omnibus-Fisher method achieves better power performance compared to its optimal version when the true disease model involves all of SNPs , RNA expressions and DNA methylations . On the other hand , the optimal Omnibus-Fisher method has better power than its regular version assuming all types of data are causal when only one type of data is actually causal . Our real data study also shows that the regular Omnibus-Fisher test has better power than the optimal test , when two or three types of data contribute to the combined p-value . Because we only consider three types of omics data in this study , assuming they all are causal could be still acceptable . However , when more omics data are analyzed , we believe that the optimal test would be more powerful for most genes than simply assuming all types of data are causal . Nevertheless , both the optimal and regular Omnibus-Fisher tests are able to detect genes with moderate separate effects , which could be overlooked by single type of data analyses . Although the optimal Omnibus-Fisher test uses perturbation to consider the correlation between omics data and search for the optimal disease model , the genome-wide data analysis could be completed within hours using multiple CPUs ( e . g . , one CPU for each chromosome ) . We adapt a stepwise manner to implement perturbation ( e . g . , more iterations for smaller p-values ) so as to save computation times when calculating large p-values . Thus , the majority of the computation time is used by genes with small p-values . However , if a large number of genes are highly associated with the phenotype , the optimal test may be infeasible due to computational intensity . In such case , the regular test is recommended . Although the regular test also involves resampling technique to calculate covariances between different types of omics data , it only requires a small number of resampling ( e . g . , the default setting is 200 times ) . Our method framework is general and flexible . Both continuous and binary traits for independent samples can be analyzed . Covariates can be easily incorporated into the model and different covariates can be used for different omics data . The regular and optimal version of Omnibus-Fisher algorithms were implemented in R ( http://www . r-project . org ) and the R package ( https://cran . r-project . org/web/packages/OmnibusFisher/index . html ) is available .
We used KM regression to calculate the gene-level p-values for association testing of a disease and SNPs , methylation markers , and expression genes . First , we test the effect of SNPs . Let there be n subjects with q genetic variants . The n × 1 vector of the continuous trait y follows a linear model: y=Xβ+Gγ+ε , when the phenotypes are binary , y follows a logistic model: logitP ( y=1 ) =Xβ+Gγ where X is an n × p covariate matrix , β is a p × 1 vector containing parameters for the fixed effects ( an intercept and p– 1 covariates ) , G is an n × q genotype matrix for the q genetic variants of interest where an additive genetic model is assumed ( i . e . , coded as 0 , 1 , or 2 representing the copies of minor alleles ) for illustration , γ is a q × 1 vector for the random effects of the q genetic variants , and ε is an n × 1 vector for the random error . The random effect γj for variant j is assumed to be normally distributed with mean zero and variance τwj; thus , the null hypothesis H0: γ = 0 is equivalent to H0: τ = 0 , which can be tested with a variance component score test [17] in the mixed model . The random variable ε is assumed to be normally distributed , and is uncorrelated with γ: γ∼N ( 0 , τW ) ε∼N ( 0 , σE2I ) , where W is a predefined q × q diagonal weight matrix for each variant and may use W = I when lacking of prior information , and σE2 is the error variance . Following the same rationale as in the derivation of the SKAT score statistic [31–33] , the test statistic is: Q= ( y−Xβ^ ) ′GWG′ ( y−Xβ^ ) /σ^E2 , when phenotypes are continuous , and Q= ( y−μ^ ) ′GWG′ ( y−μ^ ) when phenotypes are binary , where β^ is the vector of estimated fixed effects of covariates under H0 and μ^=logit−1 ( Xβ^ ) . Under the null hypothesis , the linear model is y = Xβ + ε , and the estimates are Σ^=σ^E2I=var ( y−Xβ^ ) I β^= ( X′X ) −1X′y P0=I−X ( X′X ) −1X′; the logistic model is logit P ( y = 1 ) = Xβ , and the estimates are Σ^=diag ( μ^∙ ( 1−μ^ ) ) β^= ( X′Σ^−1X ) −1X′Σ^−1y P0=Σ^−Σ^X ( X′Σ^X ) −1X′Σ^ . The statistic Q is a quadratic form and follows a mixture of chi-square distributions under H0 . Thus , Q∼∑i=1qλiχ1 , i2 , where λi are the eigenvalues of the matrix P012GWG′P012 [34] for both continuous and binary traits . The p-values can be calculated by numerical algorithms , such as Davies’ method [35] and Kuonen’s saddlepoint method [36] , which are both included in the R package . Analogously , the gene-level effects of DNA methylation markers and expression genes can be tested by replacing Gγ with Mρ and Eη , M is an n × k matrix for the k methylated loci , ρ is a k × 1 vector for the random effects of the k methylated loci , E is an n × g matrix for the RNA expression , and η is a g × 1 vector for the random effects of the RNA expression . When using microarray platform , multiple probes could map to the same gene and each probe has an expression value , which result in more than one expression value for one gene . Here , g is the number of probes for one gene . When using RNA sequencing platform , one gene can always have one expression value ( i . e . , E is an n × 1 vector and η is a scalar ) , although it is also possible to obtain the transcript ( i . e . , isoform ) level expression values . The null hypothesis is ρ = 0 for testing DNA methylation markers and η = 0 for testing expression genes . It is worth to note that all three models can have the same or different null models . In order to have one single p-value to represent the significance of a gene , we propose an approach to test if the trait is associated with any SNP , DNA methylation marker , and RNA sequencing variant . This could help researchers to screen out potentially interesting genes . Thus , after obtaining the three p-values for SNPs , DNA methylation markers , and expression genes , respectively , we used a modified Fisher’s method [37] to combine the three p-values to one . In Fisher’s method , let pi ( i = 1 , 2 , … , w ) be independent p-values obtained from n hypothesis tests . Under the null hypothesis that p-values follow a Uniform ( 0 , 1 ) distribution , the combined test statistic is equal to T=−2∑i=1wln ( pi ) that follows χ2w2 . However , within a gene , these p-values are correlated , thus the generalized Fisher’s method cannot be used directly . To address this issue , we consider a Satterthwaite approximation by approximating a scaled T statistic with a new chi-square distribution [38] . cT≈χv2 , wherec=vE ( T ) , v=2[E ( T ) ]2Var ( T ) , E ( T ) =E ( −2∑i=1wln ( pi ) ) =2wand Var ( T ) =var ( −2∑i=1wln ( pi ) ) =4w+2∑i<jcov ( −2ln ( pi ) , −2ln ( pj ) ) where w = 3 for SNPs , DNA methylation markers , and expression genes . The covariance part takes the correlations of p-values into account and can be empirically estimated by perturbations . The perturbation details are described in the following section . If the disease risk only depends on SNPs and the model with SNPs , DNA methylation markers , and expression genes is used , then the testing power will lose . Since in reality we do not know the underlying true disease model ( e . g . , only SNP effect , both SNP and RNA variant effects , or all SNP , RNA variant , and DNA methylation marker effects; totally 7 combinations ) , it is difficult to choose the correct model . Thus , it is desirable to develop a method accommodating all possible disease models to maximize power . This can be achieved by using the minimum p-value of all possible models ( 7 combinations ) as a new test statistic . Then , perturbation can be used to calculate the final p-value . The perturbation-based approach was described in Wu et al . [39] . For continuous phenotypes , with large n , under H0 the ( y−Xβ^ ) /σ^E are approximately standard normal . Then each Q= ( y−Xβ^ ) ′GWG′ ( y−Xβ^ ) /σ^E2 is essentially comprised of a vector of standard normal variables sandwiching a square matrix . Thus , we can perturb each Q by replacing ( y−Xβ^ ) /σ^E with a new , common vector of normal values to generate new score statistics . Following a similar procedure as described in Urrutia et al . [40]: 1 . Calculate the p-values for SNPs ( G ) , DNA methylation ( M ) and RNA expression ( E ) separately ( i . e . , pG ( 0 ) , pM ( 0 ) , and pE ( 0 ) ) by KM regression . 2 . For l ∈ {G , M and E} , compute Λl = diag ( λl , 1 , ⋯ , λl , ml ) , and Vl = [vl , 1 , ⋯ , vl , ml] where λl , 1 ≥ λl , 2 ≥⋯≥ λl , ml are the ml positive eigenvalues of P0l12DlWlDl′P0l12 with corresponding eigenvectors vl , 1 , ⋯ , vl , ml , where Dl ∈ {omics data matrices G , M and E} . For example , the aforementioned P012GWG′P012 is for G . 3 . Generate r ( b ) =[r1 ( b ) , ⋯ , rn ( b ) ]′ with each rj ( b ) ∼N ( 0 , 1 ) . This indicates that one subject has one rj ( b ) . If the subject has all G , M and E , the same rj ( b ) will be used for G , M and E , respectively . Thus , whether G , M and E come from the same subjects or different subjects are considered . 4 . For l ∈ {G , M and E} , rotate r ( b ) using the eigenvectors to generate rl ( b ) =Vl′r ( b ) . 5 . Compute Ql ( b ) =rl ( b ) ′Λlrl ( b ) for each l and obtain a corresponding p-value , pl ( b ) . 6 . Repeat ( 3 ) - ( 5 ) B times to obtain pG ( 1 ) , pG ( 2 ) , ⋯ , pG ( B ) , pM ( 1 ) , pM ( 2 ) , ⋯ , pM ( B ) and pE ( 1 ) , pE ( 2 ) , ⋯ , pE ( B ) for some large number B . 7 . Calculate the covariance between pG , pM and pE by using pG ( b ) , pM ( b ) , and pE ( b ) for b ∈ {0 , 1 , … , B} . 8 . Calculate the joint p-values of SNPs , DNA methylation and RNA expression ( i . e . , for b ∈ {0 , 1 , … , B} , pGM ( b ) , pGE ( b ) , pME ( b ) , and pGME ( b ) ) by Omnibus-Fisher considering p-values covariance . 9 . For l* ∈ {G , M , E , GM , GE , ME , and GME}; b ∈ {0 , 1 , … , B} , set p ( b ) =min1≤l*≤L*pl* ( b ) . 10 . The final p-value for significance is estimated as p=B−1∑b=1BI ( p ( b ) ≤p ( 0 ) )
|
In this research , we developed a statistical approach using a modified Fisher’s method ( denoted as Omnibus-Fisher ) to combine separate p-values of association testing for a trait and SNPs , DNA methylation markers , and RNA sequencing , calculated by kernel machine regression into an overall gene-level p-value to account for correlation between omics data . We further extended the method to an optimal version in order to consider all possible disease models .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"sequencing",
"techniques",
"medicine",
"and",
"health",
"sciences",
"pulmonology",
"simulation",
"and",
"modeling",
"methylation",
"epigenetics",
"dna",
"molecular",
"biology",
"techniques",
"rna",
"sequencing",
"dna",
"methylation",
"chromatin",
"asthma",
"research",
"and",
"analysis",
"methods",
"chromosome",
"biology",
"gene",
"expression",
"chemistry",
"chromatin",
"modification",
"dna",
"modification",
"molecular",
"biology",
"research",
"assessment",
"biochemistry",
"biochemical",
"simulations",
"cell",
"biology",
"nucleic",
"acids",
"genetics",
"biology",
"and",
"life",
"sciences",
"chemical",
"reactions",
"physical",
"sciences",
"computational",
"biology",
"research",
"errors"
] |
2019
|
An integrative association method for omics data based on a modified Fisher’s method with application to childhood asthma
|
Progress in modern neuroscience critically depends on our ability to observe the activity of large neuronal populations with cellular spatial and high temporal resolution . However , two bottlenecks constrain efforts towards fast imaging of large populations . First , the resulting large video data is challenging to analyze . Second , there is an explicit tradeoff between imaging speed , signal-to-noise , and field of view: with current recording technology we cannot image very large neuronal populations with simultaneously high spatial and temporal resolution . Here we describe multi-scale approaches for alleviating both of these bottlenecks . First , we show that spatial and temporal decimation techniques based on simple local averaging provide order-of-magnitude speedups in spatiotemporally demixing calcium video data into estimates of single-cell neural activity . Second , once the shapes of individual neurons have been identified at fine scale ( e . g . , after an initial phase of conventional imaging with standard temporal and spatial resolution ) , we find that the spatial/temporal resolution tradeoff shifts dramatically: after demixing we can accurately recover denoised fluorescence traces and deconvolved neural activity of each individual neuron from coarse scale data that has been spatially decimated by an order of magnitude . This offers a cheap method for compressing this large video data , and also implies that it is possible to either speed up imaging significantly , or to “zoom out” by a corresponding factor to image order-of-magnitude larger neuronal populations with minimal loss in accuracy or temporal resolution .
A major goal of neuroscience is to understand interactions within large populations of neurons , including their network dynamics and emergent behavior . This ideally requires the observation of neural activity over large volumes . Recently , light-sheet microscopy and genetically encoded indicators have enabled unprecedented whole-brain imaging of tens of thousands of neurons at cellular resolution [1] . However , light-sheet microscopy generally suffers from slow volumetric speeds ( e . g . [2] , but see also [3 , 4] ) and is usually applied to small and transparent brains . In scattering brains , current technologies with single-neuron resolution are usually based on slow , serially-scanned two-photon ( 2P ) imaging methods that can only sample from O ( 102 − 103 ) neurons simultaneously with adequate temporal resolution [5] . Recent advances have enabled faster light-sheet imaging in cortex [6] and fast volumetric 2P imaging [7] , but we must still contend with critical trade-offs between temporal and spatial resolution—and the need for even faster imaging of even larger neural populations . Another critical challenge is the sheer amount of data generated by these large-scale imaging methods . A crucial step for further neural analysis involves a transition from voxel-space to neuron-source space: i . e . , we must detect the neurons and extract and demix each neuron’s temporal activity from the video . Simple methods such as averaging voxels over distinct regions of interest ( ROIs ) are fast , but more statistically-principled methods based on constrained non-negative matrix factorization ( CNMF ) better conserve information , yield higher signal-to-noise ratio , recover more neurons , and enable the demixing of spatially overlapping neurons [8] . The methods described in [8] were not optimized for very large datasets , but NMF is a key machine learning primitive that has enjoyed more than a decade of intensive algorithmic optimization [9–12] that we can exploit here to scale the CNMF approach . We find that a very simple idea leads to order-of-magnitude speedups: by decimating the data ( i . e . , decreasing the resolution of the data by simple local averaging [13] ) , we can obtain much faster algorithms with minimal loss of accuracy . Decimation ideas do not just lead to faster computational image processing , but also offer prescriptions for faster image acquisition over larger fields of view ( FOV ) , and for observing larger neural populations . Specifically , we propose the following two-phase combined image acquisition/analysis approach . In the first phase , we use conventional imaging methods to obtain estimates of the visible neuronal locations and shapes . After this cell-identification phase is complete we switch to low-spatial-resolution imaging , which in the case of camera-based imaging simply corresponds to “zooming out” on the image , i . e . , expanding the spatial size of each voxel . This has the benefit of projecting a larger FOV onto the same number of voxels; alternatively , if the number of voxels recorded per second is a limiting factor , then recording fewer ( larger ) voxels per frame implies that we can image at higher frame-rates . We are thus effectively trading off spatial resolution for temporal resolution; if we cut the spatial resolution too much we may no longer be able to clearly identify or resolve single cells by eye in the obtained images . However , we show that , given the high-spatial-resolution information obtained in the first imaging phase , the demixing stage of CNMF can recover the temporal signals of interest even from images that have undergone radical spatial decimation ( an order of magnitude or more ) . In other words , CNMF significantly shifts the tradeoff between spatial and temporal resolution , enabling us to image larger neuronal populations at higher temporal resolution . The rest of this paper is organized as follows . We first describe how temporal and spatial decimation ( along with several other improvements ) can be used within the CNMF algorithm to gain order-of-magnitude speed-ups in calcium imaging video processing . Next we investigate how decimation can enable faster imaging of larger populations for light-sheet and 2P imaging . We show the importance of the initial cell identification phase , quantitatively illustrate how CNMF changes the tradeoff between spatial and temporal resolution , and discuss how spatially decimated imaging followed by demixing can be interpreted as a simple compression and decoding scheme . We show that good estimates of the neural shapes can be obtained on a small batch of standard-resolution data , corresponding to a short cell-identification imaging phase . Finally we demonstrate that interleaved imaging that translates the pixels by subpixel shifts on each frame further improves the fidelity of the recovered neural time series .
Constrained non-negative matrix factorization ( see Methods ) relies on the observation that the spatiotemporal fluorescence activity ( represented as a space-by-time matrix ) can be expressed in terms of a product of two matrices: a spatial matrix A that encodes the location and shape of each neuron and a temporal matrix C that characterizes the calcium concentration within each neuron over time . Placing constraints on the spatial footprint of each neuron ( e . g . , enforcing sparsity and locality of each neural shape ) and on the temporal activity ( modeling the observed calcium in terms of a filtered version of sparse , non-negative neural activity ) significantly improves the estimation of these components compared to vanilla NMF [8] . Below ( cf . Fig 1 ) , we describe a number of algorithmic improvements on the basic approach described in [8]: an iterative block-coordinate descent algorithm in which we optimize for components of A with C held fixed , then for C with A held fixed . We begin by considering imaging data obtained at low temporal resolution , specifically a whole-brain light-sheet imaging recording acquired at a rate of 2 Hz using nuclear localized GCaMP6f in zebrafish . We restricted our analysis to a representative patch shown in Fig 2A , extracted from a medial z-layer of the telencephalon ( pallium ) . ( Similar analyses were also performed on patches from midbrain and hindbrain , with similar conclusions . ) The neural centers were detected automatically using the greedy method from [8] . To ensure that the spatial components in A are localized , we constrained them to lie within spatial sub-patches ( dashed squares in Fig 2A; see also Methods ) . The first algorithmic improvement follows from the realization that some of the constraints applied in CNMF are unnecessary , at least during early iterations of the algorithm , when only crude estimates for A and C are available . Specifically , [8] imposed temporal constraints on C in each iteration: namely , C was modeled as a filtered version of a nonnegative neural activity signal S—i . e . , CG = S , for an invertible matrix G—and therefore CG is constrained to be non-negative . We found that enforcing a simpler non-negativity constraint on C instead of CG ( and then switching to impose the constraint on CG only once the estimates of A and C were closer to convergence ) led to a simpler algorithm enabling faster early iterations with no loss in accuracy . Next we found that significant additional speed-ups in this simplified problem could be obtained by simply changing the order in which the variables in this simplified block-coordinate descent scheme are updated [12] . Instead of updating the temporal activity and spatial shape of one neuron at a time ( Fig 1A , black line ) as in [8] , which is known as hierarchical alternating least squares ( HALS , [9] ) or rank-one residue iteration ( RRI , [14] ) , it turned out to be beneficial to update the activities of all neurons while keeping their shapes fixed , and then updating all shapes while keeping their activities fixed ( Fig 1A , vermilion line ) . The ensuing method is a constrained version of the fast hierarchical alternating least squares ( fast HALS , [15] ) for NMF; one major advantage of this update ordering is that in each iteration we operate on smaller matrices obtained as dot-products of the data matrix Y with A or C , and there is no need to compute the large residual matrix Y − AC ( which is of the same size as the original video ) [10 , 11] . ( In the comparisons below we computed the residual to quantify performance , but excluded the substantial time spent on its computation from the reported wall time values . ) Next we reasoned that to obtain a good preliminary estimate of the spatial shape matrix A , it is likely unnecessary to use the original data at full temporal resolution [13] . Thus we experimented with the following approach: downsample temporally by a factor of k , then run constrained fast HALS ( as described above ) for 30 iterations , and then finally return to the original ( non-downsampled ) data and run a few more iterations of fast HALS until convergence . We experimented with three different downsampling methods: 1 ) selection of the k-th frame ( this could be considered a kind of stochastic update rule , since we are forming updates based only on a subset of the data ) ; 2 ) forming a median over the data in each block of k frames ( applying the median over each pixel independently ) ; and 3 ) forming a mean over each block of k frames . The mean approach ( 3 ) led to significantly more accurate and stable results than did the subsampling approach ( 1 ) , consistent with the results of [12] , and was about an order of magnitude faster than the median approach ( 2 ) with similar accuracy , so we restrict our attention to the mean approach ( 3 ) for the remainder of this work . ( A further advantage of approach ( 3 ) relative to ( 1 ) is that ( 1 ) can miss fast activity transients . ) Fig 1B shows the results obtained for a varying number of decimation factors k; we conclude that temporal decimation provides another significant speedup over the results shown in Fig 1A . The starting point of each line is the MSE for keeping the neural shapes obtained on decimated data fixed and solving for the time series on the full data . Refining the shapes on the full data further decreases the MSE , however by less then 1% , hence good shapes are obtained even using merely decimated data . Besides downsampling methods , we also considered dimensionality reduction via structured random projections [16] or singular value decomposition ( SVD , [17] , see Methods ) . We compressed the data by the same factor k = 30 , but found that both of these dimensionality-reduction methods were less efficient than simple decimation ( Fig 1C ) . We further evaluated whether we can gain improvements by further compressing the decimated data via SVD or random projections , such that the reduced dimension is just slightly larger than the number of neurons . However , we did not obtain any improvements beyond plain decimation . We found this result to hold also for smaller patches that contained fewer neurons . Further speed gains were obtained when applying spatial decimation ( computing a mean within l × l pixel blocks ) in addition to temporal decimation over the 30 preliminary fast HALS iterations ( Fig 1C ) ; see Algorithm 1 for full details . Strikingly , spatial decimation led not only to faster but also to better solutions of the biconvex factorization problem ( where solution quality is measured by the residual sum of square errors , ||Y − AC||2 , RSS ) , apparently because the spatially-decimated solutions are near better local optima in the squared-error objective function than are the non-decimated solutions . In summary , by simplifying the early iterations of the CNMF algorithm ( by removing the temporal deconvolution constraints to use fast HALS iterations on temporally and spatially subsampled data ) , we obtained remarkable speed-ups without compromising the accuracy of the obtained solution , at least in terms of the sum-of-squares objective function . But how do these modifications affect the extracted neural shapes and activity traces ? We ran the algorithm without decimation until convergence and with decimation for 1 and 10 s respectively . Fig 2 shows the results for three neurons with overlapping patches . Both shapes and activity traces agree well even if the decimated algorithm is run for merely 1 s ( Fig 2B and 2C ) and are nearly identical if run longer ( Fig 2D ) ; hence , decimation does not impair the final obtained accuracy . Our focus has been on speeding up CNMF , one computational bottleneck of the entire processing pipeline . For completeness , we report the times spent on each step of the pipeline in Table 1 and compare to the previous CNMF version of [8] . After loading , the data was decimated temporally by a factor of 30 to speed up the detection of the neural centers using the greedy initialization method from [8] . We further decimated spatially and ran fast HALS for 30 iterations before finally returning to the whole data and performing five final fast HALS iterations . Each trace was normalized by the fluorescence at the resting state ( known as ΔF/F ) to account for baseline drift using a running percentile filter . Finally , the fluorescence traces were denoised via sparse non-negative deconvolution , using the recently developed fast method of [18] , which eliminated another computational bottleneck present in the original CNMF implementation ( last row of Table 1 ) . We have shown that decimation leads to much faster computational processing of calcium video data . More importantly , these results inspired us next to propose a method for faster image acquisition or for imaging larger neural populations . The basic idea is quite simple: if we can estimate the quantities of interest ( A and C ) well given decimated data , then why collect data at the full resolution at all ? Since spatial decimation by a factor of l conceptually reduces the number of pixels recorded over a given FOV by a factor of l2 ( though of course this situation is slightly more complex in the case of scanning two-photon imaging; we will come back to this issue below ) , we should be able to use our newly-expanded pixel budget to image more cells , or image the same population of cells faster . As we will see below , this basic idea can be improved upon significantly: if we have a good estimate for the spatial neural shape matrix A at the original spatial resolution , then we can decimate more drastically ( thus increasing this l2 factor ) with minimal loss in accuracy of the estimated activity C . This , finally , leads to the major proposal of this paper: first perform imaging with standard spatial resolution via conventional imaging protocols . Next perform the ROI detection and CNMF described above to obtain a good estimate of A . Then begin acquiring spatially l-decimated images and use the C-estimation step of CNMF to extract and demix the imaged activity . As we will see below , this two-phase imaging approach can potentially enable the accurate extraction of demixed neural activity even given quite large decimation factors l , with a correspondingly large increase in the resulting “imaging budget . ”
The basic message of this paper is that standard approaches for imaging calcium responses in large neuronal population—which have historically been optimized so that humans can clearly see cells blink in the resulting video—lead to highly redundant data , and we can exploit this redundancy in several ways . In the first part of the paper , we saw that we can decimate standard calcium imaging video data drastically , to obtain order-of-magnitude speedups in processing time with no loss ( and in some cases even some gain ) in accuracy of the recovered signals . In the second part of the paper , we saw that , once the cell shapes and locations are identified , we can drastically reduce the spatial resolution of the recording ( losing the ability to cleanly identify cells by eye in the resulting heavily-pixelated movies ) but still faithfully recover the neural activity of interest . This in turn leads naturally to a proposed two-phase imaging approach ( first , identify cell shapes and locations at standard resolution; then image at much lower spatial resolution ) that can be seen as an effort to reduce the redundancy of the resulting video data . We anticipate a number of applications of the results presented here . Regarding the first part of the paper: faster computational processing times are always welcome , of course , but more fundamentally , the faster algorithms developed here open the door towards guided experimental design , in which experimenters can obtain images , process the data quickly , and immediately use this to guide the next experiment . With more effort this closed-loop approach can potentially be implemented in real-time , whether for improving optical brain-machine interfaces [26] , or enabling closed-loop optogenetic control of neuronal population dynamics [27 , 28] . Highly redundant data streams are by definition highly compressible . The results shown in S1 and S2 Videos illustrate clearly that spatially-decimated image acquisition ( the second phase of our two-phase imaging approach ) can be seen as a computationally trivial low-loss compression scheme . Again , regarding applications of this compression viewpoint: reductions in memory usage are always welcome—but more fundamentally , this type of compression could for example help enable wireless applications in which bandwidth and power-budget limitations are currently a significant bottleneck [21 , 29–31] . Regarding applications of the proposed two-phase imaging approach: we can potentially use this approach to image either more cells , or image cells faster , or some combination of both . In most of the paper we have emphasized the first case , in which we ‘zoom out’ to image larger populations at standard temporal resolution . However , a number of applications require higher temporal resolution . One exciting example is the larval zebrafish , where it is already possible to image the whole brain , but light-sheet whole-brain volumetric imaging rates are low [1] and current efforts are focused on faster acquisition [3 , 4 , 32] . Higher temporal resolution is also needed for circuit connectivity inference [33 , 34] or the real-time closed-loop applications discussed above , where we need to detect changes in activity as quickly as possible . Finally , genetically encoded voltage indicators [35] may soon enable imaging of neuronal populations with single-cell , millisecond-scale resolution; these indicators are still undergoing intense development [36–39] but when more mature the resulting signals will be much faster than currently-employed calcium indicators and significantly higher temporal resolution will be required to capture these signals . A number of previous papers can be interpreted in terms of reducing the redundancy of the output image data . Our work can be seen as one example of the general theme of increasing the ratio N/D , with N denoting the number of imaged neurons and D the number of observations per timestep , with demixing algorithms used post hoc to separate the overlapping contributions of each cell to each observed pixel . In a compressed sensing framework , [40] proposed to image randomized projections of the spatial calcium concentration at each timestep , instead of measuring the concentration at individual locations . In [41] , [42] , and [43] , information is integrated primarily across depth , either by creating multiple foci , axially extended point spread functions ( PSFs ) , or both , respectively . In contrast to these methods , [7] instead scanned an enlarged near-isotropic PSF , generated with temporal focusing , to quickly interrogate cells in a single plane at low spatial resolution . This approach is closest in spirit to the one-phase spatially decimated imaging approach analyzed in Figs 6–8 , and could potentially be combined with our two-phase approach to achieve further speed/accuracy gains . We expect that different strategies for increasing N/D will have different advantages in different situations . One advantage of the approach developed here is its apparent simplicity—at least at a conceptual level , we just need to ‘zoom out’ without the need for radically new imaging hardware . Throughout this work we have remained deliberately agnostic regarding the physical implementation of the spatial decimation; all of the decimation results presented here were based on software decimation after acquisition of standard-resolution images . Thus to close we turn now to a discussion of potential experimental caveats . One critical assumption in our simulations is that the total recorded photon flux per frame is the same for each decimation level l . This is a reasonable assumption for light-sheet imaging ( assuming we are not limited by laser power or by the peak or average light power on the sample ) : in this case , increasing the effective pixel size could be achieved easily , either isotropically with a telescope , or anisotropically , with a cylindrical lens or anamorphic prism pair . However , faster whole-brain light-sheet imaging requires faster shifts of the light sheet and imaged focal plane . This challenge can be solved by extended depth-of-field ( EDoF ) pupil encoding [3 , 4 , 32] , remote focusing [44] , or with an electrically tunable lens [45] . Higher light-sheet imaging rates can also be obtained with swept confocally-aligned planar excitation ( SCAPE ) microscopy [6] . In short , we believe our proposed two-phase imaging approach fits well with a variety of proven light sheet methods; for similar reasons , the two-phase approach would also fit well with light-field imaging methods [46–48] . In traditional two-photon imaging the situation is more complicated . The image is created by serially sweeping a small , diffraction limited point across the sample . Along the “fast” axis , the beam moves continuously , and the integrated signal across a line is constant , regardless of detection pixelation—the signal is simply partitioned into more or fewer bins . Along the “slow” axis , however , the galvonometers are moved in discrete steps , and low pixel numbers generally mean that portions of the image are not scanned , increasing frame speed , but concomitantly these ‘missed’ areas generate no signal . This consequently reduces the total number of photons collected . Thus to achieve the same photon flux over the larger ( lower spatially sampled ) pixels , while maintaining the same SNR , we require an enlarged PSF , which maps a larger sampled volume to each pixel . This approach was recently demonstrated to be effective in [7]; alternative strategies for enlarging the PSF could involve fixed diffractive optical elements [49] or spatial light modulator ( SLM ) systems [50] . Programmable phase-only SLMs offer the additional benefit of being able to dynamically change the size and shape of the excitation PSF , even between frames , which may help disambiguate closely spaced sources , and effectively control the recorded source sparsity . In any instantiation , maximal imaging speed will be limited by the time required to collect enough photons for adequate SNR , which in turn is limited by photophysics and the light tolerance of the sample . In future work we plan to pursue both light-sheet and 2P implementations of the proposed two-phase imaging approach , to quantify the gains in speed and FOV size that can be realized in practice . We also expect techniques for denoising , demixing , and deconvolution of calcium imaging video to continue to improve in the near future , as more accurate nonlinear , non-Gaussian models for calcium signals and noise are developed; as new demixing methods become available , we can easily swap these methods in in place of the CNMF approach used here . We expect that the basic points about temporal and spatial decimation discussed in this paper will remain valid even as newer and better demixing algorithms become available .
Light-sheet imaging of zebrafish was conducted according to protocols approved by the Institutional Animal Care and Use Committee of the Howard Hughes Medical Institute , Janelia Research Campus . Two-photon imaging of mouse was carried out in accordance with animal protocols approved by the Columbia University Institutional Animal Care and Use Committee . The calcium fluorescence of the whole brain of a larval zebrafish was recorded using light-sheet imaging . It was a transgenic ( GCaMP6f ) zebrafish embedded in agarose but with the agarose around the tail removed . The fish was in a fictive swimming virtual environment as described in [51] . The closed loop setting , characterized by visual feedback being aligned with the recorded motor activity , was periodically interrupted by open loop phases . Whole-brain activity was recorded for 1 , 500 seconds with a rate of 2 volumes per second . In vivo two-photon imaging was performed in a transgenic ( GCaMP6s ) mouse through a cranial window in visual cortex . The mouse was anesthetized ( isoflurane ) and head-fixed on a Bruker Ultima in vivo microscope with resonant scanners , and spontaneous activity was recorded . The field of view extended over 350 μm × 350 μm and was recorded for 100 seconds with a resolution of 512×512 pixels at 20 frames per second . In the case of 2P imaging , the field of view contained N = 187 ROIs . It was observed for a total number of T = 2 , 000 timesteps and had a total number of D = 512×512 pixels . We restricted our analysis of the zebrafish data to a representative patch of size D = 96×96 pixels containing N = 46 ROIs , extracted from a medial z-layer of the whole-brain light-sheet imaging recording of T = 3 , 000 frames . The observations at any point in time can be vectorized in a single column vector of length D; thus all the observations can be described by a D × T matrix Y . Following [8] , we model Y as Y = A C + b f ⊤ + E ( 1 ) where A ∈ R + D × N is a spatial matrix that encodes the location and shape of each neuron , C ∈ R + N × T is a temporal matrix that characterizes the calcium concentration of each neuron over time , b ∈ R + D , f ∈ R T are nonnegative vectors encoding the background spatial structure and global intensity , respectively , and E is additive Gaussian noise with mean zero and diagonal covariance . Our model assumes a linear relationship between fluorescence and calcium concentration as well as Gaussian noise . As emphasized in [8] , more elaborate models for C can be incorporated in the alternating updates , but we did not pursue this generalization here . For the zebrafish data we ensured that the spatial components are localized , by constraining them to lie within spatial patches ( which are not large compared to the size of the cell body ) around the neuron centers , thus imposing sparsity on A by construction . Because of the low temporal resolution of these recordings , the inferred neural activity vectors are not expected to be particularly sparse , and therefore we do not impose sparsity in the temporal domain . This leads to the optimization problem minimize A , b , C , f ∥ Y - A C - b f ⊤ ∥ subject to : A , b , C ≥ 0 , A ( d , n ) = 0 ∀ d ∉ P n ( 2 ) where Pn denotes the n-th fixed spatial patch . This problem is biconvex , i . e . solving for C and f with fixed A and b is a convex problem; likewise solving for A and b with fixed C and f is convex . As discussed in the Results section , we solve this problem by block-coordinate descent , first applied to much smaller decimated data and then using this solution as a warm start for the optimization on the full data . In the resulting Algorithm 1 we appended for concision b and f as an additional column or row to A and C respectively . The matrix products A⊤Y and CY⊤ in Algorithm 1 are computationally expensive for the full data . These matrix products can also be performed on GPU instead CPU; whereas for the comparatively small 96×96 patches we did not obtain any speed-ups using a GPU , we verified on patches of size 256×256 that some modest overall speedups ( a factor of 1 . 5–2 ) can be obtained by porting this step to a GPU . For the decimated data the matrix products are cheap enough to iterate just once over all neurons and instead alternate more often between updating shapes and activities ( instead of performing many iterations within HALSactivity or HALSshape in Algorithm 1 ) . In early iterations our estimates of A and C are changing significantly and it is better to perform just one block-coordinate descent step for each neuron to update A ( and similarly for C ) ; for later iterations , and on the full data where it is more expensive to compute A⊤Y and CY⊤ , we increase the inner iterations in HALSactivity or HALSshape . Algorithm 1 is a constrained version of fast HALS . To further improve on fast HALS , [52] suggested to replace cyclic variable updates with a greedy selection scheme focusing on nonzero elements . This was unnecessary here because most nonzero elements are prespecified by the patches Pn; i . e . , we are already focusing on the nonzero elements . Because for the 2P data the observed imaging rate is much higher than the decay rate of the calcium indicator , we constrain the temporal traces C to obey the calcium indicator dynamics , to enable further denoising and deconvolution of the data . As in [8] , we approximate the calcium concentration dynamics using an autoregressive process of order 2 ( AR ( 2 ) ) , C ( n , t ) = γ 1 C ( n , t - 1 ) + γ 2 C ( n , t - 2 ) + S ( n , t ) , ( 3 ) where S ( n , t ) is the number of spikes that neuron n fired at timestep t . This equation can be conveniently expressed in matrix form as S = CG for a suitable sparse matrix G . We estimate the noise level of each pixel σd by averaging the power spectral density ( PSD ) over a range of high frequencies , and estimate the coefficients of the AR ( 2 ) process for each cell following [22] . Then we solve for A , b , C , f using the following iterative matrix updates: minimize A , b ∥ A ∥ 1 , subject to : A , b ≥ 0 , ∥ Y ( d , : ) - A ( d , : ) C - b ( d ) f ⊤ ∥ ≤ σ d T ∀ d ∈ { 1 , 2 , . . . , D } ( 4 ) minimize C , f ∥ C G ∥ 1 , subject to : C G ≥ 0 , ∥ Y ( d , : ) - A ( d , : ) C - b ( d ) f ⊤ ∥ ≤ σ d T ∀ d ∈ { 1 , 2 , . . . , D } . ( 5 ) These updates are initialized with the results from constrained fast HALS ( Alg 1 ) . They impose sparsity on the spatial as well as temporal components , using the estimates of the noise variance as hard constraints to derive a parameter-free convex program . Following the approach in [53] the spike signal S is relaxed from nonnegative integers to arbitrary nonnegative values . The basis pursuit denoising problems in Eqs ( 4 and 5 ) can be solved with one of the methods described in [8] . However , a faster update of the temporal matrix C is achieved by using OASIS [18] . Every spatial component in A was normalized to have unit ℓ2-norm , with the corresponding temporal component scaled accordingly . Following [8] we then sort the components according to the product of the maximum value that the temporal component attains and the ℓ4-norm of the corresponding spatial footprints , to penalize overly broad and/or noisy spatial shapes . In order to calculate the ΔF/F values we divided the fluorescence trace C ( n , : ) of each neuron by its baseline fluorescence that was obtained by projecting the rank-1 background bf⊤ onto the shape A ( : , n ) of the neuron , ( Δ F F ) t ≔ A ( : , n ) ⊤A ( : , n ) C ( n , t ) A ( : , n ) ⊤ b f ( t ) . While we normalized A and C such that A ( : , n ) ⊤A ( : , n ) = 1 , the ΔF/F values do not depend on this specific normalization because the scaling factor cancels out . To compress the data using truncated SVD in Fig 1C , we followed [17] and computed the eigenvectors V ∈ R M × T belonging to the M largest eigenvalues of the time by time covariance matrix Y⊤Y , which were then used to obtain the compressed data YV⊤ . This method was faster than the randomized method due to [54] . The spatial background b for the compressed data was again initialized as 20% percentile of the original data and the temporal background as f ˜ = V 1 T . Because the compressed data and temporal traces can be negative we did not enforce the non-negativity constraint of C ˜ and f ˜ , which was crucial as enforcing it indeed yielded worse results . Random compression was performed as described in [16] . We applied the structured random compression algorithm [54] to Y and Y⊤ to obtain L ∈ R D × M and R ∈ R M × T . Specifically , we drew a Gaussian random matrix Ω ∈ R T × M and performed the QR decomposition of YΩ to obtain an orthonormal basis L . Analogously we obtained R for Y⊤ . The iterated alternating fast HALS updates were C ← HALSactivity ( L⊤Y , L⊤A , C ) and A ← HALSshape ( YR⊤ , A , CR⊤ ) , with L⊤Y and YR⊤ computed once initially . Applying the code of [55] to the raw data we identified the neural shape matrix A1 and spatial background b1 . We use the convention that the presence of a lower index l signifies an estimate and its value the decimation factor , i . e . index l = 1 denotes an estimate inferred without decimation . Further , we also obtained the denoised and deconvolved traces C1 , S1 as well as f1 . To emulate imaging with lower spatial resolution , spatial decimation was performed by converting A1 back into a 512 × 512 × N tensor ( Y into a 512 × 512 × T video tensor ) and calculating the average of non-overlapping patches of size l × l or l × 1 pixels for each of the N neural shapes ( T timesteps ) . Converting the tensors back to matrices yielded the decimated neural shapes Al ( data Yl ) . We proceeded analogously for the spatial background to obtain bl . The corresponding temporal traces were estimated by solving Eq ( 5 ) ( with Yl replacing Y , Cl replacing C , etc . ) , initializing Cl and fl with the result of plain NMF that does not impose temporal constraints , i . e . solving minimizeC l , f l ∥ Y l - A l C l - b l f l ⊤ ∥ subject to Cl ≥ 0 . In order to obtain the results for 1-phase imaging without previous shape identification we solved Eq ( 4 ) for the decimated data Yl , initializing Al , bl by decimating A1 , b1 and setting the temporal components to C1 , f1 . With increasing decimation factors an increasing number of shapes got purged and absorbed in the background , reflecting the fact that it would have been difficult to detect all ROIs on low resolution data in the first place . Using the obtained remaining shapes we again solved Eq ( 5 ) as above . The correlation values for purged neurons were set to zeros for the mean values reported in Fig 6 . To obtain some form of ground truth ( Figs 6D–6F , 7C and 8C ) we generated a simulated dataset Ys by taking the inferred quantities as actual ground truth: As ≔ A1 , bs ≔ b1 , Cs ≔ C1 , fs ≔ f1 . We calculated the residual Y − AsCs − bsfs⊤ and reshuffled it randomly but signal dependent for each pixel in time . We partitioned the residual for each pixel into 200 strata according to signal size and reshuffled it within each strata , thus retaining any potential link between noise variance and signal mean . The simulated dataset Ys was obtained by adding the reshuffled residual to AsCs + bsfs⊤ and the same analysis as for the original data was performed . We performed additional control simulations that also took the inferred quantities as actual ground truth , Y * : = A 1 C 1 + b 1 f 1 ⊤ , but did not rely on a reshuffling procedure . Instead , we either added Gaussian noise , Y d , t N ∼ N ( Y d , t * , σ 2 ) , or Poisson noise ν + Y d , t P ∼ P ( ν + Y d , t * ) , where ν is the photon count that is not due to calcium fluorescence ( but rather dark counts and background light ) . Whereas the Gaussian noise had fixed variance σ2 , the Poisson model results in heteroscedastic noise because its variance grows with the mean . The variance of the Gaussian noise was chosen to be equal to the average variance of the Poisson noise . The results shown in S2 Fig agree with those obtained by reshuffling the residual and are similar for Gaussian and Poisson noise , at least on average , though there is some spread if individual traces are considered ( S2C Fig ) . Whereas the model ( Eq 1 ) assumes Gaussian noise , it nevertheless performs well under Poisson noise , consistent with the results of [53] . Another control simulation merely took A1 , b1 and f1 as ground truth . However , instead of taking the denoised fluorescence traces C1 , which by construction followed the autoregressive model , the fluorescence traces C D were obtained from two datasets that combined electrophysiological recording and calcium imaging with GCaMP6f ( 11 cells ) or GCaMP6s ( 9 cells ) [23] . The calcium response kernel k ^ for each recorded cell was determined by solving the linear regression problem k ^ = arg min k ∥ y - s * k ∥ 2 where y is the noisy fluorescence data of a neuron , s its spike counts per bin , and * denotes convolution . Because only few cells were recorded , but for a longer duration than the 100 s of our two-photon dataset , we assigned to each of our ROIs the ground truth fluorescence trace by randomly selecting a recorded cell , taking a 100 s interval of its spike train that included at least three spikes , and convolving it with the cell’s kernel k . The trace was scaled such that its maximum value was equal to the maximal value of the inferred trace C1 . Poisson noise was added , ν + Y D ∼ P ( ν + A 1 C D + b 1 f 1 ⊤ ) , as described above . Results are shown in S3 Fig . Projecting the noise of each pixel onto the neural shapes yields the noise of each neural time series . In practice the latter is estimated based on the noisy trace obtained by projecting the fluorescence data onto the shapes . For interleaved imaging ( Fig 8 ) the shape of each neuron differs between alternating frames due to the varying pixelization . Therefore , instead of using one noise level for all timesteps , we estimated two noise levels σodd and σeven based on the PSD for all odd and even frames respectively . The residuals in the noise constraint of the non-negative deconvolution were weighted accordingly by the inverse of the noise level . where y is the noisy fluorescence data of a neuron ( cell index suppressed ) obtained by subtracting the contribution of all other neurons as well as the background from the spatio-temporal raw pixel fluorescence data and projecting the odd and even frames of the result onto the considered neuron’s shapes aodd and aeven respectively . yodd and yeven denote the vectors obtained by taking only every second component of y starting with the first/second respectively . The denoised fluorescence c is denoted analogously . For simplicity we estimated the coefficients of the AR ( 2 ) process based on all frames without separating by noise level . All analyses were performed on a MacBook Pro with Intel Core i5-5257U 2 . 7 GHz CPU and 16 GB RAM . We wrote custom Python scripts that called the Python implementation [55] of CNMF [8] . Our scientific Python installation included Intel Math Kernel Library ( MKL ) Optimizations for improved performance of vectorized math routines .
|
The voxel rate of imaging systems ultimately sets the limit on the speed of data acquisition . These limits often mean that only a small fraction of the activity of large neuronal populations can be observed at high spatio-temporal resolution . For imaging of very large populations with single cell resolution , temporal resolution is typically sacrificed . Here we propose a multi-scale approach to achieve single cell precision using fast imaging at reduced spatial resolution . In the first phase the spatial location and shape of each neuron is obtained at standard spatial resolution; in the second phase imaging is performed at much lower spatial resolution . We show that we can apply a demixing algorithm to accurately recover each neuron’s activity from the low-resolution data by exploiting the high-resolution cellular maps estimated in the first imaging phase . Thus by decreasing the spatial resolution in the second phase , we can compress the video data significantly , and potentially acquire images over an order-of-magnitude larger area , or image at significantly higher temporal resolution , with minimal loss in accuracy of the recovered neuronal activity . We evaluate this approach on real data from light-sheet and 2-photon calcium imaging .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"fluorescence",
"imaging",
"statistical",
"noise",
"applied",
"mathematics",
"neuroscience",
"simulation",
"and",
"modeling",
"algorithms",
"mathematics",
"statistics",
"(mathematics)",
"computational",
"neuroscience",
"neuroimaging",
"research",
"and",
"analysis",
"methods",
"gaussian",
"noise",
"imaging",
"techniques",
"animal",
"cells",
"calcium",
"imaging",
"cellular",
"neuroscience",
"cell",
"biology",
"single",
"neuron",
"function",
"neurons",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"physical",
"sciences",
"computational",
"biology"
] |
2017
|
Multi-scale approaches for high-speed imaging and analysis of large neural populations
|
Understanding the genetic basis of adaption is a central task in biology . Populations of the honey bee Apis mellifera that inhabit the mountain forests of East Africa differ in behavior and morphology from those inhabiting the surrounding lowland savannahs , which likely reflects adaptation to these habitats . We performed whole genome sequencing on 39 samples of highland and lowland bees from two pairs of populations to determine their evolutionary affinities and identify the genetic basis of these putative adaptations . We find that in general , levels of genetic differentiation between highland and lowland populations are very low , consistent with them being a single panmictic population . However , we identify two loci on chromosomes 7 and 9 , each several hundred kilobases in length , which exhibit near fixation for different haplotypes between highland and lowland populations . The highland haplotypes at these loci are extremely rare in samples from the rest of the world . Patterns of segregation of genetic variants suggest that recombination between haplotypes at each locus is suppressed , indicating that they comprise independent structural variants . The haplotype on chromosome 7 harbors nearly all octopamine receptor genes in the honey bee genome . These have a role in learning and foraging behavior in honey bees and are strong candidates for adaptation to highland habitats . Molecular analysis of a putative breakpoint indicates that it may disrupt the coding sequence of one of these genes . Divergence between the highland and lowland haplotypes at both loci is extremely high suggesting that they are ancient balanced polymorphisms that greatly predate divergence between the extant honey bee subspecies .
Genetic adaptation to different environmental conditions is a key process in evolution and speciation . However , identifying the genetic variants involved in adaptation and the underlying regulatory networks and biological mechanisms by which they impact fitness is challenging . There are relatively few instances where the genetic basis of environmental adaptation is well understood [1] . Some examples where genetic variation has been linked to locally adaptive phenotypic differences are the pigmentation differences in the rock pocket mouse driven by variation in at least one melanocortin receptor [2] , the industrial melanism of the peppered moth driven by a transposal element in the cortex gene [3 , 4] and the adaptive evolution of populations of sticklebacks [5] , including the pelvic reduction driven by recurrent deletion of a tissue specific enhancer [6] . Studies of highland populations have proven informative for understanding the genetic basis of adaptation [7–11] . First , they inhabit different environments in close proximity to lowland populations . Genetic exchange or recent ancestry between the highland and neighboring lowland populations is therefore likely to result in low differentiation in neutrally evolving markers between highland and lowland populations , making it easier to distinguish loci involved in local adaptation . Secondly , analysis of interconnected populations spanning different habitats affords the opportunity to determine how processes such as convergent evolution [12] or adaptation from standing variation [13] have contributed to their adaptations . For example , genomic analysis of genetic adaptation of human populations living at high altitudes on three continents have revealed that convergent evolution involving selection on variants in different genes related to adaptation to hypoxia are responsible for their adaptations [8 , 14–16] . Conversely , analysis of freshwater adaptation in sticklebacks has implicated that a suite of genetic variants are present in multiple geographically distant localities , implicating selection on standing variation [5 , 17] . Most genes demonstrated to be involved in adaptation have effects on morphology or physiology [1 , 18] . However , some studies have also identified putatively adaptive variation involved in differences in fitness related to behavior [19] , such as genes that control variation in burrow architectures of Peromyscus mice [20] . Genome comparisons allow us to identify genes involved in local adaptations to different habitats where the phenotypic nature of these adaptations is not necessarily well characterized [18] . Many adaptations in social insects are likely to be behavioral [21] . In particular , honey bees have sophisticated cognitive abilities , which are needed to efficiently perform the diverse set of tasks necessary for optimal functioning of a colony . Furthermore , efficient foraging requires recognition of floral scents , location of flowers , association with a food reward and advertising food sources with a characteristic dance [22] . Optimal foraging strategies are likely to be variable between habitats and subject to selection [23] . The honey bee Apis mellifera has a large native range incorporating a wide variety of habitats . There is substantial variation in morphology , physiology and behavior across this range , which are likely to represent local adaptations [24 , 25] . The mountain regions of East Africa are highly complex in their topography with scattered high mountains , most of them of volcanic origin and comprising of three distinct vegetation belts: montane forest , subalpine heathlands and an alpine zone [26] . The average annual temperature of mountain rain forest habitats at 2600 m altitude is only 11 . 2°C , remarkably different from lowland savannah regions below 1500 m ( 20 . 8°C ) [24] . The honey bees found in the mountain forests differ in phenotype compared to the bees of the savannah . They have been designated as a separate subspecies A . m . monticola Smith 1961 [27] , whereas savannah bees have been assigned to A . m . scutellata Lepeletier de Saint Fargeau 1836 [24] . Mountain and savannah bees can be distinguished on the basis of morphometrics , although the status of monticola as a distinct subspecies from scutellata has been a matter of debate [28–32] . Bees from colonies identified as monticola tend to be darker in color , larger and less aggressive than scutellata savannah bees [24 , 27 , 33] . Measurement of mating frequencies indicates that levels of polyandry in monticola honey bees are significantly lower than in scutellata [34] . Descriptions of the behavior of monticola honey bees suggest that they can fly at lower temperatures than scutellata colonies , conserve honey stores during times of reduced nectar flow by reducing brood rearing and are less prone to abandon their nests by swarming or absconding [24 , 27 , 32 , 35 , 36] . It is therefore likely that monticola honey bees possess adaptations for life in cool mountain forests . The population history of the mountain bees is debated . The mountain refugia hypothesis proposes that mountain bees have survived as small and reproductively isolated populations for thousands of generations [29 , 31] . Such isolation can be expected to result in distinctly different patterns of genetic diversity compared to the widespread lowland bees . The results of a study based on mtDNA data supported this scenario [31] . However , a separate study of mtDNA and microsatellites did not identify genetic differences between the monticola honey bees and surrounding lowland populations [28] , which could suggest that the phenotypes observed in monticola honey bees represent phenotypic plasticity [28] or that they are determined genetically but have not led to reproductive isolation . A perennial hybridization of monticola with scutellata in a transitory zone of altitude has previously been reported [24 , 27] . In this study , we compare whole-genome sequences from 39 worker bees representing two Kenyan mountain areas that are approximately 100km apart: Mt Kenya and Mau ( Fig 1 ) . Each locality includes unmanaged bees from neighboring highland forest and lowland savannah environments that are separated by approximately 1000m in altitude . We aim to clarify the evolutionary origin of these populations and the genetic basis of their adaptation to high-altitude habitats .
We mapped genome variation in the East African mountain honey bees ( A . m . monticola ) in order to infer the evolutionary history of the population and identify loci involved in adaptation to altitude . We sequenced 39 samples previously collected from four native feral populations [28]: Mt Kenya Forest ( MKF , 2 , 300 m above sea level; n = 10 ) , Mt Kenya Savannah ( MKS , 1 , 100 m above sea level; n = 9 ) , Mau Forest ( MF , 2 , 900 m above sea level; n = 10 ) and Mau Savannah ( MS , 1 , 900 m above sea level; n = 10 ) ( Fig 1; S1A Table ) . The bees collected from the highland forest localities are referred to as A . m . monticola ( hereafter monticola ) , whereas lowland savannah samples are referred to as A . m . scutellata ( hereafter scutellata ) . Samples from the highland and lowland regions could be separated by morphometrics in a previous study [28] . We produced 490 million short-reads to generate a 463× dataset spanning the sixteen assembled nuclear chromosomes , unplaced contigs and mitochondrial genome ( see Materials and methods ) . Some unplaced contigs and the mitochondrion in particular were sequenced at extremely high coverage , inflating the average coverage . The assembled nuclear genome was sequenced to 10 . 4× per sample ( 80% of the genome was covered by >5× per sample ) and unless indicated otherwise , the results below refer to analyses of this data ( S1A Table ) . We next called single nucleotide polymorphisms ( SNPs ) across the 39 samples . 8 . 6 million biallelic SNPs were retained after filtering and imputation ( Table 1 ) . For some comparative analyses , we expanded the dataset to include previously published Kenyan honey bee genomes ( n = 11; data from [37] ) and a worldwide sample of honey bees ( n = 98; data from [25]; S1B and S1C Table ) . This makes it possible to position the Mt Kenya and Mau samples among other honey bee populations and detect uniquely divergent regions in highland genomes . The expanded dataset was produced using the same methods and spanned 13 . 6 million SNPs . The genome sequence of the Eastern honey bee A . cerana [38] was aligned against the A . mellifera reference genome in order to further facilitate assessments of divergence and distinguish between ancestral and derived variants . The corresponding A . cerana sequence was present at 78% of the A . mellifera genome . Genome-wide divergence between the two species was estimated as 6 . 9% . The mountain refugia hypothesis suggests that monticola populations are small relicts that have been reproductively isolated from lowland scutellata bees [31] . This hypothesis makes predictions about genetic variation in highland bees compared to lowland bees . Assuming that small populations have comparatively low effective population sizes ( NE ) , we can expect lower levels of neutral variation in monticola than in scutellata under equilibrium [39 , 40] . The number of SNPs detected within each of the four populations ranged between 5 . 5–5 . 9 million , corresponding to nearly identical estimates of nucleotide diversity ( π = 0 . 65–0 . 67%/bp ) , the population mutation rate ( θw = 0 . 75–0 . 78%/bp ) and effective sizes of each population ( NE = 470×103–488×103 ) ( Table 1 ) . We do not observe reduced variation in highland bees . The hypothesis also predicts that highland populations share a common ancestral population and evolutionary history separate from other bees [31] . Accordingly , we should expect highland genomes to diverge from lowland genomes . We therefore calculated genome-wide FST between populations using the Reynolds et al . estimator [41] . FST between the Mt Kenya and Mau monticola and scutellata populations range between 0 . 05 and 0 . 068 and they all group with other Kenyan and African bees ( neighbor-joining tree; Fig 2A; Table 1 ) . Among the Kenyan bees , the monticola populations cluster on a short separate branch in the population tree ( Fig 2A ) . Likewise , average pairwise genetic distances ( dXY ) split monticola and scutellata samples into different groups ( Fig 2B ) . While this could indicate limited degrees of independent evolutionary history , we note that the excess distance between monticola and scutellata samples is very small: dXY is only 1 . 02x higher between any random pair of highland and lowland samples compared with samples drawn within either habitat . We thus find that monticola populations do not diverge strongly from other Kenyan bees and that highland and lowland bees appear to be nearly undifferentiated . It should be noted that all Kenyan populations cluster to the exclusion of the Nigerian adansonii and South African scutellata and capensis populations ( the “sub Saharan” bees from [25] ) . Some of this increased divergence may be artificial and result from technical differences in short-read sequencing and mapping technologies used to assemble the extended dataset ( Illumina paired-end reads+BWA herein and in [37] vs SOLiD fragments+Lifescope in [25] ) . For instance , the mean FST between the four Kenyan populations ( Illumina ) and the three European M-group ( SOLiD ) populations is 0 . 392 , whereas it is 0 . 358 between the three African populations previously sequenced on SOLiD and the same European populations . We therefore estimate the magnitude of this “technology-bias” to be an increase in divergence of 9% , assuming that the distances should be the same . However , this bias does not affect our comparisons of the highland and lowland populations , which were all sequenced on the same technology and processed identically . To determine the relationship between the closely related highland forest and lowland savannah bees , accommodating the possibility of contradictory genealogies across the genome , we inferred FST and the corresponding population interrelationships in 10 kbp segments . In contrast to the whole-genome signal , we found that the most common pattern of relatedness in the genome groups populations by locality ( 40% of windows; Fig 2C ) . The pattern that groups populations by habitat is the least common ( 29% ) . By partitioning the inference by chromosome , we further found that the latter pattern is recovered at approximately the same frequency on all chromosomes ( 25%–31% ) except on chromosome 9 , where it is significantly enriched ( 49%; p<10−5; Fisher’s exact test; Fig 2D ) . The major pattern of relatedness across the genome is therefore consistent with exchange of genetic material between local highland and lowland populations , whereas signals that cluster the populations by environment are restricted to a smaller proportion of the genome . Taken together , these analyses suggest that the extant populations of highland and lowland honey bees have the same evolutionary origin and are not isolated from each other . Our results therefore disagree with the mountain refugia hypothesis . Genetic diversity is nearly undifferentiated between highland and lowland populations . Nevertheless , highland genomes appear to contain a small set of loci that are different from lowland genomes . These are putative targets of natural selection . Variants that are shared between the geographically separated highland populations but absent in local and closely related lowland populations could be associated with adaptation to high altitudes . We calculated FST ( Weir-Cockerham estimator [42] ) for every SNP segregating between highland samples and lowland samples in order to produce a high resolution map of differentiation across the full dataset ( ~29 bp/SNP ) and detect such loci . The result corroborates the whole-genome estimates above . Divergence is low across the genome: genome-wide FST is only 0 . 036 and 7 . 7 million ( 97% ) of SNPs have FST<0 . 1 ( Fig 3A and 3B ) . The striking exceptions are two regions on chromosomes 7 and 9 , hereafter called “r7” and “r9” . Out of the 24 , 445 SNPs that segregate with FST>0 . 5 , only 4 occur on other chromosomes or outside of these regions ( Fig 3A ) . The same divergent regions are identified when the two highland/lowland population pairs are analyzed independently ( S1 Fig ) . A strong association between highland and lowland habitats and these two chromosomal regions were detected in a genome-wide association study ( GWAS ) using PLINK , where SNPs within the r7 and r9 regions are clear outliers from the expected distribution of allele frequency differences between two groups as indicated by a Q-Q plot ( S2 Fig ) . We delineated the r7 and r9 regions by their first and last SNPs with FST>0 . 5 ( Table 2 ) , respectively . r7 bridges across four scaffolds and is composed of two blocks that together span 0 . 573 Mbp close to the start of chromosome 7 ( i and ii; Fig 3C ) . Although these blocks are separated by almost 0 . 9 Mbp and appear to be discontinuous , we suspect that the second smaller block is located on a misoriented scaffold in an ambiguous region of the current assembly . This scaffold ( 7 . 5 ) is minus oriented in the reference genome , whereas upstream scaffolds have unknown orientation . The high similarity in sample genotypes between the two blocks , as well as their shared gene family components ( see below ) , suggest that it should be reoriented . r9 spans 3 scaffolds and 1 . 639 Mbp on chromosome 9 ( Fig 3C ) . FST between highland and lowland populations is ~0 . 7 and ~0 . 3 across the r7 and r9 , respectively ( Fig 3C; Table 2 ) . They contain 111 , 161 SNPs in total , representing only 1 . 39% of the data , yet exclusion of these SNPs alone removes the split between highland and lowland samples and shifts the interrelationships of the four populations from clustering by environment to cluster by locality ( Fig 3D ) . It is therefore clear that these narrow but divergent regions stand out against the genomic background and influence the analyses of the genome-wide interrelationships ( see Fig 2B above ) . The distinct blocks of highly differentiated SNPs with clear boundaries are suggestive of the presence of non-recombining haplotypes with high levels of divergence between them ( Fig 3A ) . These patterns are unlikely to result from selective sweeps , which would be expected to result in a gradual decay of LD over shorter genetic distance . r7 contains 12 , 042 SNPS with FST of 0 . 8–0 . 85 but only 2 , 866 SNPs with FST of 0 . 5–0 . 8 ( Fig 3B ) . Likewise , r9 contains 3 , 587 SNPs with FST of 0 . 65–0 . 7 and similarly to r7 , no individual bin with FST>0 . 5 contains more SNPs than the 0 . 65–0 . 7 bin in the region ( Fig 3B ) . The regions therefore appear to be enriched for SNPs at a particular high FST bin , indicating strong association between the segregating variants . One explanation for these patterns is that the haplotypes represent structural polymorphisms , such as inversions that prevent recombination occurring between them . To further characterize r7 and r9 , we counted the genotypes of every sample at all divergent SNPs ( FST>0 . 5; Table 2 ) . Worker honey bees are diploid and at every such SNP , a sample can therefore be homozygous for the reference sequence allele ( 0/0 ) , homozygous for the non-reference allele ( 1/1 ) or heterozygous ( 0/1 ) . The reference sequence is derived from a US managed population and matches the lowland haplotype [43] . Across both regions , we found that 1/1 genotypes are significantly more frequent in highland bees than in lowland bees: 83 . 7% ( n = 267 , 947 ) vs . 2 . 7% ( n = 8 , 484 ) in r7 and 74 . 7% ( n = 126 , 136 ) vs . 13 . 4% ( n = 21 , 526 ) in r9 , respectively , showing that highland bees have haplotypes that are strongly enriched for non-reference variants ( p<10−5 for both regions; Fisher’s exact test; Fig 4 ) . Notably , we detected several samples that appear to be nearly completely heterozygous across either region: these are heterozygous for >88% of genotypes ( Fig 4 ) . We detected a few outlier samples that are homozygous for the opposite haplotype compared to the majority of samples from either environment . The same samples are heterozygous or atypical at the two physically linked r7i and r7ii sub-blocks on chromosome 7 ( S3A Fig ) , supporting the idea that they are closely located on the chromosome . This pattern pertains to other samples at the independently transmitted chromosome 9 . We performed principal component analyses ( PCAs ) using the multidimensional scaling algorithm implemented in PLINK [44] to further evaluate these divergence patterns . The PCA was carried out for all SNPs within and outside of the r7 and r9 regions , respectively . In accordance with the FST-based analyses , we found that divergence between highland and lowland samples was much higher at r7 and r9 than across the rest of the genome and that outlier and heterozygous samples clustered as predicted ( S4 Fig ) . Honey bees have among the highest recorded recombination rates of any animal [45] . Continuous megabase-scale heterozygosity suggests that they have both a lowland and highland haplotype and that meiotic recombination between them is greatly suppressed . The r7 highland haplotype ( r7h ) is completely fixed in the Mt Kenya highland samples and absent from the Mt Kenya lowland population , whereas we detect heterozygous samples or outliers in the Mau populations ( Fig 4 ) . The r9 highland haplotype ( r9h ) follows similar , albeit less extreme segregation . We estimate the average population frequencies of both r7h and r9h to be 93% across highland bees and 8% or 21% across lowland bees , respectively . Haplotype frequencies are strongly associated with environment ( p<10−5 for both regions; χ2 test ) , corresponding to FST values of 0 . 832 at r7 and 0 . 682 at r9 . We performed coalescent simulations in order to determine the probability that the extreme differences in haplotype frequencies we observe between populations could occur in the absence of natural selection on these regions . We used ms [46] to model the evolution of a highland and lowland population and test this alternative scenario . We adopted a basic split model without subsequent gene flow between descendant populations and without recombination . Inclusion of these processes would homogenize genetic variation between the descendant populations . We applied population demographic parameters inferred from the data to model the split ( see Methods ) to simulate the evolution of 1 million independent loci across the genome using the same sample size as in our dataset ( 20 vs . 19 diploids ) . We then estimated FST between populations using the same methods with the empirical dataset . The split was inferred to have occurred 28 , 410 generations ago and the average divergence between simulated populations was very close the empirical data ( 0 . 037 vs 0 . 036 ) . However , FST values as high as those observed for the r7 and r9 haplotypes ( r7: FST = 0 . 832; r9: FST = 0 . 682 ) were never observed in the simulated data , where the most divergent locus had FST = 0 . 655 ( S5 Fig ) . This indicates that such levels of divergence between two populations are highly unlikely to occur by drift alone under this scenario . It is also important to note that we observe similarly extreme levels of divergence at the r7 and r9 loci in two independent highland/lowland comparisons ( Mau forest vs . savannah and Mt . Kenya forest vs . savannah ) . There can be no direct contact between the two highland populations due to their geographic isolation , but gene flow can occur between them via the lowland populations , where frequencies of the highland haplotypes are very low . The pattern we observe , where the same haplotype variants at two loci are associated with the highland habitat in two independent comparisons is therefore indicative of selection favoring these haplotypes in highland environments . The possibility that these patterns could occur in the absence of selection can be ruled out . About 13% ( 29 Mbp ) of the honey bee reference genome is not placed on any chromosome . We assessed these sequences separately and detected 982 additional SNPs with FST>0 . 5 , distributed across 31 scaffolds/contigs ( S2A Table ) . We scanned these SNPs for genotype and haplotype patterns consistent with those in r7 and r9 ( Fig 4 ) . We find that 30 of them can be assigned to either r7 or r9 based on the pattern of segregation at the SNPs: 16 fragments spanning 28 . 6 kbp and 682 SNPs match r7 and 14 fragments covering 62 . 5 kbp and 299 SNPs match r9 ( S2A Table; S3B–S3D Fig ) . The unassigned fragment GroupUn869 contains only a single outlier SNP ( FST = 0 . 52 ) . To verify the assignments , we scanned the paired-end data for evidence of split read-pairs that could anchor the unmapped fragments to the regions using Delly2 [47] in translocation mode . All 30 fragments assigned to either r7 or r9 , but not GroupUn869 , contain evidence that place them within haplotype scaffolds or close to their borders ( S2B Table ) . Taken together , the results suggest that they may belong to the two regions , possibly extending them by up to 4–5% . We detect homozygosity for the highland haplotypes in the single monticola bee collected at Mt Elgon and sequenced by Fuller and co-workers[37] ( S6A Fig ) , and currently the only representative from a third Kenyan mountain location . The highland haplotypes appear to be absent from coastal or desert populations of Kenya . Querying the global dataset , we do not detect the r7h haplotype in any population outside of Kenya ( S6B Fig ) . We do however detect genotypes matching r9h heterozygosity in two savannah scutellata samples from South Africa ( S6B Fig ) . These samples are heterozygous for >77% of the outlier genotypes at scaffold 9 . 7 and have intermediate genetic distances to the Kenyan highland haplotypes compared with South African scutellata homozygous lowland haplotypes ( S6C and S6D Fig ) . Interestingly , this pattern is less clear on the upstream 9 . 5 scaffold , where the two samples and additional South African honey bees appear to be heterozygous for only ~50% of the outlier genotypes ( S6D Fig ) . These results suggest the presence of r9h-like haplotypes at low frequency outside of Kenya , which may include additional structural diversity . We assessed genetic differentiation between populations at the r7 and r9 regions . For this analysis , we first subsampled the Kenyan data to contain only the individuals that were homozygous for the major haplotype associated with either environment ( Fig 4 ) . We included sequence variation from other honey bee populations for either region in order to analyze the haplotypes in the context of global haplotype diversity within the species . For both regions , we find that the Kenyan lowland bees have haplotypes that are typical for African honey bees ( FST<0 . 10 against other African bees ) , while the highland haplotypes diverge strongly from African and other subspecies ( FST>0 . 5; Fig 5A and 5B; Table 3 ) . Other population interrelationships are consistent with the whole-genome analyses above and previous results ( Fig 2A above; [25] ) We next compared the genetic distance between the two haplotypes in order to estimate timing of their divergence . Across the full genome , dXY is 0 . 67% between any random pair of two haploid genomes ( Fig 3D; Table 3 ) . At r7 on the other hand , divergence at non-coding sites is 3 . 34% ( 95% CIs: 2 . 97%–3 . 49% , 2 , 000 bootstrap replicates ) between highland and lowland haplotypes , dating the split between them to about 3 . 2 ( 2 . 8–3 . 3 ) million years ago ( Fig 5C; Table 3 ) assuming a mutation rate of μ = 5 . 27×10−9 mutations per base per generation and a one-year generation time . For r9 , divergence is 1 . 34% ( 95% CIs: 1 . 28%–1 . 36% , 2 , 000 bootstrap replicates; Fig 5D ) , corresponding to the haplotypes having diverged about 1 . 28 ( 1 . 22–1 . 30 ) million years ago . These molecular clock estimates suggest that the r7 and r9 highland haplotypes have originated independently but are both very old , possibly predating the diversification of modern honey bee populations and the colonization of their current ranges by hundreds of thousands of years [25] . The r7 region has been annotated for 38 gene accessions in the current gene set , the coding regions of which span 46 kbp ( 8 . 3% of the haplotype ) , whereas r9 includes 50 accessions , many of which are uncharacterized , spanning only 23 kbp of coding sequence ( 1 . 4% of the haplotype ) ( S3 Table ) . By comparing fixed variants between haplotypes against the corresponding sites in A . cerana , it is possible to estimate the number of derived changes that have occurred on each haplotype after the split from their common ancestor . We inferred that 66% of the 9 , 941 fixed mutations have taken place on the highland haplotype , indicating higher rates of fixation in this haplotype . To assess functional evolution on each haplotype , we quantified the ratio between fixed non-synonymous and synonymous changes that has occurred on either haplotype since their common ancestor ( S3 Table ) . In the r7 region , there are 560 fixed coding differences between highland and lowland haplotypes , 57% ( n = 323 ) of which we infer to have occurred on the highland haplotype . Of the derived variants fixed on the highland haplotype , 44% of ( 142/323 ) are non-synonymous . On the lowland haplotype , only 28% ( 66/237 ) of the fixed derived variants are non-synonymous . The proportion of non-synonymous variants that are fixed on the highland haplotype is therefore significantly higher than the proportion fixed on the lowland haplotype ( Fisher's exact test; p<10−5 ) . In the r9 region we only detect 28 fixed mutations . Out of these , 21 have occurred on the highland haplotype . Of the derived variants fixed on the highland haplotypes , 62% ( n = 13 ) are non-synonymous . In the r9 lowland haplotype , only 28% ( 2 out of 7 ) variants are non-synonymous . These proportions show the same trend as the r7 haplotype , although they are not significantly different . It therefore appears that highland haplotypes have accumulated non-synonymous changes at a substantially higher rate . The two divergent regions span many divergent genes and mutations that may alter protein function , making it difficult to identify the specific targets of selection in highland bees with full certainty . Both regions contain genes that influence honey bee worker behavior that we consider to be interesting candidate genes for mediating adaptation to the montane forest habitat . The r7 region includes genes encoding four octopamine receptors: AmOctβ1R ( oa2 ) , AmOctβ3R/4R ( isoforms X2 , X1 & X3 ) and AmOctβ2R ( on r7ii ) , which together contain 12 derived non-synonymous mutations in the highland haplotypes ( Fig 6A; S3 Table ) . Octopamines are biogenic amines and essential neurotransmitters , modulators and circulatory hormones in invertebrates . They interact with specific G protein coupled receptors to increase Ca2+ or cAMP levels and modulate physiology and behavior in response to environmental stimuli [48 , 49] . In honey bee workers , octopamine increases responsiveness to sucrose and sensitivity to sensory inputs and regulates olfactory learning and memory formation [50–53] . The r9 region contains genes for encoding several isoforms of calcium/calmodulin-dependent Serine protein Kinase enzyme ( CASK; LOC411347; isoforms identified with BLAST; Fig 6B ) . CASK interacts with a second Ca2+/calmodulin kinase , CaMKII , in a fundamental pathway for memory formation that is shared between humans , fly and honey bees [54–56] . We therefore hypothesize that one or both of the divergent haplotypes contain changes to genes that underlie adaptive foraging behaviors at high altitudes . Future work could focus on identifying behavioral differences in honey bees bearing contrasting haplotypes at the r7 and r9 loci . We observed reduced mapping coverage in many locations across the r7 highland haplotype compared to the lowland haplotype ( Fig 6A ) . For r7h samples , the average genotype depth was 9 . 2× ( 0 . 90x the genome average ) , with 7% of genotypes missing in the original FreeBayes SNP call . Among FST>0 . 5 outlier SNPs , 2% of genotypes were missing . For r7l samples , the average genotype depth was 11 . 9× ( 1 . 13x the genome average ) , with 0 . 5% missing genotypes . Reduced short-read coverage can be expected due to high genetic distance between r7h and the reference genome and the difference in mapping depth is smaller for the less divergent r9 region ( Fig 6B ) . Both regions contain a few regions with very high mapping depth that are shared between both haplotypes ( Fig 6A and 6B ) . Structural variants such as duplications/deletions that segregate between populations and the reference genome can alter mapping depths and result in incorrect genotyping . To assess the influence of mapping depths for the results , we analyzed a dataset of SNPs where a set of strict filters had been applied . We masked all SNPs and base pairs in the genome which had <30% or >30% read depth , compared to the average mapping depth , or where <50% of samples had been genotyped by FreeBayes . Filters were applied per environment ( n = 20 highland vs . n = 19 lowland bees ) , for each subsample of individuals that were homozygous for either haplotype ( Fig 4 ) or across the Kenyan bee dataset as a whole . The filters retained ~80% of bases across the whole genome and full dataset ( 156 Mbp/200 Mbp ) and for the r9 region ( S7A Fig ) . At r7 , 62% of the region passed the filter in the highland subsample ( vs . 80% in the lowland subsample; S7A Fig ) . Of the original outlier SNPs with FST>0 . 5 , 7 , 205 ( 45% ) and 5 , 134 ( 61% ) were retained for r7 and r9 , respectively . For both r7h and r7l subsamples , the resulting average genotype depth was ~0 . 95x the genome average and <0 . 1% of genotypes were missing . We then re-estimated and compared levels of diversity across the whole genome and the r7 and r9 haplotypes , with or without these extra filters . For the same regions , we also compared FST between all highland and lowland bees and re-inferred haplotypes for r7 and r9 ( as in Fig 4 ) . The results are highly congruent between datasets ( S7B–S7D Fig ) . We therefore conclude that poorly mapped regions do not drive the patterns of diversity and divergence that we have inferred across r7 , r9 or the genome . Detection of haplotype breakpoints can help to disentangle the nature of putative structural variants . In fly and mosquito , inversion breakpoints have previously been linked to crossover between repeated sequence and mobile elements [57 , 58] . We scanned the genomic regions around the outermost outlier SNPs of each region ( Table 2 ) for patterns of divergent read mapping and repeated motifs . These SNPs occur close to scaffold borders and it is possible that the genome assembly is incomplete for these regions . For r7 , we find that the first outlier SNP ( pos . 11 , 056 bp ) occurs in a 500 bp region ( pos . 11 , 000–11 , 500 bp ) where scutellata reads map normally but very few monticola reads map ( S8A Fig ) . The monticola reads that contain the first outlier SNP are truncated by BWA to ~36 bp and have mate pairs that are either unmapped or mapped to scattered regions in the genome , indicating potentially aberrant alignment . Regular monticola read mapping resumes at pos . ~11 , 420 bp . The region overlaps with repetitive sequence containing three iterations of a 176 bp AluI-like monomer with starting positions 8 , 814 , 10 , 709 and 10 , 885 , respectively ( S8A Fig ) . This AluI-like element has been experimentally estimated to be a common repeat at honey bee telomeres [59 , 60] . The two latter motifs correlate with the extremely high mapping depth observed in our data ( Fig 6A ) . The average per-sample coverage across three 100 bp windows between positions 10 , 700–11 , 000 is 8 , 300× , approximately 830 times the per-sample genome average ( 10× ) . This suggests that many AluI-like short reads from across the genome have been mapped to these motifs . Within the same region , we identify a 26 bp motif with high probability ( “aattgataaaggaagggaggaagagg”; p<6 . 20 x 10−29 ) using MEME suite [61] . It is repeated 65 times between positions 8 , 942–10 , 667 and has high similarities towards a Winged Helix-turn-Helix ( HTH ) DNA binding protein motif likely containing the optix transcription factor binding site ( “tgata” , relative score = 1 ) , as predicted by JASPAR [62] . Beside their roles in transcription , proteins with HTH-binding domains are involved in recombination and may cause rearrangements[63 , 64] . The corresponding downstream region of r7 that may contain a haplotype breakpoint occurs around position 1 , 511 , 853 , where we detect the first outlier SNP in r7ii ( Table 2 ) . This SNP and the subsequent divergent SNPs are located inside the third intron of the octopamine receptor AmOctβ2R gene ( Fig 6; S8B Fig ) . Due to the strong linkage between r7i and r7ii ( S3 Fig ) , we hypothesize that the first outlier SNP represents one end of the r7 haplotype region and that the 7 . 5 scaffold should be reoriented to join the octopamine receptor gene family into a continuous block . This region does not appear to contain AluI-like fragments or HTH motifs . As the7ii border is not as repetitive as the r7i upstream border , we designed a long-range PCR experiment to amplify the third intron of AmOctβ2R . If the intron contains the second breakpoint , we expect to be able to amplify the sequence in most reference-like scutellata samples but not in the rearranged monticola samples . We successfully amplified the expected 2 kbp fragments for all scutellata samples from Mt Kenya and Mau predicted to have the reference-like haplotype ( n = 19; S8C Fig ) . Amplification failed for all monticola samples from Mt Kenya ( n = 10 ) and four of the Mau samples predicted to be homozygous for the highland haplotypes ( S8C Fig ) . However , amplification also worked for four Mau monticola samples predicted to be homozygous for the highland haplotypes . These have excess numbers of heterozygous variants for r7ii ( S3A Fig ) , and are heterozygous across the test region before they switch from lowland to highland haplotypes ( S8 Fig ) , indicating breakpoint polymorphism in this region . These results strongly suggest the presence of breakpoints within the third AmOctβ2R intron . It is possible that disruption of this octopamine receptor gene has been important for local adaptation in highland bees . The SNP-delineated borders of the r9 region occur very close to scaffold ends ( <7 kbp ) . In both cases , the outmost SNPs are located within 500 bp from short read alignment gaps ( approx . pos . 1 , 729 , 290 and 3 , 468 , 837 ) that are shared between highland and lowland bees , suggesting that the data may be incomplete for these regions and that the actual r9 breakpoints are not mapped . There is a tendency for monticola bees to be darker in color , whereas scutellata bees are more yellow , although color on its own does not distinguish highland and lowland bees and even varies within colonies [28 , 30] . We compared the color of our specimens with the identity of their haplotypes at r7 and r9 in order to determine if either of these loci controlled these differences in color ( S9 Fig ) . There was no clear association with color at either of the loci . Five of the 19 specimens collected in the lowlands have uniformly dark abdomens . Of these , three are homozygous for the lowland haplotype at both loci , which would not be predicted if the highland haplotype is associated with dark abdomens . Three of the 20 highland specimens have uniformly dark abdomens . One of these is heterozygous at the r9 locus , whereas the others are fixed for highland haplotypes at both loci . It is therefore clear that there is a much stronger distinction between highland and lowland bees in the genome at r7 and r9 than there is in their color . This suggests that , while these loci may have some subtle effect on body color , they are likely associated with adaptations to highland habitats that have greater fitness consequences than differences in color .
The presence of long genomic blocks with high divergence between highland and lowland populations with distinct boundaries indicates that the regions harbor two distinct haplotypes . This is further supported by analysis of patterns of segregation of highly differentiated SNPs in this region among individuals , which made it possible to identify individuals that were homozygous or heterozygous for the two diverged haplotypes ( Fig 4 ) . These patterns are indicative of a form of balancing selection , where haplotypes are locally adaptive coupled with repression of interallelic exchange of genetic material through recombination between haplotypes . The most likely mechanism to prevent recombination that would lead to the observed pattern of diverged haplotypes , is a structural rearrangement such as an inversion . Only one putative inversion breakpoint ( on the r7 haplotype ) mapped to known sequence in the honey bee genome assembly . We are able to amplify sequence across this breakpoint in lowland bees , but not in most highland bees , consistent with an inversion . We infer that the other breakpoint for the r7 haplotype is located close to the end of chromosome 7 . There are many Alu1-like monomers and a repeated helix-turn-helix motif in the vicinity of this region . The Alu family of mobile element has previously been associated with chromosomal rearrangements in mammals [66] . It is possible that Alu-like repeats cause chromosome instabilities and rearrangements also in other taxa and have been involved in the origin of the highland haplotype in mountain bees . We are unable to identify such sequence patterns for the r9 boundaries , which correlate more strongly with scaffold borders and incomplete mapping . There are two non-mutually-exclusive ways in which an inversion could have an effect on phenotype and fitness . The first is that the inversion mutation itself has an effect on genome function . This could be because the breakpoint disrupts a transcribed region or has an effect on gene regulation . The second is that the suppression of recombination is selectively favored because it maintains associations between co-adapted alleles . Theory suggests that inversions that capture locally adapted alleles in two populations that are connected by gene flow can quickly spread in a population [67] . Recent studies suggest that adaptation by inversions or supergenes is more common than previously thought [68 , 69] . There are many examples of structural inversions that are involved in local adaptation . For example in sticklebacks [5] several inversions govern local adaptation to fresh water habitats , and are present in many geographically separated regions . Adaptation to environmental clines in Drosophila also correlates with the frequency of cosmopolitan inversions , providing a striking example of rapid evolution [70] . It has been shown that these clines have shifted as a response to global climate change and that increased cold-tolerance may have arisen several times despite the presence of gene flow . Among butterfly species , the mimetic wing patterning of Heliconius numata provides a compelling example of chromosomal rearrangements that lead to co-adapted gene complexes involved in adaption and speciation [71] . A similar mechanism controls Batesian mimicry in the Papilio genus [72 , 73] . A polymorphic inversion governs worker behavior and reproductive strategies in the fire ant Solenopsis invicta [74 , 75] . We observe the same haplotypes associated with highland habitats in the two localities studied here ( Mau and Mt . Kenya ) and in another published dataset from Mt . Elgon [37] whereas these haplotypes appear to be rare outside of montane forest environments . This shared pattern indicates that their high frequencies in highlands is the result of selection on standing variation , rather than selection on new mutations . It is similar to the pattern of genetic adaptation observed in sticklebacks , where the same set of genetic variants , including inversions , are associated with freshwater streams and lakes across the world , but are rare in the oceans that connect them [5] . A number of genes located within the r7 and r9 regions are potential candidates for environmental adaptation in highland bees . r7 contains four out of five of the honey bee octopamine receptor genes in the genome . These are all four cAMP-inducing octopamine β-receptors ( AmOctβ1R to AmOctβ4R; Fig 6A; S3 Table ) , whereas the single Ca2+ regulating AmOctαR1/oa1 is located outside of the region , on chromosome 15 [76] . Experiments with microinjections have shown that octopamine can modify neuronal responses in different neuropils of the bee brain [52] . Octopamine signaling affects complex behaviors in bees and has a key role in social division of labor and foraging [77] . Expression of octopamine receptors correlates with worker tasks and age and experimental application of the amine induces foraging in nurse bees [78–80] . Increased octopamine levels in honey bees positively affects scouting for new food sources or new nest sites [81] . In particular , octopamine plays a major role in olfactory learning and memory formation in the honey bee [50] , important tasks for adapting to different environmental conditions . Interestingly , octopamine has also been shown to be important for stabilizing signaling integrity during hypoxic and thermal stress in other insects [82 , 83] . The four octopamine receptor genes located within r7 have many fixed differences between highland and lowland bees . Moreover , we infer that a rearrangement breakpoint is present in one of these genes that could potentially disrupt gene function ( S8 Fig ) . We therefore hypothesize that genetic variants in the octopamine receptors , that exert their effects via mediating foraging behavior , are responsible for selective advantage of the highland version of the r7 locus in montane forests . Another putative candidate for controlling honey bee behavior is the Ent2 gene ( LOC55249 ) on r7 , which has been associated with synaptic transmission and associative learning in Drosophila [84] and the Ca2+/calmodulin-dependent Serine protein Kinase ( CASK ) isoforms encoded by genes at one edge of the r9 region ( Fig 6B; S3 Table ) . CASK acts together with CaMKII to affect long-term memory formation in honey bees [56 , 85] . Differences in these genes could contribute to foraging performance in highland bees . However , we cannot rule out the possibility that the divergent haplotypes have broader functional implications as they include changes to genes with Drosophila orthologues involved in regulating chromatin ( transcriptional activator protein Pur-β; LOC72639; r7 ) , lipid ( calcium-independent phospholipase A2-gamma-like; LOC726656; r7 ) and polypeptide-folding functions ( prefold in subunit 5-like; LOC411936; r9 ) as well as muscle development ( myosin-2 heavy chain; LOC100576864; r9 ) [86] . Any of these could be important for adaptation , either independently , or as part of a co-adapted supergene complex , evolving in concert with other genes . In addition to adaptive sequence evolution , highland haplotypes appear to have been affected by more genetic drift than lowland haplotypes . First , the levels of genetic variation are much reduced on the highland haplotypes compared to the lowland haplotypes ( Table 3; Fig 6 ) : by 76% for r7h vs r7l and by 28% for r9h vs r9l . Reduced diversity among highland haplotypes likely equates to lower NE in these haplotypes compared to lowland haplotypes , probably due to them being more geographically restricted . Second , highland haplotypes have accumulated considerably more derived non-synonymous variants than lowland haplotypes ( S3 Table ) , which may indicate accumulation of slightly deleterious variants . This could reflect that NE of the mountain haplotypes may historically have been lower due to restricted distribution to particular habitats compared to the widespread lowland haplotypes and despite prevalent gene flow across the rest of the genome . High divergence between the haplotypes at r7 and r9 suggests their origin is ancient . Divergence between haplotypes at these loci is substantially higher than divergence between the major honey bee lineages found on different continents . We estimate r9 to have evolved 1 . 3 million years ago ( MYA ) and the r7 3 . 2 MYA using a molecular clock . Both of these dates are considerably older than estimations of the emergence of extant populations of A . mellifera [25] . There are two main explanations for such an ancient origin . First , it is possible that the haplotypes were present in the ancestral population of A . mellifera , before the split of extant lineages . They could have been involved in local adaptation before modern lineages came to inhabit their current ranges in Europe , Africa and the Middle East . Despite phenotypic similarities to European bees [27] , we have not detected the monticola haplotypes outside Africa ( Table 3; S6 Fig ) . A second possibility is that the highland haplotype arose from introgression with another related species in the past . This scenario is inferred to be the case with the haplotype encompassing the EPAS1 gene in humans , which is responsible for high altitude adaptation in Tibetans and inferred to have arisen by adaptive introgression from archaic humans to modern humans [11] . In the case of monticola mountain bees , a potential donor population is not known . All other Apis species are found in Asia and their native distributions have not overlapped with that of A . mellifera until the beginning of the 20th century when A . mellifera was introduced in East Asian countries . Our studies of honey bee genomes from highland and lowland populations reveal patterns that are consistent with pervasive gene flow between them , with the exception of two large and divergent blocks on chromosomes 7 and 9 . Haplotypes at these loci appear to represent long inversions that are strongly differentiated between populations from different habitats . These loci are reminiscent of supergenes that have been demonstrated to govern adaptation in several other species . Many genes within these blocks are linked to honey bee behavior . In particular , we identify a haplotype breakpoint that disrupts the transcript of an octopamine receptor , part of a family of genes involved in foraging and learning . We therefore hypothesize that these loci govern the behavioral traits that are characteristic for mountain bees and likely constitute local adaptations to the highland environment . High levels of divergence between haplotypes at both loci indicate an ancient origin , suggesting that they were involved in environmental adaptation before the dispersal of honey bees to their present geographic range .
Female worker honey bees from each highland locality in Kenya ( eastern slope of Mount Kenya and Eastern Mau Forest ) and lowland samples from neighboring locations were collected as part of a previous study [28] . The highland bees are referred to as A . m . monticola ( hereafter monticola ) , whereas lowland bees are referred to as A . m . scutellata ( hereafter scutellata ) . The monticola sampling sites were closed canopy forest areas above 2000m , whereas the corresponding lowland scutellata samples were collected in savannah vegetation or agricultural land surrounded by savannah vegetation ( Fig 1; S1A Table ) . We used the Maxwell Tissue DNA Purification Kit ( Promega ) to extract total genomic DNA from the thorax of single honey bees , each from a different colony . Images of individual abdomens were taken using a ZEISS Stereo Microscope unit Stemi 305 with an Axiocam 105 color ( Fa . Zeiss , Germany ) . The 39 DNA samples were barcoded and 2x125 bp paired-end reads were sequenced on an Illumina HiSeq 2500 sequencer . Reads were mapped against v4 . 5 of the honey bee reference genome ( Amel_4 . 5 ) [43] using the default settings in the BWA v0 . 7 . 12 aligner with the “mem” algorithm [87] . Read groups and duplicates were tagged and marked with Picard v1 . 118 . Indel-realignment and quality score recalibration was performed with GATK v3 . 3 . 0 [88] , using SNPs from Wallberg et al . [25] . These programs were used with default settings and parameters . Two datasets were produced . The first dataset contained genome data only from the 39 new libraries and was used for most analyses comparing Kenyan highland and lowland bees . For the second dataset , we used 11 short read archives from [37] ( Kenyan samples; NCBI project ID: PRJNA237819; Illumina reads ) and 98 archives from [25] ( worldwide samples; NCBI project ID: PRJNA236426; SOLiD reads ) in order to expand the population sample and facilitate additional comparative analyses . The former dataset was mapped with BWA as above , whereas the latter dataset was mapped with Lifescope™ as in [25] . See S1B and S1C Table for detailed sample information . We called single-nucleotide polymorphisms ( SNPs ) across all samples using the haplotype-based variant detector Freebayes v0 . 9 . 20–16 [89] for both datasets . We used the flags “-E 0” , “-X” and “-u” to suppress construction of short multi-nucleotide haplotypes from closely positioned polymorphisms and to avoid making composite polymorphisms . We used the flag “—theta 0 . 008” to better match the expected population mutation rate estimated in [25] , as compared to the human default value ( 0 . 001 ) . Putative SNPs were filtered for quality by accepting only biallelic SNPs with QUAL scores >50 . In addition , we removed known problematic positions where a drone closely related to the DH4 individual used to produce the reference genome had been inferred to be heterozygous in [25] . As drones are haploid , they should never be heterozygous . Such SNPs are errors that indicate problematic regions in the assembly where genotyping is unreliable . This filter removed 61 , 157 SNPs . The procedure is detailed in [25] . The restricted dataset ( n = 39 samples ) contained 8 , 593 , 016 SNPs and the expanded dataset ( n = 148 ) included 13 , 672 , 645 SNPs . BEAGLE v3 . 3 . 2 [90] was used to impute missing variants and phase haplotypes . We used the flags “iterations = 30” , “nsamples = 10” and “lowmem = true” to increase accuracy and reduce memory usage as per the recommendations in the program manual . Gene models provided by the latest official gene set ( OGSv3 . 2 ) [43] in GFF format were used to associate SNPs with genes and annotate synonymous and non-synonymous variants in coding sequences . In order to determine gene names , orthologues and putative functions , the models were cross-referenced with genes in Amel_4 . 5 , the NCBI Annotation Release 103 ( AR103 ) and a comprehensive set of Drosophila Flybase orthologues detected in [25 , 86] . Custom Perl scripts were used to parse the coordinates and label SNPs . We aligned the genome sequence of the Eastern honey bee A . cerana ( v1 . 0; from Park et al . [38] ) against the A . mellifera genome using the whole genome synteny aligner Satsuma v1 [91] . This allowed us to: i ) root phylogenetic trees; and ii ) use parsimony to distinguish between ancestral and derived variants at SNPs across the divergent haplotypes . In this method , the shared allele between A . mellifera and A . cerana is taken as ancestral and the non-shared allele is taken as derived . We used the FST statistic to estimate divergence between populations from pairwise differences in allele frequencies . For individual SNPs , we used the FST estimator of Weir and Cockerham [42] . Outlier SNPs with high FST compared to the genomic background served as a basis for detecting genomic regions segregating between highland and lowland bees . We applied the Reynolds FST estimator [41] to produce whole-genome distance matrices between all populations , allowing us to determine whether divergent regions were associated with mutations in highland or lowland bees . Reynolds FST was also used to compute divergence in 10 kbp windows along the genome in order to detect local support for conflicting genealogies between Kenyan populations . These statistics were estimated using custom Perl scripts . Population interrelationships were inferred from genetic distances using the neighbour-joining algorithm [92] as implemented in PHYLIP v3 . 696 [93] or SplitsTree v4 . 14 . 2 [94] using default settings . Average pairwise differences were computed to estimate per-base genetic distance between all samples or between the highland and lowland haplotypes detected on chromosomes 7 and 9 . These per-base genetic distances were formulated by Nei and Li [95] and are expressed as π when considering nucleotide diversity within populations and dXY when measured between populations or haplotypes . Divergence was calculated using custom Perl scripts . We applied a constant molecular clock to date the origin of the haplotypes from dXY estimated from presumably functionally neutral variants ( those located outside of gene bodies ) . For these calculations , we used a mutation rate μ = 5 . 27×10−9 mutations per base per generation , which was previously estimated from neutral divergence between A . mellifera and its sister species A . cerana [25] . We assumed one generation per year as in ref [25] . We estimated 95% confidence intervals by partitioning the haplotypes into 10 kbp windows and bootstrapping the data using per-window dXY estimates ( 2 , 000 replicates per region ) . We performed a genome-wide association study ( GWAS ) to study associations between SNPs and the environment , using lowland and highland location for each sample as the case and control phenotype ( S1 Table ) . We used the association ( —assoc ) function in PLINK v1 . 9 [44] to implement the test and plotted the observed associations over the expected assuming a random normal distribution in a Q-Q plot using the—adjust and—qq-plot flags . We also used PLINK to make a principal component analysis inside and outside the putative haplotype inversion regions using the multidimensional scaling algorithm ( —mds ) . The regions were defined by the intervals provided in Table 2 . To measure genetic diversity within populations , we estimated π ( see divergence section above ) and θw ( Watterson’s theta; the population mutation rate ) per base for each population [96] . Locally reduced genetic diversity can be a signal of selection or indicate reduced effective population size at that locus . Relative levels of genetic diversity between highland haplotypes ( h ) and lowland ( l ) haplotypes on chromosomes 7 and 9 was computed using the equation: Δπ=πh−πlπl Strongly negative values indicate reduced variation in highland haplotypes . We used the population mutation rate ( θw ) and the honey bee mutation rate ( μ ) above to estimate effective population size ( NE ) using the following equation: NE=θW3μ Because honey bees are haplodiploid , we used the inheritance scalar of 3 rather than 4 , used for diploids . We used the program ms [46] to simulate the neutral coalescent under a basic population split scenario under a Wright-Fisher model , assuming no gene flow , no recombination and constant population size . In this scenario , a hypothetical ancestral honey bee population split into a highland and a lowland population without secondary contact . We used realistic empirical estimates of the model parameters . We first estimated the generation time since the split . Under a model of neutral divergence of two populations from a common ancestor , it is possible to convert FST into an estimate of time since divergence , measured in units of scaled time , which we define as T = t/3NE , where t is the number of generations since the split . The factor of 3 is applicable for haplodiploids . T can be estimated using the following formula [41]: T=−ln ( 1−FST ) 2 The average divergence between highland and lowland honey bees ( FST = 0 . 036 ) was used to estimate T as 0 . 01833 . We next determined the number of independent loci to simulate . Excluding gaps and undetermined sites , the 16 honey bee chromosomes span 199 . 7 Mbp in total . Given the extremely high recombination rate across the genome [45] , linkage between sites is expected to decay very rapidly and the genome should therefore contain many unlinked loci . r2 was previously estimated to decay below 0 . 05 at distances greater than 200 bp in African honey bees [25] . We therefore assumed that the 200 Mbp genome would contain 1 million unlinked loci . Assuming the population mutation rate θw to be 0 . 0082 per base pair , as estimated from all data ( Table 1 ) , we simulated a coalescent process across a theoretical 1 kbp locus ( an arbitrarily chosen size ) where θw is scaled to 8 . 2 . We repeated the simulation 1 million times , instructing ms to export 1 segregating site ( SNP ) per replicate . We sampled 40 chromosomes for one descendant population ( n = 20 diploid workers ) and 38 chromosomes for the other ( n = 19 diploid workers ) . Since the coalescent tracks events back in time , the model is implemented using a population join ( -ej flag ) rather than a population split . The simulation was performed with a single command: ms 78 1000000 -t 8 . 2 -I 2 40 38 -ej 0 . 01833 2 1 -s 1 > ms . simulation . data Here , we specify to sample 78 chromosomes , repeat the analysis 1 million times , use the 1 kbp scaled mutation rate of 8 . 2 ( -t 8 . 2 ) , specify that the chromosome count for two populations is 40 and 38 , respectively ( -I 2 40 38 ) , model a population join between population 2 and 1 at generation time 0 . 01833 ( -ej 0 . 01833 ) and export one segregating locus per replicate ( -s 1 ) . This produced 1 million loci for which we calculated Weir and Cockerham FST individually , as for the empirical data . We produced a distribution of these simulated estimates by partitioning them into FST bins of 0 . 01 . Distal outlier SNPs were used delineate haplotypes in each divergent region . Read mapping was manually inspected around these coordinates using Tablet 1 . 16 . 09 . 06 [97] . Repetitive element motifs in the haplotype breakpoint region on chromosome 7 were detected with the BLAST [98] and MEME [61] web services . Some unmapped fragments ( aggregated into virtual chromosome GroupUN ) had outlier SNPs and genotype patterns consistent with those detected at the divergent haplotypes . In order to collect additional lines of evidence that those candidate fragments may belong to these regions , we performed split-read and paired-end analyses delly2 [47] in translocation mode ( -t TRA ) for the GroupUN and chromosome 7 and 9 BAM files . The program was used with default settings and the output VCF file was parsed for links between unmapped fragments and either region . In order to test a breakpoint experimentally , oligonucleotides were developed flanking the putative breakpoint located within Octβ2R ( GB49696 ) in chromosome 7 between positions 1 , 511 , 194 to 1 , 513 , 199 . Oligonucleotide Ex3in_Ex4_1fw TTTTCTTCTCCCCCTTCTTTTC and Ex3inEx4in_1rev TTCCACTATAACCGCTTTTCC were used in a standard PCR reaction setup using high-fidelity Q5 Taq polymerase ( NEB Biolabs , UK ) and the following cycle conditions: 98°C 120 sec , 33x 98°C 30 sec , 58°C 25 sec , 72° 90 sec and 72°C 4 min . PCR fragments were size separated on a 1 . 3% agarose gel ( 0 . 5x TBE ) on 140 voltage for 2 . 5 h in 0 . 5x TBE buffer . Sequence information of a subset of these PCR fragments were obtained following subsequent standard cloning procedure with pGEM-T vector system ( Promega , Germany ) and double strand sequencing of clones ( GATC Biotech , Konstanz , Germany ) .
|
Identifying the genes and genetic changes responsible for environmental adaptation is an important step towards understanding how species evolve . The honey bee Apis mellifera has adapted to a variety of habitats across its worldwide geographical distribution . Here we aim to identify the genetic basis of adaptation in honey bees living at high altitudes in the mountains of East Africa , which differ in appearance and behavior from their lowland relatives . We compare whole genome sequences from highland and lowland populations and find that , although in general they are extremely similar , there are two specific chromosomal regions ( representing 1 . 4% of the genome ) where they are strongly differentiated . These regions appear to represent structural rearrangements that are strongly correlated with altitude and contain many genes . One of these genomic regions harbors a set of octopamine receptor genes , which we hypothesize regulate differences in learning and foraging behavior between highland and lowland bees . The extremely high divergence between highland and lowland genetic variants in these regions indicates that they have an ancient origin and were likely to have been involved in environmental adaptation even before honey bees came to inhabit their current range .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"invertebrates",
"ecology",
"and",
"environmental",
"sciences",
"honey",
"bees",
"population",
"genetics",
"invertebrate",
"genomics",
"animals",
"genetic",
"mapping",
"evolutionary",
"adaptation",
"population",
"biology",
"bees",
"hymenoptera",
"ecosystems",
"animal",
"genomics",
"insects",
"genetic",
"loci",
"arthropoda",
"haplotypes",
"ecology",
"forests",
"heredity",
"genetics",
"biology",
"and",
"life",
"sciences",
"genomics",
"evolutionary",
"biology",
"evolutionary",
"processes",
"organisms",
"terrestrial",
"environments"
] |
2017
|
Two extended haplotype blocks are associated with adaptation to high altitude habitats in East African honey bees
|
Epithelial-mesenchymal-transition promotes intra-tumoral heterogeneity , by enhancing tumor cell invasiveness and promoting drug resistance . We integrated transcriptomic data for two clonal subpopulations from a prostate cancer cell line ( PC-3 ) into a genome-scale metabolic network model to explore their metabolic differences and potential vulnerabilities . In this dual cell model , PC-3/S cells express Epithelial-mesenchymal-transition markers and display high invasiveness and low metastatic potential , while PC-3/M cells present the opposite phenotype and higher proliferative rate . Model-driven analysis and experimental validations unveiled a marked metabolic reprogramming in long-chain fatty acids metabolism . While PC-3/M cells showed an enhanced entry of long-chain fatty acids into the mitochondria , PC-3/S cells used long-chain fatty acids as precursors of eicosanoid metabolism . We suggest that this metabolic reprogramming endows PC-3/M cells with augmented energy metabolism for fast proliferation and PC-3/S cells with increased eicosanoid production impacting angiogenesis , cell adhesion and invasion . PC-3/S metabolism also promotes the accumulation of docosahexaenoic acid , a long-chain fatty acid with antiproliferative effects . The potential therapeutic significance of our model was supported by a differential sensitivity of PC-3/M cells to etomoxir , an inhibitor of long-chain fatty acid transport to the mitochondria .
Prostate cancer ( PC ) is the most commonly diagnosed non-cutaneous malignancy among Western men and accounts for the second leading cause of cancer-related death [1] . In the majority of cases , PC eventually becomes independent of androgens , resuming growth after androgen-deprivation therapies in a more aggressive and therapy-refractory form [2] . The coexistence within the same tumor of a variety of cell subpopulations , featuring different phenotypes ( intra-tumoral heterogeneity ) associated with tumor evolution and progression reflects extreme plasticity and adaptation capability of neoplastic cells . This diversity is reached through genetic evolution of neoplastic cells and epigenetic and metabolic reprogramming of neoplastic and non-neoplastic tumor components that enhance tumor progression and represent a challenge for targeted therapies [3 , 4] . A major driver of intra-tumor heterogeneity is Epithelial-Mesenchymal transition ( EMT ) , which induces alterations in the intricate and large cancer cell gene regulatory and metabolic networks ( metabolic reprogramming ) [5] . However , although EMT-mediated molecular and cellular changes have been widely studied , the EMT-induced metabolic changes remain poorly understood . In this sense , it is widely accepted that metabolic reprogramming is one of the ten hallmarks of cancer [6] which endows cancer cells with a phenotype characterized by a rapid and continuous proliferation , metastasis , invasion , and treatment resistance . Thus , study of the metabolism in these heterogeneous cellular populations is of special interest and must be approached from a global perspective integrating global metabolism with consideration of different subpopulations . In this context , integration of omics data from high-throughput technologies , such as transcriptomics , into a genome-scale metabolic network reconstruction analysis , has been successfully used to study the metabolic mechanisms underlying different cancer types [7 , 8] . However , the differences in metabolic physiology between intra-tumoral subpopulations have not yet been taken into account in these computational approaches . Here , we have built comparative genome-scale metabolic network models based on transcriptomic data for two clonal sub-populations isolated and separated from an established prostate cancer cell line ( PC-3 ) : i ) a Cancer Stem Cell subpopulation -CSC- with high metastatic potential , low invasiveness and a higher proliferation rate ( PC-3/M cells ) and ii ) a non-CSC subpopulation expressing EMT markers with high invasiveness and low metastatic potential ( PC-3/S cells ) [9] . These neoplastic cell sub-populations , capturing extreme epithelial vs . mesenchymal phenotypes , were derived from the same tumor cell line and represent an excellent cellular model to study how intra-tumoral heterogeneity and the different phenotypes endowed by the different subpopulations provides advantages to the tumor in terms of metastatic capability and drug resistance . Our computational analysis has unveiled several subpopulation-specific metabolic alterations associated with long-chain fatty acids ( LCFA ) metabolism . First , we have identified an increased transport activity of LCFA into the mitochondria via Carnitine palmitoyl transferase I ( CPT1 ) , suggesting an increased β-oxidation , which could enhance proliferation in the PC-3/M subpopulation . Second , PC-3/S cells are predicted to have enhanced conversion of LCFA to arachidonic acid ( AA ) , the precursor of a variety of eicosanoids that enhance angiogenesis , cell adhesion and invasion . Finally , the lower CPT1 activity predicted in PC-3/S cells leads to Docosahexaenoic acid ( DHA ) accumulation , a LCFA with antiproliferative effects [10] . The latter prediction is consistent with the lower proliferation rate observed in PC-3/S cells . Next , using targeted metabolomics measurements , we experimentally confirmed these predictions and demonstrated that: i ) the low proliferative rate of PC-3/S cells compared with PC-3/M subpopulation is consistent with higher intracellular concentrations of DHA ii ) PC-3/M cells presents a higher CPT1 and β-oxidation activity that can be associated with their observed higher proliferative rate and iii ) in PC-3/S , the reported cell adhesion and enhanced invasive capability can be explained by higher levels of AA and eicosanoids , PGE2 and 12S-HETE . Finally , we experimentally showed that the low efficacy of etomoxir ( a CPT1 inhibitor ) in metastatic PC tumors is conferred by the low sensitivity of non-metastatic subpopulation ( PC-3/S ) towards this drug in contrast with the high sensitivity showed by metastatic subpopulation ( PC-3/M ) and can be explained by altered LCFA transport activity facilitated by CPT1 . The approach presented hereby provides a tool to unveil key metabolic nodes and vulnerabilities specific to distinct cancer cell subpopulations and opens new avenues in the development of more specific and efficient anti-tumoral therapies .
To infer the activity states of the metabolic networks of PC-3/S and PC-3/M subpopulations , we used previously generated transcriptomic data for PC-3/S and PC-3/M cells based on microarray technology [9] which was integrated into a genome-scale reconstruction of the human metabolic network [11] . In brief , this integrative method defines an upper threshold above which the genes are considered highly expressed and a lower threshold below which the genes are considered lowly expressed and seeks a network activity state in which the number of active reactions associated with highly expressed genes and the number of inactive reactions associated with lowly expressed genes are maximized [12 , 13] . In other words , this approach defines an objective function intended to maximize the similarity between gene expression and the activity state of the metabolic network rather than predefine an objective function that may not properly describe the cellular phenotype ( i . e . biomass maximization ) . Next , we identified a set of reactions whose activity state was unambiguously different between subpopulations using sensitivity analysis ( see Methods and Supplementary information S1 File and S1 Text ) . Finally , our computational analysis permitted to infer , for each subpopulation , a set of active reactions , either intracellular or nutrients uptake/secretion between cell and media ( see Fig 1 and Supplementary information S1 Text ) . The predicted metabolic uptake/secretion rates were in accordance with experimental measurements ( true positive rate of 70% with an associated p-value < 0 . 001; [14] ) . For instance , our model-driven analysis successfully predicted the consumption / production patterns of most of the bioactive amino-acids , as well as glucose consumption and lactate production in both subpopulations ( Supplementary information S1 Text ) . This model-driven analysis revealed two major differences between the two subpopulations . First , the activity of fatty acid oxidation predicted by the model is higher in PC-3/M than in PC-3/S cells . The oxidation of fatty acids in the mitochondria produces NADH , FADH2 and acetyl-CoA that fuels the production of energy via TCA cycle and the electron transport chain . Most of the reactions differentially activated in this pathway involve carnitine palmitoyl transferase 1 ( CPT1 ) . This mitochondrial membrane protein actively transports LCFA from cytosol into the mitochondria [15] . Our analysis provided a set of eight cytosolic LCFAs that were predicted to be substrates of CPT1 exclusively in PC-3/M cells . Interestingly , it has been reported that five of these eight LCFAs , including DHA , have antiproliferative effects [15–20] . Second , the analysis predicted an increased activity of eicosanoid metabolism in PC-3/S cells that is associated with angiogenesis , cell invasion and adhesion [21–23] . Arachidoic acid metabolism ( AA ) is the main precursor of this pathway that in turn is fueled by a fraction of the LCFA previously mentioned . Finally , our analysis revealed other significant differences between the metabolic activities of the two PC-3 subpopulations that are consistent with published evidences . For example , Vitamin D3 metabolism was predicted to be more active in the PC-3/S subpopulation . This molecule controls proliferation in prostate cells [24] and has antiproliferative effects on a number of cancer cell lines , being PC-3 cells ( parental cell line ) one of the few cell lines insensitive to this drug [25] . Taken together , the in silico analysis suggests the occurrence of metabolic alterations that correlate with a more proliferative phenotype in PC-3/M cells and with a less proliferative and more invasive phenotype in PC-3/S cells . These predictions are consistent with the reported phenotypes of both PC3 subpopulations [9] . Our computational analysis predicts both a higher LCFA entry into the mitochondria via CPT1 and a more active LCFA β-oxidation in PC-3/M cells . LCFA must be imported into mitochondria to be degraded via β-oxidation and CPT1 , a mitochondrial membrane enzyme that plays a critical role in its transport into the mitochondrial matrix [15 , 26] . To experimentally verify that CPT1 protein levels differ between PC-3/M and PC-3/S cells , we compared CPT1 levels in PC-3/M and PC-3/S subpopulations by western blotting ( Fig 2C ) . In line with the computational inference , we found that PC-3M cells express 30% higher levels of CPT1 than PC-3/S cells . As CPT1 mRNA levels are not significantly different between the two subpopulations ( 7 . 68±0 . 22 A . U in PC-3/M cells and 7 . 13±0 . 033 A . U . in PC-3/S cells and p-value > 0 . 01 using T-test ) , this highlights the power of the data integration approach to infer metabolic alterations even if mRNA and protein-level changes do no match owing to post-transcriptional regulation [13] . To experimentally determine whether β-oxidation activity is higher in PC-3/M , we measured acylcarnitine levels . These molecules are CPT1 activity intermediates in the entry of acyl‐CoAs into the mitochondrial matrix and are used to experimentally infer the β-oxidation activity [27] . Here we found that acylcarnitine levels were significantly higher in PC-3/M compared with PC-3/S cells ( Fig 2E and Supplementary information S6 File ) , which is in accordance with our computational predictions and supports the hypothesis that PC-3/M cells present a more active β-oxidation . Etomoxir is a CPT1 inhibitor that consequently inhibits β-oxidation and its associated oxygen consumption [15 , 26] . Since in many cancer types , tumor onset and progression relies more on lipid fuel than on aerobic glycolysis , this compound is widely used in cancer research [15] but not in clinical practice due to its hepatotoxicity . However , as stated above , its antiproliferative efficacy on PC-3 cells is the lowest among the different PC cell lines studied [15] . Based on these evidences and the results of our analysis , we hypothesized that this could be explained by a low activity of CPT1 in the PC-3/S subpopulation . To experimentally test this hypothesis , we measured the oxygen consumption rate ( OCR ) before and after exposure to etomoxir of both subpopulations ( Fig 2B , see Methods ) . We found that PC3/M cells show a 30% higher sensitivity to CPT1 inhibition than PC-3/S cells , implying that CPT1 and hence β-oxidation is more rate-limiting in the PC-3/M subpopulation . In addition , we studied the dose-effect relation between etomoxir and the proliferation of PC-3/M , PC-3/S and parental PC-3 cells . This experiment also showed a higher sensitivity of PC-3/M cells towards the antiproliferative effects of etomoxir and the difference in the proliferative rate between subpopulations increased with higher concentrations of etomoxir ( Fig 2F and Supplementary information S5 File ) . These experimental observations are in accordance with our computational predictions . Furthermore , PC-3 cells that represent a heterogeneous population containing cells with “PC-3/M-like” and “PC-3/S-like” phenotypes showed an intermediate sensitivity which supports our hypothesis that the low sensitivity of PC-3 cells to etomoxir may be conferred by tumor cell subpopulations with PC-3/S-like metabolic features . We also hypothesized that , as a side effect of the relatively low CPT1 activity in PC-3/S cells , the levels of some LCFAs would be higher in this subpopulation compared to PC-3/M cells . More specifically , we focused on determining the levels of docosahexaenoic acid ( DHA ) , a compound with antiproliferative effects in cancer [28–31] . To this aim , we used a targeted approach based on the Biocrates platform Assay [32] to quantitatively measure the concentration of DHA in both PC-3 subpopulations . We found that the concentration of DHA was 262% higher in PC-3/S cells compared to PC-3/M cells ( Fig 2D , see Methods ) , which supports our model-driven predictions and is consistent with the lower proliferation rate reported in the PC-3/S subpopulation . A further key prediction of our model-driven analysis is that the eicosanoid metabolism is more active in PC-3/S cells . Arachidonic acid ( AA ) is the precursor of this pathway , which in turn is metabolized from some of the eight LCFAs that are predicted to be transported by CPT1 into the mitochondria exclusively in PC-3/M cells . Based on these results , we hypothesized that the levels of AA and other eicosanoids were higher in PC-3/S than in PC-3/M cells . To validate this hypothesis , we applied a targeted approach using the Biocrates platform Assay [32] , which allows quantitative measurements of arachidonic acid , eicosanoids and other oxidation products of polyunsaturated Fatty Acids ( PUFAs ) . Among all the metabolites measured by this platform ( Supplementary information S4 File ) , AA , Prostaglandin E2 ( PGE2 ) and 12 ( S ) -hydroxy-5Z , 8Z , 10E , 14Z-eicosatetraenoic acid ( 12S-HETE ) were significantly higher in PC-3/S than in PC3-M cells ( Fig 3 ) . 12-HETE is the dominant AA metabolite in PC3 cells and its levels in human prostate cancer tissues exceed by > 9-fold its levels in normal human prostate tissue [34] . Furthermore , in PC3 cells , 12 ( S ) -HETE increases the expression of ITGA3V gene , which is associated with cell adhesion [20] and promotes PGE2 metabolism in cultured PC3 cells [35] . High PGE2 levels are associated with cancer [36–38] and affects different mechanisms that have been shown to play a role in carcinogenesis such as cell invasion via the Akt signaling pathway [22] or angiogenesis by over-expressing the VEGF gene [23 , 39] . Consistent with these regulatory effects of AA metabolites , the expression level of ITGA3V , VEGF and AKT genes were significantly higher in PC-3/S than in PC-3/M cells ( Log2 FC of 1 . 24 ± 0 . 27 , 0 . 79 ± 0 . 25 , 1 . 46 ± 0 . 4 respectively ) . Taken together , these results indicate a higher activity of eicosanoid metabolism in PC-3/S cells leading to higher levels of eicosanoid pathway intermediates and the upregulation of genes associated with angiogenesis , cell adhesion and invasion , all of which are consistent with the phenotype observed in this subpopulation .
Intratumoral heterogeneity is key to understanding the hierarchical and functional relationships between different neoplastic cell populations within a given tumor , with direct implications on tumor dynamics and progression [40] . Here we have focused on the study of the metabolic profiles of two clonal cell sub-populations isolated from an established prostate cancer cell line ( PC-3 ) : PC-3/M and PC-3/S cells [9] . These sub-populations were derived from the same PC cell line and thus they can be assumed to coexist within the same tumor , representing an excellent model to study how intra-tumoral heterogeneity benefits the tumor in terms of invasiveness , metastatic capability and drug resistance . This fact also enables us to investigate the relationships between gene expression and metabolism with tumor-initiating cell or mesenchymal-like attributes in neoplastic cells . Here , we inferred the metabolic activity states of these PC-3 subpopulations by integrating transcriptomic data with a genome-scale metabolic network reconstruction . The applied constraint-based method treats gene expression levels merely as cues for the likelihood that their associated reactions carry metabolic fluxes and hence allowing for potential post-transcriptional regulatory effects . For example , a reaction associated with a highly expressed gene does not necessarily entail a high flux . As a consequence , the method allows us to infer metabolic activity patterns that go beyond conventional gene expression analysis . Indeed , some of the inferred metabolic differences between PC-3/M and PC-3/S cells do not correlate with the expression patterns of the underlying genes . One example is provided by the expression levels of genes associated with the fatty-acyl-ACP hydrolase reaction that participates in the oxidation of fatty acids . The transcriptomic profiles suggest that these genes are more active in PC-3/S cells , which is in contrast to our computational analysis identifying the corresponding reaction to be active only in the PC-3/M subpopulation ( See Supplementary information S3 File ) . Importantly , we have provided experimental support for this prediction by observing a significantly higher sensitivity of PC-3/M cells to CPT1 and β-oxidation inhibition by etomoxir compared to PC-3/S cells , thereby demonstrating the importance of considering the network context when inferring metabolic changes from transcriptomic data . Overall , our approach has revealed two major metabolic differences at the level of LCFA utilization with relevance for tumor proliferation , invasiveness and metastasis in our dual cell model . First , it has unveiled an increased CPT1 activity in PC-3/M cells . CPT1 has numerous cytosolic substrates , including cervonyl coenzyme A ( DHA precursor ) , eicosatetranoyl coenzyme A , arachidyl coenzyme A ( arachidonic acid precursor ) , trans-2-octadecenoyl-CoA ( 4- ) , palmitate , Malonyl-CoA , linoelaidyl coenzyme A ( linoleic acid precursor ) and vaccenyl coenzyme A . Interestingly , it has been reported that five of these eight LCFAs have antiproliferative effects [16–20] . Thus , the higher CPT1 and β-oxidation activities in PC-3/M cells may have two roles: i ) first , and probably the most evident , is to maintain the energetic requirements imposed by the high proliferation rates of PC-3/M cells and ii ) to eliminate LCFAs with antiproliferative effects . The predicted differences in CPT1 activity were supported experimentally by the finding of 33% higher CPT1 protein levels in PC-3/M than PC-3/S cells . Further , we demonstrated that β-oxidation activity was more sensitive to the inhibition of CPT1with etomoxir in PC-3/M than in PC-3/S cells ( see Fig 2B ) which highlights the importance of this enzyme in the energy metabolism of this subpopulation . This is of special interest since fatty acid oxidation plays a key role as source of NADH , NADPH , ATP and FADH2 , all providing survival advantage to cancer cells [41] . Finally , we showed that the concentration of DHA is significantly higher in PC-3/S than in PC-3/M cells ( Fig 2D ) . Several studies have reported anti-proliferative effects of DHA in tumors , consistent with the high proliferative rate observed in PC-3/M cells and support the hypothesis that an increased activity of CPT1 is also necessary to eliminate anti-proliferative molecules in PC-3/M cells . Taken together , our findings supports the key role played by CPT1 to sustain the high proliferation rate of PC-3/M cells by degrading LCFAs through energy metabolism while avoiding their antiproliferative effects . Secondly , our analysis predicted a higher activity of the eicosanoid metabolism in PC-3/S cells . Most of the LCFAs described in this study are AA precursors that in turn fuel this pathway . Eicosanoid metabolism produces a variety of molecules with reported tumorogenic activity in prostate cancer [42] . Importantly , these processes are predicted to occur in the lysosome , in which is reported that may produce pro-oncogenic alterations [43] . Here we validated this prediction by using metabolomic measurements which revealed higher levels of AA in PC-3/S ( see methods and Fig 3 ) . In addition , in line with the computational predictions , the concentrations of several products of eicosanoid metabolism , such as 12S-HETEand PGE2 , were significantly higher in the PC-3/S subpopulation . Prior evidence suggests that these compounds are associated with the upregulation of ITGA3V , VEGF and AKT [21–23 , 39] , which promote cell adhesion , angiogenesis and cell invasion . Importantly , we have found that these genes are indeed upregulated in PC-3/S cells . Thus , our findings suggest that higher levels of AA , 12S-HETE and PGE2 , associated with a more active eicosanoid metabolism in PC-3/S cells , contribute to increased angiogenesis , cell adhesion and invasion potentials of these cells . Overall , our findings support the view that the relatively high activity of CPT1 in PC-3/M cells increases the entry of LCFAs into the mitochondria to be oxidized and to produce energy to sustain a high proliferation rate ( Fig 4 ) . This process decreases the levels of LCFAs such as DHA , thus preventing their antiproliferative effects . In contrast , the lower CPT1 activity in PC-3/S cells would lead to an accumulation of anti-proliferative LCFAs in the cytosol , thereby reducing the growth rate in PC-3/S cells and increasing the availability of substrates for eicosanoid synthesis ( Fig 4A ) . Summing up , we propose that the metabolic reprogramming involving LCFA utilization enhances the metastatic potential and proliferation in PC-3/M cells while in PC-3/S subpopulation increases cell adhesion , invasion and angiogenic capability and promotes DHA accumulation which reduces its proliferation . The model-driven analysis employed here has provided additional insights into metabolic changes linked to cancer phenotypes . For example , our analysis further predicted acid ceramidase ( ASAH1 ) to be active predominantly in PC-3/M cells , a prediction consistent with experimental evidences showing a higher ASAH1 enzymatic activity in this subpopulation [44] . Our analysis also predicts that calcitriol metabolism is mainly active in PC-3/S cells . This molecule has antiproliferative activity in a variety of human cancer cells [25] which is consistent with the low proliferative rate of PC-3/S cells compared with PC-3/M cells [9] . Our model prediction also suggests that the reported low sensitivity of PC-3 cells towards Vitamin D3 [24] could be conferred by the low Vitamin D metabolism activity in the PC-3/M subpopulation . In addition , this prediction is consistent with the lower proliferative rate observed in PC-3/S cells , which would be more sensitive to Vitamin D3 anti-proliferative effects . Our analysis of a dual-cell model representing distinct and opposing neoplastic phenotypes allows us to propose subpopulation-specific and complementary therapeutic interventions . The results of the experiment determining the dose-effect relation between etomoxir concentration and cell proliferation showed that PC-3/M cells are more sensitive to etomoxir than PC-3/S cells , and that the parental PC-3 cell line presents an intermediate sensitivity . Thus , the poor performance of etomoxir at inhibiting the growth of PC-3 cells compared to other prostate cell lines may be explained by the low metabolic dependence of the PC-3/S subpopulation on CPT1 . In other words , androgen-independent prostate cancer cells with CSC attributes similar to PC-3/M cells would be sensitive to etomoxir , while this drug would be less efficient in tumor cell subpopulations with EMT attributes similar to PC-3/S cells with a phenotype characterized by high cell invasion and adhesion and angiogenic capability . Finally , it has been reported that the cox-2 reaction , which produces PGE2 and is over-expressed in prostate cancer [45] , is activated by 12S-HETE [46] which is in turn metabolized by the 12-LOX reaction . Interestingly , a number of drugs such as cinnamyl-3 , 4-dihydroxy-alpha-cyanocinnamate ( CDC ) or baicalein , that inhibit 12-LOX activity , have been shown to present strong anti-tumoral effects in prostate cancer [47] . Our study represents a novel approach to discern metabolic vulnerabilities associated with heterogeneous tumor cell populations . However , future studies measuring the effects of single and combinatorial drug treatments affecting subpopulation-specific targets on heterogeneous co-culture of non-CSC ( PC-3/S ) and CSC ( PC-3/M ) subpopulations are needed to determine the significance of these findings . For instance , the combinatorial effect of CDC or baicalein with drugs such as oxfenicine or perhexiline ( CPT1 inhibitors without the hepatotoxicity of etomoxir–[48] ) could be tested as potential anti-tumoral drug treatments targeting the key metabolic processes preferentially active in PC-3/S or PC-3/M cells , respectively . Additionally , as gene networks associated with progression and metastasis in our PC-3 dual model is significantly correlated with those in other tumor types [14] , the metabolic reprogramming proposed here could be extrapolated to different cancer types . Our findings will facilitate a better understanding of the EMT-induced metabolic changes and their role in tumor heterogeneity and opens new avenues for the development of new subpopulation-specific anti-cancer therapies .
Transcriptomic data: Gene expression levels of each cell subpopulation ( three replicates per subpopulation GSE24868 , [9] ) by microarray analysis ( Affymetrix genechip u133a 2 . 0 ) and normalized by RMA [49] . Transcriptomic data was integrated into a genome-scale metabolic network reconstruction analysis to infer the activity state of the metabolic reactions in both subpopulations . Consumption and production of metabolites: Additionally , we used the measured consumption and production of some metabolites [14] to assess the reliability of model predictions ( Supplementary information S1 Text ) . These metabolites were: glucose , lactate , pyruvate , glutamate and aminoacids . To obtain accurate cell-specific genome-scale metabolic models of the PC-3 subpopulations , we performed a subpopulation-specific genome-scale network reconstruction analysis by integrating the transcriptomic data into the most recent reconstruction of human metabolism ( Recon2 ) [11] . Recon2 is a genome-scale stoichiometric model that represents the entire network of human metabolic reactions . This generic genome-scale metabolic model provides the appropriate transcript-protein-reaction associations that permit the integration of the previously mentioned transcriptomic data for which we used a widely tested constraint-based method [12] . In order to reduce the computational time necessary to perform the analysis , the metabolic model ( Recon2 ) [11] was reduced . The reduction was done by removing the blocked reactions from the model . These reactions are those incapable of carrying any metabolic flux in steady state [50] . To this aim we first performed a Flux Variability Analysis ( FVA ) [51–53] using Fasimu software [54] . This analysis computes minimal and maximal flux in each reaction . Each analysis evaluates the feasibility of the simulation . The reactions in which their maximization and minimization simulations were not obtained any feasible solution were considered as blocked reactions . In order to ensure that the reduced model was able to consume/produce the experimentally measured extracellular metabolites , we forced the corresponding exchange reactions to be always active . It was achieved by splitting all the exchange reactions in a forward and a backward reaction and the lower/upper bounds of the reactions associated to the experimentally measured metabolites were fixed at 0 . 001/1000 in the forward reactions and at -1000/-0 . 001 in the backward reactions . Once determined , the blocked reactions were removed from the model , as well as those metabolites that were neither products nor substrates of any reaction . We integrated the transcriptomic data into Recon2 [11] by using the gene-protein-reaction ( GPR ) associations included in the model . These associations are “and/or” logical sentences that establish a relation between the metabolic reactions and the genes encoding the enzymes that catalyze them . GPR associations include information related with isoenzymes ( using the logical “or” ) , complexes ( using the logical “and” ) or direct gene-reaction relations ( i . e . the activity of Reaction1 depends on: “ ( geneA and geneB ) or ( geneC and geneD ) ” ) . To integrate the gene expression data from PC-3/M and PC-3/S subpopulations into Recon2 and determine the gene expression level associated to the metabolic reactions in each subpopulation , we substituted the logical “and” and “or” by “minimum” and “maximum” . Thus , for example , if the activity of a given reaction depends on the expression of different genes and it is defined by the following logical expression “ ( geneA and geneB ) or ( geneC and geneD ) ” , and the expression of the gene A , B , C and D are 0 . 5 , 3 , 1 and 0 . 1 respectively . Then , by integrating the gene expression levels into the logical sentence and replacing the logical operators by “minimum” and “maximum” we obtained the following expression: “max ( min ( 0 . 5 , 3 ) , min ( 1 , 0 . 1 ) ) ” . Thus , based on the transcriptomic data and the GPR association , the gene expression associated with the reaction is 0 . 5 . Finally , we obtained a numerical value for each reaction indicating the level of expression of their corresponding associated genes . We used gene expression levels associated with the metabolic reactions to infer the activity states of reactions in the network by using a recently developed constraint-based method [12] . This method solves a mixed integer linear programming ( MILP ) problem to obtain a flux distribution in which the number of reactions associated with highly expressed genes is maximized ( RH ) , and the number of reactions associated with lowly expressed genes is minimized ( RL ) while satisfying the thermodynamic and stoichiometric constrains imposed by the model: maxv , y+ , y−= ( ∑i∈RH ( yi++yi− ) +∑i∈RLyi+ ) S∈v=0 ( 1 ) The mass balance constraint: where v is the flux vector and S is a n x m stoichiometric matrix , in which n is the number of metabolites and m is the number of reactions . Thermodynamic constraints , that restrict flow direction , are imposed by setting vmin and vmax as lower and upper bounds respectively . The Boolean variables y+ and y– . In RH reactions represent whether the reaction is active or not respectively . In RL y+ represents the reaction is not active . A highly expressed reaction is considered to be active if it carries a significant positive flux that is greater than a positive threshold Ɛ . In our study Ɛ = 1 . Consequently the ith reaction is active if: vi ≥ 1 vi+yi− ( vmin , i+ε ) ≤vmax , i , i∈RH or has a significant negative flux <–Ɛ ( as our model didn't consider reversible reactions it cannot occur ) vmin ( 1−yi+ ) ≤vi≤vmax , i ( 1−yi+ ) , i∈RL ( 5 ) Lowly expressed reactions are considered to be inactive if they carry zero metabolic flux , though changing Eq ( 5 ) to enable these reactions to carry a low metabolite flux ( that is , with an upper bound lower than Ɛ ) and still be considered inactive provides qualitatively similar results . The Fig 5 illustrates the process . This method defines an upper threshold above which the expression of a given gene is considered high and another threshold below which gene expression is considered low . In our study , the chosen upper and lower thresholds were those symmetric percentiles that maximize the cases where the number of reactions associated with highly expressed genes in one subpopulation were associated with lowly expressed gene in the other subpopulation and vice versa . Thus , we defined the upper threshold at the 66th percentile and the lower threshold at the 33th percentile . The method also uses the parameter that represents the flux above which a given reaction is considered to carry a significant metabolic flux . As is defined in [12] we gave to Ɛ a value of 1 . Once the thresholds were fixed , we performed the expression-based activity prediction analysis with Fasimu software by applying “compute-FBA–xs” option ( See Fasimu tutorial [54] ) . In the Expression-based activity prediction analysis we found an optimal solution in terms of the objective function maximization , although this solution may not be unique . A space may exist of alternative optimal solutions that represent alternative steady-state flux distributions yielding the same similarity with the gene expression data ( the same objective function value ) . To account for these alternative solutions , we employed Sensitivity analysis [12] . This is performed by solving two MILP problems ( as is described in Expression-based activity prediction ) for each reaction to find the maximal attainable similarity with the expression data when the reaction is: ( i ) forced to be activated and ( ii ) forced to be inactivated . Thus , a reaction is considered to be active if a higher similarity with the expression data is achieved when the reaction is forced active than when it is inactive ( the objective function is higher when the reaction is active ) . Conversely , it is considered to be inactive if the similarity is higher when the reaction is forced to be inactive . If the similarities with the experimental data are equal in both cases the activity state of the reaction is considered to be undetermined . From this analysis we could infer which pathways are more active in each subpopulation . By analyzing the predicted activity state of the exchange reaction we can infer which metabolites are consumed and/or produced . In order to determine the goodness of our model predictions we compared qualitatively the consumption and production of some experimentally measured metabolites [14] with the corresponding model predictions ( Supplementary information S1 Text ) . This comparison was done by constructing a 2x2 contingency matrix and the levels of significance were determined using Fisher exact test ( Supplementary information S1 Text ) . The algorithm used to integrate the information from gene expression levels into a Genome-scale metabolic network reconstruction defines a threshold above which gene expression levels are considered high and a second threshold below which they are considered low . It calls for the performance of a robustness analysis in order to demonstrate the lack of dependency of our predictions on the thresholds used in the analysis . In order to determine the robustness of our prediction , we performed the analysis previously described in sensitivity analysis defining different sets of thresholds: Thereby , we defined a set of reactions that were predicted to be active , inactive or undetermined ( the method cannot predict their activity state ) independently of the thresholds . Cells were seeded in XF24-well microplates ( Seahorse Bioscience ) at 4 . 5·104 cells/well and 9 . 0·104 cells/well , respectively , in 100 μL of growth medium , adding 100 μL more after 3–5 h , and then incubated at 37°C with 5% CO2 overnight . After overnight incubation and 1 h before the assay , growth media was replaced by basal media ( unbuffered DMEM; Sigma-Aldrich ) with 3 mM glucose and 5 mM carnitine . The sensor cartridge was loaded with etomoxir and calibrated prior to the start of the experiment . Determinations were performed on a XF24 Extracellular Flux Analyzer ( Seahorse Bioscience ) . Responses to etomoxir ( Signma-Aldrich ) treatment ( final concentration 30 μM ) were expressed as LOG2 to indicate the fold change comparing the measured point immediately after and before the corresponding injection . To determine eicosanoids and oxidation products of polyunsaturate fatty acids levels in PC-3/M and PC-3/S cells we used Biocrates triple quadrupole MS-based platforms [32] . This platform enables the systematic quantification of relevant biological metabolites . The method is a quantitative screen of selected metabolites detected with multiple reaction monitoring , neutral loss and precursor ion scans . Metabolites are then quantified by comparison to structurally similar molecules labeled with stable isotopes added to the samples in defined concentrations as internal standards . The process is controlled by MetIDQ Software which controls sample management , data collection , data validation , and analysis . Cell extracts were obtained from frozen cell pellets using RIPA buffer ( 50 mM Tris , pH 8 . 0 , 150 mM NaCl , 0 . 1% SDS , 1% Triton X-100 and 0 . 5% sodium deoxycholate ) supplemented with protease inhibitor cocktail ( Sigma-Aldrich ) . Protein concentrations from the supernatant were determined by the BCA assay . Thirty-five mg of protein per sample were loaded and separated by 10% SDS-PAGE and transferred to PVDF membranes . Membranes were blocked by incubation with PBS-Tween ( 0 . 1% ( v/v ) ) containing 5% non-fat dried milk for 1 hour at room temperature . Then , membranes were incubated with CPT1 primary antibody ( Sigma-Aldrich , SAB1410234 , 1/200 ) , rinsed with PBS-Tween ( 0 . 1% ( v/v ) ) and finally incubated with the secondary antibody anti-rabbit ( Amersham Biosciences , NA934V , 1/3000 ) for 1 hour at room temperature . Blots were treated with the Immobilon ECL Western Blotting Detection Kit Reagent ( Millipore ) and developed after exposure to Fujifilm X-ray film . For sample acquisition and processing , 5 106 cells of PC-3M and PC-3S cell lines were tripsinized and washed twice with ice-cold PBS prior to snap-freezing in liquid nitrogen . Cell pellets were stored at -80°C until measure . Right before measuring , cell pellets were thawed at room temperature and resuspended in 70 μL of 85:15 EtOH:PBS solution . Cells were disrupted by two sonication/freezing/defreezing cycles using titanium probe ( VibraCell , Sonics & Materials Inc . , Tune: 50 , Output: 25 ) , liquid N2 and a 95°C heat block . Cell lysates were then centrifuged at 20 , 000 rcf for 5 minutes at 4°C . Supernatants were collected into new tubes and total protein content was determined by Bichinconinic acid ( BCA ) assay ( Thermo Fisher Scientific , Waltham , MA USA ) . Then , standards , internal standards , quality controls ( 10 μL of each ) , and samples ( 30 μL ) were loaded into the Biocrates AbsoluteIDQ® p180 plates ( BIOCRATES Life Sciences AG , Innsbruck , Austria ) , processed according to manufacturer instructions and measured by FIA-MS/MS using a SCIEX 4000 QTRAP mass spectrometer . Concentrations for metabolites were determined using the MetIDQ™ software package , which is an integral part of the AbsoluteIDQ® kit . The obtained metabolite concentrations were corrected considering the loaded volume of cell lysates and normalized by protein content .
|
The coexistence within the same tumor of a variety of subpopulations , featuring different phenotypes ( intra-tumoral heterogeneity ) represents a challenge for diagnosis , prognosis and targeted therapies . In this work , we have explored the metabolic differences underlying tumor heterogeneity by building cell-type-specific genome-scale metabolic models that integrate transcriptome and metabolome data of two clonal subpopulations derived from the same prostate cancer cell line ( PC-3 ) . These subpopulations display either highly proliferative , cancer stem cell ( PC-3/M ) or highly invasive , epithelial-mesenchymal-transition-like phenotypes ( PC-3/S ) . Our model-driven analysis and experimental validations have unveiled a differential utilization of the long-chain fatty acids pool in both subpopulations . More specifically , our findings show an enhanced entry of long-chain fatty acids into the mitochondria in PC-3/M cells , while in PC-3/S cells , long-chain fatty acids are used as precursors of eicosanoid metabolism . The different utilization of long-chain fatty acids between subpopulations endows PC-3/M cells with a highly proliferative phenotype while enhances PC-3/S invasive phenotype . The present work provides a tool to unveil key metabolic nodes associated with tumor heterogeneity and highlights potential subpopulation-specific targets with important therapeutic implications .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"cell",
"physiology",
"medicine",
"and",
"health",
"sciences",
"neurochemistry",
"eicosanoids",
"metabolic",
"networks",
"cell",
"metabolism",
"neuroscience",
"metabolites",
"network",
"analysis",
"genome",
"analysis",
"pharmacology",
"drug",
"metabolism",
"computer",
"and",
"information",
"sciences",
"lipids",
"genomics",
"neurochemicals",
"gene",
"expression",
"pharmacokinetics",
"biochemistry",
"cell",
"biology",
"genetics",
"transcriptome",
"analysis",
"biology",
"and",
"life",
"sciences",
"fatty",
"acids",
"computational",
"biology",
"metabolism"
] |
2018
|
Model-driven discovery of long-chain fatty acid metabolic reprogramming in heterogeneous prostate cancer cells
|
The 16p11 . 2 600 kb BP4-BP5 deletion and duplication syndromes have been associated with developmental delay; autism spectrum disorders; and reciprocal effects on the body mass index , head circumference and brain volumes . Here , we explored these relationships using novel engineered mouse models carrying a deletion ( Del/+ ) or a duplication ( Dup/+ ) of the Sult1a1-Spn region homologous to the human 16p11 . 2 BP4-BP5 locus . On a C57BL/6N inbred genetic background , Del/+ mice exhibited reduced weight and impaired adipogenesis , hyperactivity , repetitive behaviors , and recognition memory deficits . In contrast , Dup/+ mice showed largely opposite phenotypes . On a F1 C57BL/6N × C3B hybrid genetic background , we also observed alterations in social interaction in the Del/+ and the Dup/+ animals , with other robust phenotypes affecting recognition memory and weight . To explore the dosage effect of the 16p11 . 2 genes on metabolism , Del/+ and Dup/+ models were challenged with high fat and high sugar diet , which revealed opposite energy imbalance . Transcriptomic analysis revealed that the majority of the genes located in the Sult1a1-Spn region were sensitive to dosage with a major effect on several pathways associated with neurocognitive and metabolic phenotypes . Whereas the behavioral consequence of the 16p11 region genetic dosage was similar in mice and humans with activity and memory alterations , the metabolic defects were opposite: adult Del/+ mice are lean in comparison to the human obese phenotype and the Dup/+ mice are overweight in comparison to the human underweight phenotype . Together , these data indicate that the dosage imbalance at the 16p11 . 2 locus perturbs the expression of modifiers outside the CNV that can modulate the penetrance , expressivity and direction of effects in both humans and mice .
Understanding the interactions between genes that determine brain activity , metabolism and behavior is crucial to decipher changes in underlying neurocognitive disorders . For syndromes caused by copy number variation at the 16p11 . 2 locus , causal gene discovery is particularly challenging . The region shows a high density of genes , the majority of which are expressed in the brain and are potentially important for not only the normal development of the nervous system but they also seem to impact body mass index . The 16p11 . 2 600 kb BP4-BP5 breakpoint ( BP ) deletions and reciprocal duplications both have a population prevalence of approximately 1/1000 [1] , and both genetic lesions are found in almost 1% of all cases with intellectual disability ( ID ) [2] and autism spectrum disorders ( ASD ) [3–7] . Phonological processing and language disorders are affected in 56% ( deletion ) and 46% ( duplication ) of the 16p11 . 2 patients [7] connected with brain anatomic alterations of the auditory and language systems [8] , but no hearing impairment has been reported in these individuals so far [7 , 9] . In addition to ID and ASD , both rearrangements have also been associated with epilepsy [10–13] , whereas the duplication has been linked to schizophrenia , bipolar disorder and depression [14–16] . Opposite effects of 16p11 . 2 deletions and duplications on body mass index ( BMI ) and head size have also been reported in 16p11 . 2 CNVs carriers [1 , 11 , 12 , 17 , 18] . We showed recently that the brain volume and the specific cortico-striatal structures were similarly correlated with the number of copies of the 600 kb region [19] . The BP4-BP5 recurrent rearrangements contain 28 “unique” genes ( SPN , QPRT , C16orf54 , ZG16 , KIF22 , MAZ , PRRT2 , PAGR1 ( C16orf53 ) , MVP , CDIPT , CDIPT-AS , SEZ6L2 , ASPHD1 , KCTD13 , TMEM219 , TAOK2 , HIRIP3 , INO80E , DOC2A , C16orf92 , FAM57B , ALDOA , PP4C , TBX6 , YPEL3 , GDPD3 , MAPK3 , CORO1A ) and multiple copies of BOLA2/2B , SLX1A/1B , SULT1A3/4 and NPIP . Transcriptome profiling of lymphoblastoid cell lines of deletion and duplication carriers and control individuals showed that all genes with detectable expression within the CNV , in particular the BOLA2/2B , SLX1A/1B , SULT1A3/4 and SLX1A/1B-SULT1A3/A4 read-through transcripts , were correlated positively with gene dosage . Further , these analyses implicated a potential role for ciliary dysfunction in the 16p11 . 2 600 kb BP4-BP5 pathology [20] . The reciprocal impacts on BMI as well as brain volume and structures involved in reward , language and social cognition indicated that the phenotypes could have mirror aetiologies depending on the observed changes in gene transcript levels . To probe the mechanistic basis of the human genetic data , mouse models have been generated to study the correlation between phenotype and genotype . The first published mouse models of the 16p11 . 2 syndromes carry deletion and duplication of the Slx1b-Sept1 region and are reported to display locomotor activity alterations and ventral midbrain volume changes [21] . Despite the presence of compatible phenotypes , the interval targeted in this model includes four genes outside the human BP4-BP5 interval , potentially posing interpretive challenges . In addition , the consequence of the duplication was not investigated comprehensively in this report . More recently , a second 16p11 . 2 mouse model carrying the deletion of the Coro1a-Spn region was generated , mutant animals display circuit defects in the basal ganglia [22]; these animals have a hearing deficit whereas no particular hearing defect has been reported in humans with 16p11 . 2 CNVs [7] or in the first mouse model [21] . Given the possibility that 1 ) such condition might alter the response to behavioral tests and 2 ) the second mouse model might carry a second mutation causing deafness , we decided to engineer and characterize novel 16p11 . 2 CNV mouse models that carry rearrangements of the Sult1a1-Spn interval on an inbred C57BL/6N genetic background , which corresponded to the BP4-BP5 syntenic region in humans . We reasoned that evaluation of the breakpoint region in the engineered deletion or duplication mouse would allow us to decipher if changes in the number of copies have an impact on the phenotypes compared to the two published 16p11 . 2 models . To understand better the phenotypic variability in individuals carrying the 16p11 . 2 deletion or duplication , we also posited that evaluation of our model on an inbred C57BL/6N background would allow us to study the genetic background influence between the three mouse model phenotypes . Metabolic response [23] is an exemplar of a phenotype strongly influenced by the inbred genetic background . We thus performed the phenotypic characterization on inbred ( C57BL/6N ) and hybrid ( F1 C57BL/6NxC3B ) genetic backgrounds . We first utilized a comprehensive set of tests to assess the impact of the genetic rearrangements on behavior and hippocampal synaptic function to determine whether the models recapitulate some of the cognitive deficits and autistic traits associated with the 16p11 . 2 syndromes . Second , since humans with deletion or duplication of the 16p11 . 2 locus display body weight changes ( respectively obesity and leanness ) , we explored metabolism phenotypes in our mouse model on either a normal chow diet or with a high fat high sugar diet to challenge energy balance . Third , and to complete the phenotypic characterization of our models , we performed an analysis of craniofacial structure of the mutant mice . Finally we performed transcriptional analyses of different brain regions and the liver to understand better the consequences of the disease on brain and metabolic functions .
The deletion ( Del/+ ) and duplication ( Dup/+ ) carrier mice were generated on the C57BL/6N ( B6N ) genetic background ( see S1 File; Fig 1 ) . The segregation of the Del allele ( 30 . 7% ) was reduced considerably compared to the Dup allele ( 45 . 8%; Table 1 ) . To confirm these results and characterize the mice carrying both the deletion and the duplication , we crossed Del/+ with Dup/+ animals . We confirmed that the low transmission of the Del/+ allele was compensated in the Del/Dup carriers , thus demonstrating that lethality is associated with the deletion on the B6N background ( Table 1 ) . To evaluate the influence of the genetic background , we crossed B6N Del/+ and Dup/+ mice with sighted C3H/HeH ( C3B ) wild-type ( wt ) mice [24] and observed normal segregation of the Del ( 49 . 1% ) and Dup ( 49 . 2% ) alleles on the hybrid F1C57BL/6NxC3B ( F1B6C3B ) background ( Table 1 ) . To determine whether B6N Del/+ died in utero or postnatally , the fetuses of pregnant wt females sired by Del/+ males were collected at embryonic day 18 . 5 ( E18 . 5 ) , a few hours before natural delivery . Among the 69 fetuses extracted by caesarean section , we found a normal Mendelian ratio of 37 wt and 32 Del/+ foetuses . Overall , Del/+ fetuses required more time to oxygenize , and 3 mutant animals died within several minutes after a caesarean section . Furthermore , Del/+ were underweight in comparison with wt littermates , an observation which has been reported in human 16p11 . 2 del newborns [9] ( Del/+: 1 . 08 ± 0 . 02 g , n = 32; wt: 1 . 18 ± 0 . 01 g , n = 37; F ( 1 , 67 ) = 27 . 318 , P < 0 . 001 , S1A Fig ) . In contrast , weight was not affected by the Dup/+ mutation ( Dup/+: 1 . 22 ± 0 . 02 g , n = 27; wt: 1 . 22 ± 0 . 01 g , n = 24; F ( 1 , 49 ) = 0 . 047 , P = 0 . 829 , S1A Fig ) . Furthermore , the Del/+ allele on the C57BL/6N genetic background induced a developmental delay with weight defect that led to the death of approximately 55% of the Del/+ neonates between birth and weaning . We first studied the effects of the Sult1a1-Spn deletion and duplication on behaviour with independent cohorts of young adult mice bred on an inbred C57BL/6N genetic background . Complete description of behavioral data is reported in the supplementary information ( S1 and S2 Tables , S2 Fig ) . To confirm these data and to study the behavior of mice carrying both deletion and duplication of the 16p11 . 2 syntenic region , we generated and characterized a compound Del-Dup cohort with littermates of 4 genotypes: Del/+ , wt , Del/Dup , and Dup/+ ( S3 and S4 Tables ) . Our behavioural tests revealed an opposite effect of Sult1a1-Spn deletion and duplication on several traits , including locomotor activity ( Fig 2A and 2B ) , repetitive behaviors ( Fig 2C ) , and learning and memory performance ( Fig 2E ) . Del/+ and Dup/+ mice had normal circadian activity and feeding behaviors ( S3 Fig ) , but during the dark phase , Del/+ mice showed increased spontaneous locomotor activity and rearing behavior compared to wt ( Kruskal-Wallis analysis and Mann-Whitney U-test H ( 3 , 51 ) = 15 . 803 , P = 0 . 001; Del/+ vs wt: P = 0 . 039; Fig 2A ) . In contrast , Dup/+ mice manifested reduced locomotor activity during the light phase in comparison to wt animals ( one-way ANOVA and Tukey’s post-hoc test F ( 3 , 51 ) = 5 . 724 , P = 0 . 002; Dup/+ vs wt: P = 0 . 003 ) and reduced vertical ( rearing ) activity during the dark phase ( H ( 3 , 51 ) = 15 . 803 , P = 0 . 001 , Dup/+ vs wt: P = 0 . 015 ) and light ( H ( 3 , 51 ) = 11 . 119 , P = 0 . 011; Dup/+ vs wt: P = 0 . 008;Fig 2A ) . The water and pellet consumption were normal but the pellet loss was increased in Del/+ mice from separated cohort ( S4E Fig ) . When tested in the open field , only Dup/+ mice had significantly different locomotor activity scores from those of wt animals ( F ( 3 , 58 ) = 8 . 920 , P < 0 . 001; Del/+ vs wt: P = 0 . 089; Dup/+ vs wt: P = 0 . 013; Fig 2B ) . The time spent in the center of the arena , a measure of emotional behavior , was increased in Del/+ mice and decreased in Dup/+ mice ( H ( 3 , 58 ) = 23 . 758 , P < 0 . 001; Del/+ vs wt: P = 0 . 001; Dup/+ vs wt: P = 0 . 029; Fig 2B ) . The Del/+ mice also spent more time in the center of the arena in the last 5 min of the test compared to the first 5 min ( S4 Fig ) . Visual observations of the animals in their home cages revealed a range of abnormal repetitive behaviors in the Del/+ mice ( Fig 2C ) . We recorded a marked increase in rearing ( F ( 3 , 41 ) = 6 . 156 , P = 0 . 001; Del/+ vs wt: P = 0 . 010 ) and jumping ( H ( 3 , 41 ) = 14 . 394 , P = 0 . 002; Del/+ vs wt: P = 0 . 016 ) behaviors in these mice , whereas no behavioral abnormalities were detected in the Dup/+ animals . Del/Dup mice carrying one copy of both rearrangements behaved similarly to wt animals ( P > 0 . 05 for all measures; Fig 2A–2C ) , suggesting that abnormal activity phenotypes of Del/+ and Dup/+ mice were due to altered gene dosage from the Sult1a1-Spn region . In the social interaction test , we observed no difference in sniffing or following time between genotypes ( Fig 2D ) . Next , we evaluated our models in the novel object recognition test , the most common assay of the various facets of recognition memory in rodents . We first investigated whether Del/+ and Dup/+ mice could discriminate a novel object from a previously explored object after a short retention delay of 30 min ( Fig 2E ) . During the acquisition session , mice of all genotypes spent an equal amount of time exploring the sample object ( S4 Table ) . In the subsequent choice session , Del/+ mice displayed a significant memory impairment compared to wt , whereas Dup/+ mice tended to display a memory improvement ( F ( 3 , 49 ) = 8 . 080 , P < 0 . 001; Del/+ vs wt: P = 0 . 004 ) . To explore this cognitive phenotype further , we extended the retention delay to 3 h ( Fig 2E ) . The Del/+ mice again displayed a poor recognition performance , whereas Dup/+ mice showed a memory improvement compared to wt mice ( H ( 3 , 51 ) = 16 . 014 , P = 0 . 001; Del/+ vs wt: P = 0 . 043 , Dup/+ vs wt: P = 0 . 021 , Del/Dup vs wt: P = 0 . 048 ) . Interestingly , Del/Dup also displayed improved performance in this test , similar to Dup/+ mice ( P < 0 . 05 ) . We employed a series of assays to investigate potential alterations in sensory and motor functions . No significant differences in motor coordination and motor learning were detected between mutant and wt mice in the rotarod test ( Fig 2F ) . In the notched bar test , the Del/+ mice made a significantly higher number of errors compared to wt , whereas Dup/+ and Del/Dup mice performed normally ( H ( 3 , 60 ) = 14 . 044 , P = 0 . 003; Del/+ vs wt: P = 0 . 007; Fig 2F ) . In the grip test , Del/+ and Dup/+ mice showed stronger and weaker grip strength respectively , compared to wt ( F ( 3 , 60 ) = 23 . 598 , P < 0 . 001; Del/+ vs wt: P = 0 . 002; Dup/+ vs wt: P < 0 . 001; S2 Table ) . Nevertheless , no difference between wt and mutant mice was observed in measuring in situ tibialis anterior ( TA ) isometric contraction in response to nerve stimulation ( S55C–S5E Fig ) . In addition , we observed normal fiber size and normal succinate dehydrogenase ( SDH ) staining in TA muscle ( S5F Fig ) . Finally , we evaluated the hearing discrimination of animals in the ABR test and detected no abnormal phenotypes in the mutant mice ( S6A and S6B Fig ) . Humans with 16p11 CNVs present autistic traits . Although hyperactive and repetitive behaviors were found during the study on the C57BL/6N genetic background , we asked whether the observed behavioral phenotypes were sensitive to the genetic background . Thus , we repeated our behavioral analyses on mutant F1 animals with a C57BL/6N×C3B genetic background ( Fig 3 , S6 and S7 Tables ) . Sep2arate cohorts of Del/+ , Dup/+ and their corresponding wt littermates were tested . As expected , Del/+ and Dup/+ mice on a hybrid genetic background showed a range of behavioral abnormalities compared to their wt counterparts . Opposite phenotypes were noted for rearing in the circadian activity test ( Fig 3A ) , climbing behavior ( Fig 3C ) , and cognitive capacity ( Fig 3E ) . Del/+ mice demonstrated recognition memory deficits in the novel object recognition task with a 3 h retention delay as well as lack of object-place recognition memory ( Fig 3E , S7 Table ) . Recognition memory improvements were also confirmed for Dup/+ mice ( Fig 3E ) . Compared to wt mice , social interaction was reduced in both Del/+ ( F ( 1 , 9 ) = 6 . 799 , P = 0 . 028 ) and Dup/+ ( F ( 1 , 10 ) = 5 . 594 , P = 0 . 040 ) mutant mice ( Fig 3D ) . In the three-chambered social procedure , mutant and wt mice displayed similar exploration time of the first stranger ( S7 Table ) . However , in comparison to wt , Del/+ mice showed deficits of social preference for the second stranger ( F ( 1 , 14 ) = 12 . 849 , P = 0 . 003 ) . To validate these social phenotypes , we verified the olfactory discrimination of Del/+ and Dup/+ mice after the exposure to social and non-social odors on the C57BL/6N genetic background . No genotype effect was found for these parameters ( S6C and S6D Fig ) . Our behavioral studies revealed strong opposite phenotypes of recognition memory in both inbred and hybrid genetic backgrounds . To investigate the potential electrophysiological underpinnings of the observed phenotypes , we studied synaptic transmission and its plasticity in synapses between Schaffer collaterals and apical dendrites of CA1 pyramidal neurons in the hippocampus , the principal region implicated in spatial memory [25–27] . This experiment was performed on hippocampal slices of wt , Del/+ , Dup/+ and Del/Dup mice on an inbred C57BL/6N genetic background . The analysis of input-output relationships ( Fig 4A ) showed that there was no significant interaction between genotype and stimulus ( repeated measures ANOVA F ( 27 , 468 ) = 0 . 98; P = 0 . 498 ) . Although we observed nominally decreased slopes of field excitatory postsynaptic potentials ( fEPSPs ) in Del/+ and Dup/+ mutants , especially in response to higher stimulus strengths ( Fig 4A ) , a separate two-way nested ANOVA performed on maximum fEPSP values similarly failed to reveal a significant effect of genotype ( F ( 3 , 35 ) = 1 . 229 , P = 0 . 314 ) . Likewise , genotype did not influence significantly the values of paired-pulse facilitation , a model of short-term synaptic plasticity ( F ( 3 , 35 ) = 0 . 0298; P = 0 . 993; Fig 4B ) . To investigate long-term synaptic plasticity , we induced long-term potentiation ( LTP ) of fEPSPs in Schaffer collaterals-CA1 synapses by theta-burst stimulation ( Fig 4C ) . A two-way nested ANOVA demonstrated a significant effect of genotype on LTP values ( F ( 3 , 35 ) = 3 . 43; P = 0 . 027 ) . Post hoc Dunnett’s Multiple Comparison test performed on individual slice values demonstrated that LTP was significantly smaller ( Q = 2 . 441; P < 0 . 05 ) in slices from Dup/+ mice ( 140 ± 9% , slice number n = 10 , animal number N = 4 ) compared to wt slices ( 177 ± 7% , n = 18; N = 8 ) , while levels of LTP in Del/+ ( 160 ± 12% , n = 10; N = 4 ) and Del/Dup slices ( 182 ± 11% , n = 19; N = 6 ) did not differ significantly from wt levels . We concluded that long-term synaptic plasticity is sensitive to the duplication of the Sult1a1-Spn region , whereas neither short- nor long-term plasticity are affected by deletion or deletion/duplication rearrangements . We also observed a trend to nominally lower fEPSP slopes in all three mutants compared to wt values . The latter observation may require further experiments with larger cohorts . Individuals with 16p11 CNVs present with mirror BMI phenotypes [28] . To determine whether our mouse models recapitulated this phenomenon , body weight of the adult animals was recorded once a week during behavioral analyses . In the first separated cohorts , C57BL/6N Del/+ mice were significantly underweight ( two-way ANOVA genotype effect F ( 1 , 120 ) = 88 . 115 , P < 0 . 001; S1B Fig ) , whereas C57BL/6N Dup/+ mice were significantly overweight compared to wt littermates ( two-way ANOVA genotype effect F ( 1 , 99 ) = 8 . 391 , P = 0 . 009; S1B Fig ) . Similar results were observed when we generated the two genotypes at the same time in the Del-Dup cohort ( Fig 5A ) . In comparison with wt and Del/Dup mice , Del/+ littermates were underweight , whereas Dup/+ mice showed trends for overweight ( two-way ANOVA genotype effect F ( 3 , 780 ) = 9 . 954; P < 0 . 001; Del/+ vs wt: P = 0 . 002; Dup/+ vs wt: P = 0 . 211 ) . Feeding behaviors evaluated during the circadian activity test were similar in mutant and wt animals ( S3C and S3D Fig ) . In addition , no correlation was observed between the activity and body mass for the different genotypes . We evaluated body fat percentage of live animals by qNMR ( quantitative nuclear magnetic resonance ) and found a decrease in adiposity in the Del/+ mice ( H ( 3 , 60 ) = 22 . 770 , P < 0 . 001; Del/+ vs wt: P < 0 . 001; Fig 5B ) ; these data were confirmed in sacrificed animals which displayed a lack of visceral fat pads ( Fig 5D ) . Body size was reduced in the Del/+ mice ( F ( 3 , 35 ) = 6 . 834 , P < 0 . 001; Del/+ vs wt: P = 0 . 027; Fig 5C ) , which correlated with the animal weight ( Pearson correlation coefficient , ρ = 0 . 893 , P = 0 . 001 ) . No weight , body fat , or size modifications were noted in Del/Dup mice . On an F1 ( C57BL/6N × C3B ) hybrid background , Del/+ animals were still underweight in comparison to wt littermates ( two-way ANOVA genotype effect F ( 1 , 456 ) = 16 . 620 , P < 0 . 001; S1D Fig ) , whereas no body weight phenotype was observed in the Dup/+ animals . Adiposities were similar between mutant and control mice ( S1E Fig ) . A body size reduction was confirmed in the Del/+ mice ( F ( 1 , 33 ) = 17 . 137 , P < 0 . 001; S1F Fig ) , but no correlation between body size and body weight was observed . To characterize further weight and adiposity changes , we performed a metabolic analysis of the C57BL/6N Del/+ and Dup/+ separate cohorts challenged with a high-fat diet between 5 and 15 weeks of age ( Fig 5E–5I , S8 Table ) . In comparison to wt animals , the Del/+ mice were underweight ( two-way ANOVA genotype effect F ( 1 , 102 ) = 48 . 671 , P < 0 . 001; Fig 5E ) , whereas the Dup/+ mice were overweight ( two-way ANOVA genotype effect F ( 1 , 112 ) = 8 . 674 , P = 0 . 010 ) . Body size was decreased in Del/+ mice ( F ( 1 , 17 ) = 51 . 771 , P < 0 . 001; S8 Table ) and increased in Dup/+ mice ( F ( 1 , 16 ) = 29 . 229 , P < 0 . 001 ) . Body composition analysis revealed a decrease in fat percentage in the Del/+ mice ( F ( 1 , 16 ) = 32 . 138 , P < 0 . 001; Fig 5F ) and an increase in the Dup/+ mice ( F ( 1 , 16 ) = 8 . 621 , P = 0 . 010 ) . No correlation was found between weight and adiposity in any of the genotypes . The energy expenditure ( EE ) of the animals was analysed by indirect calorimetric measure during the dark and light phases ( Fig 5G ) . The Del/+ mice showed a higher EE during the dark phase ( F ( 1 , 16 ) = 5 . 313 , P = 0 . 035 ) , whereas the Dup/+ mice showed a lower EE during the dark ( F ( 1 , 15 ) = 12 . 994 , P = 0 . 003 ) and light phases ( F ( 1 , 15 ) = 13 . 997 , P = 0 . 002 ) . These results are consistent with endogenous activity phenotypes of mutant mice observed in the circadian activity test . The intraperitoneal glucose-tolerance test ( IPGTT ) indicated a faster glucose clearance in the Del/+ mice ( F ( 1 , 15 ) = 23 . 396 , P < 0 . 001; Fig 5H ) and hyperglycemia in the Dup/+ animals ( F ( 1 , 16 ) = 8 . 220 , P = 0 . 011 ) . Consistent with the body fat composition data , endocrinology analysis ( Fig 5I ) revealed decreased blood levels of leptin ( F ( 1 , 16 ) = 7 . 059 , P = 0 . 017 ) and adiponectin ( F ( 1 , 17 ) = 6 . 790 , P = 0 . 018 ) in the Del/+ mice and an increase of leptin blood levels in the Dup/+ mice ( F ( 1 , 15 ) = 5 . 350 , P = 0 . 035 ) . A nominal increase in insulin blood levels , which failed to reach statistical significance , was also noticed in the Dup/+ mice . Blood chemistry analysis did not reveal gross hematology changes except for a decrease of free fatty acid levels in the Del/+ mice ( F ( 1 , 17 ) = 5 . 333 , P = 0 . 034; S8 Table ) . Finally , we repeated the high-fat diet protocol in the Del/+ animals on the hybrid C57BL/6N×C3B background ( S7 Fig , S9 Table ) . We observed that F1 Del/+ mice showed similar weight , fat , energy expenditure , glucose clearance and endocrinology phenotypes in comparison to the Del/+ mice on the C57BL/6N background ( S7A–S7E Fig ) . C57BL/6N×C3B Del/+ animals also presented with a basal hyperglycemia , higher blood levels of calcium and lower blood levels of total cholesterol and glucose ( S7F Fig ) . Bomb calorimetric analysis of mouse feces revealed a similar level of food energy usage between the wt and the Del/+ mice ( S7G Fig ) . In addition to neuropsychiatric and BMI features , patients with 16p11 . 2 CNVs also present with brain volume changes and craniofacial malformations [19] . We studied the influence of 16p11 . 2-homologous CNVs on mouse craniofacial structure by analysing computed tomography ( CT ) cranial scans of animal heads combined with ulterior reconstruction of 3D skull images using 39 cranial landmarks ( S8A Fig ) . Separate Del/+ and Dup/+ cohorts of females were used for the Euclidian distance matrix analysis [29] . A global skull effect was observed in the Del/+ females with a significantly reduced skull size ( T = 0 . 009; S8B Fig ) and altered skull shape ( Z = 0 . 046; S8C Fig ) . Most affected regions corresponded to the premaxilla , maxilla and zygomatic human bones , indicating an important alteration of Del/+ facial features . Although the skull size of the Dup/+ animals was unchanged ( T = 0 . 297; S8D Fig ) , their skull shape was altered significantly ( Z = 0 . 037; S8E Fig ) . To identify altered pathways in the 16p11 . 2 mouse models , we performed transcriptomic analysis of three brain regions ( hippocampus , striatum , and cerebellum ) and of one peripheral tissue ( liver ) in Del/+ , Dup/+ and wt animals ( data accessible at NCBI GEO database[30] , accession GSE66468 ) . We assessed expression of the genes encompassed by the Sult1a1-Spn region boundaries and found that the majority of them showed mRNA levels proportional to the CNV copy number ( 5 ) . The expression of the genes in the Sult1a1-Spn region was largely affected by the CNVs in all three brain regions and the liver , with the following exceptions: Kif22 , Qprt , Spn , and Zg16 appeared to be under dosage compensation in all tissues studied , whereas Aldoa , Coro1a , Fam57b , Prrt2 , and Tbx6 expression levels were only compensated in the liver ( Fig 6A–6D ) . Furthermore , principal component analysis suggested that the effect of the CNV on the expression of Sult1a1-Spn interval genes was stronger in the Del/+ animals ( Fig 6E ) . At a 5% false discovery rate ( FDR ) , between 7 and 14 probe sets mapping elsewhere on the genome were associated with the CNV dosage ( 9–32 at 10% FDR ) . We then performed gene set enrichment analysis ( GSEA ) on the genes ranked by the results of the differential expression analysis to identify the sets of genes that were positively or negatively associated with the Sult1a1-Spn copy number . Our GSEA analysis considers the pre-ranked gene list according to the LIMMA results based on the dosage model . We used the MSigDB C2 gene set database ( c2 . all . v4 . 0 . symbols . gmt ) , which collects curated gene sets from online pathways , and publications in PubMed . The data with an FDR below 25% are reported in S10 Table . In the hippocampus , two gene sets were dosage-dependently upregulated ( upregulated in the Dup/+ mice and downregulated in the Del/+ mice ) including genes from midbrain markers [31] and Akt1 signalling via mTOR [32] . In contrast , six gene sets were downregulated in a dose-dependent manner ( upregulated in the Del/+ mice and downregulated in the Dup/+ mice ) including the targets of Creb1 ( a transcription factor induced in brain reward pathways after chronic exposure to recreational drugs [33] ) as well as the genes involved in resistance to Ifna2 [34] , in the response to TSA and butyrate [35 , 36] , and in the PID P38alpha/beta pathway [37] . A total of 22 gene sets were affected in the cerebellum including gene sets that were downregulated exclusively in a dosage-dependent fashion . In the striatum , 240 gene sets were dose-dependently upregulated . Six gene sets were dose-dependently downregulated , including genes involved in cocaine reward [33] , involved in the regulation of inhibitory GABAergic synapse through NpaS4 [38] , regulated by exercise [39] , or regulated during the EGF response [40] . In the liver , a total of 163 gene sets were dose-dependently upregulated , including genes involved in adipogenesis [41 , 42] and targets of PPARg [43] . Six gene sets were dose-dependently downregulated , including genes involved in the cholesterol biosynthesis and genes involved in lipid and carbohydrate metabolism regulated by HNF1a [44] or by Klf10 [45] . Fourteen pathways were found to be misregulated in different tissue types , including 12 pathways deregulated in the striatum and the liver and two pathways deregulated in common between the hippocampus and the striatum ( Fig 6F ) .
To achieve improved phenotypic resolution of 16p11 . 2 BP4-BP5 rearrangement syndromes , we engineered and characterized novel mouse models carrying a deletion and/or duplication of the Sult1a1-Spn region and evaluated the genetic background effect on phenotype . Compared with mouse models described previously [21 , 22] and as shown in Fig 1 , our mouse models carry accurate rearrangements corresponding to the 16p11 . 2 BP4-BP5 syntenic region and do not show hearing deficits . We carried out a comprehensive battery of tests with these mouse models and unravelled the impact of genetic rearrangements on several functions affected in human patients , including cognition , repetitive behaviors , synaptic function , skull volume , and metabolism without a major impact of the genetic background . Compared to previous studies , we addressed new behavioral and metabolic paradigms , and highlighted the occurrence of alterations with opposite directions associated with the deletion or the duplication of the 16p11 . 2 syntenic region . Most of them were restored to normal levels in the pseudo-disomic mice carrying the deletion and the duplication ( Del/Dup , Fig 2 ) . The Del/+ and Dup/+ animals were first generated separately on an inbred C57BL/6N genetic background . Independent cohorts of animals were subjected to two complementary behavioral pipelines to evaluate a range of parameters that can be related to human neuropsychiatric disorders . The Del/+ mice displayed hyperactivity and recognition memory deficits , whereas the Dup/+ mice displayed reciprocal phenotypes . Interestingly , Del/+ mice also showed stereotypic behaviors ( jumping and excessive climbing ) commonly found in mouse models for autism [46] . No phenotypes related to schizophrenia , depression , or epilepsy were detected in our mouse models; mutant and control littermates showed similar capacities in the prepulse inhibition , sucrose preference , and pentylenetetrazol sensitivity tests ( S2 and S6 Tables ) . We generated an "all-mutant" cohort by crossing Del/+ with Dup/+ animals to compare the Del/+ and Dup/+ mutant mice with a single group of wt animals and also to characterize the Del/Dup animals carrying two copies of the Sult1a1-Spn region on a single chromosome . Using this cohort , we confirmed the phenotypes for activity and recognition memory observed previously in the two groups of mutant mice bred separately ( Fig 2 , S4 and S5 Tables ) . Motor coordination and grip strength phenotypes were observed in both Del/+ and Dup/+ animals . During the evaluation of in situ TA isometric contraction in response to nerve stimulation , similar muscle force was measured between wt and mutant mice with ( S8C–S8E Fig ) . In addition , we observed normal fiber size and normal succinate dehydrogenase ( SDH ) activities in TA muscle ( S8F Fig ) suggesting that these grip strength phenotypes in both genotypes were not due to severe muscular dysfunction . The ambiguity of Del/+ results with improvement trends in the rotarod test and obvious deficits in the notched bar test suggested an important bias related to activity alterations of mutant mice . We were particularly interested in altered learning and memory observed in mutant mice , which can be linked to intellectual disability and memory impairment found in humans carrying 16p11 . 2 BP4-BP5 CNVs [7 , 9] . Thus , we performed the novel object recognition task using two retention delays: 30 min and 3 h . With the shorter delay , the Del/+ mice showed deficits in recognition memory . With the longer delay , similar to the results in two separate mutant cohorts , the Del/+ mice and the Dup/+ mice displayed deficits and improvements , respectively , of novel object discrimination . Notably , the Del/Dup animals had similar phenotypes as the Dup/+ animals and showed recognition memory improvements in comparison with wt mice . This result suggested that the Dup/+ memory improvement phenotype might be linked to DNA structure changes associated with the duplication of the Sult1a1-Spn region . Except for this assay , the Del/Dup compound mutants did not display changes in any of the other phenotypes compared to the controls . Among the series of synaptic transmission parameters measured in acutely prepared hippocampal slices , LTP was the only parameter to be statistically lower in Dup/+ mice ( Fig 4 ) . We did not observe trends for opposite phenotypes in slices from Dup/+ and Del/+ . Moreover , LTP in Del/+ mice was also nominally lower than in wt animals . Similarly , amplitudes of fEPSPs obtained in response to strong stimuli in slices from both Del/+ and Dup/+ mice were nominally lower than in slices from wt animals . The lack of statistically significant alterations in the input-output relationship and paired-pulse facilitation in the three lines of mutant mice suggested that despite the observation that several genes were affected by the 16p11 . 2 copy-number variations , basal synaptic transmission and short-term synaptic plasticity in a conventional hippocampal synapse were resistant to these changes . Of note , a similar lack of these phenotypes was noted recently in 16p11 . 2 Del/+ mice by Tian et al . ( 2015 ) [47] . Our electrophysiological data were obtained on a modest number of animals and larger cohorts will be required to confirm our observations , particularly with respect to enhanced hippocampal LTP phenotype in Dup/+ mice . At the same time , it is conceivable that behavioral and cognitive alterations in mice with 16p11 . 2 CNVs may be accounted for by changes in brain circuits other than the hippocampus . For instance , changes in dopaminergic signaling and striatal electrophysiology may underlie the lack of habituation observed in reaction to novel objects [22] . Differing from previous studies , we examined the influence of genetic background on the observed phenotypes . In addition to the characterization of inbred mouse models on a C57BL6/N genetic background , we generated and characterized Del/+ and Dup/+ animals on a F1 C57BL/6N×C3B hybrid genetic background . The presence of one copy of C3B alleles did not alter the main phenotypes observed in the 16p11 . 2 CNV models generated in the study ( Figs 2 and 3 ) . The nocturnal and diurnal hyperactivity , object recognition defect and repetitive behaviors of Del/+ mouse models have been observed in both the C57BL/6N and C57BL/6N ×C3B genetic backgrounds and across different laboratories ( Fig 7 ) . Those phenotypes are highly reproducible , as they have been observed at least in two independent research centers . We propose that they represent robust and consistent core phenotypes of the 16p11 . 2 deletion mouse models . The higher and lower levels of climbing activity in the Del/+ and Dup/+ mice respectively were found on both inbred and hybrid genetic backgrounds . Horev et al . found similar climbing phenotypes in their Del/+ and Dup/+ models that were maintained on a C57BL/6N×129Sv hybrid background [21] ( Fig 7 ) . Notably , 24% of children with the 16p11 . 2 BP4-BP5 deletion had a diagnosis of ASD with some level of restricted and repetitive behavior reported [7]; such phenotypes were also observed in the Del/+ mouse models . Some individuals carrying 16p11 . 2 CNVs present with attention-deficit/hyperactivity disorder [7 , 11] and bipolar disorders [48 , 49] , which support a relation between 16p11 . 2 copy number and the regulation of activity and attention . The use of a C57BL/6N×C3B hybrid genetic background increased considerably memory capacities of wt animals . Similar to what we observed for the inbred genetic background , we found recognition memory deficits for Del/+ mice and improvements for Dup/+ mice on the hybrid background in the new object recognition task with 3 hours of retention delay . Since the use of the hybrid background reduces the activity phenotypes of Del/+ and Dup/+ mice , these results indicated that the recognition memory phenotypes seen on an inbred background were not due to an activity alteration , but rather , were the consequence of a diminished interest for the objects . In addition , Dup/+ mice also displayed enhanced working memory performance in the Y-maze alternation task ( S7 Table ) , suggesting that duplication of the Sult1a1-Spn region impacts primarily short-term memory processes . Portmann et al . performed the same recognition memory experiment with a retention delay of 1 hour . The authors also found an important deficit in their Del/+ model maintained on a N5-7 C57BL/6N×129P2 hybrid genetic background [22] ( Fig 7 ) . Taken together , these experiments performed by independent research groups confirm the robustness of the activity and recognition memory phenotypes of the 16p11 . 2 CNV mouse models . Moreover , a recent study looking at the neuropsychological profile of 16p11 . 2 BP4-BP5 CNVs carriers shows that duplication carriers outperform intrafamilial controls on a verbal memory task , with the same IQ level [50] . This observation correlates with the higher performance of the mouse duplication carrier in the object recognition paradigm . Comparison of the phenotypes induced by the 16p11 . 2 deletion on inbred C57BL/6N and hybrid C57BL/6N×C3B genetic backgrounds revealed striking differences . The embryonic lethality observed in the transmission of the deleted chromosome was completely abolished in the F1 cross ( Table 1 ) . The body weight of deletion carrier neonates at birth was normal in the hybrid background ( S1C Fig ) but the retarded growth was still observed later with a reduced body weight of adult Del/+ animals observed in both backgrounds ( S1D Fig ) . The loss of the neonatal lethality in the C57BL/6N×C3B Del/+ mice has already been observed in other studies [51] . We compared the polymorphisms in the Sult1a1-Spn region between several mouse strains but we did not find significant differences , suggesting a possible contribution of a recessive allele located in trans outside of the 16p11 . 2 homologous region . An attenuation of the horizontal exploration activity was also observed in the F1 genetic background with the loss of phenotypes in the open field observed in animals with the inbred background ( S4A and S4B Fig ) . In the C57BL/6N×C3B hybrid genetic background , the level of social behavior of wt animals was significantly increased , which allowed us to observe an obvious diminution of social interaction in both Del/+ and the Dup/+ mice ( S4D Fig ) . As such , the C57BL/6N genetic background is less permissive to allow the detection of social interaction phenotypes . To confirm that the mutant animals did not present any olfactory deficits , we performed social odor and non-social odor tests and we found no difference between wt and mutant animals ( S6C and S6D Fig ) . Social interaction deficits found in Del/+ and Dup/+ can be considered as a read-out for the autistic traits found in humans carrying deletions or duplications of the 16p11 . 2 BP4-BP5 region [3 , 5] . The reduction in social interaction observed for mutant mice on the C57BL/6N×C3B is the first phenotype following the same direction for both Del/+ and Dup/+ animals . The presence of this phenotype on the C57BL/6N×C3B hybrid background but not on the C57BL/6N inbred background nor on the C57BL/6N×129Sv or C57BL/6N×P2 hybrid genetic backgrounds reported by Horev et al . and Portmann et al . respectively [21 , 22] , suggests that the influence of the genetic background is contributory to the manifestation of social behavior deficits induced by 16p11 . 2 CNV . Therefore , the genetic background has to be taken into account for the design of the in vivo modeling of other neuropsychiatric disorders [52 , 53] . Because individuals carrying rearrangements at the 16p11 . 2 BP4-BP5 locus display extreme BMI alterations ( lean or obese ) , we evaluated the influence of Sult1a1-Spn copy number on the weight and metabolism in our mutant animals . On an inbred genetic background , we observed a significant reduction of body weight , size and adipogenesis in the Del/+ animals under a standard diet , which is consistent with previous studies [21 , 22] . During the high-fat diet challenge , the Del/+ metabolic phenotypes were similar to the normal diet , and we also observed opposite phenotypes in the Dup/+ mice , including an increase in body weight and adipogenesis ( Fig 5 ) . Moreover , analysis of glucose clearance and endocrinological parameters revealed alterations of metabolism in the Del/+ and Dup/+ mice . Interestingly , on a C57BL/6N×C3B hybrid background , we observed similar phenotypes under standard ( S7 Fig ) and high-fat diets ( S9 Table ) . The opposite effects of Del/+ and Dup/+ animals on body weight and adipogenesis are independent of the diet and of the genetic background , which suggests that rearrangements of the 16p11 . 2 homologous region induce robust metabolic alterations . Strikingly , our data indicated that weight and adiposity phenotypes are opposite between humans and mice carrying the 16p11 . 2 BP4-BP5 rearrangements . In humans , the 16p11 . 2 BP4-BP5 deletion is associated with obesity and hyperphagia [17] , whereas reciprocal duplication is associated with lower weight and restrictive eating behavior [1] . More recently , data collected from young individuals with 16p11 . 2 deletion report that satiety response is altered before the onset of obesity [54] , and should be a strong contributor to the energy imbalance in 16p11 . 2 CNVs in human . In comparison with their wt littermates , Del/+ and Dup/+ mice displayed normal food and water consumption ( S2 Fig ) . No deregulation of the Sh2b1 gene , a candidate located upstream of the BP4-BP5 region , was found in the liver , the cerebellum , the striatum or the hippocampus of the Del/+ and Dup/+ animals . In addition , we did not find any significant expression level changes for the genes located around the Sult1a1-Spn rearranged region . Similarly , both Del/+ and Dup/+ mouse models show skull shape alterations , however , microcephaly is observed in the Del/+ mice ( S8 Fig ) , in contrast to humans carrying the 16p11 . 2 deletion , who display macrocephaly . Looking at the 16p11 . 2 regions in the human and the mouse genomes , several changes occurred during primate evolution with the duplication of the low copy repeat region where the breakpoints take place and the general organization of the neighboring regions which differs from mouse ( https://genome . ucsc . edu/ ) . Such local changes may influence the outcome of the deletion and the duplication of the 16p11 . 2 region in the two organisms . The opposite phenotypes described in this study suggest the implication of dosage-sensitive gene ( s ) within and/or outside the BP4-BP5 region that could influence mouse activity , recognition memory and metabolism . To test this possibility , we analysed transcriptome profiles of Del/+ , Dup/+ and wt animals in three brain regions and the liver . Overall , we found that the majority of genes within the Sult1a1-Spn region were sensitive to gene dosage . Rearrangements of Sult1a1-Spn region had a low impact on the whole genome transcriptome in the hippocampus , suggesting that dosage-sensitive genes leading to opposite phenotypes may be located in the 16p11 . 2 syntenic region . Principal component analysis suggested further that the effect of the deletion on the expression of the genes in the Sult1a1-Spn interval was stronger , as reported previously [21] . This result may indicate why behavioral , anatomic , and metabolic phenotypes are stronger in Del/+ animals compared to the opposite phenotypes of the Dup/+ mice . Our GSEA analysis revealed that the genes found “upregulated in the nucleus accumbens after cocaine treatment” were also upregulated in the striatum of Del/+ mice , whereas they were downregulated in the striatum of Dup/+ mice . These results thus indicated an alteration in the reward system , which could contribute to the behavioral phenotypes observed in our study . A large number of pathways were found altered in the striatum compared to other brain regions , suggesting a major impact of the 16p11 . 2 genetic dosage on the striatum that could explain repetitive behaviors and cognition defects observed in the 16p11 . 2 mutant mice . In addition , the analysis pinpointed a misregulation of genes involved in adipogenesis and obesity [55 , 56] and contributing to the Reactome cholesterol biosynthesis in the liver of the Del/+ and Dup/+ mice which could explain the weight phenotypes observed in the mutant animals . We also found that the genes “down-regulated in liver tissue upon knockout of Hnf1alpha” were upregulated in the liver of Del/+ mice , whereas they were downregulated in the liver of Dup/+ mice . HNF1A is known to regulate numerous hepatic genes and heterozygous HNF1A mutations cause pancreatic-islet beta-cell dysfunction and monogenic diabetes [57 , 58] . Hnf1a deficiency in mice leads to highly-specific changes in the expression of genes involved in key functions of both islets and liver [44 , 59] . In humans HNF1A deficiency is the genetic cause Maturity-Onset Diabetes of the Young type 3 ( MODY3 ) . Such changes in the Hnf1a-dependent pathway most likely contribute to the metabolic phenotypes found in Del/+ and Dup/+ animals . Interestingly , differences exist in MODY3 humans versus mouse models; in the mouse , ablation of two Hnf1a alleles are required to induce a phenotype whereas only one copy of HNF1A is affected in human cases . These differences might explain the differences observed in the obesity phenotypes found in mouse compared to human . Together , our results demonstrated that pathways mediating intellectual disability in carriers of the 16p11 . 2 deletion or duplication might be linked directly to altered expression of genes encompassed within the 16p11 . 2 BP4-BP5 region . This region contains several candidate genes for intellectual disability and neuropsychiatric disorders including DOC2A , KCTD13 , SEZ6L2 , and TAOK2 . DOC2A ( Double C2-like domains Alpha ) encodes a synaptic vesicle-associated Ca2+ -binding protein . Mouse studies revealed that Doc2α interacts with Munc13 , and is implicated in Ca2+-dependent neurotransmitter release [60] . Doc2α mutant mice show impairment in long-term potentiation and passive avoidance tasks , pinpointing a potential contribution of Doc2α to memory formation [61] . KCTD13 ( Potassium Channel Tetramerization Domain containing 13 ) encodes PDIP1 ( polymerase delta-interacting protein 1 ) which interacts with the proliferating cell nuclear antigen and therefore might have a role in the regulation of cell cycle during neurogenesis [62] . A study in zebrafish models revealed that overexpression of the human KCTD13 transcript in embryos induces microcephaly , whereas knockdown of endogenous kctd13 by morpholino antisense oligonucleotides leads to macrocephaly , recapitulating the mirrored phenotype ( i . e . head circumference ) seen in 16p11 . 2 humans [63] . SEZ6L2 ( Seizure related 6 homolog ( mouse ) -Like 2 ) encodes a cell surface protein that has a strong homology with SRPX2 ( Sushi-repeat-containing protein , X-linked ) , mutations in which cause epilepsy and language disorders [64] . In addition , association between a SEZ6L2 coding variant and ASD was suggested but is still controversial [65] . TAOK2 ( Thousand-And-One-amino acid Kinase 2 ) encodes a serine/threonine kinase that activates mitogen-activated protein kinase ( MAPK ) pathways to regulate gene transcription . TAOK2 interacts with semaphorin 3A receptor neuropilin 1 , which regulates basal dendrite arborization . Recently , TAOK2 has been shown to play a role in basal dendrite formation in mouse cortical neurons [66] . Taken together these data suggest a possible contribution of these genes in the recognition memory and activity phenotypes observed in Del/+ and Dup/+ mouse models . To test this possibility , Dup/+ mice crossed with each of the heterozygous knockout mouse models for each of these candidate genes would help to determine whether the restoration of a normal copy number rescues the recognition memory and hypoactivity of Dup/+ mouse model . Although elegant and definitive , this approach is not time and cost-efficient . While most of the defects were stable and robust in our conditions or in other studies [21 , 22] , the 16p11 . 2 CNV affects the transcriptional profile of multiple pathways ( a total of 246 and 169 misregulated in the striatum and in the liver , respectively; S10 Table ) supporting the hypothesis of a broader genome wide transcriptional effect of the 16p11 . 2 gene dosage imbalance . Our study highlights the benefit of combining inbred and hybrid genetic background studies in mouse models of human syndromes , particularly when cases show symptoms of incomplete penetrance . In fact , the C57BL/6N background potentiates the activity and weight phenotypes of Del/+ and Dup/+ mice whereas the mixed C57BL/6N and C3B backgrounds reveals recognition memory phenotypes and social interaction deficits . We showed for the first time social interaction deficits of Del/+ and Dup/+ mice on a C57BL/6N×C3B hybrid background , which is notable due to the observation that 16p11 . 2 BP4-BP5 rearrangements are one the most common CNVs found in individuals with autism [3–7] . These novel mouse models for the 16p11 . 2 CNV will thus facilitate the identification of candidate genes responsible for memory and activity symptoms observed in 16p11 . 2 CNV carriers and could eventually be utilized to test small molecules of therapeutic benefit .
16p11 . 2Yah mouse models were generated through Cre-LoxP in vivo recombination using a mouse line carrying two loxP sites inserted upstream of Sult1a1 and downstream from Spn genes in a trans configuration . Deletion of the Sult1a1–Spn region , Del ( 7Sult1a1-Spn ) 6Yah , referred as Del/+ , was identified by PCR using primers Fwd1 ( 5’-CCTGTGTGTATTCTCAGCCTCAGGATG-3’ ) and Rev2 ( 5’-GGACACACAGGAGAGCTATCCAGGTC-3’ ) . Duplication of the same region , Dp ( 7Sult1a1-Spn ) 7Yah ( here Dup/+ ) was identified using primers Fwd2 ( 5’-ACTGCAGCCCGTCACCTAACTTCTT-3’ ) and Rev1 ( 5’-GGACACACAGGAGAGCTATCCAGGTC-3’ ) . The wt allele was identified using Fwd1 and Rev1 primers . The PCR reactions gave deletion , duplication and wt products of 500 bp , 461 bp and 330 bp , respectively ( Fig 1C ) . All mice were genotyped by PCR using the following program: 95°C /5 min; 35× ( 95°C/30 s , 65°C/30 s , 70°C/1 min ) , 70°C/5 min . All mouse experimental procedures were approved by the local ethics committee , Com’Eth ( n°17 ) under the accreditation number 2012–069 . YH was the principal investigator of this study ( accreditation 67–369 ) . The mouse lines are available through the INFRAFRONTIER/European Mouse Mutant Archive ( EM:06133 and EM:06134 ) . Viability tests were performed to study causes of Del/+ lethality and to evaluate the weight of animals at birth . Fetuses were collected at embryonic day 18 . 5 ( E18 . 5 ) to monitor their capacity to survive at birth . The fetuses were weighed , placed on a warm plate at 37°C and rolled gently to stimulate them to breathe . At 30 min after extraction , the numbers of breathing animals versus cyanotic and lethargic animals were counted . Tail samples were collected for genotyping . We isolated 37 wt and 32 Del/+ fetuses on the C57BL/6N background from 7 pregnant wt females mated to Del/+ males . 24 wt and 30 Del/+ fetuses on the C57BL/6NxC3B background were isolated from 7 pregnant wt C3B females mated to Del/+ C57BL/6N males . Finally , we also studied 24 wt and 27 Dup/+ fetuses isolated from 6 pregnant wt females crossed to Dup/+ males . An independent Del-Dup cohort including males and females was used for electrophysiology analysis ( comprising 4 Del/+ , 8 wt , 6 Del/Dup , and 4 Dup/+ animals ) . All animals were 8–9 months old on the day of dissection . Investigators were blind to the genotype of mice until the end of all experiments within the groups . Acute hippocampal slices were used to record field excitatory post-synaptic potentials ( fEPSPs ) , by the MEA60 electrophysiological suite ( Multi Channel Systems , Reutlingen , FRG ) as described [72 , 73] . Further details are provided in the supplementary information ( S1 File ) . Since males were used for behavioral analyses , metabolic and biochemistry analysis were performed on females to both reduce the number of animals used and also because female mice are more prone to weight gain after metabolic challenges with an enriched diet . The separate Del/+ and Dup/+ cohorts on the B6N genetic background ( n = 11 wt and 9 Del/+; n = 8 wt and 10 Dup/+ ) and a Del/+ cohort on the C57BL/6NxC3B background ( n = 8 wt and 10 Del/+ ) were put on a high-fat diet for metabolism and biochemical analyses . Mice were transferred from the animal facility into the phenotyping area at the age of 4 weeks . Two to four animals were housed in one cage and fed with a standard chow diet ( D04 , Safe , USA ) until the age of 5 weeks , when the diet was switched to a high fat/high carbohydrate diet ( HFHCD , RD 12492 , Research Diet , USA ) until the end of the study ( 16 weeks ) . Body weights were recorded once a week from the age of 6 to 13 weeks . At the age of 11 weeks , mice were put individually into the TSE cages for 24 hours for the measurement of energy expenditure by indirect calorimetry . At the age of 12 weeks , an intra-peritoneal glucose tolerance test ( IPGTT ) was performed after overnight fasting . Body composition ( lean , fat and free body fluid content ) was evaluated in conscious mice by quantitative nuclear magnetic resonance ( qNMR ) at the age of 13 weeks . At the age of 15 weeks , blood was collected by retro-orbital puncture under isoflurane anesthesia for biochemistry , haematology and endocrine analysis . At the age of 16 weeks , animals of the Del/+ cohort on the C57BL/6NxC3B background were put in individual cages for 48 h to collect feces for the measurement of energy content by bomb calorimetry . Body composition analysis by qNMR was performed to give a precise assessment of the body composition for fat content , lean tissues and free body fluid by nuclear magnetic resonance apparatus and Minispec+ analyser ( Bruker , Germany ) . The test was conducted during the light period in conscious , fed mice . Energy expenditure was evaluated by indirect calorimetry that measured oxygen consumption with an open flow respirometric system ( TSE System , Labmaster , Germany ) . Sensors measured the difference in CO2 and O2 concentrations in the air volumes flowing through control or animal cages . The amount of oxygen consumed over a given period of time could be calculated , as long as the air flow through the cage was known . The gas exchange data are expressed in Kcal/h/kg0 . 75 . The system also monitored CO2 production , and we calculated the respiratory exchange ratio ( RER = VCO2/VO2 ) , which defines fuel preference between glucose vs lipid metabolism and heat production ( Kcal/h/kg0 . 75 ) . Following a 3 h acclimatization period , the experiment was performed over 21 h: from 17:00 on the first day to 11:00 on the day after , at an ambient temperature ( 21°C ± 2 ) under a 7:00/19:00 light/dark cycle . The intraperitoneal glucose tolerance test ( IPGTT ) was used to assess the regulation of glycaemia after induced hyperglycaemia by injecting a standardized glucose bolus ( 2 g/kg ) . A glucose solution was administered by intraperitoneal injection . Blood glucose was measured in a drop of blood collected from the tail at different time points over 120 min after injection using a blood glucose monitor and glucose test strips ( Roche Diagnostics , Accu-Chek , France ) . The test was conducted during the light period after 16 h of overnight fasting . Blood was collected by retro-orbital puncture under isoflurane anaesthesia after 4 h fasting . Blood chemistry analysis was performed using a chemistry analyser ( Olympus AU-400 , USA ) and commercial reagents ( Olympus Diagnostica GmbH , Lismeehan , Ireland ) . The following parameters were measured: total cholesterol , HDL and LDL cholesterol , triglycerides , free fatty acids , total proteins , albumin , calcium , phosphorus , glucose , urea , creatinine , total proteins , and albumin . Plasma insulin and leptin levels were measured with a BioPlex analyser using Milliplex beads Panel ( Millipore , USA ) . Plasma adiponectin levels were measured using a mouse adiponectin ELISA kit f ( R&D Systems , USA ) . Fecal energy content was estimated by bomb calorimetry . The feces were collected over 48 h from mice that were housed individually . The energy content of fecal samples was evaluated in a bomb calorimeter ( C503 control , IKA , USA ) . The samples were burned in an oxygen-rich atmosphere inside a sealed chamber surrounded by a jacket containing a known volume of water . The rise in the water temperature was recorded and used to calculate the amount of heat produced . The assay was performed on feces to evaluate indirectly the energy digested by mice and the intestinal function . The energy digested was calculated by the difference between the total calories ingested and excreted in feces . Craniums of 13 week-old female mice ( n = 10 wt and 10 Del/+; n = 8 wt and 8 Dup/+ ) of C57BL/6N genetic background were stored in 100% ethanol . Three-dimensional coordinates of 39 relevant cranial landmarks were recorded using Landmark software , and posterior comparisons were performed using the Euclidean distance matrix analysis ( EDMA ) with the WinEDMA software ( version 1 . 0 . 1 beta ) . Further details are provided in supplementary information ( S1 File ) . Expression profiling after RNA extraction was performed using an Affymetrix Mouse Gene 2 . 1 ST 24-Array Plate following the manufacturer's protocol . All arrays passed the standard Affymetrix quality control checks . Expression signals from 12 liver samples and 44 brain samples ( 12 cerebellum , 16 striatum , and 16 hippocampus ) were quantile normalized separately . We applied a non-specific filter to discard probe sets with low signals ( maximum signal lower than the median signal for all probe sets ) or with low variance ( lowest 25th percentile of variance ) . Differentially expressed genes were defined using linear models as implemented in LIMMA v3 . 18 . 13 [74] . We fitted a dosage model where the 16p11 . 2 copy number variants ( CNVs ) were considered as a numerical variable , i . e . , we assumed a dosage effect of the CNVs: gene_i ~ b0+ b1CNV CNV = -1 , 0 , 1 . Gene Ontology ( GO ) analysis was performed using the library topGO v2 . 14 . 0 . Gene Set Enrichment Analysis ( GSEA ) was conducted using the Broad Institute algorithm v2 . 1 . 0 [75 , 76] . We used pre-ranked gene lists defined according to the LIMMA results based on the dosage model , and we utilized MSigDB C2 ( c2 . all . v4 . 0 . symbols . gmt ) to collect curated gene sets from online pathways , and publications in PubMed . The data are accessible through the NCBI’s Gene Expression Omnibus with the GEO Series accession number GSE66468 . We assessed statistical significance using Sigma Plot software ( Sigma ) . All acquired behavioral data were analysed using one-way or two-way ANOVA with repeated measures followed by Student’s t-test or Tukey’s post-hoc test as appropriate . Otherwise , the non-parametric Kruskal-Wallis analysis and Mann-Whitney U-test were used . Student’s t-tests were used to compare recognition index values to the chance level ( 50% ) . A Pearson’s chi-square test was used to evaluate mutant allele transmission . The data are represented as the mean ± standard error of the mean ( s . e . m . ) and the significance threshold was P < 0 . 05 unless otherwise indicated . In electrophysiological experiments , input-output relationships were compared initially by mixed model repeated-measures ANOVA and the Bonferroni post hoc test implemented in Prism 5 ( GraphPad Software , Inc . , San Diego , CA , USA ) using individual slice data as independent observations . Because several slices were recorded routinely from every mouse , the fEPSPmax , PPF and LTP values between wt and mutant mice were also compared using a two-way nested ANOVA design with genotype ( group ) and mice ( sub-group ) as factors ( STATISTICA v . 10 , StatSoft , Inc . , Tulsa , OK , USA ) . To compare data obtained from Del/+ , Dup/+ and Del/Dup mice to their litter-matched wt counterparts , we used post hoc Dunnett’s Multiple Comparison test on data from individual slices . Statistical effects were considered significant if P < 0 . 05 . Graph plots and normalization were performed using OriginPro 8 . 5 software ( OriginLab , Northampton , MA , USA ) . Throughout the text , the electrophysiological data are presented as the mean ± s . e . m . with n and N indicating the number of slices and mice , respectively .
|
The 16p11 . 2 BP4-BP5 deletion and duplication syndromes are frequent copy number variants in humans , and are associated with developmental delay and autism spectrum disorders , with a reciprocal effect on head circumference and body mass index . Here we explored gene dosage effect in mouse models and found that the deletion and duplication induced opposite behavioral phenotypes . Notably , we observed that some behavioral phenotypes , such as social interaction , were sensitive to the genetic background . For the metabolism , the energy imbalance and adipocyte phenotypes were mirrored in the deletion and duplication carriers but opposite to the human phenotypes , the deletion mouse carriers were lean whereas the individuals with the deletion were obese . The main cause of the phenotypic features is the copy number variation of the 16p11 . 2 region with many genetic pathways altered in the striatum and the liver . Thus the final consequences of the rearrangement are likely governed by the interplay between many cellular pathways in both human cases and mouse models .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"vertebrates",
"social",
"sciences",
"mice",
"neuroscience",
"animals",
"mammals",
"learning",
"and",
"memory",
"diet",
"biological",
"locomotion",
"animal",
"models",
"perception",
"model",
"organisms",
"cognitive",
"psychology",
"animal",
"behavior",
"nutrition",
"cognition",
"experimental",
"organism",
"systems",
"memory",
"zoology",
"research",
"and",
"analysis",
"methods",
"behavior",
"object",
"recognition",
"mouse",
"models",
"rodents",
"psychology",
"eukaryota",
"phenotypes",
"physiology",
"genetics",
"biology",
"and",
"life",
"sciences",
"cognitive",
"science",
"amniotes",
"organisms"
] |
2016
|
Reciprocal Effects on Neurocognitive and Metabolic Phenotypes in Mouse Models of 16p11.2 Deletion and Duplication Syndromes
|
The increasing burden of dengue fever ( DF ) in the Americas , and the current epidemic in previously unaffected countries , generate major costs for national healthcare systems . There is a need to quantify the average cost per DF case . In Mexico , few data are available on costs , despite DF being endemic in some areas . Extrapolations from studies in other countries may prove unreliable and are complicated by the two main Mexican healthcare systems ( the Secretariat of Health [SS] and the Mexican Social Security Institute [IMSS] ) . The present study aimed to generate specific average DF cost-per-case data for Mexico using a micro-costing approach . Expected medical costs associated with an ideal management protocol for DF ( denoted ´ideal costs´ ) were compared with the medical costs of current treatment practice ( denoted ´real costs´ ) in 2012 . Real cost data were derived from chart review of DF cases and interviews with patients and key personnel from 64 selected hospitals and ambulatory care units in 16 states for IMSS and SS . In both institutions , ideal and real costs were estimated using the program , actions , activities , tasks , inputs ( PAATI ) approach , a micro-costing technique developed by us . Clinical pathways were obtained for 1 , 168 patients following review of 1 , 293 charts . Ideal and real costs for SS patients were US$165 . 72 and US$32 . 60 , respectively , in the outpatient setting , and US$587 . 77 and US$490 . 93 , respectively , in the hospital setting . For IMSS patients , ideal and real costs were US$337 . 50 and US$92 . 03 , respectively , in the outpatient setting , and US$2 , 042 . 54 and US$1 , 644 . 69 in the hospital setting . The markedly higher ideal versus real costs may indicate deficiencies in the actual care of patients with DF . It may be necessary to derive better estimates with micro-costing techniques and compare the ideal protocol with current practice when calculating these costs , as patients do not always receive optimal care .
Dengue fever ( DF ) is caused by infection with the dengue virus , a single-stranded positive-sense RNA virus of the Flaviviridae family [1] . The virus is transmitted almost exclusively by the mosquito vector Aedes aegypti , and humans are the only known reservoir for the virus [2 , 3] . Clinical presentation varies , with signs and symptoms ranging from uncomplicated fever in the case of simple DF , to bleeding and low platelet counts in the case of dengue hemorrhagic fever ( DHF ) . According to an estimate recently published by the World Health Organization ( WHO ) , between 50 and 100 million infections occur every year [4] . Based on a cartographic approach , Bhatt et al . [2] estimated the number of annual worldwide dengue infections to be 390 million ( 95% credible interval 284–528 million ) . The WHO estimated that 500 , 000 people require hospitalization each year and about 125 , 000 of those affected die [4] . As in most of Latin America , the disease is widespread in Mexico , although the incidence rate has varied since its reappearance in the 1970s . Peaks in the number of cases occurred in 1980 , 1997 , and 2009 , when more than 130 , 000 cases were reported nationwide [5] . Between 1995 and 2011 , a cumulative total of almost 600 , 000 cases were reported , with 11% corresponding to DHF . According to the Sub-Directorate General for Epidemiology in Mexico , 62 , 330 , 32 , 021 , and 26 , 665 cases were reported for the years 2013 , 2014 , and 2015 , respectively [6] . Given these high incidences , the potential seriousness of infections and the considerable disease burden , it is important to have accurate estimates of the average costs associated with the disease . Such estimates will enable efficient allocation of finite healthcare resources [7] , for example for vaccination [8 , 9] , vector control [10–12] , and integrated control of dengue through vaccination and vector control [13] . When the present study was initiated , some studies of the economic consequences of dengue had been conducted in Latin American countries including Brazil [14] , Colombia [15] , and Panama [16] . However , substantial variations in the underlying healthcare context exist in these countries and different methodologies were used in each analysis , such that we cannot extrapolate from these data to other settings . It is thus necessary to standardize and improve the reliability of the methodology and available estimates of the costs of dengue [17 , 18] . In Mexico , available data on the costs of the disease are limited . In one of the most comprehensive international studies of the cost of DF , Shepard et al . [19] used specific information from index countries ( Venezuela , El Salvador , Guatemala , Panama , Brazil , and Puerto Rico ) and estimated the cost of a DF case in countries of the Americas . The cost of an ambulatory case of DF in Mexico was estimated to be US$486 ( of which US$264 corresponded to direct medical costs ) , while the estimated cost of a hospitalized case was US$1 , 209 ( of which US$502 corresponded to direct medical costs ) . Clearly , the assumptions used when making such an estimate might not hold for Mexico , given the segmentation of the healthcare system and the different organization of each sector within the healthcare system . Secondly , in Mexico there are no cost centers in any of the local healthcare units or in most hospitals . It is therefore not possible to simply aggregate the costs incurred from these types of sources [20 , 21] . Thirdly , the two largest public healthcare systems in Mexico are very different in terms of the package of benefits provided , organization of the medical units , as well as the level of resources available and quality of the services provided . The Mexican Social Security Institute ( Instituto Mexicano del Seguro Social [IMSS] ) provides coverage to all industrial workers and their families . The Secretariat of Health ( Secretaría de Salud [SS] ) provides coverage to uninsured individuals who do not qualify for IMSS coverage or cannot pay private insurance , and tend to be in a lower income bracket [22 , 23] . The proportion of the Mexican population covered by each system is roughly equal ( 30 . 4% and 36 . 6% , respectively ) [24] . The co-existence of two different main public healthcare delivery systems ( IMSS and SS ) is not the only factor that may hamper extrapolation of cost data . In Mexico , the effectiveness of the healthcare sectors has been questioned after multiple reforms have been enacted [25] , with patients not always receiving optimal treatment . In particular , the SS healthcare sector , which provides coverage through ‘Popular Insurance’ , has suffered from suboptimal funding and fragmented organization ( in reality there are 32 different healthcare systems–one for each state in the country ) [26] . A recently published study of the economic and disease burden of dengue in Mexico estimated an average cost for DF . The authors used primary data from four major hospitals in the states of Quintana Roo , Morelos , and Tabasco in Mexico [27] . Estimates of DF costs for the two healthcare systems using a larger number of medical units and data from states with endemic DF may be more accurate and will allow comparisons between the two providers . The main objective of this study was therefore to assess the associated medical costs and cost to the individual with DF using a micro-costing approach to overcome the lack of cost center data . Chart review and interviews with patients and key personnel from 64 medical units in 16 states with endemic DF provided data on costs associated with actual treatment of the disease ( ‘real costs’ ) . For further comparison we estimated the costs that would be incurred if an ideal treatment protocol were followed ( ‘ideal costs’ ) .
The study was approved by the National Commission of Scientific Research for the IMSS , register number: R-2012 785–070 . All participants provided written informed consent and all patient data were anonymized . The cost per case of DF was calculated by summing the following costs: ( 1 ) direct medical costs incurred by healthcare units ( including professional services , 127 medical inputs , medical drugs and related products , as well as laboratory tests ) ; ( 2 ) costs of dengue from the patient’s perspective ( direct medical costs not covered by the public healthcare services and direct nonmedical costs , eg travel expenses ) ; and ( 3 ) indirect costs ( to the patient and their family ) resulting from loss of productivity and loss of earnings . The second and third components include medical and nonmedical direct costs and also indirect costs , such as loss of productivity , caregiver’s costs , etc . All costs were calculated in Mexican pesos and converted to US$ using the exchange rate on September 12 , 2012 ( 1 US$ = 13 . 03 Mexican pesos ) and adjusted for inflation to December 2014 when the analysis was finished . A hierarchical micro-costing approach was used to calculate direct costs incurred by the healthcare services ( known as a program , actions , activities , tasks , inputs [PAATI] analysis ) [26 , 27] . This type of analysis has previously been used for economic assessments of malaria [28] and hemodialysis in Mexico [26] . The method essentially involves identifying the tasks and inputs from an ideal protocol or current treatment practice for DF and then assigning a unit cost to each . The unit costs were obtained from official IMSS and SS databases . Treatment costs are different for each institution because they have different unit costs for the inputs ( see Table 1; full details of the individual unit costs included in the calculation are available in the S1 Appendix ) . The tasks and inputs included in the analysis were defined in two ways . In one scenario , an ‘ideal’ cost was calculated according to the tasks and inputs included in an ideal management protocol for patients with DF . Fig 1 outlines the protocol , which was validated in an expert consensus meeting held on June 5 , 2012 . The protocol was prepared using a systematic review of the literature , as well as guidelines available for Mexico and Latin America . These materials were also used in an expert group discussion of the issue sponsored by the Ministry of Health , in which the authors of this paper actively participated ( see Betancourt-Cravioto et al . [29] for further details ) . Essentially , the clinical protocol is divided into three sections ( diagnosis and case identification , classification and notification , and treatment ) , and each section is assigned a series of activities , which in turn are assigned tasks and inputs . In the second scenario , treatment costs associated with current practice ( which we denote ‘real costs’ ) in the healthcare services were determined by calculating costs for the tasks and inputs actually performed . These tasks and inputs were determined by chart review and patient interviews . Where data from the medical units were missing or incomplete , the information was supplemented as far as possible from interviews with treating physicians and hospital administrators . The PAATI approach is summarized in Fig 2 . The cost of each activity was calculated using the average use reported per input ( ie resource or cost type ) . We looked at each separate type of cost incurred , which allows control of the variability for each component of the healthcare process , rather than using other methodologies where the average overall cost per patient is calculated and then assigned to each activity . For example , using the patient’s clinical chart , the number of blood biometry procedures the patient had undergone was assessed and the average per patient was calculated , which in this case was 1 . 6 times ± 1 . 1 per patient . This average was then adjusted by the number of patients in each care setting ( outpatient , hospitalized , or intensive care unit [ICU] ) who used that resource . The resulting adjusted average resource use was finally multiplied by the unit resource cost . In this way the average cost per case was adjusted for variability by taking into account the proportion of patients who reported consumption of inputs within the sample of patients with DF , in both the patient files reviewed and the patient interviews . Given the different structures and attributes of the two principal sectors of the healthcare system , costs were calculated for both of them . In addition , unit costs in the private sector were used to estimate the cost of the ideal protocol only , as we had no way of evaluating current treatment practice in the private sector , given the difficulties of selecting a sample from the myriad of private healthcare units in the 16 states with endemic DF . Therefore , in the present study , we only assessed operational costs or variable costs , and no estimates were made of sunk costs or other costs , such as overhead costs ( eg utilities , administration , etc . ) . The indirect costs from the patient´s perspective were calculated based on information derived from interviews with patients . For these data , the methodology used took into account costs reported by the patients , and these were used to calculate the average cost per case , as well as confidence intervals . A bootstrap analysis was performed to assess variability . Where possible , the patients interviewed were the same patients whose charts were reviewed . Indirect cost was defined as loss of earnings associated with the disease ( patient and/or caregiver ) , with information taken either from direct questions or an estimate based on the number of days of work lost multiplied by the net loss of earnings per day , separately for each institution . For example , in patients who reported being employed , the average length of hospital stay in days was multiplied by the average daily salary . The same approach was used for caregivers . A cost was not assigned if the patient or caregiver did not report any income or they were students . For patients under 16 years of age , only loss of earnings for the caregiver was calculated . We selected a non-probabilistic sample . The study sample was drawn from 32 hospitals and 32 ambulatory care units in 16 Mexican states with endemic DF . Within each state , the hospitals and ambulatory care units with the highest number of cases were selected and up to a maximum of 40 cases per hospital or unit were selected at random for chart review . To avoid selection bias reviewers were instructed to use random numbers to select records . For units with fewer than 40 cases , a census was carried out , and all cases were included in the review . Only patients with an infection occurring in 2012 were considered for chart review . A sample size sufficient to provide reasonably robust data was calculated using the formula: n= ( Nσ2Z2 ) / ( [N−1]e2+σ2Z2 ) Where n = sample size , N = total population , σ = standard deviation , e = acceptable sampling error limit ( 0 . 05 ) , and Z = 1 . 96 ( for a 95% confidence interval ) . With these parameters , a sample size of 1 , 440 subjects was obtained . As discussed above , costs were calculated for both the SS and IMSS systems , and so the sample size calculation was applied to both systems separately , resulting in double the overall number of patients . This sample would be representative of overall proportions in the SS and IMSS systems at a national level , although it does not take into account state-level stratification . For indirect costs to the patient , a bootstrap sensitivity analysis was conducted to assess the robustness of the estimation of the average input ( based on the method proposed by Efron [30] ) . Bootstrap analyses are used to estimate the distribution , bias , or variance in a statistical sample or analysis , and to estimate confidence intervals or test hypotheses on parameters of interest when the true distribution of those parameters is unknown . For each type of cost the original sample was sampled to get 5 , 000 bootstrap resamples to provide estimates and 95% confidence intervals for comparison with the original estimates .
As shown in Table 2 , the medical cost differed according to setting ( based on outpatients , hospitalized patients , and patients in the ICU ) , regardless of whether ideal or real costs were considered or which healthcare system ( SS or IMSS ) was used . Of particular note were the marked differences between the ideal and real costs apparent in both systems . The main factor influencing these differences was the cost of professional services ( which accounts for approximately 90% of the differences in the case of outpatients and almost 100% in the case of hospitalized patients and those in the ICU ) . Professional services also accounted for the largest proportion of the medical cost , while the contribution of expenditure on medicines to the overall medical cost was limited . Nevertheless , of the cost types considered , the cost of medicines showed the largest difference ( in terms of percentages rather than absolute costs ) between the ideal scenario and the real current treatment practice . In both the ideal and real scenarios , the costs to the IMSS were greater ( by a factor of 2–4 ) than the costs to the SS across all patient settings . As for differences between ideal and real costs , the main driver was the cost of professional services . Expenditure on medical consumables and drugs and related products showed very little variation between the two systems , while expenditure on laboratory tests was somewhat higher for the IMSS . The overall costs of the ideal management of patients in the private sector was higher than the corresponding ideal costs in the IMSS for ambulatory patients ( US$487 . 39 vs US$337 . 50 , respectively ) and hospitalized patients ( US$4 , 077 . 81 vs US$2 , 042 . 54 ) , while costs for patients in the ICU were similar ( US$23 , 753 . 19 vs US$23 , 452 . 63 ) . Expenses from the patient’s perspective are presented in Table 3 . As can be seen , here we report direct medical costs incurred by the patient as well as direct costs reported by the healthcare units . These costs increased when the patient received care in the hospital setting . The difference was not so marked between hospitalized patients and those who received care in the ICU , with the exception of patients in the SS system . In the outpatient setting , the direct medical care costs are almost double for IMSS patients than for SS patients . It is important to remember that these costs are independent from what was spent by the healthcare unit in the treatment of the patient . We estimated the loss of productivity based on the number of days reported in the patient interviews for general hospitalization or ICU care . The average cost for a hospitalized patient is therefore not the same as for an ICU patient because , among other factors , the average length of stay is longer for a patient who receives ICU care . Data from the national census of population and national health surveys indicate that , on average , IMSS patients are more affluent than SS patients [22 , 23] . The study did not collect data on the average income of patients in the two healthcare systems , although it is important to bear in mind possible differences when interpreting the data . The questionnaire did not address loss of income for outpatients and these data are indicated as not available in Table 3 . We consider this a limitation of the study . In general , indirect costs appeared to be higher for patients in the IMSS system than for those in the SS system , with the exception of costs for patients in the ICU , which tended to be higher for SS patients ( Table 3 ) . The indirect costs for patients with DF corresponding to loss of earnings ( not related to medical costs for the patient and/or their caregivers ) for patients admitted to the ICU were actually lower compared with hospitalized patients ( Table 4 ) .
Despite the fact that a non-probabilistic sampling method was employed in this analysis , the extensive chart review and direct interviewing techniques provided a robust estimation of the average cost of treatment of DF in Mexico . The study also provided information according to type of healthcare system used , thus enabling qualitative comparisons . In this study we tried to address the limitations of other estimations of average cost per case of DF . The available cost estimates in the literature showed great variability in their methodology , using primary data , macro-costing data , patient questionnaires , administrative data , or most commonly , a combination of data sources including some primary data in a highly restricted population . In our study , the use of micro-costing provides a more detailed understanding of the direct medical costs of dengue , and that is one of its major strengths . Although we used average costs , we did not want to make the assumption that patients from a single medical unit reflect the whole experience of a country . Therefore , in our study the average costs per case were calculated from data collected from medical audits of 64 healthcare units ( 1 , 293 chart reviews ) and 1 , 168 patient interviews in 16 states where DF is endemic . To compare our results with other studies , here we present the original costs and within brackets the dollars adjusted for inflation using the national consumer price index by country and the official exchange rate , annual average for The World Bank for 2014 . For ambulatory cases , the direct real medical cost of 2012 US$32 . 60 ( 2014 US$35 . 93 ) in the SS was lower than the direct medical costs reported by Sheppard et al . for Brazil 2010 US$49 ( 2014 US$54 . 05 ) [19] ) , and more recently Colombia 2012 US$67 ( 2014 US$65 . 18 [15] ) , whereas the IMSS direct real medical cost of 2012 US$92 . 0 ( 2014 US$101 . 38 ) was lower than reported in Venezuela 2010 US$118 ( 2014 US$130 . 15 ) [19] ) and Panama 2005 US$332 ( 2014 US$501 . 65 ) [16] ) . With a micro-costing approach , a multicenter Brazilian study reported a confidence interval of US$31 to US$89 2013 , ( 2014 US$33 . 81 to 2014 US$97 . 06 ) [14] , similar to the SS and IMSS ambulatory costs in our study . After we had performed our data collection , Undurraga et al . [27] published an estimation of the ambulatory costs associated with DF in Mexico , where derived costs per episode were calculated by combining patient interviews in four major hospitals in the states of Quintana Roo , Morelos , and Tabasco , macro-costing data from two major public hospitals in Tabasco , MoH health and surveillance data , WHO-CHOICE estimates for Mexico , and previous literature on dengue burden . Indirect costs were obtained based on productivity losses by age , considering both the patient and the patient’s caregivers . These authors reported a cost of 2012 US$65 . 53 ( 2014 US$72 . 22 ) per outpatient visit and the average cost for ambulatory patients was 2012 $451 ( 2014 US$497 . 05 ) thus lying between the SS and the IMSS costs in our study . The direct medical costs of hospitalized patients , in relative terms , are higher in our study at 2012 US$490 ( 2014 US$ 539 . 99 ) and 2012 US$1 , 644 ( 2014 US$1811 . 71 ) for the SS and IMSS , respectively , compared with the estimates for Brazil 2013 US$318 ( 2014 US$346 . 80 ) [19] ) , Colombia 2012 US$330 . 6 ( 2014 US$321 . 63 ) [15] ) and in the new multicenter study for Brazil 2013 US$238–479 ( 2014 US$259 . 55–522 . 38 ) [14] ) . Costs reported for Venezuela 2010 US$864 ( 2014 US$952 . 99 ) [19] ) and Panama 2005 US$1065 ( 2014 US$1609 . 22 ) [16] ) are higher than the SS costs but lower than the cost to the IMSS . Although comparisons with other countries may be illustrative , firm conclusions cannot be drawn given the differences in economic development , population size , and healthcare systems , as well as the methodology used for the estimates . In the case of hospitalized patients , comparisons are more difficult because no distinction is made between hospitalized and ICU settings in most Latin American studies . Interestingly , the ideal costs estimated in our study , 2012 US$587 ( 2014 US$646 . 88 ) for SS and US$2 , 042 ( 2014 US$2 , 250 . 31 ) for IMSS were closer to the extrapolated costs reported by Shepard et al . [19] for total hospitalized cost , 2010 US$1 , 209 ( 2014 US$1 , 333 . 53 ) in Mexico . Undurraga et al . [27] reported that the hospital cost for Mexico was 2012 US$240 . 04 ( 2014 US$261 . 48 ) per bed day and "the average cost per non-fatal dengue episode was $1 , 327 for hospitalized patients ( 2014 US$1 , 445 . 53 ) ( direct medical: $1 , 010 ( 2014 US$1 , 100 . 22 ) ; direct non-medical: $174 ( 2014 US$189 . 54 ) ; indirect: $143 ( 2014 US$155 . 77 ) and $451 for ambulatory patients ( 2014 US$491 . 29 ) ( direct medical: $253 ( 2014 US$275 . 60 ) ; direct non-medical: $92 ( 2014 US$100 . 22 ) ; indirect: $106 ( 2014 US$115 . 47 ) " . Thus , these data are between the SS and the IMSS costs in our study . Finally , we are not interested in presenting the virtues or deficiencies of the Undurraga approach; we simply want to clarify the differences with our approach . Indeed , a particularly striking feature of our results are the differences between the direct medical costs of actual treatment in clinical practice and those that would be generated if an ideal protocol , validated by experts in the field , were followed . The difference was particularly marked for outpatients . The first implication is that cost studies based on an ideal treatment protocol may not reflect clinical reality , at least in Mexico . The difference in personnel costs may reflect systematic differences in either the productivity of the personnel , their training , and treatment standards among the medical units of the institutions included in the study . The second implication is that there may be shortcomings in the Mexican healthcare systems , particularly in the outpatient setting , despite extensive reform in recent years with a view to improving the quality of care [20 , 21 , 31] . The main driver of the difference between real and ideal costs is the cost of professional services and the use of laboratory tests ( including confirmatory tests ) . This may indicate that the treating physicians are not dedicating sufficient time to their patients , nor providing optimal laboratory follow-up for patients . Furthermore , we note that most of the data were collected in 2012 , which was not an epidemic year . In epidemic years , it is likely that overburdening of the system would further accentuate the differences between real and ideal costs , as physicians would be forced to dedicate less time to their patients , and the demand for laboratory tests would be higher . Another noteworthy feature of the results presented here is the higher direct medical costs incurred within the IMSS system compared with the SS system . Another Mexican study utilizing chart reviews and direct patient interviews estimated the direct medical costs in the IMSS system for patients with osteoporosis and hip fracture [24] . Compared with the SS system , they also found considerably higher cost for IMSS ( US$3 , 891 . 20 vs US$1 , 590 . 70 , respectively ) . The main driver of this difference in the present study was personnel costs , which may reflect differences in pay between the two systems . The IMSS was set up over 60 years ago and has strong unions , which may be responsible for higher staff costs . In addition , there may be a tendency to use the private sector as a guide when setting prices , particularly in the ICU setting , where the IMSS and private sector prices are very similar . As noted by Clark et al . [24] in their study of hip fracture , the SS system , which provides medical care to the lowest income groups , receives larger subsidies that are not otherwise reflected in the costs generated by a micro-costing analysis such as the present one . It is important to note that the results in our study are closer to the multicenter study in Brazil [14] that used a micro-costing approach , even though we have clear differences between our healthcare systems . The greater availability of treatments in the IMSS system may explain the slightly higher treatment costs in the outpatient and hospitalized settings , although the more fragmented nature of the SS system may increase procurement costs . There is evidence that the heterogeneity of the state healthcare services ( SS system ) may lead to small variations between areas , resulting in care below that recommended in clinical guidelines . This substandard care is associated with lower costs . Of note were the expenses reported by the patient beyond those incurred by the healthcare institutions for hospitalized patients and patients in the ICU . Expenses incurred by patients were higher within the SS system , suggesting that these patients had to supplement the care provided by the SS to a greater extent than IMSS patients . Any comparison of the costs in the IMSS and SS systems should bear in mind that the cost of each activity was calculated using the average use reported per input and the unit cost reported for every organization and state . In general , the unit costs in the IMSS system are more stable and better registered at the central level of the organization than costs within the SS system . A limitation of our study is that loss of earnings due to days off work was only registered for patients admitted to hospital or ICU and thus was not captured for outpatients . Indirect costs were greater for patients who attended through the IMSS , possibly reflecting their higher socioeconomic status . Interestingly , the difference was greater for hospitalized patients than those admitted to the ICU . This observation can be explained by the fact that family members may be allowed to stay with patients in a hospital ward but not with patients in an ICU setting and they will thus incur more expenses . A second limitation was that we do not include data on management , electricity , and other utilities in the cost estimation . However , it was considered that the relevant cost is variable cost , and given the lack of data on management , electricity and other utilities , we considered that it would be misleading and add more uncertainty to the cost estimation . As noted above , a principal weakness of the study is possible bias resulting from the non-probabilistic sampling method employed in this study . The states where DF is endemic were chosen , and , within those states , the centers with the highest number of cases were selected . Furthermore , patients requiring hospitalization and those with DHF would be more likely to attend larger reference centers , so this approach may lead to an over-representation of patients with more severe disease . A further bias may have been introduced by not ensuring that all patients interviewed corresponded to those whose charts were reviewed . However , given that 80% of the patients interviewed also had a chart review , differences would have limited impact on the final average results . Nonetheless , by concentrating on larger centers with a larger number of cases , it was possible to collect extensive primary data that would otherwise be hard to acquire from probabilistic samples . In essence , statistical robustness was sacrificed to enable the data collected to reflect more accurately clinical practice in the sample , in contrast to many studies of costs of DF in the Latin American region . Additionally , the data available from chart review were not complete ( and may not have been entirely reliable ) and had to be complemented with other data sources . While this may also compromise the integrity of the data , we believe that chart review , despite its inherent limitations , ultimately provides the best approximation to what actually happens in clinical practice . It is also possible that recall bias was present in the interviews , but for most patients , DF is an event that they are unlikely to forget . In conclusion , this study pointed to real costs associated with DF in line with those reported in other Latin American countries ( with the caveat that direct comparisons should be treated with caution given the differences between countries in economic development , healthcare systems , size , and the cost methodology employed in studies ) . Of particular note were the large differences in the real costs derived from patient records and the ideal cost calculated from an ideal treatment protocol . This difference perhaps points to deficiencies in the care of patients with DF in Mexico . It also indicates the importance of collecting primary data when calculating DF costs to guide health policy decisions in this and other diseases that lack an appropriate estimate of its cost to the system , which poses a major hurdle for healthcare planning .
|
Dengue fever ( DF ) is caused by infection with the dengue virus , which is spread by the Aedes aegypti mosquito . Although the effects of DF are usually mild , in some cases serious illness and even death may result . The average costs per case when extrapolated to society may therefore be high , particularly given the large number of people infected during an endemic year . In Mexico , relatively little is known about the average cost per case ( from either the healthcare system or the patient perspective ) . Such information is important to guide decisions about health policy , e . g . vaccination or public education . We aimed to quantify the average cost per case of DF using a micro-costing approach , both for DF treatment according to an ideal protocol for the management of the patient ( ´ideal costs´ ) and according to current treatment practice in the health services ( ´real costs´ ) . Our results were largely consistent with findings from other international studies , but showed higher ideal costs compared with real costs . We think this may point to inadequate use of laboratory tests and treatments for patients with DF in Mexico . Our cost data will be used in a subsequent publication regarding the economic impact of DF in Mexico .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"infographics",
"medicine",
"and",
"health",
"sciences",
"public",
"policy",
"geographical",
"locations",
"tropical",
"diseases",
"social",
"sciences",
"indirect",
"costs",
"sociology",
"charts",
"health",
"care",
"ethnicities",
"north",
"america",
"political",
"science",
"welfare",
"(social",
"security)",
"neglected",
"tropical",
"diseases",
"patients",
"infectious",
"diseases",
"computer",
"and",
"information",
"sciences",
"health",
"economics",
"dengue",
"fever",
"social",
"welfare",
"economics",
"people",
"and",
"places",
"finance",
"outpatients",
"data",
"visualization",
"population",
"groupings",
"mexican",
"people",
"viral",
"diseases",
"financial",
"management",
"mexico"
] |
2016
|
Calculation of the Average Cost per Case of Dengue Fever in Mexico Using a Micro-Costing Approach
|
To investigate the stability and functional role of long-residence water molecules in the Q61H variant of the signaling protein K-ras , we analyzed all available Ras crystal structures and conformers derived from a series of independent explicit solvent molecular dynamics ( MD ) simulations totaling 1 . 76 µs . We show that the protein samples a different region of phase space in the presence and absence of several crystallographically conserved and buried water molecules . The dynamics of these waters is coupled with the local as well as the global motions of the protein , in contrast to less buried waters whose exchange with bulk is only loosely coupled with the motion of loops in their vicinity . Aided by two novel reaction coordinates involving the distance ( d ) between the Cα atoms of G60 at switch 2 and G10 at the P-loop and the N-Cα-C-O dihedral ( ξ ) of G60 , we further show that three water molecules located in lobe1 , at the interface between the lobes and at lobe2 , are involved in the relative motion of residues at the two lobes of Q61H K-ras . Moreover , a d/ξ plot classifies the available Ras x-ray structures and MD-derived K-ras conformers into active GTP- , intermediate GTP- , inactive GDP-bound , and nucleotide-free conformational states . The population of these states and the transition between them is modulated by water-mediated correlated motions involving the functionally critical switch 2 , P-loop and helix 3 . These results suggest that water molecules act as allosteric ligands to induce a population shift among distinct switch 2 conformations that differ in effector recognition .
The role of solvent on the structure and function of proteins has been the subject of numerous previous studies [1]–[6] . For instance , it has been shown that waters in the dimer interface of hemoglobin play an important role in its transition between deoxy and oxy forms [7] , [8] . Water molecules serve as “lubricants” to facilitate conformational inter-conversion [9] , [10] or as “adhesives” in binding interfaces [11] . Dehydrated biomolecules therefore lose their biological activity and have suppressed dynamics . Experiments by Frauenfelder and colleagues using Mossbauer and neutron scattering techniques demonstrated that internal fluctuations of proteins are linked to the dynamics of the surrounding solvent [12] , [13] , leading to the idea that protein dynamics is ‘slaved’ by [13]–[17] ( or coupled to [18] ) solvent dynamics . Many other experimental and theoretical studies arrived at similar conclusions [19]–[24] . However , it is less clear how buried solvent molecules might modulate an allosteric coupling between spatially distant regions in multi-domain or monomeric proteins [25]–[28] . A better understanding of how protein-bound waters modulate coupled motions and allostery can lead to new strategies for ligand and protein design . Water molecules reported in crystallographic structures , particularly those located at domain boundaries , interfacial regions or near active sites , often have a structural or functional role [29]–[34] . In contrast , the role of buried waters located far away from the active or ligand-binding site of a monomeric protein is not always obvious . In some cases , these waters may participate in long-range hydrogen bond networks that connect distal protein segments to functionally important regions . Recent reports indicate that this applies to H-ras [35]–[36] , a close homologue of the subject of this study , K-ras . Both H- and K-ras belong to the Ras family of molecular switches that play a central role in a variety of signaling pathways [37]–[39] . Ras is turned on when bound to guanosine triphosphate ( GTP ) and off when bound to guanosine diphosphate ( GDP ) [40] . The population of the ‘on’ and ‘off’ states is regulated by guanine nucleotide exchange factors ( GEFs ) , which increase the dissociation rate of GDP , and GTPase-activating proteins ( GAPs ) , which accelerate the slow intrinsic rate of GTP hydrolysis [39] , [41] , [42] . Oncogenic mutations that impair intrinsic Ras function and/or GAP action are found in ∼15% of all human tumors and in up to 90% of cases in specific tumor types [38] , [43] , [44] . The catalytic domain of K-ras is bi-lobal , with lobe1 ( residues 1–86 ) , but not lobe2 ( residues 87–167 ) , being evolutionarily conserved among the Ras family [45]–[48] . The major structural difference between the GDP- and GTP-bound forms involves two switch regions located in lobe1 ( S1: residues 25–40 and S2: 57–75; see Fig . 1 ) . Among several key residues known to have distinct interactions in the inactive and active forms are T35 on S1 and G60 on S2 ( See Fig . 1B ) , which interact with Mg2+ and the γ-phosphate of GTP; GTP hydrolysis leads to the loss of these interactions and relaxation of the switches to an open conformation [39] , [49] . Furthermore , a recent 31P-NMR [50] , [51] study indicated the presence of two conformational states in GTP-bound H-ras , state1 and state2 [52] . The two states are characterized by different chemical shifts in the α- and γ-phosphorous atoms of the GTP and by the ability of state2 to interact with effector proteins while state1 is stabilized by GEFs [51] . To investigate the role of buried water molecules on the distribution of functionally relevant conformational states of K-ras , we analyzed all available Ras x-ray structures and conformers derived from seven sets of twenty explicit solvent multi-copy MD simulations of Q61H K-ras . The simulations were performed with and without restraints on selected water molecules and protein segments , as well as in the presence and absence of two conserved protein-bound water molecules . Using a pair of simple , previously uncharacterized , reaction coordinates we grouped both the experimental and simulated structures into active GTP-bound , intermediate-like GTP-bound , inactive GDP-bound and nucleotide-free conformational states . We describe the effect of selected protein-bound waters on the distribution of these states .
As of 2010 , the protein data bank has 53 entries of Ras crystal structures comprised of 65 chains ( see supplementary Table S1 ) . Of these , 45 chains with a resolution of 2 . 8 Å or better contain crystallographic waters . Five water molecules were found conserved in over 70% of these structures . Fig . 1A shows the location of these waters in the protein based on their nearest neighbor , which is defined as the nearest of any backbone nitrogen or oxygen atom that is within 3 . 5 Å of a water oxygen atom . Two of the five water molecules are well-known for their interaction with the bound nucleotide [53] . However , little attention has been given to the remaining three waters , despite their apparent role in stabilizing the functionally critical nucleotide binding switches and the P-loop ( Fig . 1A ) . These include W1 coordinated by H27 and V29 at S1 and A18 at helix H1 , W3 interacting with G10 at the P-loop as well as side-chain atoms of T58 and R68 at S2 , and W4 which interacts with backbone atoms of A11 at the P-loop , V81 at β4 and S89 at H3 . The average reported B-factors for W3 and W4 are 22 and 19 Å2 , suggesting limited thermal fluctuation . The average backbone solvent accessible surface areas ( SASA ) of their nearest neighbors G10 and A11 are 2 . 2 and 1 . 2 Å2 . In contrast , W1 is relatively dynamic with a B-factor of 24 Å2 , and surface exposed with a backbone SASA of 2 . 9 Å2 for its nearest neighbor A18 . W3 and W4 are curiously missing in several Ras structures with mutations at position 12 ( P-loop , e . g . , PDB: 621P ) , 38 ( effector binding loop , e . g . , PDB: 221P ) , and 61 ( S2 , PDB: 721P ) . In addition , W3 ( but not W4 ) is missing in a G12D variant ( PDB: 1AGP ) and a Ras-Sos complex ( PDB: 1NVW ) , whereas W4 is missing in G12R ( PDB: 421P ) , another Ras-Sos complex ( PDB: 1HE8 ) and a Ras-PI3K complex ( PDB: 1BKD ) . It is important to note that oncogenic mutations frequently occur at positions 12 and 61 , and mutations at position 38 often impede effector binding . Moreover , W4 is located at the interface between the two lobes of Ras , connecting the P-loop in lobe1 with H3 in lobe2 via the interfacial strand β4 ( Fig . 1A ) . These observations prompted us to undertake a systematic MD analysis of the structural and functional roles of these waters . The time- and ensemble-averaged water diffusion coefficient , D , ( see Methods ) calculated as a function of distance from the protein surface , r , shows that water molecules within 2 . 5–5 . 0 Å of the protein diffuse very slowly ( D≈0 . 12–0 . 33 Å2/ps; see Fig . 2A ) . D progressively increases and stabilizes after r≈10 Å to ∼0 . 4 Å2/ps , which corresponds to the bulk-water diffusion coefficient of the TIP3P water model used in this work [54] . The reduction in translational motion of the protein-bound waters is coupled with a change in their orientation , as <cosθ> ranges between −0 . 3 and 0 . 1 in the region 2 . 4≤r≤6 Å before stabilizing near zero ( Fig . 2B ) . Further information about the influence of the protein-solvent interaction on the diffusive property of waters in different hydration shells can be obtained from the autocorrelation function of the electrical dipole moment , μ . Plots of this function for r values of 3–11 Å at an increment of 2 Å clearly show that the reorientation dynamics of the water dipoles in the first hydration shell ( r = 3 . 0 Å ) is much slower than those in the second hydration shell ( r = 5 . 0 Å ) , which in turn is much slower than those in bulk ( Fig . 2C ) . Previous studies have shown that protein-solvent interaction , especially hydrogen bonding , is the primary reason for the retarded dynamics of protein-bound waters [55] . The affinity of these hydration waters for the protein can be quantified by their survival probability , Nw ( t ) . A plot of Nw ( t ) versus time indicates that up to 6 waters have a survival probability of at least 8–10 ns in the three unrestrained trajectories ( Fig . 2D ) . Of these , 2–3 waters survived for the whole 100 ns duration of the simulations . This indicates that some of the conserved crystal waters described in the previous section are stably bound to the protein in solution and can be considered an integral part of the protein structure . The stability of these water molecules is remarkable given the higher diffusive property of TIP3P waters compared with some of the other more popular water models [56]–[59] . In sum , our results about the hydration behavior of Q61H K-ras are consistent with many previous reports on the hydration of other proteins , such as lysozyme [55] and plastocyanins [60] . Six water molecules were found to have tmean ( α ) ≥1 ns ( see eq . 2 and Table 2 , Fig . 3A ) . Each of these waters interacts with 2–3 backbone or side-chain atoms of nearby residues ( Fig . 3B ) . Nearest-neighbor analysis ( see Methods ) identified A18:O , G10:N , and A11:N as the primary interaction partners of W1 , W3 , and W4 , respectively . These are the same water-binding sites found in the majority of the x-ray structures ( Fig . 1A ) , indicating their preservation during the simulations . The remaining three waters W2 , W5 and W6 at sites K16:O , A83:O and D126:O were not present in the starting structure or the majority of the Ras x-ray structures . W1 has a mean residence time at site A18:O ( tmean ( A18:O ) ) of 17 ns ( Table 2 ) and occupancy of 30% ( see Methods ) . The other interaction partners of W1 are the backbone polar atoms of H27 in lobe1 and occasionally A146 in lobe2 . The latter interaction often occurs through another water molecule and appears to facilitate hydrogen bonding between A146 carbonyl and the nucleotide base . In this context , it is important to mention that A146T K-ras mutation has been recently implicated in colorectal cancer [61] . W2 , which bridges the backbone of K16 and the side-chain of D57 , has a mean residence time ( tmean ( K16:O ) ) and occupancy of 11 ns and 20% , respectively . K16 stabilizes the phosphate of the nucleotide [39] , while D57 is part of the conserved D57TAGQ61 motif [62] , [63] . W3 and W4 have occupancies of 72 and 84% at their respective sites G10:N and A11:N . They are tightly bound to these atoms with a mean square displacement ( MSD ) of 0 . 48 and 0 . 39 Å2 . W3 is located near the highly flexible S2 and is further stabilized by interactions with the side chain functional groups of T58 and R68 . As a result of S2's dynamics , W3 frequently exchanges with bulk water ( tmean ( G10:N ) ≈3 ns ) , suggesting that water binding at this site is entropically favored . In contrast , the more buried W4 rarely exchanges with bulk water ( tmean ( A11:N ) ≈41 ns ) , suggesting a strong enthalpic binding that involves accepting a hydrogen bond from A11:N-H and donating to V81:O and S89:O . This internal water thus couples the P-loop to H3 . Note that the smaller tmean ( A11:N ) is due to the exchange and eventual loss of W4 in one of the three trajectories ( tres , j = 100 ns in the other two ) . The non-crystallographic waters W5 and W6 also have significant tmean ( 2 . 0 and 1 . 5 ns ) at their respective sites A83:O and D126:O . Overall , these six long-residence waters stabilize functionally critical regions , such as the P-loop and the nucleotide binding switches , or physically connect the evolutionarily conserved lobe1 with the variable lobe2 . The latter is similar to previous reports on protein-kinase A where several water molecules participate in an extended network of inter-domain interactions [19] , [33] . To further investigate the functional role of the long-residence waters discussed above , it is important to identify features of the protein that may change with the presence and absence of these waters . In previous reports , we used principal component analysis ( PCA ) -based collective coordinates to discriminate between active and inactive conformations of Ras [45] , [46] , [64] , [65] . We found that the wild-type GDP- and GTP-bound structures form distinct clusters while mutant structures group into separate clusters that are intermediate between the two major clusters [45] , [46] . Though dominated by the switch regions , PCA contains contribution from the overall dynamics of the protein . Since the effect of individual water molecules may be localized near their binding site , we searched for reaction coordinates that only involve protein segments in the vicinity of W3 and W4 . Ideally , such reaction coordinates should also be able to discriminate between the fully active GTP- , less active GTP- , inactive GDP-bound and nucleotide-free conformations . ( For brevity , we will refer to these states as the GTP , intermediate , GDP and free state . ) The GTP and GDP states are very well characterized and covered in numerous excellent reviews ( e . g . , [26] , [39] , [66] , [67] ) . Nucleotide-free conformations , often found complexed with Sos [68] , are characterized by wide-open switch conformations . The intermediate state is comparatively less well-characterized , but several structural and biochemical studies have shown that some mutations on S1 ( e . g . , T35S ) and S2 ( e . g . , G60A ) lead to intermediate GTP-bound structures [62] , [63] , [69] , [70] that are defective in effector recognition [51] , [52] . The adjacent P-loop residues G10 and A11 interact in two different directions via W3 and W4: with T58 & R68 involving the conserved D57TAGQ61 motif in lobe1 and with V81 & S89 in lobe2 , respectively . Considering the role of G60 in stabilizing the γ-phosphate of GTP via its amide group ( Fig . 1B ) and the displacement of S2 away from the P-loop in the GDP state [39] , we reasoned that the distance between the Cα atoms of G10 and G60 ( d ) and the G60 N-Cα-C-O dihedral ( ξ ) may prove useful ( Fig . 1B ) . Indeed , Fig . 4A shows that analysis of the x-ray structures in our data set in terms of d led to two groups populated by GTP-bound ( d≤7 . 5 Å ) and GDP-bound plus nucleotide-free conformers ( d>7 . 5 Å ) . A similar analysis in terms of ξ produced two other groupings populated by GDP plus intermediate conformers ( ξ≤0 . 0° ) on the one hand , and GTP plus nucleotide-free conformers ( ξ>0 . 0° ) on the other . A scatter plot of d versus ξ yielded four distinct clusters populated by structures in the GTP , intermediate , GDP and nucleotide-free state ( Fig . 4A ) . It is remarkable that these simple reaction coordinates enable such a neat discrimination among states , including the intermediate GTP-bound sub-state populated by mutant structures ( e . g . , A59G , PDB: 1FL0 ) and few Ras-Sos complexes ( e . g . , PDB: 1NVX ) ( see Supplementary Material ) . The active and inactive clusters largely coincide with the previous PCA results [45] , [46] , [64] , [65] , apart from the GDP-bound G12V structure ( PDB: 2Q21 ) that , unlike in the PCA analysis , now clusters with the rest of the inactive structures . This is to be expected because the current classification emphasizes S2 while it was the uniquely open S1 conformation that separated 2Q21 from the main inactive cluster [45] , [46] . Overall , our d/ξ plot centered on S2 and the P-loop effectively captures the nucleotide dependent conformational dynamics of Ras . Projection of the MD-derived d/ξ values into the d/ξ space defined by the crystal structures shows that only 45% ( see Methods ) of the MD conformers sample from the GTP state ( Fig . 4B ) . Another 25% went to the intermediate state while the remaining conformers are distributed in the GDP and nucleotide-free states . Clearly , Q61H K-ras samples all four conformational states at room temperature and in the presence of GTP , with the active state being the most favored , followed by the intermediate state . This result thus re-enforces our previous finding that Ras exists in different conformational sub-states in solution and predominantly operates via a conformational selection mechanism [64] . Projection of conformers with and without W3 at G10:N ( Fig . S1 , A&B ) indicates no correlation with the water-occupancy of this site and the protein conformation . This implies that dynamics of the flexible loop of S2 is not directly dependent on W3 , as one would expect from the interaction of this water with S2 residues T58 and R68 . Interestingly , however , escape of the distal W4 , which occurred during one of the three F series of trajectories and represents about 17% of the conformers , shifted the equilibrium towards the more open inactive state ( Fig . 4B ) . This is surprising because , unlike W3 , W4 does not directly interact with S2 whose motion dominates the inter-state dynamics captured by the d/ξ plot . This can be understood by noting that the S2 loop around G60 moved away from H3 in conformations lacking W4 , which allowed for W4 to escape from its deeply buried position between the two lobes . This resulted in an increased solvation of G10 , as shown by the water coordination number of its carbonyl oxygen from 1–2 in the presence of W4 to 3–4 in its absence ( Fig . 4C ) . Thus , W4 influences the solvation behavior of W3's nearest neighbor even if the dynamics of the two waters is not strictly coupled . While waters W1 and W6 do not seem to affect the dynamics of S2 , entry of W2 and W5 in the latter half of the simulations resulted in significant changes in the conformational distribution ( Fig . S1 , C–F ) . In the presence of W2 the intermediate state is favored over the GTP state . In contrast , conformers containing W5 preferentially sample the nucleotide-free region . Taken together , the decoupled dynamics of the protein with some waters , such as W3 , and its strong coupling with others , such as W4 , implies that a global master-slave relationship of water and protein dynamics is not always applicable at the level of individual water molecules . The relationship between water and protein motion appears to be modulated by subtle differences in binding sites , such as whether a given structural water is entirely coordinated by the less flexible backbone atoms instead of side chains . Our results also underscore that the dynamics of individual water molecules can influence protein fluctuation locally or at a distance . To assess the role of backbone dynamics on the fluctuation of individual protein-bound waters and evaluate the extent of their coupling , we ran four independent sets of additional simulations . In the unrestrained simulations , W5-containing conformers ( representing 4% of the total ) are characterized by ξ>0°/d>7 . 5 Å and preferentially populate the nucleotide-free region ( Fig . 4B ) . In contrast , W5-containing conformers derived from the simulations are characterized by ξ<0°/d<7 . 5 Å and populate the intermediate state ( Fig . 5F ) . The reason for this discrepancy is not clear , but it may be related to the recently proposed multi-pathway nature of the allosteric effect [71] where an ensemble of states ( instead of a single propagation pathway ) defines a particular allosteric interaction . In both cases , however , the GTP state is disfavored , and entrance of W5 to the N-terminus of H3 coincides with a large conformational change of S2 , as shown by superposing the average structures with and without W5 ( Fig . 6 ) . As an example , A83:O and S89:OH were ∼3 . 6 Å apart in the absence of W5 but a partial loss of a helical turn at the N-terminus of H3 upon the entry of W5 led to increased solvation of S89 . That this local conformational change at the N-terminus of H3 is felt by S2 suggests their allosteric coupling , with W5 acting as a ligand . Though we are not aware of a report linking S2 dynamics with the N-terminus of H3 , previous studies have found a water-mediated interaction between Y32 at S1 and N86 at the N-terminus of H3 [35] , as well as a correlated motion between S2 and the C-terminus of H3 [46] . Although the full implication of this result to Ras function warrants further investigation , it underscores the potential of multiple/long simulations to uncover mechanistic insights hidden in crystallographic average structures . Nonetheless , the following observations highlight how water-mediated allostery might underlie the differential sampling of phase space by Q61H K-ras in the presence and absence of specific water molecules . Recent studies by Buhrman et al . [35] , [36] showed that a conformational change at H3/loop7 allosterically modulates S2 and thereby positions Q61 for a direct action in GTP hydrolysis . One of the two hydrogen bond networks responsible for the allosteric coupling involves Y96 , Q99 and R102 . In our simulations Y96 underwent major displacement away from S2 , thereby increasing the solvent accessibility of key P-loop and S2 backbone atoms ( Fig . 7 ) . The distribution of the distance between G60:O and Y96:OH ( Fig . 7A ) shows that the phenoxy group of Y96 is hydrogen bonded , often through a water molecule , with the carbonyl of G60 in the GTP-like conformers ( with a mean distance of 3 . 5 Å for the active and intermediate states ) . This hydrogen bond is broken as Y96 reorients away from S2 in the GDP and nucleotide-free states ( average G60:O-Y96:OH distance of 6 . 5 Å ) . These mean values closely match the average G60:O-Y96:OH distances obtained from GDP- and GTP-bound x-ray structures ( ca . 4 . 0 and 7 . 6 Å , Fig . 7A ) . Consistent with the previously observed positive correlation between the dynamics of S2 and H3 [46] , the reorientation of Y96 is coupled with the displacement of S2 away from the P-loop , as suggested by a strong correlation ( R = 0 . 78 ) between distances d and Y96:OH-G60:O . The reorientation of Y96 side chain toward H3 was analyzed by the angle between a vector connecting the Cα atoms of S89 and Y96 that approximately traces the helical axis , and a vector from the Cγ to the OH atoms of Y96 along the plane of the ring ( Fig . 7B ) . The angle decreased from ∼60° in the GTP/intermediate states to as low as 30° in the GDP state . The decrease is accompanied by the expulsion of W4 and a significant increase in the water coordination number of G10 , i . e . , the site of W3 binding ( Fig . 4C & 7C ) . Entry of waters such as W5 at the N-terminus of H3 did not result in a similar Y96 reorientation , but is associated with a larger opening of S2 and entry of other solvent molecules . These observations led us to conclude that the way in which Q61H K-ras samples conformational states , and the population of those states , is modulated in various ways by protein bound water molecules . The observed water-mediated correlated motion among S2 , the P-loop and H3 is very sensitive to perturbation and may be altered by ligands designed to interfere with the inter-lobe communication of Ras proteins . The aim of the current work was to investigate the solution stability and role of buried water molecules on the overall dynamics of Q61H K-ras , as well as to assess if crystallographic waters have a role in the inter-lobe communication and allosteric behavior of Ras proteins . To this end , extensive restrained and free MD simulations , involving seven sets of 20 multi-copy runs for a total of 1 . 76 µs , were carried out in the presence and absence of selected crystallographic water molecules . Analysis of hydration waters indicated that the dynamics of K-ras-bound waters is dramatically different from the bulk behavior . Moreover , the dynamics of several deeply buried long-residence water molecules , such as W4 , is coupled with the local as well as global motions of the protein . As a result , the protein was able to sample different regions of phase space in the presence and absence of some of these waters . Three water molecules , namely , W1 ( located in lobe1 ) , W4 ( at the interface between the lobes ) and W5 ( at lobe2 ) facilitate the relative motion of residues located at the two lobes of Q61H K-ras . On the other hand , the fluctuation of less buried waters is only loosely coupled with the global motion of the protein even in cases where their exchange is gated by the fluctuation of functionally important protein segments , such as the S2 loop . Our analysis was facilitated by two simple reaction coordinates: distance ( d ) between the Cα atoms of G60 at S2 and G10 at the P-loop and the N-Cα-C-O dihedral ( ξ ) of G60 . A d/ξ scatter plot allowed us to classify the available Ras x-ray structures into active GTP-bound , intermediate GTP-bound , inactive GDP-bound , and nucleotide-free states . We emphasize that the structures classified here as intermediates are deficient in effector binding and contain mutations or are derived from complexes with GEFs . Projection of the MD-derived conformers on the d/ξ space derived from the x-ray structures enabled us to associate functionally distinct conformational states with the presence and absence of conserved high-residence water molecules . Moreover , we found a water-mediated correlated motion involving S2 , P-loop and H3 . This motion is mediated by specific residues that received little attention in previous studies . Notably , the reorientation of Y96 side chain away from S2 in GDP-like structures leads to increased solvation of G10 at the P-loop . The presence and absence of this interaction led to differences in the conformation of S2 , suggesting a water-mediated modulation of S2 by H3 . Thus , water molecules act as allosteric ligands to induce a population shift in the conformational states of the canonical switches .
53 x-ray structures ( 65 chains ) of Ras were downloaded from the Protein Data Bank ( PDB [72] ) and analyzed for their water content , focusing only on waters within the first hydration shell and interacting with the protein backbone , defined by a distance cutoff of 3 . 5 Å between water oxygen and backbone oxygen or nitrogen atoms . Analysis of the fraction of crystal structures containing one or more water molecules bound to a given protein site identified two particularly interesting waters , W3 and W4 ( see Results and Discussion ) . These waters were selected for a more detailed analysis by MD . The 2 . 27 Å resolution crystal structure of GTP-bound Q61H K-ras ( PDB id: 3GFT ) was used to perform seven sets of multi-copy MD simulations ( Table 1 ) . An oncogenic mutant of H-ras G12V has been shown to sample a wide range of conformational space [45] , [46] . We therefore reasoned that the oncogenic Q61H K-ras , the only available K-ras structure , may also sample a large conformational space during classical MD simulation . In the three simulations named F ( i . e . , free ) , all crystal waters were kept and no restraints were applied . In simulations and , restraints were applied ( force constant k = 10 kcal/mol/Å2 ) on the Cα atoms of switch 2 ( S2: residues 57–75 ) or part of helix 3 ( H3: residues 87–95 ) . Using the same force constant , conserved waters W3 and W4 were positionally restrained in simulations and , their partial charges were removed in simulations and they were excluded at the start of the simulations ( Table 1 ) . In each case , the system setup involved assignment of charges assuming neutral pH ( D , E and C-terminus de-protonated and K , R and N-terminus protonated ) . The protein was solvated in a cubic box of side 60 . 4 Å containing TIP3 waters , allowing a minimum of 10 Å distance between the edge of the box and the protein . The system was neutralized by adding 12 Na+ ions , and an additional 30 Na+ and 30 Cl− ions were added to achieve a 150 mM ionic strength . Energy minimization was then carried out for 2000 steps with the protein heavy atoms fixed and for another 5000 steps with all atoms set free . During the initial 200 ps equilibration , a harmonic restraint of k = 4 kcal/mol/Å2 was applied on the Cα atoms , which was then progressively reduced by 1 kcal/mol/Å2 every 100 ps . All equilibration steps used a 1 fs time step , which was increased to 2 fs in the production phase with SHAKE [73] applied to covalent bonds involving hydrogens . The temperature of the system was maintained at the physiological value of 310 K using Langevin dynamics with a damping coefficient of 2 ps−1 . The Nose-Hoover Langevin piston method was used to maintain constant pressure at 1atm with a piston period of 100 fs and decay time of 50 fs . The short-range van der Waals interactions were switched off gradually between 8 . 5 and 10 Å with a 12 Å cutoff used for non-bonded list updates . Long-range electrostatic interactions were calculated using the Particle Mesh Ewald ( PME ) method [74] . Each simulation was run for either 60 or 100 ns , yielding an aggregate simulation time of 1 . 76 µs . All simulations were performed with the NAMD program [75] and the CHARMM27 force field [76] . For the purpose of this work , hydration waters are defined as waters whose oxygen atom is within 3 . 5 Å of any non-hydrogen protein atom .
|
K-ras belongs to the Ras family of G-proteins that regulate cell proliferation and development . To execute its function , K-ras adopts different conformational states when it is active and inactive . In addition to these two states , it samples many transient intermediate conformations as it makes the transition from one state to the other . Mutations that affect the population of these states can cause cancer or developmental disorder . Using simulation approaches , here we show that a number of water molecules buried within the structure of an oncogenic K-ras protein modulate the distribution of its conformational states . Moreover , a detailed analysis based on two novel structural parameters revealed the existence of long-range water-mediated interactions that facilitate a dynamic coupling between the two lobes of the protein . These findings pave the way for a dynamics-guided strategy to inhibit abnormal Ras signaling .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Methods"
] |
[
"biology",
"computational",
"biology",
"biophysics"
] |
2012
|
The Role of Conserved Waters in Conformational Transitions of Q61H K-ras
|
Clonorchiasis is prevalent in the Far East , and a major health problem in endemic areas . Infected persons may experience , if not treated , serious complications such as bile stone formation , pyogenic cholangitis , and even cholangiocarcinoma . Early diagnosis and treatment are important to prevent serious complications and , therefore , the simple and reliable diagnostic method is necessary to control clonorchiasis in endemic areas , where resources for the diagnosis are limited . The loop-mediated isothermal amplification ( LAMP ) assay has been applied for the detection of Clonorchis sinensis DNA . Six primers targeting eight locations on the cytochrome c oxidase subunit 1 gene of C . sinensis were designed for species-specific amplification using the LAMP assay . The LAMP assay was sensitive enough to detect as little as 100 fg of C . sinensis genomic DNA and the detection limit in 100 mg of stool was as low as one egg . The assay was highly specific because no cross-reactivity was observed with the DNA of other helminths , protozoa or Escherichia coli . Then , LAMP assay was applied to human fecal samples collected from an endemic area of clonorchiasis in Korea . Using samples showing consistent results by both Kato-Katz method and real-time PCR as reference standards , the LAMP assay showed 97 . 1% ( 95% CI , 90 . 1–99 . 2 ) of sensitivity and 100% ( 95% CI , 92 . 9–100 ) of specificity . In stool samples with more than 100 eggs per gram of feces , the sensitivity achieved 100% . To detect C . sinensis in human fecal samples , the LAMP assay was applied and achieved high sensitivity and specificity . The LAMP assay can be utilized in field laboratories as a powerful tool for diagnosis and epidemiological survey of clonorchiasis .
Clonorchiasis is an important human parasitic infection and is highly prevalent in eastern Asian countries , including China , Korea , and Vietnam [1 , 2] . In Korea , the Clonorchis sinensis egg positive rate in the general population is 1 . 9% , and approximately 1 million people are estimated to be infected [3] . However , in some endemic provinces and river basins , the infection rate is reported to be more than 10% [4] . Most infected people have no symptoms , but chronic infection induces some clinical manifestations , including abdominal or epigastric discomfort , fatigue , jaundice , vomiting , fever , and diarrhea [5] . The most important and serious complication of C . sinensis infection is cholangiocarcinoma , and the parasite has been classified as a group 1 biological carcinogen [6] . The specific diagnosis of C . sinensis is important for successful treatment and control of the infection . The Kato-Katz ( KK ) method and/or formalin-ether concentration technique are commonly used for clonorchiasis diagnosis [7] . However , stool examinations are not highly effective because lightly infected cases can be missed [7] . Moreover , due to morphological similarities , the eggs of C . sinensis are easily confused with the eggs of other flukes ( e . g . , Heterophyidae or Opisthorchiidae ) . For this reason , specific diagnosis of C . sinensis eggs in the feces is sometimes difficult by KK method especially for the lightly infected cases [1] . ELISAs for serodiagnosis of clonorchiasis are widely used , but they cannot differentiate between past and current infections [8 , 9] . Recently , a sensitive , specific ELISA has been developed to detect C . sinensis antigens directly from stool samples , but its application in the field has not yet been evaluated [10] . Several PCR assays have been developed to detect C . sinensis DNA in stools , but the sensitivities and specificities of these PCR assays vary depending on their target genes [11–16] . Moreover , these PCR methods require sophisticated equipment , such as a thermal cycler , which has prevented the widespread use of these techniques in low resource areas . An alternative DNA amplification technique known as loop-mediated isothermal amplification ( LAMP ) has been developed [17] . The LAMP assay can be performed under isothermal conditions ( 60°C to 65°C ) ; therefore , a simple water bath or block heater is sufficient to amplify the specific DNA [17] . Moreover , the assay allows visual detection of DNA amplification through the addition of fluorescent dyes [18] . It has several advantages over conventional PCRs such as higher sensitivity and specificity , rapid and simple procedures , and is a good candidate approach to detect many pathogens including parasites in field conditions , without the use of expensive equipment [19] . Shorter reaction time with visual judgment of positivity without requiring sophisticated equipments makes it an attractive diagnostic method for field application . The LAMP assay has been applied successfully to detect various parasitic infections , including opisthorchiasis [20–22] , schistosomiasis [23] , paragonimiasis [24] , fascioliasis [25] , and taeniasis [26] . Recently , a LAMP technique has been developed to detect C . sinensis DNA in freshwater snails [27] and fish [19] , as a tool of control and prevention of clonorchiasis in endemic areas; however , there is still no report of C . sinensis DNA detection in human fecal samples . In the present study , a highly sensitive and specific LAMP assay has been developed to detect C . sinensis DNA in human stool samples . The LAMP assay was evaluated using the fecal samples collected from a clonorchiasis endemic area of Korea and compared with the combined results of KK method and real-time PCR .
The design of this study was reviewed by the institutional review board of Seoul National University Hospital ( IRB approval number E-1512-075-727 ) . Institutional review of this study was waivered because this study used anonymized stool samples that were randomly selected from the pool of stool samples of the residents of an endemic area of clonorchiasis in Korea , which had been obtained from the previous studies [28 , 29] . The animal experiment was reviewed and approved by the Institutional Animal Care and Use Committee ( IACUC ) of Seoul National University , Seoul , Korea and followed the National Institutes of Health ( NIH ) guideline for the care and use of laboratory animals ( ISBN 0-309-05377-3 ) . Metacercariae of C . sinensis were collected from naturally infected freshwater fishes in Korea . Adult worms were recovered from the bile ducts of Sprague-Dawley rats after 2 months of infection with metacercariae . DNA was extracted from adult worms and used as a positive control for the LAMP assay . For analysis of the LAMP assay , stool samples were randomly selected from the pool of stool samples of the residents of Sancheong county in Korea , where clonorchiasis is endemic , and risk factors and incidence of cholangiocarcinoma among this resident were investigated since 2006 [28 , 29] . For each stool sample , two KK smears and one real-time PCR were performed . For the KK smear , 41 . 7 mg of feces was examined by microscopy and multiplied by 24 to convert to eggs per gram of feces ( EPG ) [7] . Specific diagnosis of C . sinensis eggs in the feces is sometimes difficult by KK method due to the morphological similarities of eggs of other flukes [1] . To confirm whether the infection is due to the C . sinensis , in addition to KK , our laboratory-developed sensitive and specific real-time PCR was performed according to procedures previously described [16] . The real-time PCR reaction was performed in 4 μl of extracted stool DNA in a total volume of 25 μl . Then , stool samples positive by both method were considered as a positive reference sample and negative samples by both method were considered as a negative reference sample . Based on the results of two KK smears and one real-time PCR for each stool sample , 70 C . sinensis-positive and 50 C . sinensis-negative samples were selected . The selected samples were from 120 subjects ( age range , 31–80 years; median age , 60 . 5 years; 71 males ) . By evaluating these samples with the LAMP assay , the diagnostic accuracy of the assay was compared with the combined results of KK method and real-time PCR . The genomic DNA from adult C . sinensis and other parasites was extracted using QIAamp Tissue Kit ( QIAgen ) following the manufacturer’s instruction . DNA from stool samples was extracted using the procedures previously described [30] . Briefly , 100 mg stool was washed twice with 1 ml PBS and centrifuged at 8000 rpm for 5 min . After centrifugation , the pellet was resuspended in 200 μl of a 2% polyvinylpolypyrolidone ( Sigma , St . Louis , MO , USA ) solution and then heated in a heat block at 100°C for 10 min . After treatment with sodium dodecyl sulfate–proteinase K for 2 hr at 55°C , the DNA was eluted by adding 100 μl elution buffer through QIAamp Tissue Kit spin columns ( Qiagen , Hilden , Germany ) according to the manufacturer’s instructions . Sets of forward and backward external primers ( F3 and B3 ) , forward and backward internal primers ( FIP and BIP ) and forward and backward loop primers ( LF and LB ) were designed based on the sequence of the cytochrome c oxidase subunit 1 ( cox1 ) gene of C . sinensis ( GenBank accession no . AF181889 ) . The LAMP primers were designed using the software ‘PrimerExplorer V4’ ( http://primerexplorer . jp/e ) . The list of primers and their locations in cox1 gene are shown in Fig 1 . The LAMP assay was performed in a total reaction mixture volume of 25 μl , containing 2 . 5 μl of 10x ThermoPol reaction buffer ( New England BioLabs , Ipswich , MA , USA ) , 6 mM MgSO4 , 1 M betaine , 1 . 4 mM dNTP mix , 0 . 2 μM each of F3 and B3 primers , 1 . 6 μM each of FIP and BIP primers , 0 . 8 μM each of LF and LB primers , 8 U Bst DNA polymerase ( New England BioLabs ) and 4 μl of the template DNA . The reaction mixture was incubated at 64°C for 60 min in a heat block and followed by incubation at 82°C for 2 min to terminate the reaction . Amplified LAMP products were detected by adding 1 . 0 μl of 1:10 diluted 10 , 000x concentration of SYBR Green I ( Invitrogen , Carlsbad , CA , USA ) to each tube . The amplicon was observed directly either by the naked eye or by placing the reaction tube under UV light ( Gel documentation system , UVItech , Cambridge , UK ) . In addition , 5 . 0 μl of the LAMP products was examined by electrophoresis on a 2% agarose gel , followed by ethidium bromide staining and visualization under UV light . Sensitivity of the assay was determined by amplifying 10-fold serial dilutions of genomic DNA of the C . sinensis adult worm from 1 ng to 1 fg . Also , egg-negative feces were experimentally spiked with 10 , 000 , 1 , 000 , 100 , 10 and 1 egg ( s ) of C . sinensis . Eggs were collected by tearing the uterus of adult C . sinensis under a stereomicroscope . Numbers of eggs were counted and 10-fold serial dilutions were made from 10 , 000 to 100 eggs . For the accuracy of the assay , 10 and 1 egg ( s ) were collected under a stereomicroscope using fine tip glass Pasteur pipette and spiked in the negative feces . DNA was extracted from each spiked feces and amplified by the LAMP assay for determination of the minimum detectable number of C . sinensis eggs in feces . Specificity of the LAMP assay was evaluated using DNA isolated from trematodes ( Metagonimus yokogawai , Opisthorchis viverrini and Fasciola gigantica ) , cestodes ( Spirometra erinacei and Diphyllobothrium latum ) , nematodes ( Ascaris lumbricoides , Ascaris suum , Necator americanus and Trichuris trichiura ) , protozoa ( Cryptosporidium parvum , Entamoeba histolytica and Giardia lamblia ) and Escherichia coli . For each reaction , the same amount of DNA , 1 ng , was utilized . PCR was performed with two LAMP external primers ( F3 and B3 ) to compare the sensitivity to that of LAMP and to confirm that LAMP correctly amplified the target . The PCR reaction was conducted in a 25 μl reaction mixture containing 2 . 5 μl of 10x PCR buffer , 0 . 2 mM dNTP mix , 0 . 4 μM each of F3 and B3 primers , 1 . 5 U Ex Taq polymerase ( Takara , Otsu , Shiga , Japan ) and 4 μl of the template DNA . Ten-fold serial dilutions of C . sinensis genomic DNA , starting from 1 ng down to 1 fg , were used as PCR templates .
LAMP products were visually detected by the naked eye after adding SYBR Green I dye to each reaction tube . The color of the reaction solution was green in the presence of C . sinensis DNA ( positive LAMP reaction ) ; otherwise , in the absence of C . sinensis DNA , it remained orange ( negative reaction ) ( Fig 2A ) . The LAMP products were also visualized by placing the reaction tube under UV light ( Fig 2B ) . Upon gel electrophoresis , the LAMP products were observed as typical ladder-like bands ( Fig 2C ) . When PCR was performed using C . sinensis genomic DNA with F3 and B3 primers , an expected PCR product of 210 bp was obtained ( Fig 2D ) . To determine the detection limit of the LAMP assay , a 10-fold serial dilution of genomic DNA of the C . sinensis adult worm was amplified by LAMP . The assay detected as little as 100 fg C . sinensis genomic DNA ( Fig 2A , 2B and 2C ) , whereas PCR amplified as little as 10 pg C . sinensis genomic DNA ( Fig 2D ) . The assay also detected DNA from feces experimentally spiked with a series of known numbers of C . sinensis eggs; the minimum detectable number of eggs was one in 100 mg of feces ( Fig 3 ) . However , the LAMP assay did not amplify DNA from other helminths , protozoa or E . coli ( Fig 4 ) . The diagnostic efficiency of the LAMP assay was evaluated using 120 stool samples , including 70 positive and 50 negative samples , confirmed by both KK and real-time PCR . The sensitivity and specificity of the LAMP assay relative to the combined result of KK and real-time PCR was 97 . 1% ( 68/70 ) and 100% ( 50/50 ) , respectively ( Table 1 and S1 File ) . When the LAMP results were compared with the KK results , the LAMP assay could detect C . sinensis DNA in 41 ( 100% ) of 41 stool samples with EPGs more than 100 . However , the LAMP assay detected C . sinensis DNA in 27 ( 93 . 1% ) of 29 stool samples with EPGs less than 100 ( Table 2 ) . The lowest EPG detected by the LAMP assay was 24 . The two samples that were positive using KK and real-time PCR , but negative by the LAMP assay , had an EPG of 12 .
Despite the development of various sensitive and specific molecular tools , clonorchiasis is still diagnosed by stool examinations [1] . However , the microscopic approach can miss lightly-infected cases , and differential diagnosis with other minute intestinal flukes is often difficult and requires expertise for accurate diagnosis [5] . Because extremely low burden cases are more prevalent in endemic areas in Korea , where extensive control measures are executed by health care workers , it is still a challenge to diagnose cases of clonorchiasis [1] . To advance more useful clonorchiasis diagnostic approaches , we have developed a LAMP assay and evaluated its diagnostic efficacy in human fecal samples . This is the first report on the detection of C . sinensis DNA in human fecal samples using the LAMP assay . In this study , the LAMP assay was highly sensitive and detected as little as 100 fg of C . sinensis genomic DNA , which is 100 times more sensitive than the PCR performed using the external LAMP primers . The lower detection limit of LAMP assay than that of conventional PCR is consistent with other studies with C . sinensis [19 , 27] . The high sensitivity comes from its ability to detect as few as six copies of DNA in the reaction mixture [17] . Another reason for this high sensitivity is that LAMP primers have been designed based on the mitochondrial gene ( cox1 ) of C . sinensis , which is present in hundreds to thousands of copies per cell [31] . LAMP assay targeting the mitochondrial gene ( NADH dehydrogenase subunit 1 , nad1 ) of O . viverrini was also highly sensitive to detect as little as 100 fg of genomic DNA , which is the same as our developed assay for C . sinensis [21] but another study targeting ribosomal DNA ( internal transcribed spacer , ITS1 ) of O . viverrini showed less sensitivity with a detection limit around 1 pg DNA/μl [22] . In spite of high sensitivity with genomic DNA , human stool samples are highly challenging for DNA amplification . Even after the DNA extraction , remaining contaminants and the presence of DNA polymerase inhibitors in fecal constituents can inhibit DNA polymerase [32] . The inhibition is considered to be associated with false-negative results of stool samples evaluated by PCR [15 , 33] . Besides the inhibitors , stool samples contain egg stage only , thus successful disruption of the egg shell is required to give access to DNA [12] . Although LAMP assays to detect C . sinensis were already applied to various samples such as adult worms , metacercariae , infected fish muscle [19] and freshwater snails [27] , it is noteworthy that we achieved successful application of LAMP assay for human stool samples in the present study . Despite challenges mentioned above , our LAMP assay on stool samples spiked with known number of eggs revealed the minimum detection limit as low as one egg in 100 mg stool that corresponds to 10 EPGs , and this is the lowest detection limit that can be achieved in our experiment settings with 100 mg of feces . It is important to lower minimum detection limit down to 12 EPGs because the presence of one egg in 2 KK examinations will have a result of 12 EPGs . With low enough detection limit , our LAMP assay could achieve sensitivity as high as 97 . 1% , but not 100% due to the two false-negatives in the stool samples with an EPG of 12 . The inability of the LAMP assay to detect these two stool samples might be due to the presence of low numbers of eggs . Eggs , if present in low number , may not be homogenously distributed in the stool and therefore , a sampling of the stool for DNA extraction might have missed the eggs leading to the false negative result . Less detection sensitivity in the stool samples with low infection intensity is in agreement with a previous study , in which real-time PCR for C . sinensis failed to detect 6 out of 70 samples with an EPG below 100 [14] . Due to limited amount of stools utilized for LAMP assay and the stochastic nature of egg distribution in a fecal sample , more experiments would be needed to determine the true detection limit of the LAMP assay in comparison with KK and real-time PCR . Although there is no previous report on LAMP method for detection of C . sinensis from stool samples , there are few reports for detection of O . viverrini , and only one had evaluated its sensitivity compared with KK method [22] . The sensitivity of the LAMP assay was 100% for detection of O . viverrini DNA in egg positive stool samples . The amount of stool prepared for LAMP assay was twice as much as the amount we prepared for , and the less amount utilized in this study could led to less sensitivity in low infection intensity . The false negative result that LAMP showed in this study , is also expected for KK methods in such a low infection intensity where eggs of similar morphology from other intestinal flukes may exist . However , due to high specificity , the LAMP assay would be able to detect specifically C . sinensis from such samples . Further diagnostic performance of the test will be followed using unknown samples to optimize the assay for application in the real field condition . The less sensitivity is tolerable because not all clonorchiasis cases especially the low worm burden cases would proceed to cholangiocarcinoma if untreated [34] . However , it is still important to identify and treat such low burden cases of clonorchiasis in those countries where the elimination program is in close proximity . Because patients with low EPG does not always represent the real situation of worm burden as they may harbor juvenile worms that might not have yet started laying eggs [35] . Application of this robust and highly sensitive LAMP assay for detection of such lightly infected cases in endemic countries would be really useful to make the controlling program success . The sensitivity can be influenced by the volume of buffer used for final elution and the amount of template DNA added . Another study with stool samples utilized 50 μl of elution buffer on 200 mg of stool while this study used 100 μl of elution buffer on 100 mg of stool [22] . More densely eluted samples in the study can be one of the reasons for 100% sensitivity even in stool samples with less than 100 EPGs [22] in comparison with 93 . 1% sensitivity in the corresponding stool samples eluted by recommended methods in this study . Less elution buffer ( 50 μl ) than the amount recommended by the manufacturer ( 200 μl ) can increase the density of template DNA to help reaction [22] , but , at the same time , can decrease the elution efficiency resulting in less sensitivity . Instead we increased the amount of template DNA to 4 μl to help the reaction . Prior to application to field samples , this kind of adjustment should be optimized using stool samples spiked by known number of eggs . In addition , the choice of target genes can also influence sensitivity . As the amplification is sequence specific , the target genes should be conserved throughout the species . According to the investigation on DNA variations among three isolates of C . sinensis from Korea and China , few intraspecific nucleotide substitutions were found in 18S , ITS1 , ITS2 and cox1 sequences [36] . Among them , the nucleotide gap ( insertion , deletion ) differences were slightly larger for ITS1 [36] , which showed less sensitivity than IST2 in PCR assays [15] . In this study , we designed the LAMP assay to target cox1 gene , which shows very low level of intraspecific variation . The specificity of LAMP is generally high because it uses four primers that recognize six locations on the target DNA [17] . In our study , in addition to four primers ( two externals and two internals ) , two additional loop primers were included , resulting in recognition of a total of eight locations on the cox1 gene of C . sinensis . Thus , the LAMP assay in the present study was highly specific and did not cross-react with the DNA of other helminths , protozoa or E . coli including the closely related liver fluke O . viverrini . Hence , the assay must be robust enough to specifically detect C . sinensis infection in areas where mixed parasite infections occur . Despite inherent specificity of LAMP assay , the choice of target gene can influence the specificity as well . PCR assays targeting ITS regions of C . sinensis showed cross-reaction with the DNA of other liver flukes such as O . viverrini [12 , 15] or Opisthorchis felineus [14] . PCR assays targeting other genes such as cox1 and nad1 have been developed for differential diagnosis of C . sinensis and O . viverrini [11 , 13] . LAMP assays also encountered cross-reaction between O . viverrini and O . felineus , the researchers changed the target from ITS1 to microsatellites of O . viverrini and achieved the specificity [20] . Our target , cox1 gene , did not cross-react with the DNA of other liver fluke O . viverrini . The LAMP assay has several advantages in detecting C . sinensis DNA from stool samples compared to our previously developed real-time PCR assay [16] and commonly used conventional PCR . As the Bst DNA polymerase utilized for the LAMP assay is considered to be more resistant to the inhibitors [37] , the LAMP assay have superiority for analyzing stool samples . In the aspect of efficiency , it can amplify large quantities of the target DNA under isothermal conditions with less time for DNA amplification [17] . Compared to the commonly used conventional PCR method , LAMP saves minimum 2 hours as it does not require thermocycler and gel electrophoresis . The cost of the LAMP assay is relatively less because it does not require sophisticated instruments like thermocycler for amplification of target gene , gel electrophoresis unit and gel documentation system for detection of amplified PCR products . A heat block or water bath is all that is required to perform the entire reaction successfully . Moreover , visual detection of the LAMP reaction is possible by simply adding fluorescent dyes , such as SYBR Green I , to the tube [18] . These features make the LAMP assay suitable for application to field laboratories in clonorchiasis endemic areas . Although the reaction of LAMP assay has been evaluated by naked eye with SYBR Green I , visualization under UV light , and agarose gel electrophoresis; in field condition visual judgment by naked eye with fluorescent dyes would be the appropriate approach of detection because it is achieved by simply adding dyes and does not require additional equipment or complicated procedures as we have demonstrated . Longer time to read results and the requirement of equipment and materials for DNA extraction would prevent the use of LAMP assay at this level of development as a point-of-care diagnostic in low resource clinical settings . To enhance field applicability , more simple and user-friendly DNA preparations and a formulated ready-to-use reaction mixture should be available . Recently , a high throughput LAMP detection system has been developed for the diagnosis of malaria parasites which takes less than 2 hours from DNA extraction to reading results [38] . The system consists of a portable box which allows parallel processing of a large number of samples for DNA extraction and detection in such a way without the need of pipetting and centrifugation . In the future , a similar approach should be incorporated with our currently developed LAMP assay to promote the assay in poorly equipped laboratories . In conclusion , we report a highly sensitive and specific LAMP assay for detection of C . sinensis DNA in human fecal samples . Due to the shorter reaction time and better visual judgment of positivity without requiring sophisticated instruments , the LAMP assay can be more easily applied in field laboratories than PCR as a powerful tool for more specific and reliable diagnosis of clonorchiasis , thereby improving both treatment and control programs .
|
Clonorchiasis , a parasitic infection by Clonorchis sinensis , is one of the major public health problems in eastern Asian countries , including China , Korea , and Vietnam , causing cholangiocarcinoma if not treated . The definitive diagnosis is important for successful treatment and prevention of the infection in endemic areas . The fecal examination is a standard diagnostic method and mainly based on microscopic observation of eggs in feces . However , they sometimes miss lightly infected cases and can lead to misidentification at the species level because of morphological similarity between the eggs of liver flukes and minute intestinal flukes . To overcome the diagnostic limitations of stool examination , ELISAs and PCRs have been developed as a sensitive and specific diagnostic method , but they require sophisticated equipment such as thermal cycler and spectrophotometry , which has prevented the widespread use of these techniques in low resource areas . We developed a LAMP assay targeting the cytochrome c oxidase subunit 1 gene of C . sinensis for rapid and simple detection of C . sinensis DNA in human fecal samples . The LAMP assay was highly specific and sensitive enough to detect as little as 100 fg of C . sinensis genomic DNA and to detect a single egg of C . sinensis in 100 mg of stool samples . Then , LAMP assay was applied to human fecal samples collected from an endemic area of clonorchiasis in Korea . Using samples showing consistent results by both Kato-Katz method and real-time PCR as reference standards , the LAMP assay showed 97 . 1% ( 95% CI , 90 . 1–99 . 2 ) of sensitivity and 100% ( 95% CI , 92 . 9–100 ) of specificity . In stool samples with more than 100 eggs per gram of feces , the sensitivity achieved 100% . Due to the sensitive and specific detection of C . sinensis DNA in fecal samples , the LAMP assay can be applied in field laboratories as a powerful tool for diagnosis and epidemiological survey of clonorchiasis .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] |
[
"invertebrates",
"medicine",
"and",
"health",
"sciences",
"helminths",
"tropical",
"diseases",
"geographical",
"locations",
"parasitic",
"diseases",
"animals",
"invertebrate",
"genomics",
"trematodes",
"organisms",
"clonorchis",
"sinensis",
"foodborne",
"trematodiases",
"neglected",
"tropical",
"diseases",
"molecular",
"biology",
"techniques",
"extraction",
"techniques",
"research",
"and",
"analysis",
"methods",
"korea",
"artificial",
"gene",
"amplification",
"and",
"extension",
"flatworms",
"clonorchiasis",
"molecular",
"biology",
"animal",
"genomics",
"people",
"and",
"places",
"helminth",
"infections",
"dna",
"extraction",
"eukaryota",
"asia",
"clonorchis",
"genetics",
"biology",
"and",
"life",
"sciences",
"genomics",
"gene",
"amplification",
"polymerase",
"chain",
"reaction"
] |
2017
|
Application of a loop-mediated isothermal amplification (LAMP) assay targeting cox1 gene for the detection of Clonorchis sinensis in human fecal samples
|
Malaria parasites elude eradication attempts both within the human host and across nations . At the individual level , parasites evade the host immune responses through antigenic variation . At the global level , parasites escape drug pressure through single nucleotide variants and gene copy amplification events conferring drug resistance . Despite their importance to global health , the rates at which these genomic alterations emerge have not been determined . We studied the complete genomes of different Plasmodium falciparum clones that had been propagated asexually over one year in the presence and absence of drug pressure . A combination of whole-genome microarray analysis and next-generation deep resequencing ( totaling 14 terabases ) revealed a stable core genome with only 38 novel single nucleotide variants appearing in seventeen evolved clones ( avg . 5 . 4 per clone ) . In clones exposed to atovaquone , we found cytochrome b mutations as well as an amplification event encompassing the P . falciparum multidrug resistance associated protein ( mrp1 ) on chromosome 1 . We observed 18 large-scale ( >1 kb on average ) deletions of telomere-proximal regions encoding multigene families , involved in immune evasion ( 9 . 5×10−6 structural variants per base pair per generation ) . Six of these deletions were associated with chromosomal crossovers generated during mitosis . We found only minor differences in rates between genetically distinct strains and between parasites cultured in the presence or absence of drug . Using these derived mutation rates for P . falciparum ( 1 . 0–9 . 7×10−9 mutations per base pair per generation ) , we can now model the frequency at which drug or immune resistance alleles will emerge under a well-defined set of assumptions . Further , the detection of mitotic recombination events in var gene families illustrates how multigene families can arise and change over time in P . falciparum . These results will help improve our understanding of how P . falciparum evolves to evade control efforts within both the individual hosts and large populations .
Although the global burden of malaria has declined over the last few years to 216 million cases and 655 , 000 deaths in 2010 [1] , the overall goal of global eradication is still out of reach . Emerging resistance to artemisinin , a frontline chemotherapeutic for which resistance is not widespread , has recently been reported along the Thai-Cambodia border ( reviewed in [2] ) . Furthermore , RTS , S , the most advanced vaccine candidate in development , is only minimally effective and does not induce long-lived sterile immunity [3] . A primary reason why malaria is difficult to control is its genome's ability to recombine and/or mutate away from a protective immune response or drug pressure . For example , the development of an effective vaccine has been hampered by the prevalence of strain-specific immunity , where vaccination with one antigenic haplotype protects for only one specific variant [4] . To date , this has been attributed to pre-existing genetic diversity; however , it may also be that escape mutants emerge in vaccinated individuals . Plasticity of the Plasmodium genome can also contribute to the evolution of resistance against anti-malarial drugs . Single nucleotide variants ( SNVs ) and copy number variants ( CNVs ) in target and resistance genes allow the parasites to evade drug pressure . Most notably , the emergence of chloroquine-resistant parasites ultimately caused a huge resurgence in the number of malaria cases in the 1990s . Although these two mechanisms are well described , it is not understood how often variation arises during mitotic asexual growth or how quickly SNVs accumulate in the absence of selection pressure . In addition to diversity at the population level , there is also variability within the individual parasite . Multigene families , where only one or few members are expressed , provide antigenetic diversity and allow the parasite to persist in a host . Recombination events which occur in meiosis [5] , [6] as well as mitosis [7] give rise to new variants in these already diverse families . This genetic variability in parasites , both in an individual host and on a population level , allows the parasite to evade the host immune system even in the absence of transmission ( i . e . during dry seasons ) . Given this remarkable genetic diversity , it is not surprising that naturally-infected patients often carry multiple , genetically distinct parasite clones . The multiplicity of infection ( MOI ) has traditionally been estimated with a handful of genetic markers , which may encode proteins under strong selection by the host immune system . However , these methods are not comprehensive enough to measure true parasite heterogeneity and many variants are missed within an individual [8]–[10] . It is unclear whether parasite heterogeneity is created through multiple infectious mosquito bites , heterogeneity in a single mosquito inoculation , or evolution of new genetic changes ( SNVs and structural variants ) within the human host . Knowing the rate at which genetic changes occur is critical to understanding the emergence of drug resistance , the evolution of antigen polymorphisms and multigene families , and the patterns of malaria transmission . It is not possible to study neutral parasite mutation rates in humans due to the influence of selection pressure of the host immune response and genetic host-to-host variability . Previous quantifications of the mutation rate have focused only on single genes under drug selection [11] . The basal rate at which mutations drive Plasmodium evolution has therefore never been measured at the whole-genome level and much must be inferred . Using whole-genome sequencing as well as whole-genome tiling arrays , we have determined the rates at which genetic changes occur in clonal parasite populations in the absence of immune and drug pressure in vitro . In addition to the accumulation of SNVs in the core genome , we observed major deletions in the subtelomeric regions and identified seven mitotic recombination events . The rates of these events were not changed by the addition of atovaquone , a commonly used anti-malarial drug .
To study the natural genomic plasticity of P . falciparum , we investigated how the parasite genome changes over time . A single clonal parent was split into six lines . To investigate the effect of selection pressure on the mutation rate , five parasite lines were exposed to atovaquone ( ATQ ) , a hydroxy-1 , 4-naphthoquinone that targets the mitochondria-encoded cytochrome bc 1 ( CYTbc 1 ) complex of the electron transport chain of Plasmodium parasites [12] . ATQ is a component of Malarone , a traveler's medicine drug combination currently used to prevent or treat malaria . Resistance mutations are known to arise quickly [13] . Five lines were exposed to various concentrations of ATQ ( R1 , R2 and R3: 2 nM , R4: 20 nM , R5: 50 nM ) and one line was cultured without drug pressure ( S1 ) for up to a year ( Figure 1A ) . For each line , two clones were selected ( three clones for the drug-free line ) . The four clones of lines R1 and R2 were retained in culture and cloned again ( R1a/b and R2a/b , generation 1 and 2 ) . Growth inhibition dose-response assays confirmed that ATQ-resistant clones had indeed acquired at least a 9-fold increase in EC50 values for ATQ compared to the 3D7 parent ( Figure 1B ) . Hence , we were able to evolve drug-resistant parasite clones that are 10-fold more tolerant to ATQ than their parent . To facilitate the analysis of different parasites lines on a whole-genome level , all lines were cloned by limiting dilution before being expanded to isolate DNA for further analysis and were as genetically homogenous as could be expected after this process . Next we identified the number of genetic changes by comparing the genomes of all seventeen clones to the original 3D7 parental clone using comparative genomic hybridization analysis . The whole-genome tiling microarray we used covers 76% of the coding regions and 41% of the non-coding regions and has a SNV discovery rate of 91% and a false discovery rate of 11% [14] . The PfGenominator software [14] was used to analyze the microarray data and predicted 15 polymorphisms . Capillary sequencing revealed that two of these polymorphisms were deletions , while two were false positives and eliminated from further analysis ( Table S1 ) . As the microarray does not cover the whole genome , the parent and sixteen clones were further analyzed by whole-genome sequencing ( WGS ) using paired-end 60 bp reads with an average 145 bp insert size . On average , 91 . 8% of the P . falciparum genome was covered by five or more reads , with clones having between 73 . 5% and 99 . 9% of the genome covered by fivefold or greater coverage ( Table 1 ) . To assess the clonality of our haploid parasite populations , we calculated the number of positions where a significant amount of nucleotides were different from the most prevalent nucleotide . On average , only 235 positions were detected throughout the whole genome and we thus deemed coverage by five or more high quality reads adequate to call SNVs accurately . Areas with less than fivefold coverage included highly repetitive regions such as the telomere repeats and the flanking telomere associated regions ( TARs ) as well as certain conserved regions within gene families such as the var , rifin , and stevor families . It is therefore possible that some genetic changes in these hard-to-align and poorly annotated regions were not detected . The WGS data was analyzed with the PlaTypUS 0 . 12 software ( M . Manary et al . , manuscript in preparation ) , which integrates many community-developed tools into a pipeline to first align the reads to the reference genome and then detect SNVs . To decide on the characteristics of a true SNV , a computer-learning algorithm was trained on a set of 10 , 500 known SNVs . The WGS data confirmed ten of the fifteen polymorphisms detected by the microarray and identified an additional 25 SNVs across all clones ( Figure 2A and Table S2 ) . To verify that a reasonable cutoff was set to call SNVs in PlaTypUS , six SNVs that did not make the cutoff were analyzed by capillary sequencing ( Table S1 ) . Four events located next to a poly A or poly T stretch and sequencing confirmed that all clones , including the parent , had the same sequence . The other two events were in fact small deletions but were discarded by the PlaTypUS software as it is designed to detect SNVs only . In summary , 38 SNVs were detected by microarray and WGS ( Table S2 ) and the construction of a cladogram with all polymorphisms in the genome ( 38 SNVs , nineteen deletions , and one duplication ) confirmed the known evolutionary relationship between the clones ( Figure 2C ) . To estimate how closely related our 3D7 parent is to the reference 3D7 genome , we compared the WGS data to the 3D7 reference sequence from PlasmoDB v9 . 1 . While on average only 5 . 4 SNVs distinguished a single clone from the parent , we detected 58 SNVs in the parental 3D7 clone relative to the 3D7 reference ( Figure 2A and Table S3 ) . It is not clear whether these differences are true SNVs or due to sequencing/assembly errors from the lower coverage reference genome sequencing efforts [15] , platform differences , or from the different sources of genomic DNA used in the sequencing projects; however , they highlight the importance of having a recent , isogenic reference genome for such comparative studies . We compared the genetic changes acquired in the sensitive clones to the ATQ-resistant clones . As expected , all ATQ-resistant clones acquired SNVs in cytochrome b , the target of the drug , while the sensitive clones did not ( Figure 2 and Figure 3 ) . Mutations were observed at amino acid position 133 ( M133V in R1 pairs and M133I in R2 , R3 , and R4 pairs ) , as well as an L144S change in R4 and an F267V mutation in the clones R5 . In addition , we found a single amplification event on chromosome 1 in the same R5 sister clones ( Figure 3B ) . Interestingly , this ∼220 kb amplified region encompasses PFA0590w ( PF3D7_0112200 ) , which encodes the P . falciparum multidrug resistance associated protein 1 ( PfMRP1 ) , a protein that has been implicated in parasite resistance to chloroquine and quinine but has not been associated with naphthoquinone resistance [16] , [17] . In order to determine if this amplification leads to cross-resistance , we tested our mutants against a series of diverse anti-malarial drugs ( Figure 3C ) . Only the strains that contained the amplification ( R5 ) were cross-resistant to decoquinate , a compound that also targets cytochrome b [18] ( ∼10X , p<0 . 0001 , one way ANOVA followed by a Dunnett posttest ) . In contrast , the R4 clones , which showed the strongest EC50 shift for ATQ compared to the parent , were no more resistant to decoquinate than the parent or the other ATQ-resistant clones . While the presence of ATQ selected for SNVs in cytochrome b , it had little influence on the overall number of genetic changes accumulated in ATQ-resistant clones ( 0 . 02 genetic changes per days in culture ) compared to parasite clones cultured without drug pressure ( 0 . 01 genetic changes per day in culture , p = 0 . 08 , unpaired student t test ) . Our data suggest that external influences such as drug pressure do not appreciably increase the genetic variability of P . falciparum . Having identified the number of isolated genetic changes present in each clone , we calculated the mutation rate per base pair per generation based on the number of mutations detected within each clone and the total time in culture for each clone ( Figure 1 and Figure 2 ) [19] , [20] . To accurately calculate the mutation rate , we needed to include not only the observed SNVs but also those SNVs that were lethal/deleterious , which were unobserved due to loss of mutants from the population . The selection force against nonsynonymous mutations causes the appearance of fewer mutants than actually occurred; thus , using observed mutations alone would cause a downward bias in the mutation rate estimate . To correct for this selection bias , we calculated the dN/dS ratio of P . falciparum to be 0 . 59 , which confirmed the presence of selection ( Materials and Methods ) . This indicates that 40% ( 35 . 5-21 ) /35 . 5 of nonsynonymous mutations are deleterious or lethal . Thus , our observed 21 nonsynonymous exonic mutations most likely came from a population of approximately 35 . 5 true exonic nonsynonymous SNVs . Having accounted for selection , we used a generalized linear model with a linear link function and Poisson distributed error to estimate the mutation rate for each individual clone ( Table S2 ) . We calculated an average mutation rate of 1 . 7×10−9 ( SD: 1 . 2×10−9 ) per base pair per generation for the 3D7 lines in the absence of drug pressure and 4 . 6×10−9 ( SD: 2 . 5×10−9 , student t test alpha >0 . 05 ) for the ATQ-resistant clones . These per base pair per generation mutation rates are comparable to other organisms such as yeast ( 3 . 3×10−10 ) [21] , Drosophila melanogaster ( 8 . 4 10−9 ) [22] , and humans ( 1 . 1 to 2 . 5×10−8 ) [23] , [24] . If we assume mutation rates of 1 . 7×10−9 per base pair per generation and a selection disadvantage against nonsynonymous mutations of 40% , this would result in 0 . 04 mutations per surviving daughter parasite on average . Thus , after 25 generations , every surviving parasite would be expected to have accumulated one mutation relative to its parent . Deletions or amplifications of large chromosomal stretches have been observed in long-term in vitro P . falciparum cultures as well as in field isolates [10] , [25]–[29] , with amplifications typically being associated with drug pressure [30]–[32] . We therefore examined the data for structural variants using both microarray ( manifested as a substantial loss of hybridization for multiple consecutive probes , or an increase in signal for a block of probes ) and WGS ( a lack of aligned reads or an increase in read pileup , Figure 4 ) . All derived clones were compared to our parental 3D7 clone . In addition to the above mentioned single amplification event in two sister clones exposed to ATQ ( R5a and R5b ) , we detected eighteen independent large-scale ( >1000 bp ) deletions of subtelomeric regions ( within ∼60 kb of the telomere ) and one chromosome internal deletion in the seventeen clones analyzed by microarray as well as WGS ( Figure 2 , Figure 4 , and Figure 5 ) . Comparison of our 3D7 parent to the available 3D7 genome from PlasmoDB v9 . 1 revealed a deletion on chromosome 2 in our parental clone . To quantify the appearance of these structural variation events , we calculated a structural variation rate in the same manner as the mutation rates but for the total number of nucleotides deleted or amplified , without the correction for the dN/dS ratio . A structural variation rate of 4 . 7×10−6 per base pair per generation ( SD: 1 . 7×10−6 ) was determined in the absence of drug and 1 . 0×10−5 per base pair per generation ( SD: 9 . 0×10−6 ) in the presence of drug , with the assumption that deletions are fitness neutral . P . falciparum subtelomeric regions encode multigene families that are implicated in antigenic variation . Over 80% of the genes involved in structural variation events were members of multigene families with 11 var , 14 rifin , and 4 stevor genes , showing genetic changes that were so substantial that they could no longer be found by WGS or microarray ( Figure 5 ) . We confirmed that exon 1 from var gene PFB0010c on chromosome 2 was deleted by Southern blot analysis ( Figure 4 ) . The identification of deletions at the telomeres confirmed earlier observations from our laboratory where deletions were also common in field isolates [29] . Structural variants can be the result of a variety of events including double-strand breaks and mitotic recombination . To investigate if the observed deletions in multigene families were associated with mitotic recombination events , we further analyzed the boundaries of the detected structural variants in the same paired-end read WGS data set used for variant discovery . Since read pairs are generated on the same fragment of DNA , they normally map to the same chromosome . However , at the edge of a deletion , we found that reads aligning towards the deletion often had read pairs that aligned to a different chromosome , indicating a likely recombination event . The read coverage at the position of the read mate on the other chromosome was often twice the expected number , suggesting a gene conversion event where DNA from one chromosome is added onto another , thereby doubling the donor sequence , while the original sequence in the recipient chromosome is lost . To test this hypothesis , we extracted the reads that aligned within 1000 bp of each predicted deletion as well as the reads within 1000 bp of their pair mates on a different chromosome . A read from the deletion site was used as a seed to generate a de novo assembly of all these extracted reads . Seven contigs could be assembled that spanned between 1000 and 4000 bp of the sequence next to the deleted region as well as the sequence from a different chromosome indicating a recombination event ( Figure 6 ) . Sanger sequencing of three predicted recombination events confirmed the presence of two new chimeric var genes consisting of PFF0010w ( PF3D7_0600200 ) and MAL13P1 . 1 ( PF3D7_1300100 ) as well as PFA0005w ( PF3D7_0100100 ) and PF10_0001 ( PF3D7_1000100 ) . The third recombination was already present in the parent and involved intergenic regions close to var genes on chromosome 2 and 12 . A previous study also showed gene conversion events between var genes from E5 and 3D7 , two different clones present in the NF54 isolate; short , 100 bp stretches of a 3D7 var gene sequence were found in the context of var gene sequences unique to E5 [33] . In contrast , the gene conversions described here are longer and no mosaic sequences were observed . As the parasite is haploid during the asexual blood stage phase of its cycle , the rearrangements of the genome observed here are results of rearrangements during mitotic growth , while the gene conversions described by Frank et al . could be either due to meiotic or mitotic recombination [33] . Not all of the observed deletions were associated with translocation of DNA from one chromosome to another . Broken telomere ends can also be repaired by the addition of telomere repeats to the broken chromosome arm [34] , [35] . Therefore , we wanted to investigate if deletions induce a general increase of the telomere repeat length ( TRL ) on all chromosomes ( general increase in the average TRL ) or if the telomere repeat size is only increased on broken telomere arms ( presence of two populations of TRL ) . We measured the average TRL by the terminal restriction fragment Southern blot method , where the peak TRL is determined . Of the tested clones , only S1a ( containing three deletions ) had two peaks indicative of a second population of extended telomere repeats ( Figure S1 ) . All other clones had only one peak that was shorter than in the parent , and the average TRL varied between clones from different lines . There was no correlation between the number of deletions and the average TRL of a clone ( Figure S1 ) . In addition , clones R1a and R1b from the first generation had longer average TRL than their offspring in generation 2 that had been retained in culture and subcloned again . This indicates that the average TRL fluctuates over time . This also explains the differences observed between the average TRL in this study and an earlier report for 3D7 and Dd2 [34] . These findings show that deletions at the telomeres , in combination with the fluctuation of the average TRL and mitotic recombination events in the asexual erythrocytic cycle can accelerate the evolution of genetic diversity in gene families exposed to the host immune system . Based on our findings of genetic variation arising in long-term culture , we predicted substantial changes in the genomes of different circulating clones of 3D7 . Three clones from different laboratories were compared to our parental 3D7 by microarray . As expected , all clones contained at least one deletion in the subtelomeric region and only seven small genomic changes in core chromosomal regions were detected in total ( Figure S2 and Table S4 ) . Comparing our parental 3D7 to the different 3D7 clones by microarray confirmed a deletion on chromosome 2 in the parent detected by WGS when compared to the reference genome . To determine if these findings were specific to the 3D7 line , which is susceptible to most drugs , we also analyzed the mutation rate in a P . falciparum Dd2 clone and its eight offspring clones that had been selected for resistance to spiroindolones [30] . Dd2 is a multidrug-resistant clone originally derived from W2 , a multidrug resistant patient isolate from Southeast Asia , and has been reported to acquire drug resistance at higher rates than other strains ( Accelerated Resistance to Multiple Drugs or ARMD [36] ) . Microarray and WGS analysis detected that the core genome was stable in all Dd2 clones regardless of drug pressure ( 29 SNVs in six resistant clones and 15 SNVs in two sensitive control clones ) . Calculation of the mutation rates as described above suggested that Dd2 does not acquire resistance through an intrinsically higher average mutation rate ( 3 . 2×10−9 and 2 . 4×10−9 , in the absence or presence of drug , respectively ) ( Figure S3 and Table S5 ) , although we cannot rule out that some clones in the W2 strain might have this phenotype . We could detect only one deletion in the eight Dd2 clones that involved a var gene , but two clones showed large hybridization differences in subtelomeric regions . The high level of genetic diversity between 3D7 and Dd2 subtelomereric regions makes hybridization analysis with microarray designed for 3D7 , or WGS aligned to 3D7 , more difficult to interpret; deletions and SNVs could be under- or overestimated in Dd2 . Even though the two compounds tested are chemically different and target different pathways , our observations suggest that the mutation rates are similar between different P . falciparum strains with different geographical origins .
Using controlled in vitro conditions , we were able to reproduce genetic changes ( SNVs , deletions in subtelomeric regions and CNVs ) usually observed in field isolates and estimate a mutation as well as a structural variation rate for P . falciparum . The majority of the 38 SNVs ( 39% ) mapped to intergenic regions or uncharacterized conserved proteins ( 21% ) . The appearance of deletions in subtelomeric regions has mainly been attributed to the absence of immune pressure and therefore the lack of counter-selection within in vitro cultures . The majority of the genes located in these regions are members of multigene families such as the 60 var genes that encode different versions of the P . falciparum erythrocyte membrane protein 1 ( PfEMP1s ) . Var genes are further sub-grouped into three major groups ( A , B and C ) based on chromosomal location , orientation and sequence composition of their coding and non-coding upstream regions [37] . Using our new methodology to identify genetic recombination events , we were able to characterize seven recombination events during mitotic growth . Every recombination event observed here occurred either between two group B var genes or two group B var gene upstream regions , suggesting that recombination events within the same group are more frequent . PfEMP1 , an adhesive molecule that is exported to the surface of infected human erythrocytes [38] , is a key pathogenicity protein involved in immune evasion . While this adhesive molecule is obsolete in in vitro cultures , previous studies showed similar structural variants in the subtelomeric regions of Peruvian field isolates indicating that these deletions are not an artifact of in vitro cultures [29] . While we were able to detect structural variants with WGS and microarray analysis , the mechanisms by which they occur remain unclear . The recombination events observed here probably resulted from DNA breaks that were repaired either by direct joining of the broken DNA ends ( resulting in a deletion ) or by homologous recombination with a closely related var gene on a different chromosome ( resulting in a gene conversion ) . While the essential proteins needed for homologous recombination are present in the P . falciparum genome , the key determinants for non-homologous end joining ( NHEJ ) are not [15] . Alternative end joining mechanisms exist in other eukaryotes ( reviewed in [39] ) , though further studies are needed to identify the full repertoire of DNA repair pathways present in P . falciparum . Although we observed deletions in individual clones , it may be that no loss occurs due to DNA translocation to another chromosome in a daughter cell on a population level . Because deletions are much easier to detect than duplications , we might have underestimated the number of the latter . The generation of new chimeric variants of var genes could allow the parasite to further evade the immune system , thereby extending its persistence in the patient . This is especially important for the survival of the parasite in semi-endemic areas where there are no meiotic recombination events because no transmission occurs during the dry season . The generation of new variants in polymorphic genes such as the merozoite surface protein 1 ( msp1 ) [6] or multigene families has largely been attributed to meiotic rather than mitotic recombination . However , our data indicate that ectopic recombination and rearrangement during asexual growth is common between members of antigenic gene families located at the telomeres and could extend to other genes bearing repeated sequence motifs . An interesting feature of var genes is that only one of the several dozen copies is transcribed in any given parasite and transcriptional switching between members occurs [40] , [41] . Because humans develop antibodies against this surface-exposed protein , the parasite is believed to use this mutually exclusive expression mechanism to evade the immune response and persist in its human host . In vitro switching between members of the var gene family has been estimated to occur at a rate of 2% of the parasites per generation in clones derived from IT 4/25/5 [42] , [43] . Although this rate is higher than the rate of mitotic change detected here , it may be that genetic change contributes or at least confounds the analysis of var gene switching . For example , recombination events and deletions in the subtelomeric regions might interfere with the tethering of chromosome termini to the nuclear periphery [5] , [34] or the recruitment of proteins initiating the spread of chromatin , thereby repressing var gene transcription ( telomere positioning effect ) [34] , [44] , [45] . Differences in average TRL between different strains of P . falciparum were reported earlier [34] , [35] . We observed alterations in the average TRL of individual clones as well as fluctuation of the average TRL over time , suggesting that the telomere ends are highly dynamic . Evidence that the telomere length could have an influence on the expression of nearby genes was found in Trypanosomes , where the telomere adjacent to the active expression site of VSGs was truncated [46] . All of these events could eventually result in a switch of var gene expression . In addition , the changes could further contribute to an evolutionary arms race whereby the host immune response and parasite antigens evolve in parallel to constantly outcompete each other . The acquisition of new SNVs ( most likely due to polymerase errors ) and the recombination of the telomeres during mitosis provide the parasite with two mechanisms that generate an enormous amount of genetic diversity . We used a Poisson model to estimate the mutation rate from the number of SNVs accumulated over time in continuous culture ( 1 . 7×10−9 for 3D7; 4 . 6×10−9 for the ATQ-resistant clones ) . Our use of the dN/dS-corrected mutation rate from continuous culture provides an estimate that is unaffected by the population bottlenecks caused by dilutions ( Materials and Methods ) . Given that a single human has 1013 red blood cells , by the time parasites are visible by thin smear ( 1% parasitemia ) more than 85% of the parasites will have acquired at least one novel genetic change , of which over 53% will be in the repetitive regions of the genome . In chronic infections ( more than 30 days ) , most parasites will differ from one another assuming there is no selection for particular variants . This is supported by a PCR product length analysis of var genes in line E5 , a clone of a P . falciparum patient isolate , NF54 that is also the parental strain of the 3D7 clone . E5 shows at least 13 var gene differences to 3D7 [33] . Although we did not detect any differences in major antigens such as merozoite surface protein 1 or the circumsporozoite protein , these genes do bear repeat regions and given that they show high levels of genetic diversity in populations [47] , they may also have higher rates of mitotic recombination . These high recombination rates may partially explain why people can be continually re-infected with closely related strains . While it is accepted in the field that most malaria infections are polyclonal , the actual diversity is still vastly underestimated [9] , [10] . The WHO advises the use of only three markers ( merozoite surface proteins 1 and 2 , and glutamine rich protein ) to establish the MOI in clinical trials on antimalarial drug efficacy [48] . All of the additional genetic changes throughout the parasite genome go unnoticed . Many of these genetic changes might not have an immediate selection advantage to the parasite . However , these unnoticed mutations might become important when a new drug with a new mode of action is introduced and a selection advantage is suddenly introduced . Our results provide the baseline information needed to design diagnostic studies , whole-genome population genetic studies or drug treatment studies .
3D7 ( MRA-151 ) was obtained from the Malaria Research and Reference Reagent Resource Center ( MR4; American Type Culture Collection deposited by D . Walliker ) and cloned in our laboratory at The Scripps Research Institute ( TSRI ) , La Jolla [49] . The San Diego 3D7 clone was derived from the same 3D7 stock as the parental 3D7 propagated at TSRI but was cultured long term at the Genomics Institute of the Novartis Research Foundation before cloning . The St Louis 3D7 clone was obtained from Dan Goldberg and is likely related to MRA-102 ( Bei Resources , Plasmodium falciparum 3D7 , deposited by D . J . Carucci and obtained from D . Walliker ) [49] , [50] and was used in the genome-sequencing project [51] . The New York strain from David Fidock , Columbia University , was obtained directly from David Walliker . Parasites were propagated in human erythrocytes depleted of white blood cells by filtration using leukocyte filters ( Fenwal Inc . , Lake Zurich ) with medium containing 5% human serum and 1 . 2% GIBCO AlbuMAX II ( Invitrogen , Grand Island ) [52] . Parasites were split 1∶20 when they reached parasitemias >8% . To obtain single clones , each line was cloned by limiting dilution ( 0 . 25 , 0 . 5 , 1 , 10 and 30 parasites per well ) in 96-well plates and two clones from each line ( three clones from the sensitive line ) were chosen for further analysis . Clones R1a and b and R2a and b were kept in culture for another 118 ( R2a and b ) or 128 ( R1a and b ) days and subcloned again . The original clones were termed generation 1 ( R1aG1 , R1bG1 , R2aG1 , and R2bG1 ) and the offspring were termed generation 2 ( R1aG2 , R1bG2 , R2aG2 , and R2bG2 ) . Genomic DNA was purified using a standard phenol-chloroform extraction method [53] . EC50 assays were performed using a 384-well format as previously described [54] . Each experiment was repeated three times . The averages of the EC50 were calculated for each experiment and log transformed . A one-way ANOVA followed by a Dunnet post-test with the 3D7 parent as reference was performed in Graphpad Prism . Microarray analysis was completed following the protocol in Dharia et al . [14] . In brief , fifteen micrograms of genomic DNA and 2 . 5 ng each of Bio B , Bio C , Bio D , and Cre Affymetrix control plasmids ( Affymetrix Inc . , Santa Clara ) were fragmented with DNaseI and end-labeled with biotin . The samples were hybridized to the microarrays overnight at 45°C in Affymetrix buffers , washed , and then scanned using a modified protocol with wash temperatures of 23°C to account for the high AT content of P . falciparum . Briefly , for polymorphism detection we scanned for sets of three overlapping probes that had significantly lower hybridization in the sample when compared to the parental 3D7 . Polymorphisms termed SFPs ( single feature polymorphism ) were detected by z-tests using an empirically derived standard deviation of the normal distribution and a p-value cutoff of 1×10−8 . Genomic DNA libraries were prepared for WGS using the standard Illumina protocol of fragmentation , end-polish , and adapter ligation ( v . 2011 , Illumina , Inc . , San Diego ) . The final PCR enrichment step was conducted with the KAPA HiFi polymerase ( KapaBiosystems , Inc . , Boston ) . This enzyme has been shown to more accurately amplify DNA regions with extremely high AT content [55] . DNA libraries were clustered and run on an Illumina Genome Analyzer II , according to manufacturer's instructions . Base calls were made using CASAVA v1 . 8+ ( Illumina , Inc . , San Diego ) . We have found that different amounts of RNA remaining after the DNA extraction process can give rise to hybridization differences in probes to noncoding RNA ( ncRNA ) genes . These changes are characterized by hybridization loss or gain over a wide range of probes throughout the gene . Altogether , 12 . 6% of the 1779 SFPs detected between the parental 3D7 and all clones mapped to ncRNA genes such as tRNAs ( 15 ) , rRNAs ( 68 ) , snRNAs ( 4 ) the signal recognition particle RNA ( 3 ) , and MAL8P1 . 310 , which showed similar hybridization patterns and was suspected of encoding an ncRNA . These SFPs were excluded from further analysis . After excluding these ncRNA genes , we identified a total of 46 SFPs suggestive of SNVs that mapped to 23 different locations in the genome . Blemishes on the microarray surface or nonfunctional probes can give rise to false positives , especially for SNVs with only one or two probes . We assumed that a true SNV would not only be detected between hybridizations of the parent and the clone but also when compared to the other clones . Therefore all clones were also compared to one another to confirm or contradict the initial SFPs detected between the parent and the clone . In the end fifteen SNVs were considered real , of which WGS confirmed ten and capillary sequencing detected two false positives that were excluded from further analysis . An internally developed WGS pipeline , dubbed PlaTypUS ( M . Manary et al . , in preparation ) , was used to align and analyze all WGS data . The program , compiled as a standalone executable , integrates many community-developed tools into its processing pipeline . In order to generate alignments , PlaTypUS follows a multi-step alignment process to generate and execute quality-control measures on an alignment map . FASTQ files obtained from sequencing were aligned to the Pf3D7 reference ( PlasmoDB v9 . 1 ) using BWA v0 . 6 . 1 with soft clipping of bases with quality score 2 and below [56] . Those reads that did not map to the reference genome were excluded from further analysis using SAMTOOLS v0 . 1 . 18 [57] . PCR duplicates were next identified and removed using Picard v1 . 48 MarkDuplicates ( http://picard . sourceforge . net ) . Aligned reads were then realigned around indels and areas of high entropy using GATK v1 . 4+ IndelRealigner . The base quality scores of realigned reads were then recalibrated using GATK TableRecalibration [58] . After realignment and recalibration , the samples were considered ready for use in downstream analysis . For SNV detection , the PlaTypUS uses a computer-learning algorithm to decide on the characteristics of a true SNV . WGS data for 10 , 500 known SNVs were obtained from PlasmoDB , and each of these positions was analyzed with regard to 26 metrics . In this way , the profile of a true SNV was discovered . The following criteria were found to be indicative of a true SNV , and were used to filter the set of raw SNVs from our samples . Genetic variants were identified with the GATK UnifiedGenotyper with a minimum base quality score threshold of 20 and subsequently filtered using GATK VariantFiltration with custom filters designed for Plasmodium spp [58] , [59] . SNVs were excluded if they failed one or more of the following filters: strand bias ( p<0 . 00001 , Fisher's exact test ) , minimum depth of coverage ( <5 reads ) , variant quality ( <100 phred scaled ) , likelihood estimate of genotype being correct ( <5 phred scaled ) , mapping quality bias ( p<0 . 10 using Mann-Whitney Rank Sum test ) , base quality bias ( p<0 . 10 using Mann-Whitney Rank Sum test ) , variant quality as a function of depth ( <5 per polymorphism ) , and percentage of non-uniquely mapped reads covering the variant ( >10% of total depth with a minimum of 4 ) . Variant annotations to which filters are applied are thoroughly explained on the GATK website ( http://www . broadinstitute . org/gsa/ ) . 19 , 532 raw variant calls across all seventeen clones were filtered down to 38 high quality SNVs . Gene annotations for high quality SNVs were generated using snpEff v . 2 . 0+ ( www . snpeff . sourceforge . net ) . The PlaTypUS also integrates a novel CNV-calling algorithm , which combines Weierstrass convolution of depth of coverage data with Canny edge detection to identify the break points of copy number events . 18 possible large deletions in the subtelomeric regions were identified for follow-up . We assumed that a number of lethal or deleterious mutations might arise during long-term culture that would not be detected , as the corresponding parasites would be lost; therefore , we tested for evidence of selection by calculating the dN/dS ratio and corrected our observed SNV ratio for this assumed loss . To do this we first calculated the codon potential ratio ( the likelihood that a single nucleotide replacement results in a nonsynonymous mutation ) from the list of all sequences from the 3D7 reference genome . We utilized the Phylogenetic Analysis by Maximum Likelihood ( PAML ) software suite [60] ( yn00 , default settings ) to generate a nonsynonymous to synonymous codon potential ratio of 4 . 63 . An uncorrected dN/dS ratio was generated by computing the total number of nonsynonymous SNVs over the length of all possible sites containing those SNVs ( the exonic genome ) and dividing by the total number of non-nonsynonymous SNVs over their entire domain ( the length of the genome ) . This gave an uncorrected dN/dS ratio of 2 . 73 , which corresponds to a true dN/dS ratio of 0 . 59 , indicating a deficit of nonsynonymous SNVs . In the absence of other information , the assumption was made that the actual ratio should approach 1 in the absence of deleterious mutations , and that only nonsynonymous mutations are deleterious . We used a generalized linear model with a linear link function and Poisson distributed errors to model the mutation rates for the various lines with and without drug . We assumed that the data followed the modelwhere Mi is the number of mutations scaled by the dN/dS factor; Xi = [gi li di]′ where gi , is the number of generations in culture , li , is a categorical variable indicating the parasite line , di , is a categorical variable indicating the presence of drug treatment; m is the number of parasite lines; and θ is the vector of coefficients to be determined by the model . We used this generalized linear regression to calculate the mutation rate per clone , using the number of SNVs and time to acquire them for each clone separately . We assumed that the mutation rate regression does not include a constant because there are no mutations at baseline . The expected number of mutations per experiment is thus We applied a Poisson-distributed number of mutations with the assumption that each parasite line accumulated mutations with a constant probability per generation . The probability of a mutant occurring in each generation is approximately binomially distributed B ( n , p ) with n the number of nucleotides and p the mutation rate . After g generations the distribution of the number of mutations per daughter parasite is B ( n , pg ) , assuming replication is independent and identically distributed . These replications are identically distributed because we assume that the accumulation of mutations neither quickens nor slows the mutation rate . By the Poisson theorem , we can approximate B ( n , pg ) as a Poisson distribution and thus use the regression model as above . The reason that we can assume a constant mutation rate ( i . e . a constant probability of SNVs occurring ) is that we have corrected for the selection bias against deleterious mutations using the dN/dS ratio . The dN/dS-corrected number of mutations is an unbiased estimator of the number of mutations that would have occurred in the absence of selection . The number of dilutions of parasite cultures during culture maintenance will not affect this quantity , because we assume that mutations are accumulating in both the remaining and disposed culture at the same constant rate . The number of dilutions would affect the mutation rate if we were calculating the rate using the fraction of mutants observed after culturing , because population bottlenecks may increase the variance of the fraction of mutants . Calculating the mutation rate from the proportion of mutants versus wild type organisms after parallel culturing is known as the ‘mutation rate problem’ [20] and the experiments used to calculate these rates are called fluctuation tests . However , we avoid these issues by calculating rates from continuous cultures after correcting for selection pressure [61] , [62] . To run these regressions we used the glm command ( MATLAB R2012b , The Mathworks ) and then divided by the number of base pairs in the P . falciparum genome ( 2 . 35×107 ) to determine the mutation rates with and without drug . To confirm the partial deletion of PFB0010w on chromosome 2 in clones R2a and b; G1 and 2 , R3a and b and S1a and c , 8×108 parasites of the 3D7 parent and the S1a clones were prepared , digested and run on a gel according to previous methods [63] , [64] . The chromosome 2 probe was PCR amplified from 3D7 gDNA ( chr2_PFB0015cF2: 5′-TACCAACATCGAAAAATACCAAACG-3′ and chr2_PFB0015cR: 5′-TGGCGGAGAGATTTGATGATATTG-3′ ) and labeled with [32P]-αdCTP ( 3000Ci/mmol ) . The membrane was incubated with the probe overnight at 42°C in ECL gold hybridization buffer ( GE Healthcare ) and washed according to the manufacturer's instructions . To determine the average TRL , mixed-stage parasites cultures of the parental 3D7 , all first generation ( G1 ) clones , second generation ( G2 ) clones R1a and R1b , as well as a clonal Dd2 strain were lysed with saponin and the pellets were resuspended in agarose to generate plugs . Agarose plugs containing 8×107 parasites were digested with AluI , DdeI , MboII , RsaI . The digested gDNA was run on an agarose gel and transferred to a membrane . The probe for the telomeres ( 5′-GGGTTTAGGGTTTAGGGTTTA-3′ ) was labeled with [32P]-γATP and hybridized in Church wash ( 40 mM NaPi pH 7 . 2 , 1 mM EDTA pH 8 , 1% SDS ) overnight at 55°C and washed . The membranes were placed against a phosphor storage screen for 2 days and then scanned in a phosphor imager . The average TRL was estimated by quantifying the signal intensities for each lane using QuantityOne 1-D analysis software ( BioRad , Hercules ) . The area under the curve was then calculated using Prism ( GraphPad Software , Inc . , La Jolla ) and the peak was reported . To confirm mutations predicted by microarray and WGS or to confirm accurate rejection of some mutations that did not make the cut off with the PlaTyPus software , 16 predicted SNVs were PCR-amplified with Phusion polymerase ( Finnzymes Inc . , Woburn ) using genomic DNA in a 100 µL PCR reaction volume for 35–40 reaction cycles . Genomic DNA from the 3D7 parent , R1aG2 , R2aG2 , R3b , R4b and R5b was used as templates . In addition , three predicted gene conversion events were also analyzed by Sanger sequencing . All PCR products were sequenced directly ( Retrogen , Inc . , San Diego ) . The primer sequences and results are summarized in Table S1 . A hierarchical lineage cladogram was constructed from the profile of detected mutations ( 38 SNVs , 18 deletions , and one duplication ) for each clone . Due to the small total number of mutations and the essentially linear process in which they were acquired , an un-weighted pair group method with arithmetic mean was used to generate distances between nodes in the software program Mesquite [65] , and a tree diagram was constructed from these calculated distances . Mesquite uses a heuristic algorithm to generate the tree of minimum complexity from the mutation data given as categorical changes ( booleans ) , and then minimizes the number of total number of changes in each tree . Each tree was then re-rooted at the parental strain . For each suspected recombination event , a de novo assembly of a contig spanning the deletion/recombination event was attempted , using the reads spanning 10 , 000 bp around the expected site of recombination from both the donor and recipient chromosomes . The PRICE Genome Assembly program ( http://derisilab . ucsf . edu/software/price/index . html ) was seeded with a single read from the region next to the predicted deletion , whose mate pair aligned to another chromosome , and then run for twenty cycles with otherwise default settings . The largest contig ( range 1399 bp to 5405 bp ) from each assembly was searched against the entire Plasmodium genome using BLAST and was then aligned using ClustalW2 [66] to the two chromosomal regions with the highest identity score and trimmed .
|
Malaria is one of the six diseases that together are responsible for 90% of all infectious disease deaths throughout the world . The five species of Plasmodium that cause human malaria take over 655 , 000 lives each year . Parasites evade the immune response through antigenic variation and develop resistance to anti-malarial drugs through genetic changes in either the drug target or genes conferring resistance . We used whole-genome sequencing and microarray techniques to study evolution in P . falciparum parasites propagated in vitro for up to 180 generations . We determined the mutation rate and found that the core genome of a single clone is stable , while the subtelomeric regions are prone to acquire structural variants . These changes occur mainly in multigene families involved in immune evasion . Our findings indicate that the parasite specifically increases the sequence variability in multigene families through mitotic recombination . This high plasticity of the parasite genome suggests that multiple haplotypes will be present in a natural infection initiated by a single parasite .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genetic",
"mutation",
"genome",
"evolution",
"mitosis",
"telomeres",
"molecular",
"genetics",
"chromosome",
"biology",
"biology",
"evolutionary",
"genetics",
"mutagenesis",
"genetics",
"gene",
"duplication",
"genomics",
"evolutionary",
"biology",
"genomic",
"evolution",
"genetics",
"and",
"genomics"
] |
2013
|
Mitotic Evolution of Plasmodium falciparum Shows a Stable Core Genome but Recombination in Antigen Families
|
The bromodomain protein Brd4 promotes HIV-1 latency by competitively inhibiting P-TEFb-mediated transcription induced by the virus-encoded Tat protein . Brd4 is recruited to the HIV LTR by interactions with acetyl-histones3 ( AcH3 ) and AcH4 . However , the precise modification pattern that it reads and the writer for generating this pattern are unknown . By examining a pool of latently infected proviruses with diverse integration sites , we found that the LTR characteristically has low AcH3 but high AcH4 content . This unusual acetylation profile attracts Brd4 to suppress the interaction of Tat with the host super elongation complex ( SEC ) that is essential for productive HIV transcription and latency reversal . KAT5 ( lysine acetyltransferase 5 ) , but not its paralogs KAT7 and KAT8 , is found to promote HIV latency through acetylating H4 on the provirus . Antagonizing KAT5 removes AcH4 and Brd4 from the LTR , enhances the SEC loading , and reverses as well as delays , the establishment of latency . The pro-latency effect of KAT5 is confirmed in a primary CD4+ T cell latency model as well as cells from ART-treated patients . Our data thus indicate the KAT5-AcH4-Brd4 axis as a key regulator of latency and a potential therapeutic target to reactivate latent HIV reservoirs for eradication .
The recent development of novel immunotherapeutic agents such as the bispecific dual-affinity retargeting ( DART ) antibodies that can engage a bound cytotoxic T cell to destroy an HIV-infected cell through recognizing the viral envelope proteins on the latter’s surface has fueled new optimism for finding a cure for HIV/AIDS[1] . However , a major impediment to the cure effort is the latent viral reservoirs in long-lived CD4+ T cells that do not express any HIV protein/RNA , and thus cannot be recognized either by DART or by the host immune system[2] . HIV latency is the result of silenced proviral transcription due to multiple complementary mechanisms[3] . To expose the latent reservoirs for accelerated clearance , numerous “latency-reversing agents” ( LRAs ) have been identified that target specific stages of the HIV transcription cycle[4] . However , extensive ex vivo studies suggest that individual LRAs are not very potent and that combinations of LRAs will be required to effectively purge the latent reservoirs[5] . Among the successful combinatorial ex vivo trials , inhibition of the BET bromodomain protein Brd4 with JQ1 strongly synergized with other LRAs to reverse viral latency[6 , 7] . Although these ex vivo studies provide an important proof of concept , the available LRAs are either highly toxic or yet to be proven efficacious in clinical settings[3] . Thus , better and safer LRAs are urgently needed . Brd4 is known to use its two bromodomains to bind to acetylated histones H3 and H4 ( AcH3 and AcH4 ) [8] and its C-terminal PID ( P-TEFb-interacting domain ) to recruit the human positive transcription elongation factor b ( P-TEFb ) to chromatin[9 , 10] to promote transcription of cellular primary response genes[11] . Counterintuitively , during activation of HIV transcription by the viral-encoded Tat protein , Brd4 acts as a potent inhibitor[10 , 12] . This is because Brd4 , which is highly abundant in vivo , can directly compete with Tat for binding to the limited cellular supply of P-TEFb[10 , 12] , the protein kinase that is an integral component of the multi-subunit Super Elongation Complex ( SEC ) used by Tat as a host cofactor for HIV transactivation[13 , 14] . By removing Brd4 from the HIV long terminal repeat ( LTR ) , JQ1 and other bromodomain inhibitors have been shown to reverse HIV latency by promoting the Tat-SEC formation , and consequently , Tat-transactivation[15 , 16] . Antagonizing the Tat-SEC complex may not be the only way that Brd4 controls HIV latency . A recent study has shown that the short isoform of Brd4 ( Brd4S ) that lacks the PID can instead recruit the repressive SWI/SNF chromatin-remodeling complexes onto the latent HIV-1 promoter to repress transcription[17] . Given that targeting Brd4 can be a very effective strategy to purge latent HIV reservoirs[6 , 7] , it is important to identify the prevailing histone acetylation pattern that allows Brd4 to be recruited and persistent on the HIV provirus . Knowing the exact Brd4 target on HIV chromatin and how it is created during latency establishment may help us devise more specific and efficient methods to displace Brd4 from the LTR for latency reversal . This will also help resolve the paradox that histone acetylation , which is typically associated with relaxed chromatin structure and HIV transactivation[18] , can be used to attract Brd4 to silence proviral transcription . We therefore examined the status of AcH3 and AcH4 present on the LTR of a pool of latent HIV proviruses that had diverse integration sites . Our data indicate that upon entering latency , these proviruses displayed elevated levels of AcH4 and Brd4 but drastically reduced levels of AcH3 on their LTR , raising the possibility that AcH4 is the primary histone modification responsible for attracting and retaining Brd4 on the silenced proviruses . Three major histone acetyltransferases ( HATs ) are known to modify H4: KAT5 ( Tip60 ) is known to acetylate lysines 5 , 8 , 12 and 16[19]; KAT7 targets three out of the four positions; and KAT8 ( MOF ) modifies lysine16 exclusively[20] . Our data indicate that downregulating the expression or activity of KAT5 , but not KAT7 and KAT8 , removed AcH4 and Brd4 from the HIV LTR , leading to the enhanced SEC loading , Tat-transactivation and escape from latency . Consistent with these results , our data further show that silencing KAT5 interfered with the establishment of latency . Together , these results have demonstrated a critical role for KAT5 and its acetylation of H4 in promoting Brd4’s recruitment to the HIV LTR and establishment of latency . Furthermore , they reveal that inhibiting the KAT5-AcH4-Brd4 axis is potentially an effective strategy for activating the latent proviruses for subsequent eradication .
Given the importance of Brd4 in modulating HIV latency , we determined the precise histone acetylation pattern on the viral LTR , which is targeted by Brd4 to sequester P-TEFb away from the Tat-SEC complex . Toward this goal , we first examined the levels of AcH3 and AcH4 on the LTR of latent HIV proviruses . The original observations indicating the importance of Brd4 in HIV latency were made by using isolated clones of Jurkat T cells ( e . g . the J-Lat cell lines and the 2D10 system[12 , 15 , 16] ) that contain only a single viral integration site within each cell population . To rule out site-specific integration effects , we created a pool of latently infected cells that contain a diverse array of all possible integration sites . This was achieved by adapting a protocol from Pearson et al . [21] to progressively establish HIV latency in Jurkat cells that were first infected on day 0 with a GFP-encoding HIV virus ( Fig 1A ) . The freshly infected cells containing actively replicating HIV were then sorted by fluorescence-activated cell sorting ( FACS ) on day 2 into a GFP ( + ) population , which was then cultured for additional 6 weeks ( Fig 1A ) . This population of cells , called the total pool , was examined for the percentages of GFP ( + ) cells on various days post infection ( d . p . i . ) . The results showed that on 43 d . p . i . , 95 . 3% of the cells still remained GFP ( + ) , while the rest have reverted to the GFP ( - ) state ( Fig 1B and 1C ) . To enrich the latently infected cells , a portion of the GFP ( + ) cells that was originally isolated by FACS on 2 d . p . i . was subjected to sorting again on 29 d . p . i . to isolate the GFP ( - ) cells , which were then allowed to propagate until 43 d . p . i . ( Fig 1A ) . This population , called the latent pool , had 89 . 8% of cells that remained GFP ( - ) on 43 d . p . i . , while the rest had undergone spontaneous reactivation to become GFP ( + ) again ( Fig 1C ) . It is important to point out that the HIV proviruses in the latent pool could readily be reactivated by conventional LRAs such as JQ1 and prostratin , demonstrating that the pool retained transcriptionally functional proviruses ( Fig 1D ) . Western analysis of cell lysates indicates no major difference in the overall levels of AcH3 , AcH4 , total H3 and total H4 between the total and latent pools ( Fig 1E ) . However , examination by chromatin immunoprecipitation ( ChIP ) demonstrates a 2 to 3-fold increase in the levels of total H3 and H4 on the HIV LTR in the latent pool compared to the total pool ( Fig 1F ) . This is consistent with the previous finding that the latent , transcriptionally silent HIV proviruses are more likely to be occupied by nucleosome 1 ( nuc-1 ) situated immediately downstream of the transcription start site[22] . Despite the increased total H3 level on the viral LTR in the latent pool , the acetylation of H3 decreased by 2-fold ( Fig 1F ) . This observation agrees with the general view that deacetylated H3 provides a marker for repressed and compact chromatin structure[23] as well as the result obtained previously using the Jurkat 2D10 cells , a widely used post-integrative HIV latency model[21] . In contrast to AcH3 , the AcH4 level on the LTR increased when the pool of proviruses entered latency ( Fig 1F ) . In fact , AcH4 increased at about the same extent as total H4 during this process . Similarly , more Brd4 was found on the LTR in the latent pool than in the total pool ( Fig 1F ) , likely due to the increased AcH4 level . Collectively , these data indicate that during the establishment of HIV latency , the incoming nucleosomes assembled on the viral LTR had very low AcH3 content , but maintained a high AcH4 level , which in turn increased the recruitment of Brd4 to inhibit HIV gene expression . Previously , it has been reported that Brd4 recognizes both AcH3 and AcH4 to interact with a chromatin template[8] . However , the surprising finding that AcH4 , but not AcH3 , exists in high concentration on the silenced HIV LTR raises the possibility that AcH4 is the primary histone modification responsible for recruiting Brd4 onto the LTR . Lysines 5 , 8 , 12 and 16 ( H4K5/8/12/16 ) are the primary acetylation sites found at the N-terminus of H4 [24] and acetylation of these sites has been shown to promote the binding of Brd4 to H4 both in vivo and in vitro [8] . Three major histone acetyltransferases ( HATs ) are known to modify H4 . While KAT5 ( Tip60 ) is the only one capable of acetylating all four H4K5/8/12/16 positions , the other KATs are more selective [20 , 24] . For example , KAT7 acetylates H4K5 , 8 and 12 , whereas KAT8 only acetylates H4K16 . We therefore examined the impact of silencing the expression of KAT5 , KAT7 or KAT8 on HIV transcription and latency . We used the doxycycline ( Dox ) -inducible CRISPRi system[25] to suppress the expression of KAT5 , KAT7 or KAT8 in the Jurkat-based 2D10 cell line , a widely used post-integrative HIV latency model containing the GFP-coding sequence in place of the viral Nef gene[21] . An sgRNA sequence ( sg1 ) that specifically targets the promoter region of the KAT5 gene was found to reduce the KAT5 mRNA level by ~80% in the engineered CRISPRi-KAT5-sg1 cells upon exposure to Dox ( Fig 2A , left panel ) . Western analysis of the cell lysates showed a corresponding decrease in the KAT5 protein level as well as a marked reduction of the AcH4 but not AcH3 level ( Fig 2A , right panel ) . The global AcH4 reduction agrees well with the demonstrated role of KAT5 as a promiscuous acetyltransferase for H4[19] . Using GFP induction as an indicator of latency reversal in the engineered 2D10 cells , the FACS analyses show that the Dox-induced inhibition of KAT5 expression by CRISPRi caused ~15% of the cell population to become GFP ( + ) ( Fig 2B ) . In addition , CRISPRi against KAT5 also strongly synergized with JQ1 , prostratin , and SAHA , the three well-studied conventional LRAs , to enhance GFP production across a broad range of drug concentrations ( Fig 2B ) . Very similar results were obtained in another 2D10-based CRISPRi-KAT5 cell pool , in which the inhibition of KAT5 expression was achieved by using sgRNA #2 ( sg2 ) that targets a distinct KAT5 promoter sequence ( S1 Fig ) , thus effectively ruling out a potential off-target effect caused by CRISPRi . Finally , inhibition of the catalytic activity of KAT5 with a selective inhibitor MG-149[26] also reversed HIV latency and potentiated traditional LRAs in a dosage-dependent manner ( Fig 2C ) . In contrast to the above results showing a key role of KAT5 in silencing HIV transcription , the CRISPRi-mediated inhibition of the KAT7 or KAT8 expression in 2D10 cells ( Fig 2D–2G ) neither activated GFP expression by itself nor promoted the effects of the three conventional LRAs ( Fig 2E and 2G ) . In fact , the inhibition of KAT7 even slightly decreased the levels of both basal and the LRA-induced GFP production ( Fig 2G ) , a result that was also observed in another CRISPRi-KAT7 cell pool generated with a different sgRNA ( S2 Fig ) . Thus , unlike KAT5 , KAT7 appears to play a small but positive role in HIV transcription . Consistent with the demonstration of KAT8 as a specific H4K16 acetyltransferase and KAT7 as a HAT for three out of the four lysine positions in the H4 N-terminus[20 , 24] , the global AcH4 level was only slightly decreased in the CRISPRi-KAT8 cells but more prominently decreased in the CRISPRi-KAT7-sg1 cells ( Fig 2D & 2F ) . Collectively , these data indicate that KAT5 , but not its two paralogs KAT7 and KAT8 , is required to maintain the HIV provirus in a transcriptionally silent state in latency . In the majority of cases , acetyltransferases activate gene expression through acetylating histone tails , which de-condenses chromatin to facilitate transcriptional initiation and elongation[23] . The Brd4-P-TEFb complex that is recruited to a chromatin template through the interaction between the Brd4 bromodomains and acetyl-histones plays a critical role in promoting transcriptional elongation of many cellular genes , especially those that are involved in primary responses[27 , 28] . In light of our surprising finding that KAT5 did not activate HIV , but rather repressed HIV gene expression to keep the virus in latency , we investigated how antagonizing KAT5 might affect the expression of non-HIV genes such as MYC , FOS and JUNB , which are well-known cellular primary response genes[11 , 29] . In agreement with the expectation that the pan acetyltransferase KAT5 should act to promote gene expression , inhibiting KAT5’s expression by CRISPRi ( Fig 3A ) or activity by MG-149 ( Fig 3B ) was found to decrease the mRNA levels of all the three genes . To confirm that the inhibition of HIV gene expression by KAT5 was indeed working through the viral LTR and at the transcription level , we first examined the effect of the shRNA-induced KAT5 knockdown ( KD ) on the ability of an integrated HIV-1 LTR to drive the expression of the luciferase reporter gene in Jurkat 1G5 cells[30] . Interestingly , while the KD slightly decreased the basal LTR activity in the absence of Tat , it significantly increased luciferase expression when Tat was present in the cells ( Fig 3C ) . The differential effect on basal versus Tat-dependent HIV LTR activity was also observed when the catalytic activity of KAT5 was inhibited by MG-149 ( Fig 3D ) . In contrast to the effects caused by antagonizing KAT5 , the overexpression of KAT5 significantly repressed the Tat-dependent , but not -independent LTR activity ( Fig 3E ) . Finally , using a qRT-PCR-based assay that can distinguish between the processes of transcription initiation and elongation[31] , the inhibitory effect of KAT5 on Tat-transactivation was deemed to be primarily at the elongation stage ( S3 Fig ) . All together , these data strongly suggest that while KAT5 plays a stimulatory role in promoting the expression of cellular Brd4-dependent primary response genes , it can efficiently inhibit the Tat-dependent HIV transcriptional elongation . To determine whether the global reduction of the AcH4 but not AcH3 level observed in the CRISPRi-KAT5-sg1 cell lysates ( Fig 2A ) would also result in a decreased AcH4 but not AcH3 level on both HIV and cellular gene promoters , we conducted the anti-AcH4 and -AcH3 ChIP assay . Indeed , suppressing KAT5’s expression by CRISPRi ( Fig 4A ) or activity by MG-149 ( Fig 4B ) reduced the level of AcH4 detected on the HIV provirus at both the viral LTR and Env gene . In addition , a marked reduction in the AcH4 level was also detected at the promoters of two cellular genes , MYC and IκBα , as well as at the endogenous retroviral element HERVK[32] and an intergenic region upon the inhibition of KAT5 ( Fig 4A and 4B ) . In contrast , CRISPRi against KAT5 did not significantly affect the AcH3 levels at these HIV and non-HIV locations , with the only exception seen at the MYC promoter , where a small reduction that could be a secondary effect of the diminished transcription was detected ( Fig 4C ) . Furthermore , on the silent HIV provirus in 2D10 cells , the overall levels of AcH3 relative to total H3 were lower than those detected on the two cellular genes MYC and IκBα ( Fig 4C ) , a result that echoes the observation made in a whole population of latently infected Jurkat cells in Fig 1F . Finally , the poorly transcribed HERVK and the intergenic genomic region displayed generally low levels of AcH3 and AcH4 when compared with the robustly expressed HIV , MYC and IκBα genes ( Fig 4C & 4B ) . While the relatively low level of AcH3 on the latent HIV LTR has been reported previously [21 , 33] and is likely due to a diminished recruitment of H3 acetyltransferase p300 [34] , the unusually high level of AcH4 on the LTR ( Figs 1F , 4A and 4B ) is yet to be explained . To this end , we compared the levels of KAT5 on the HIV LTR and the MYC promoter in infected cells . The ChIP data demonstrate that the LTR had a mildly higher level of KAT5 than did the MYC promoter in the total pool of infected cells . However , the difference between the two became more pronounced and statistically significant upon the establishment of viral latency ( Fig 4D and 4E ) . Thus , the unbalanced loadings of p300 and KAT5 likely contribute to the unusually low AcH3 but high AcH4 level on the latent HIV LTR . Since inhibition of KAT5 led to an overall reduction of the AcH4 level at both the HIV and non-HIV gene promoters , we asked how the reduced AcH4 level could promote Tat-dependent HIV transcription while at the same time suppress the expression of cellular primary response genes . To answer this question , we investigated the consequence of inhibiting KAT5 on the binding of Brd4 and the human Super Elongation Complex ( SEC ) to the HIV and non-HIV loci . The ChIP assay conducted in the inducible CRISPRi-KAT5-sg1 cells indicates that upon the Dox-induced down-regulation of KAT5 , the Brd4 level significantly decreased at the HIV LTR , HIV Env , and the MYC promoter region ( Fig 5A ) . This implicates AcH4 as the primary route of recruitment for Brd4 at these locations . In contrast , although the Brd4 level was quite high at the IκBα gene promoter before the Dox treatment , it displayed little change after CRISPRi silencing of KAT5 ( Fig 5A ) , probably because Brd4 is recruited to this locus by recognizing mostly AcH3 but not AcH4 . The Brd4 levels at the poorly transcribed HERVK and the intergenic region were relatively low and remained fairly constant upon CRISPRi silencing of KAT5 ( Fig 5A ) . Importantly , a very similar change in the Brd4 distribution pattern was also observed when the catalytic activity of KAT5 was inhibited by MG-149 ( Fig 5B ) . A recent study by the Ott laboratory demonstrates that both the long and short isoforms of Brd4 can promote HIV latency , albeit through distinct mechanisms [17] . To assess the roles of the two Brd4 isoform in transducing the signal downstream of the KAT5-AcH4 axis , we compared the effect of MG-149 on the bindings of the long ( simply labeled as Brd4 throughout the current study ) and short form of Brd4 ( Brd4S ) to the integrated HIV-1 LTR . The ChIP data show that when expressed at a similar level , more Brd4 bound to the LTR than did Brd4S ( S4 Fig ) . Furthermore , compared to Brd4S , the binding of Brd4 to the LTR was also more sensitive to the MG-149-induced AcH4 reduction . Thus , between the two Brd4 isoforms , the long form appears to be more responsive to changes in the level of AcH4 on the LTR and thus more likely to be a key target of the KAT5-AcH4 axis . Because the abundant Brd4 long isoform containing the C-terminal P-TEFb-interacting domain is a direct competitor of HIV Tat for the limited cellular supply of P-TEFb[10 , 12] , the decreased Brd4 occupancy on the HIV provirus in KAT5-inhibited cells is expected to free up P-TEFb for its incorporation into the Tat-SEC complex , which is very efficiently recruited to the provirus through the TAR RNA route [16 , 35] . Consistent with this mechanism , the ChIP analysis showed that inhibiting KAT5’s expression by CRISPRi ( Fig 5C ) or activity by MG-149 ( Fig 5D ) significantly increased the level of AFF1 , the scaffolding subunit of the SEC , on both the HIV LTR and Env gene . In addition , MG-149 also significantly enhanced the loading of another key SEC subunit ELL2 onto the HIV provirus in an engineered Jurkat cell line ΔE2-R2 ( Fig 5E ) , in which the endogenous ELL2 gene has been knocked out and ELL2-FLAG is expressed at a similar level from an integrated vector[31] . By contrast , the inhibition of KAT5 decreased the SEC level on the MYC promoter ( Fig 5C–5E ) , which agrees with the diminished MYC expression in the KAT5-inhibited cells ( Fig 3A & 3B ) . Although the precise function of SEC in MYC transcription is yet to be revealed , our data suggest a likely role for Brd4 in recruiting SEC to this important primary response gene . Finally , very low levels of AFF1 and ELL2-FLAG were detected at the IκBα gene promoter , HERVK , and the intergenic region ( Fig 5C–5E ) , implicating that the SEC plays little role in their transcription . The data presented thus far are consistent with the hypothesis that abundant AcH4 on the HIV provirus is used to locally recruit Brd4 , which then competes with Tat for host P-TEFb . KAT5 enhances the levels of AcH4 , and consequently it behaves as an inhibitor of the Tat-SEC formation on HIV LTR to suppress Tat-transactivation and enforce viral latency . Based on this notion , we hypothesized that silencing the expression of KAT5 would therefore make it more difficult for HIV to establish latency in the progressive latency establishment assay outlined in Fig 6A . To test this hypothesis , we generated stable pools of Jurkat cells that express either a non-targeting scrambled shRNA ( shScramble ) or the KAT5-specific shRNA ( shKAT5 ) . The latter was shown to cause ~70% reduction in the cellular level of KAT5 ( Fig 6B ) . When tested in the progressive latency establishment assay , the shKAT5 pool displayed a significantly delayed dynamics in reverting to the GFP ( - ) status , i . e . transcriptionally silent or latent state , when compared with the shScrambled pool ( Fig 6C & 6D ) . This result confirms the prediction that KAT5 and its acetylation of H4 play a critical role in allowing HIV to efficiently establish latency . In primary T-cells , P-TEFb levels are dramatically reduced when the cells enter quiescence , and this in turn , forces HIV into latency . To evaluate the role of KAT5 in maintaining HIV in a transcriptionally silent state in primary CD4+ T cells , we used our recently described Th17 cell latency model [36] . Briefly , polarized and expanding Th17 cells were infected with a VSVG-pseudotyped HIV-1 virus HIV-Nef-CD8a/eGFP and then forced to enter quiescence by culturing in a restrictive cytokine environment ( Fig 7A ) . Proviral gene expression , as assessed by immunofluorescence staining for Nef , was low in the quiescent cells , but could be induced ~4 to 6-fold upon activation of the T-cell receptor ( TCR ) by antibodies to CD3 and CD28 ( S5 Fig and Fig 7C ) . In each of these experiments , the EGFP signal remained high in Th17 cells containing latent HIV due to the prolonged stability of the membrane-bound CD8a-EGFP fusion protein . As expected , MG-149 significantly and preferentially decreased the AcH4 level in primary CD4+ T cells ( Fig 7B ) . In comparison , the levels of AcH3 and total H4 were only mildly reduced , probably due to an indirect effect of global inhibition of gene expression in the treated cells . When used alone , MG-149 ( at 10 or 50 μM ) and JQ1 ( at 1 or 5 μM ) produced no , or only very modest , stimulatory effects on proviral gene induction in our Th17 cell latency model ( Fig 7C and 7D ) . However , exposure of the cells to 1 μM JQ1 plus 10 μM MG-149 for 48 hr increased the HIV-expressing cell population by more than 2-fold ( Fig 7C ) . Similarly , while the treatment with 5 μM JQ1 alone slightly elevated HIV expression , the addition of 50 μM MG-149 further enhanced the stimulatory effect of 5 μM JQ1 by ~1 . 8-fold ( Fig 7D ) . At 10 to 50 μM , MG-149 produced very little enhancing effect on HIV reactivation in cells that were treated with SAHA , the PKC agonist Ingenol [6 , 37] , or TCR-activating signals ( S5 Fig and Fig 7C ) . Thus , the enhancing effect of MG-149 appears to be restricted to JQ1 in this primary T cell model of latency . Of note , both MG-149 alone and in combination with other LRAs did not cause global T-cell activation as indicated by minimal changes in CD25 and CD69 immunostaining after the treatment ( S6 Fig ) . To further investigate the impact of KAT5 inhibition by MG-149 in a more clinically relevant setting , we performed an ex vivo experiment to test the effect of MG-149 alone or in combination with JQ1 to release virions from primary T cells that are isolated from ART-treated patients ( Fig 7E and S1 Table ) . The data reveal a statistically significant positive effect displayed by MG-149 alone ( 6 . 5-fold increase on average of HIV RNA released into supernatant ) as well as by the JQ1 plus MG-149 combination ( 5 . 3-fold increase on average ) . In comparison , JQ1 alone failed to produce a statistically significant effect . It is interesting to note that a previous report shows that both the PKC-agonist and PMA plus ionomycin ( PMA/I ) , but not JQ1 or HDACi alone , could efficiently increase virion release from latently infected primary cells [7] . Our ex vivo data confirm these observations about PMA/I and JQ1 , and more importantly , show that MG-149 was almost as effective as PMA/I to single-handedly increase the virion release ( Fig 7E ) . This effect of MG-149 could be important for maximally exposing the latently infected cells for immuno-recognition and clearance and thus implicate the potential utility of KAT5 inhibition as an effective latency reversal strategy .
In contrast to the well-characterized positive effects of histone H3 acetyltransferases p300/CBP and P/CAF on HIV transcription [4] , the roles played by the major H4 acetyltransferases in this process are unknown . In this study , we report that KAT5 , but not KAT7 and KAT8 , is a host factor that promotes HIV latency establishment , and inhibits latency reversal , by acetylating H4 on the viral LTR . The unique pattern of histone acetylation found at the LTR permits recruitment of Brd4 to the HIV promoter where it competes with Tat for P-TEFb , blocks Tat-SEC formation , and ultimately inhibits Tat-transactivation ( Fig 7F ) . When KAT5 is antagonized by either CRISPRi or MG-149 , the loss of AcH4 on the LTR dissociates Brd4 and allows P-TEFb to join Tat in the Tat-SEC complex to enhance HIV transcriptional elongation and latency reversal ( Fig 7F ) . Importantly , the strong correlation between the shutdown of HIV transcription during latency establishment and elevated levels of AcH4 and Brd4 on the LTR was demonstrated not only with the extensively characterized Jurkat clone 2D10 cell model , but also in a pool of latently infected cells with diverse integration sites . Thus , the negative effects exerted by the KAT5-Ac-H4-Brd4 axis on HIV transcription are independent of the proviral integration site and the cellular chromatin context . Although inhibitory to Tat-induced HIV transactivation , the KAT5-AcH4-Brd4 axis appears to play a positive role in stimulating the expression of cellular genes , especially those involved in primary response . The differential effects of KAT5 on transcription of the HIV versus host genes implicate this histone acetyltransferase as a promising candidate that can be selectively targeted to reactivate latent HIV without globally activating numerous host genes at the same time . The high selectivity toward HIV is likely to be an important attribute for LRAs used in the “Shock and Kill” strategy for eradicating latent HIV reservoirs [3] . In contrast to KAT5 , KAT7 and KAT8 produced no inhibitory effects on HIV transcription and latency reversal ( Fig 2D–2G ) . We ascribe this to KAT5’s broader substrate specificity toward four H4 lysines ( K5 , K8 , K12 , and K16 ) versus KAT8’s sole targeting of H4K16 and KAT7’s more restrictive target specificity [19 , 20 , 24] . It has been shown that the poly-acetylated K5 , K8 , K12 is essential for Brd4’s binding to H4 , whereas the acetylation of K16 only minimally increases Brd4’s affinity to H4 peptides already acetylated at K5 , K8 , and K12 [8] . Interestingly , KAT5 was first identified as an HIV Tat-interacting protein of 60 kDa , hence originally named as Tip60 [38] . It was subsequently found to be inhibited by Tat [39] . Thus , in addition to the inhibition of KAT5 by MG-149 , which functions as a LRA to kick-start the initial rounds of HIV transcription to produce viral proteins including Tat , the accumulated Tat protein may further inhibit KAT5 to decrease AcH4 and antagonize Brd4’s action , thus fueling another positive feedback loop besides the Tat-SEC axis to expedite the reversal of latency . Proviral transcription of latent HIV is silenced by multiple mechanisms including heterochromatinization of the LTR [40] . In the present study , we found that the AcH3 level is very low on the LTR in 2D10 cells as well in a diverse population of latently infected Jurkat cells . In contrast , the AcH4 level on the LTR is relatively high and comparable to those on the cellular genes . The activity of AcH4 is therefore distinct from that of AcH3 , which is an activating signal since it prevents H3K9/K27-methylations that recruit the chromatin-compacting proteins HP1 and Polycomb [23] . By contrast , AcH4 is compatible with heterochromatin [41] and can even induce heterochromatinization in certain cases [42] . Thus , our data point to a plausible scenario where latent HIV proviruses are low in AcH3 but high in AcH4 , which helps keep the LTR heterochromatinized but still capable of retaining Brd4 to suppress Tat-transactivation . This also resolves an apparent paradox concerning how histone acetylation , which is generally considered pro-euchromatin and promote HIV transcription [18] , can be exploited by Brd4 to further silence Tat-transactivation . Finally , it is worth noting that prior to the current study , the KAT5-AcH4-Brd4 axis has already been reported as a restriction mechanism for papillomaviral and adenoviral gene transcription through retaining or interacting with negative transcription factors on the viral promoters [43 , 44] . In light of these findings , our current work adds HIV to the growing list of viral infections that employ KAT5 as a crucial regulator . Due to the presence of multiple additional restrictions imposed on proviral transcription in quiescent primary T cells , it is considerably more difficult to reverse HIV latency in these cells than in transformed cell lines [45] . Nevertheless , the observation that the KAT5 inhibitor MG-149 cooperates with JQ1 to promote HIV latency reversal not only in activated cell lines but also in quiescent primary T cells derived from ART-treated HIV patients underscores the clinical relevance of targeting the KAT5-AcH4-Brd4 axis . Future studies are necessary to explore this potential for therapeutic intervention and eventual eradication of HIV/AIDS .
The part of this study utilizing specimens from HIV-infected individuals has been approved by the Bioethics Review Committee of the Shenzhen Center for Disease Control and Prevention in China . All research participants gave written informed consent , and all subject data and specimens were coded to protect confidentiality . To downregulate KAT5 , KAT7 and KAT8 in Jurkat 2D10 cells ( previously generated by Karn lab based on human CD4+ T cells Jurkat line[21] ) , vectors pHR-TRE3G-Krab-dCas9-P2A-mCherry[25] and pLVX-advanced-TetOn ( gift from IGI , UC Berkeley ) were packaged separately using a 3rd generation lentiviral packaging system and co-infected into Jurkat 2D10 cells . A clone ( 2D10-iI ) expressing Krab-dCas9-HA-P2A-mCherry in response to doxycycline ( Dox ) treatment was picked by FACS , and verified by Western blot . DNA oligos containing sgKAT5#1 ( sg1; 5’-GACTCAGTAGACCGCCAC-3’ ) , sgKAT5#2 ( sg2; 5’-GCCTCAGGCCGAGCCCTAGG-3’ ) , sgKAT7#1 ( sg1; 5'-GTGTATCAGTCCCAATCCTG-3' ) , sgKAT7#2 ( sg2; 5'-GGGATCGTCCGCAGGATT-3' ) , and sgKAT8 ( 5’-GAGAGACGCGGCCCGGGGAT-3’ ) were synthesized and cloned separately into the pSico-BFP-puro vector[25] , which was then packaged and transduced into the 2D10-iI cells . After 1 μg/ml puromycin selection for 4 days , the stable CRISPRi-KAT5-sg1 , CRISPRi-KAT5-sg2 , CRISPRi-KAT7-sg1 , CRISPRi-KAT7-sg2 , and CRISPRi-KAT8 cell pools were treated with 1 μg/ml Dox for 72 hours and checked by RT-qPCR and Western blot with KAT5 ( Invitrogen ) , KAT7 ( Bethyl ) and KAT8 antibodies ( Invitrogen ) for CRISPRi efficiency . Total RNA from ~5×106 cells were extracted by RNeasy kit ( Qiagen ) and reverse transcribed using the M-MLV Reverse Transcriptase ( VWR ) with random hexamers ( Invitrogen ) . The cDNAs were subjected to qPCR using a DyNAmo HS SYBR Green qPCR kit ( Fisher ) on a CFX96 machine ( Bio-Rad ) with the following primers: qKAT5-F ( 5'-AACCAGGACAACGAAGATGAG-3' ) , qKAT5-R ( 5'-GTCACCCATTCATCCAGACG-3' ) ; qKAT7-F ( 5'-AGCCCTTCCTGTTCTATGTTATG-3' ) , qKAT7-R ( 5'-CATAGCCCTGTCTCATGTACTG-3' ) , qKAT8-F ( 5'-GGGAAAGAGATCTACCGCAAG-3' ) , qKAT8-R ( 5'-TCCACGTCAAAGTACAGTGTC-3' ) ; qActB-F ( 5'-AGAGCTACGAGCTGCCTGAC-3' ) , qActB-R ( 5'-AGCACTGTGTTGGCGTACAG-3' ) ; qLTR-F ( 5’-GGGTCTCTCTGGTTAGACCAG-3’ ) , qLTR-59R ( 5’-GGGTTCCCTAGTTAGCCAGAG-3’ ) , qLTR-190R ( 5’-CTGCTAGAGATTTTCCACACTGAC-3’ ) ; qMYC-F ( 5'-TTCGGGTAGTGGAAAACCAG-3' ) , qMYC-R ( 5'-AGTAGAAATACGGCTGCACC-3' ) ; qFOS-F ( 5'-TTGTGAAGACCATGACAGGAG-3' ) , qFOS-R ( 5'-CCATCTTATTCCTTTCCCTTCGG-3' ) ; qJUNB-F ( 5'-AGCCCAAACTAACCTCACG-3' ) , qJUNB-R ( 5'-GGGCATCGTCATAGAAGGTC-3' ) . All reactions were carried out in triplicates . The PCR signals were normalized to those of ActB and displayed . The CRISPRi-KAT5-sg1/sg2 , CRISPRi-KAT8 , or parental Jurkat 2D10 cells were first treated with the various LRAs or MG-149 ( APExBIO ) at the indicated concentrations for 20 hr , and then subjected to flow cytometry on a BD Bioscience LSR Fortessa X20 cytometer for GFP fluorescence . The data were analyzed using Flowjo software . Each drug treatment was done in 200 μl RMPI medium ( Invitrogen ) with 10% FBS ( Gemini 900108 Lot A96C ) . To induce CRISPRi-mediated downregulation of KAT5 or KAT8 , the CRISPRi-KAT5-sg1/sg2 , CRISPRi-KAT8 cells were pre-treated for 48 hr with 1 μg/ml Dox before the LRA treatments . For control groups , 0 . 1% DMSO was used . Naïve CD4+ T cells isolated from a previously frozen healthy donor leukapheresis pack ( previously generated by Karn lab[36] ) by negative selection using the EasySep Naïve CD4 T-cell isolation kit ( 19155RF; Stem Cell ) were simultaneously treated with T-cell receptor activator magnetic beads and a Th17 polarization cytokine cocktail for 6 days as previously described[36] . On Day 4 , IL-2 was added to the cells at 60 IU/ml . Following Th17 polarization and during T-cell expansion , cells were either infected or not with a VSVG-pseudotyped pHR’-Nef+-CD8a/GFP viral construct by spinoculation . HIV-infected cells were later positively selected for by anti-CD8a magnetic separation . To generate quiescent T cells , infected cell populations were cultured in medium containing reduced concentrations of IL-2 ( 15 IU/ml ) and IL-23 ( 12 . 5 μg/ml ) for at least 2 weeks . Proviral latency was monitored by assessing for the expression of Nef and gp120 Env by immunofluorescence flow cytometry before and after T-cell receptor reactivation for 18 hr . Achievement of resting T cells was monitored by immunostaining for Ki67 , cyclin D3 , and pSer175 CDK9 under the same conditions . Primary resting Th17 cells , latently infected with HIV as described above , were pretreated or not for 30 min with MG-149 at varying concentrations ( 10 or 50 μM ) prior to 24 or 48 hr challenge with either of the following stimuli: T-cell receptor activator anti-CD3/anti-CD28 antibody cocktail , JQ1 ( at 1 μM or 5 μM ) , SAHA ( 500 nM ) , or ingenol ( 20 nM ) . Cells were washed once with 1X PBS and fixed in 4% formaldehyde for 15 min at room temperature prior to permeabilization with 1X BD Perm/Wash buffer ( BD Biosciences ) . After blocking with a non-specific IgG for 15 min , cells were immunofluorescently stained for HIV Nef . Thereafter , these cells were washed three times with 1X permeabilization buffer and subjected to flow cytometry analysis using the LSR Fortessa instrument ( BD Biosciences ) equipped with the appropriate laser and filters . Approximately 2 . 4 × 107 memory CD4+ T cells were isolated by negative selection from a healthy donor leukapheresis pack containing 1 . 5 × 108 PBMCs . Following the isolation , cells were allowed to recover overnight in 10 ml complete RPMI media supplemented with 60 IU/ml IL-2 and 1 mM SAHA and then divided equally into a 24-well plate and treated with varying concentrations of MG-149 for 24 hr . Whole cell extracts were mixed with 50 ml of a 2X SDS-PAGE sample buffer prior to microtip sonication to mechanically shear viscous DNA . Thereafter , the samples were boiled at 95°C for 10 min and subjected to SDS-PAGE and immunoblotting analysis for N-terminally acetylated histone H4 , total histone H4 , and N-terminally acetylated histone H3 . Densitometry analysis was performed using Quantity One software ( Bio-Rad ) . HIV-1–infected individuals enrolled in the study were based on the criteria of good response to suppressive ART and undetectable plasma HIV-1 RNA levels . The PBMCs from the EDTA+ blood were isolated by Ficoll centrifugation . Approximately 10 × 106 PBMCs were cultured in 2 ml RPMI1640 medium containing 10% FBS . The cells were then treated with 0 . 2% DMSO , 50 ng PMA + 1 μM Ionomycin ( PMA/I ) as a positive control , 50 μM MG-149 , 2 μM JQ1 , or 50 μM MG-149 + 2 μM JQ1 for 24 hr . After spinning down the cells , supernatants ( 1 ml each ) were collected and subjected to HIV-1 RNA quantification with the Roche COBAS AmpliPrep/TaqMan HIV-1 Qualitative Test system . Each sample was tested twice , and the average viral copy numbers were normalized to the DMSO group and displayed in a scatter plot . The Jurkat 1G5[30] and 1G5+Tat cells[46] ( both kind gifts from Melanie Ott lab , Gladstone Institutes , San Francisco ) were infected with the pLKO . 1-puro-based lentiviral vector containing shScramble 5’-CCTAAGGTTAAGTCGCCCTCG-3’ , or shKAT5 5'-CCTTGACCATAAGACACTGTA-3' sequences . Two days after infection , the cells were selected by 1 μg/ml puromycin for 5 days until the stable pools were obtained . The KD efficiencies of the pools were examined by Western blot . Two days before the assay , HEK 293T cells ( from American Type Culture Collection ) at ~60% confluency in 6-well plates were transfected using Polyethylenimine in triplicates by 0 . 1 μg HIV-1 LTR-luciferase construct combined with 1 . 85 μg pcDNA3-FLAG-KAT5[47] and/or 0 . 05 μg pRK5-Tat-HA . The total amount of DNA in each transfection was brought to 2 μg by empty pcDNA3 vector when necessary . For luciferase assays in WT Jurkat 1G5 or 1G5+Tat cells , 1 ml cells at the concentration of 0 . 5 million/ml were aliquoted into 12-well plates in triplicates and treated with the indicated concentrations of MG-149 for 18 hr . For luciferase assays in 1G5-/+Tat stable shRNA cell pools , the cells were counted on a haemocytometer and about 1 . 7 million cells from each pool were aliquoted in triplicates . The cells were harvested and lysed in 200 μl 1 × Reporter Lysis Buffer ( Promega ) , and centrifuged at maximum speed on a benchtop centrifuge for 30 seconds . The supernatants were subjected to luciferase activity measurement by using the Luciferase Assay System ( Promega ) on a Lumat LB 9501 luminometer . For each group , 10 μl cell lysate was aliquoted before centrifugation from each repeat and pooled for Western blots to check the indicated protein levels . The assay was based on Nelson et al . [48] with some modifications . Briefly , 10 million cells were fixed by 1% formaldehyde for 10 min at room temperature , and then quenched for 5 min by adding glycine to 125 mM final concentration . After washing with PBS and lysed in the IP buffer[48] , the chromatin was sheared using the Covaris S220 System with 200 Watt peak energy for 30 cycles ( 30 sec ON , 20 sec OFF ) to an average size of 0 . 5–1 kb DNA fragments . The sheared chromatin was centrifuged at 20 , 800 × g for 10 min at 4°C . Each ChIP reaction was carried out with 300 μl supernatant and 1 μg each of the following antibodies: anti-Ac-H3 ( Millipore , 06–599 ) , anti-Ac-H4 ( Millipore , 06–598 ) , anti-H3 ( Invitrogen , PA5-31954 ) , anti-H4 ( Invitrogen , 720166 ) , anti-FLAG ( Sigma , F1804 ) , anti-Brd4 [10] , normal rabbit total IgG ( Santa Cruz , sc-2027 ) , and normal mouse total IgG ( Santa Cruz , sc-2025 ) . After overnight incubation with rotation at 4°C , the reactions were centrifuged at 20 , 800 × g for 10 min , and 90% of each supernatant was combined with 20 μl Salmon sperm DNA ( Invitrogen ) -blocked Protein A Agarose ( Invitrogen ) . After 45 min rotation , the beads were washed 6 times with 1 ml IP buffer and DNA fragments were extracted by boiling in Chelex ( Bio-Rad ) . qPCRs were carried out with the following primers: HIV-nuc1-F ( 5'-CTGGGAGCTCTCTGGCTAACTA-3' ) , HIV-nuc1-R ( 5'-TTACCAGAGTCACACAACAGACG-3' ) ; HIV-env-F ( 5'-TGAGGGACAATTGGAGAAGTGA-3' ) , HIV-env-R ( 5'-TCTGCACCACTCTTCTCTTTGC-3' ) ; MYC-C-F ( 5’-GCGCGCCCATTAATACCCTTCTTT-3’ ) , MYC-C-R: ( 5’-ATAAATCATCGCAGGCGGAACAGC-3’ ) ; IκBα 5’end-F ( 5'-AAGAAGGAGCGGCTACTGGAC-3' ) , IκBα 5’end -R ( 5'-TCCTTGACCATCTGCTCGTACT-3' ) ; HERVK-LTR-F ( 5'-GGGCAGCAATACTGCTTTGT-3' ) , HERVK-LTR-R ( 5'-CAATAGTGGGGAGAGGGTCA-3' ) ; intergenic-F ( 5'-CTCCCAAATTGCTGGGATTA-3' ) , intergenic-R ( 5'-ATTCCAGGCACCACAAAAAG-3' ) . For ChIP in NH1 cells ( previously generated by Zhou lab based on human HeLa cell line [49] ) containing an integrated HIV-1 LTR-luciferase reporter construct , the cells seeded in 6-well plates were first transfected in duplicate by either 2 μg empty vector or pCMV2-based vectors containing N-terminally FLAG-tagged Brd4 long isoform ( Brd4L , isoform A , amino acids [aa] 1–1 , 362 ) , or Brd4 short isoform ( Brd4S , isoform C , aa 1–722 ) . Both Brd4 vectors were kind gifts from Melanie Ott lab ( Gladstone Institutes , San Francisco , [50] ) . 31 hours after transfection , DMSO or MG-149 were added to the medium to the final concentrations of 0 . 1% and 30 μM respectively . After 18 hr of drug treatment , about 106 cells/well were subjected to ChIP as described above . The ChIP was conducted using anti-FLAG beads ( Sigma , A2220 ) , and the qPCRs were carried out with the following primers: LTRS-1 ( 5’-GTTAGACCAGATCTGAGCCT-3’ ) , and LTRS-2 ( 5’-GTGGGTTCCCTAGTTAGCCA-3’ ) . Signals obtained by qPCR were normalized to those of input DNA , and the averages from triplicated qPCR reactions were shown with the error bars representing standard deviations . Two-tailed Student's t-tests were conducted and the different significance levels were marked by 1 to 3 asterisks . The HIV-1 infection and progressive latency establishment assay were based on Pearson et al . [21] with some modifications . Briefly , HIV-1 virions were produced by transfecting HEK 293T cells with the wild-type pHR’-p-d2EGFP plasmid[21] using a 3rd generation lentiviral packaging system and then used to spin-infect wild-type Jurkat ( from American Type Culture Collection ) or Jurkat pools expressing the indicted shRNAs in the presence of 6 μg/ml polybrene ( Santa Cruz ) . Two days post infection ( d . p . i . ) , GFP ( + ) cells ( of the HIV-infected Jurkat cells described above ) were enriched by flow cytometry and allowed to propagate in culture . The percentages of GFP ( + ) cells in this population ( called the total pool in Fig 1 ) were monitored continuously until the indicated d . p . i . On 29 d . p . i . , a population of GFP ( - ) cells were enriched by flow cytometry from the original GFP ( + ) cells selected on 2 d . p . i . and maintained as the latent pool in Fig 1 . On 43 d . p . i . , the latent pool was subjected to latency reversal assays as described above with the indicated concentrations of LRAs , and both the total and latent pools were subjected to Western blots and ChIP assays for the indicated proteins .
|
A major impediment to the cure of HIV/AIDS is the viral latency . Previous studies have identified the bromodomain protein Brd4 as a promoter of HIV latency by binding to the viral LTR to inhibit Tat-induced transcription . Here , we discover that the LTR of latent HIV has low acetylated histone H3 ( AcH3 ) but high AcH4 content , which recruits Brd4 to inhibit Tat-transactivation . Furthermore , the lysine acetyltransferase KAT5 but not the paralogs KAT7 and KAT8 promotes latency through acetylating H4 on the provirus . Antagonizing KAT5 removes AcH4 and Brd4 from the LTR , enhances loading of the Super Elongation Complex , and interferes with the establishment of latency . Thus , the KAT5-AcH4-Brd4 axis is a key regulator of HIV latency and a potential therapeutic target for eradicating latent HIV reservoirs .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"blood",
"cells",
"flow",
"cytometry",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"pathology",
"and",
"laboratory",
"medicine",
"evolutionary",
"biology",
"pathogens",
"immunology",
"population",
"genetics",
"microbiology",
"dna-binding",
"proteins",
"gene",
"pool",
"dna",
"transcription",
"retroviruses",
"viruses",
"immunodeficiency",
"viruses",
"rna",
"viruses",
"population",
"biology",
"research",
"and",
"analysis",
"methods",
"spectrum",
"analysis",
"techniques",
"white",
"blood",
"cells",
"animal",
"cells",
"proteins",
"medical",
"microbiology",
"hiv",
"gene",
"expression",
"microbial",
"pathogens",
"t",
"cells",
"hiv-1",
"histones",
"spectrophotometry",
"cytophotometry",
"biochemistry",
"cell",
"biology",
"viral",
"persistence",
"and",
"latency",
"virology",
"viral",
"pathogens",
"genetics",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"lentivirus",
"organisms"
] |
2018
|
The KAT5-Acetyl-Histone4-Brd4 axis silences HIV-1 transcription and promotes viral latency
|
Reorganization of the microtubule network is important for the fast isodiametric expansion of giant-feeding cells induced by root-knot nematodes . The efficiency of microtubule reorganization depends on the nucleation of new microtubules , their elongation rate and activity of microtubule severing factors . New microtubules in plants are nucleated by cytoplasmic or microtubule-bound γ-tubulin ring complexes . Here we investigate the requirement of γ-tubulin complexes for giant feeding cells development using the interaction between Arabidopsis and Meloidogyne spp . as a model system . Immunocytochemical analyses demonstrate that γ-tubulin localizes to both cortical cytoplasm and mitotic microtubule arrays of the giant cells where it can associate with microtubules . The transcripts of two Arabidopsis γ-tubulin ( TUBG1 and TUBG2 ) and two γ-tubulin complex proteins genes ( GCP3 and GCP4 ) are upregulated in galls . Electron microscopy demonstrates association of GCP3 and γ-tubulin as part of a complex in the cytoplasm of giant cells . Knockout of either or both γ-tubulin genes results in the gene dose-dependent alteration of the morphology of feeding site and failure of nematode life cycle completion . We conclude that the γ-tubulin complex is essential for the control of microtubular network remodelling in the course of initiation and development of giant-feeding cells , and for the successful reproduction of nematodes in their plant hosts .
Root-knot nematodes ( RKN ) Meloidogyne spp . are minuscule worms which are widespread in the soil . They are obligate sedentary phyto-endoparasites known to infect above 3000 plant species . In the course of a compatible interaction , the nematodes of the genera Meloidogyne are able to alter the host plant metabolic pathways to their own benefit [1] . The parasitic cycle commences when the motile second-stage juvenile ( J2 ) penetrates a root in the elongation zone [2] . This infective stage nematode migrates via intercellular space of the root cortex towards the root tip and then moves acropetally along a xylem pole to the differentiation zone of the root vascular tissue where it establishes the feeding site by altering the developmental and metabolic program of the vascular parenchymal cells [3] . A typical nematode feeding site ( NFS ) consists of 6 to 8 hypertrophic cells , named giant cells , with dense cytoplasm containing numerous organelles and characterised by high metabolic activity [4] . These cells serve as the exclusive source of nutrients for the nematode until their reproduction . In the course of differentiation giant cells undergo karyokinesis followed by incomplete cytokinesis as well as endoreduplication cycles , resulting in the accumulation of multiple enlarged nuclei [5] , [6] . This is accompanied by the partial depolymerisation/fragmentation of both main components of plant cytoskeleton: microtubules ( MTs ) and actin filaments [7] . The reorganization of the cytoskeleton is essential for establishment of the feeding site and successful nematode reproduction [7]–[9] . Microtubules are dynamic filaments formed by polymerization of heterodimeric protein α-/β-tubulin . They are essential for the spatial organization of the cytoplasm , establishment of the cell shape and polarity , cell division , intracellular transport and cell wall deposition . In plants MTs form four functionally specialized arrays: 1 ) interphase cortical network regulates the cell architecture including the direction of cell expansion; 2 ) preprophase band ( PPB ) during G2/M transition predicts site of the forthcoming division [3 , 10 ) mitotic spindle separates daughter chromatides; 4 ) phragmoplast mediates trafficking of components required for the cell-plate synthesis during cytokinesis . The organization of these arrays requires initiation of new MTs , their elongation , shrinking , severing and bundling with other MTs . Initiation of new MTs occurs on structures called MT-organizing centres ( MTOCs ) [11] . In animals , centrosomes serve as MTOC during both interphase and cell division . Higher plants lack a conspicuous MTOC and new MTs are nucleated from multiple dispersed sites [12] . A key component of MTOCs is γ-tubulin , an evolutionary conserved homologue protein of α- and β-tubulin [13]–[16] . γ-Tubulin localizes to the MT nucleation sites of interphase and dividing plant cells . There are two γ-tubulin genes in the genome of Arabidopsis thaliana and their transcripts were observed in seedlings , roots , flowers and tissue culture cells [17] . Using heterologous expression in fission yeast Horio and Oakley [18] have shown that Arabidopsis γ-tubulin was targeted to MTOCs and was able to nucleate MTs . Downregulation or knockout of both genes causes disorganization of cortical microtubule network , spindle and phragmoplast [19] , [20] . Thus , plant γ-tubulin plays an essential role in MT organization at all stages of the plant cell cycle . In active MTOCs , γ-tubulin associates with five proteins forming the γ-tubulin ring complex , or γTuRC [11] . Collectively , six proteins are called γ-tubulin complex proteins ( GCPs ) , with γ-tubulin itself being GCP1 . The γTuRC binds to MT minus ends and prevents it from depolymerisation [21] . The Arabidopsis thaliana genome contains orthologues of all components of mammalian γTuRC: two γ-tubulin genes ( TUBG1 and TUBG2 ) and γ-tubulin complex protein genes GCP2 to GCP6 [22] . GCP2-GCP6 proteins may function as a scaffold for the interaction between 13 γ-tubulin molecules and the MT minus end . Electron microscopy revealed an open ring structure containing γ-tubulin clusters and similar clusters have been found on the minus ends of MTs [23] . In fungal and animal cells , components of the γTuRC preferentially appear at the spindle pole body and the centrosome [24] . In plant cells , γ-tubulin is spatially not restricted to MT ends but also colocalizes along MTs [25] , [26] where it nucleates new MTs [27] . Plant GCP proteins are also required for nucleation of MTs . For example two core γTuRC components GCP2 and GCP3 decorate the nuclear envelope of tobacco ( Nicotiana tabacum ) BY-2 suspension cells and are required for MT nucleation and cell division [28] , [22] , [29] . Recently it has been shown that GCP4 is associated in vivo with γ-tubulin in Arabidopsis thaliana being an essential component for the function of γ-tubulin in MT nucleation and organization in plant cells [30] . However , the role of other GCP proteins in functional plant γTuRC remains unknown . The precise coordination of the feeding site establishment and microtubule reorganization suggests that RKN can control the host's cytoskeleton . An example of this control is synchronous assembly of multiple disorganised and enlarged spindles and misaligned phragmoplasts [7] . These phragmoplasts fail to assemble a cell plate and consequently result in the formation of multinucleate cells . Whether these abnormal arrays are simply the remnants of a prematurely aborted cytokinesis or the result of specific rearrangement of MTs and microfilaments in response to parasite factors requires further investigation . However , all data demonstrate that abnormal cytokinesis is principal event for the establishment of the feeding site and successful completion of the nematode life cycle [8] . In order to investigate the role of γ-tubulin in the rearrangements of the cytoskeleton in nematode feeding cells [7] we have carried out nematode infection tests on roots of γ-tubulin mutant lines and show here that knockouts of either gene delays feeding site development . Immunocytochemical analysis show that γ-tubulin protein localizes to the cell cortex and cytoplasm of feeding cells , the nuclear surface and along malformed phragmoplasts . Moreover , γ-tubulin co-localizes and makes a complex with a component of γTuRC , GCP3 [22] . Our data demonstrate that accelerated growth of giant cells and establishment of functional nematode feeding site requires functional γTuRC .
To explore the role of γTuRC for gall development , we determined transcription levels of four key members ( TUBG1 , TUBG2 , GCP3 and GCP4 ) by quantitative reverse transcriptase-mediated real-time PCR ( qRT-PCR ) . The total RNA for the assays was extracted from Arabidopsis roots infected with M . incognita at three stages of gall development: at young stage ( 7 days after inoculation-DAI ) , intermediate stage ( 14 DAI ) , at a mature stage of gall development ( 21 DAI ) and uninfected roots . Transcription levels augmented in galls for both γ-tubulin genes ( TUBG1 and TUBG2 ) preferentially at early developmental stage ( 7 DAI ) and for two GCPs ( GCP3 and GCP4 ) at intermediate stages ( 14 DAI ) of gall development ( Figure 1 ) . To address the functional significance of the upregulation of γTuRC genes during feeding site development , we investigated the morphology of galls at different developmental stages in 4 mutant lines ( tubg1-1 , tubg2-1 , tubg1-1 tubg2-2 and amiR-GCP4-9 ) ( Figure 2 ) . Infection tests were performed to evaluate the competence of nematode development and reproduction in mutant lines ( Figure 3 ) . Cortical , epidermal and root hair cells of uninfected roots , of tubg1-1 and tubg2-1 of mature seedlings ( 40 days after sowing-DAS ) were swollen , expanding isotropically ( Figure 2A , 2E and 2I ) but no perceptible phenotype was seen in the vascular tissue where galls develop . This phenotype was not detected in young roots ( 14 DAS ) used for nematode infection ( Figure S1; wild-type root 1A and mutant lines 1B to 1D ) and nematode penetration or infection occurred normally ( Figure S1E ) . Lateral root development also appeared similar in both wild-type and mutant lines . A lack of root phenotype in roots of both γ-tubulin mutant lines was also observed by Pastuglia et al . [20] . Based on these observations we predict that the mild phenotype present in mature uninfected roots ( 40 DAS ) should not influence on gall development . The abnormalities of root morphology and cells expansion were more pronounced in the double mutant ( Figure 2M ) . Although , knockdown of γ-tubulin genes had no discernable effect on the ability of nematodes to penetrate , migrate , and induce giant cells , feeding sites development was delayed at 7 DAI in both mutant lines ( Figures 2B , 2F and 2J ) . At 14 DAI giant cells in γ-tubulin mutants were smaller and contained enlarged vacuoles ( Figures 2C , 2G and 2K ) . At this stage nematodes often remained vermiform at stage 2 juvenile ( J2 ) whereas in wild-type parasitic J2 were larger . At 21 DAI , wild-type plants contained typical multinucleated cells with dense cytoplasm ( Figure 2D ) . In contrast , the infected roots of γ-tubulin mutant lines had smaller feeding cells ( Figures 2H and 2L ) with less nuclei and large vacuoles . The development of a fraction of nematodes was arrested at the parasitic juvenile stages . In the double mutant line ( tubg1-1 tubg2-2 ) infection process resulted in tiny giant cells containing an average of two nuclei ( Figures 2N ) . Uninfected roots of the GCP4 mutant line did not show any evident phenotype ( Figure 2O ) except for shorter root hairs . Young giant cells in the amiR-GCP4-9 line were smaller than in wild-type , contained a reduced number of nuclei and enlarged vacuoles ( Figure 2P ) . At later stages of development ( 14 DAI and 21DAI ) nematode development was delayed and giant cells remained small ( Figure 2Q at 14 DAI ) . Examination of giant cell morphology correlates with nematode reproduction . Assessment of the nematode life cycle in γ-tubulin mutant lines ( tubg1-1 and tubg2-1 ) showed arrest of half of galls development and delayed nematode maturation . Consequently , production of egg masses was significantly reduced ( Figure 3 ) . A similar analysis could not be performed with the double mutant since the plants are viable for only 3 weeks . The localization of γ-tubulin in galls of wild-type and mutant lines was analysed by immunocytochemistry ( Figure 4 ) using a polyclonal antiserum [20] . In order to locate both γ-tubulin proteins separately in galls , we used the mutant tubg1-1 for TUBG2 localization and , tubg2-1 for TUBG1 detection . In wild-type and mutant roots ( tubg1-1 , tubg2-1 and tubg1-1 tubg2-2 ) γ-tubulin staining was observed in all root cells ( Figure 4D ) . At 14 DAI galls of wild-type seedlings , γ-tubulin protein was localized throughout the giant-feeding cells and fewer label was detected in the neighbouring cells ( Figure 4A and Figure S2A and S2D ) . At the same stage of infection , roots of tubg1-1 revealed γ-tubulin staining in the giant-cell cortex and less in the cytoplasm ( Figure 4B and Figure S2B and S2E ) while in tubg2-1 mutant γ-tubulin was localized along the giant cell cortex and around the nuclei ( Figure 4C and Figure S2C and S2F ) . Weak γ-tubulin expression was seen in giant cells of tubg1-1 tubg2-2 ( Figure 4E ) . Immuno-gold analysis of the infected wild-type plants ( 14 DAI ) by electron microscopy demonstrate that γ-tubulin co-localize with MTs ( α-tubulin ) in the giant cell cortex ( Figure 4F ) , cytoplasm ( Figure 4G ) , at the nuclear surface ( Figure S3B and S3B' ) and with the phragmoplast during mitosis ( Figure 4H to 4H' ) . Cell wall fragments were visible at the MT ends as dark patches of severely misaligned phragmoplasts ( Figure 4H and 4H' ) . Free cytosolic γ-tubulin is apparent in Figure S3A . Crosses between γ-tubulin mutant lines ( tubg1-1 and tubg2-1 ) and marker lines expressing Pro35S-MBD:GFP and nuclear histone H2B:YFP were used to study microtubule organization in the mutant background ( Figure S4 ) . Mitotic spindles were bowed and chromosomes were often misaligned in both γ-tubulin mutant lines ( Figure S4A for tubg1-1 and 4B , 4C for tubg2-1 ) in comparison with wild-type root cells ( Figure 4D ) . GCP3 localization was observed in giant cells of wild-type ( Figure 5A ) as well as tubg1-1 ( Figure 5B ) and tubg2-1 ( Figure 5C ) plants concentrating around the nuclei and cell cortex . Co-localization of GCP3 and γ-tubulin was analysed using electron microscopy . Both proteins were found in the cytoplasm ( Figure 5D ) , at the nuclear surface ( Figure 5E and 5E' ) , and at the cell cortex ( Figure S5A and S5A' ) . Commonly , multiple colloidal gold particles corresponding to γ-tubulin and GCP3 grouped together forming compact clusters in the cytoplasm ( Figure 5D ) . We have measured the distance between closest gold particles and found that majority of them were in the proximity of less than 10 nm ( Figure 5F ) . These data provide compelling evidence for the interaction between GCP3 and γ-tubulin in giant cells as part of a single multi-protein complex . 35Spro:TUBG1-GFP and 35Spro:GFP-TUBG1 were transiently expressed in tobacco leaf cells using Agrobacterium infiltration . GFP fluorescence was observed in the Arabidopsis cytoplasm and in the nucleus for both constructs ( Figure S6 ) and discrete fluorescent dots were apparent in the cytoplasm and cell cortex . Arabidopsis seedlings were transformed with 35Spro:TUBG1-GFP and 35Spro:GFP-TUBG1 constructs and the F1 generation was analyzed for GFP expression . Seedlings containing the C-terminal fusion ( TUBG1-GFP ) showed the best germination efficiency . The roots of seedlings germinated on the vertical plates exhibit skewing to the left if observed from the top side of the plate ( Figure S6A ) and the leaves were curling ( Figure S7C ) compared to wild-type ( Figure S7B and S7D respectively ) . Confocal microscopy imaging of roots revealed a left-handed twisting phenotype ( Figure 6A and 6B ) . The twisting is likely to result from the miss-shaping of the root cells and disorganization of the root tissues as shown in a cross section ( Figure 6F compared to 6G ) . Sections of a shoot apex ( Figure 6H compared to 6I ) showed right-handed displacement of the cells . To verify if twisting was caused by the altered properties of MTs we treated γ-tubulin overexpressing seedlings with MT polymerisation inhibitors propyzamide and oryzalin . Both treatments inhibited root skewing and left-handed twist of the epidermal cell layers implying that ectopic expression of γ-tubulin increase stability of the microtubules leading to the organ twisting phenotype ( Figure 6C , 6D compared to a twisted root in 6E ) . Plants expressing γ-tubulin-GFP lacking a twisted phenotype but still showing GFP-fluorescence were used for localization studies . Transgenic root apical meristem and lateral root meristem displayed a patchy expression pattern ( Figure 6J and 6K ) that varied between different roots suggesting a post-transcriptional control of γ-tubulin expression . In mitotic cells γ-tubulin was localized to the spindle ( Figure 6L ) and the phragmoplast ( Figure 6M ) . During the interphase , multiple discrete foci and homogeneous fluorescence were observed in the cytoplasm ( Figure 6L inset and 6M ) . We investigated the mobility of γ-tubulin in the cytoplasm using fluorescence loss in photobleaching ( FLIP ) analysis . Most of the γ-tubulin-GFP fusion protein was found to be highly mobile ( Figure S8 ) suggesting that only a minor fraction of total γ-tubulin is utilised in the nucleation of microtubules . γ-Tubulin-GFP fluorescence was also observed in the cells of uninfected root vasculature ( Figure 7A ) as well as throughout all gall tissues ( Figure 7B ) . Young giant cells ( 3 DAI and 5 DAI ) showed concentration of γ-tubulin around the nuclei and diffuse fluorescence in the cytoplasm ( Figure 7C and 7D ) . At the later stages ( 7 DAI and 10 DAI ) γ-tubulin expression was observed in all gall and giant cells ( Figure 7E , 7E' and 7F ) and a speckled fluorescence was notable around the perinuclear cytoplasm ( Figure 7G ) . Young ( 7 DAI; Figure 7H ) as well as transitional and mature giant cells ( 14 DAI and 21 DAI; Figure 7I and 7J ) remained small and contained more nuclei than wild-type control ( Figure 7K ) . Mitotic chromosomes were detectable in giant cells of γ-tubulin overexpressing lines ( Figure 7I ) , but not in the wild-type ( Figure 7K ) . Area measurements on giant cells confirmed a decreased size compared to wild-type ( Figure 8A ) . Nuclei counts per section at the core of giant cells ( 7 DAI ) validated the observation of a larger number of nuclei per giant cell when under ectopic γ-tubulin expression ( Figure 8B ) . Infection tests have shown a decrease in gall number and egg mass production in γ-tubulin overexpressing lines ( Figure 8C ) .
Transcription analyses demonstrate augmentation of the γ-tubulin ( TUBG1 and TUBG2 ) and two γ-tubulin-complex protein genes ( GCP3 and GCP4 ) in the course of infection with root-knot nematodes in galls . This corroborates with in situ hybridization analysis which has shown the transcriptional activation of the γ-tubulin genes in giant feeding cells as well as in neighbouring tissues that derived from M . incognita targeted vascular root cells [7] . The transcription levels of TUBG1 and TUBG2 were higher at an early stage of gall development ( 7 DAI ) while GCP3 and GCP4 transcription increased at an intermediate stage ( 14 DAI ) . The transcription of three genes ( TUBG1 , TUBG2 and GCP3 ) declined in mature giant cells ( 21 DAI ) coinciding with the end of the mitotic activity in giant-feeding cells . In addition to γ-tubulin proteins , both GCP3 and GCP4 are indispensable components for the MT nucleating activity of γTuRC in plant cells [22] , [30] . This suggests that an optimum level of γ-tubulin and GCP in galls is necessary to provide a sufficient number of new microtubule nucleation sites for the remodelling of the MT network in giant-feeding cells and to support recurrent ongoing mitotic activity in both giant- and neighbouring cells [7] . Analysis of the T-DNA knockout lines demonstrates the importance of γ-tubulin for the establishment of the feeding site and completion of the parasite life cycle . No visible phenotype was observed in the vasculature of single knockout plants before infection by the root-knot nematodes in agreement with the partial functional redundancy of TUBG1 and TUBG2 proposed by Pastuglia et al . [20] and Binarova et al . [19] . Both double mutant tubg1-1-tubg2-2 and RNAi knockdown plants exhibit disruption of anisotropic cell expansion [19] , [20] in a similar manner to the drug-induced reduction of the number of nucleation sites and randomization of cortical MTs [31] . Although nematodes were able to penetrate , migrate , and induce gall formation in roots of tubg1-1 and tubg2-2 lines , there was a significant delay in feeding site formation , indicating that both γ-tubulin genes are required for proper nematode feeding site development . Consequently , the number of nematodes that could complete their life cycle and reproduce was significantly lower in both T-DNA lines as compared to wild-type . It has been demonstrated in γ-tubulin knockout lines that reduced MT nucleation can delay chromosome separation and nuclear proliferation , thereby inhibiting cell division and impairing growth polarity [32] . We have observed that residual levels of γ-tubulin in the double knockout line is sufficient for gall induction in agreement with the dose-dependent effect of γ-tubulin on the early feeding site development and underlines the functional significance of a rise of γ-tubulin gene transcription . Localization of γ-tubulin throughout giant cell differentiation illustrates its role during nematode infection . At the early stages of gall development ( 3 DAI ) γ-tubulin is mainly localized at the nuclear surface in giant-feeding cells as well as neighbouring cells . As giant cells matured , a scattered GFP fluorescence was also detected in the cytoplasm and cell cortex . Accumulation of γ-tubulin close to the nematode head suggests reorganization of the microtubule network in the region proximal to the secretion and/or feeding processes , consistent with the accumulation of ER and other organelles at this site ( de Almeida Engler , unpublished data ) . At this stage , nematodes are alternately injecting secretions and feeding on the cell cytoplasm . It is known that the cytoskeleton acts differently depending on the plant host and the invading pathogen , and that the MT response to infection is variable between different plant-microbe interactions [33] , [34] . So far , the molecular mechanisms regulating MT dynamics during host/parasite interactions is not well understood and our data suggest that increased γ-tubulin expression might be part of this apparatus . Immunolocalization experiments demonstrate that γ-tubulin and GCP3 co-localize each other and MTs around the nuclei , in the cell cortex and in mitotic MT arrays . Therefore dispersed free or MT associated γTuRCs in giant cells might provide nucleation of new MTs required for fast array reorganization in giant cells . γ-Tubulin was shown to associate with preprophase bands ( PPBs ) , spindle , phragmoplast and cortical cytoplasm of soybean , onion , Arabidopsis and cells of other species [17] , [25] , [35] , [36] , [37] . Each MT array appears to have multiple sites for the nucleation of new MTs which can be located along the lattice of extant MTs , resulting in branching of cortical MTs . This activity may result in the array composed of randomly oriented bundles observed in the giant cell cortex . In addition , γ-tubulin cooperates with other known regulators of MT organization including MOR1/GEM1 and MAP65 to regulate MT organization in giant cells [8] , [38] , [39] , [40] , [41] . Although γ-TuRC components are conserved in plant genomes , their association in a functional complex has not yet been proven . The pioneering study of Seltzer et al . has provided biochemical evidence for the association of Arabidopsis GCP2 , GCP3 and γ-tubulin in the cytoplasmic soluble complex [22] . Recently Nakamura et al . [42] have shown that all six members of the Arabidopsis γ-TuRC immunoprecipitate jointly . Immunogold electron microscopy shows that γ-tubulin and GCP3 are located less then 10 nm apart from each other proving further evidence for the existence of γ-TuRC in vivo and its close association with the MT lattice . Both GCP2 and GCP3 localize at the nuclear envelope and play a role in MT nucleation [22] , [28] and GCP2 can control organization of cortical MTs by positioning the γ-tubulin-containing complex on pre-existing MTs [29] , [43] . The presence of abnormal spindles in uninfected Arabidopsis roots of tubg1-1 and tubg2-2 lines agrees with the previous observations of collapsed or defective spindles and chromosome segregation defects in γ-tubulin mutants of several species such as S . pombe , S . cerevisiae and Drosophila melanogaster [14] , [44] , [45] , [46] , [47] . In addition , γ-tubulin depletion in A . nidulans abolishes nucleation of spindle MTs [48] . This suggests that γ-tubulins have conserved functions in organizing the spindle in phylogenetically distant organisms . Furthermore , a reduction of the γ-tubulin signal seen on spindle and phragmoplast in amiR-GCP4 cells suggests their interaction with these arrays during mitosis in plant cells [30] . The localization of γ-tubulin-GFP in living root cells corroborates with results obtained by immunostaining . Indeed , a strong GFP signal was observed on mitotic spindles and phragmoplasts during mitosis as well as during interphase around nuclei , a known site of MT nucleation . The apparent speckled fluorescence in the cytoplasm indicates the presence of discrete γTuRCs . A similar distribution pattern has been observed in BY-2 cells transiently expressing a GFP-γ-tubulin fusion protein . There , at the end of cell division γ-tubulin was firstly accumulated at the daughter nuclear surfaces and evenly spread along the cell cortex [49] , [50] . In animals , γ-tubulin is also important for the coordination of late mitotic events [51] and has a MT-independent function in mitotic checkpoint control [52] . Whether plant γ-tubulin performs similar functions remains unknown , but it can not be ruled out that reduction of nuclei number in the giant cells of the double mutant results from the deficiency in these activities . Our FRAP analysis indicate that the absolute majority of γ-tubulin-GFP freely distributes in the cytoplasm supporting the idea that microtubule nucleation events in plants are driven by γ-tubulin-containing complexes released from the nucleation sites and redistributed around the cell . Electron microscopy observations confirm the cytoplasmic distribution of γ-tubulin and the presence of a large number of closely associated γ-tubulin and GCP3 proteins pointing towards their interaction . Analysis of the ectopic γ-tubulin expression in plants shows a curling phenotype in leaves and skewing of roots , suggesting an alteration in dynamics and/or organization of MT arrays . Treatment of seedlings with oryzalin and propyzamide restored the normal pattern indicating that overexpression of γ-tubulin leads to MT stabilization or excessive nucleation . It is also plausible that ectopic expression of γ-tubulin in giant cells provokes higher mitotic activity by accelerating nuclear division . In addition , excessive nucleation of MTs might impede giant cell expansion , nematode feeding and interfere with gall development and eggmass production . Our previous reports suggested that cytoskeleton stabilization by drug treatment ( taxol ) or ADF2 downregulation can disturb gall development and consequently nematode maturation [7] , [9] . Taking into account present and previous data [7] we propose a model for MT dynamics in giant cells ( Figure 9 ) . The transcription analysis shows that two components of plant MTOCs ( γ-tubulin and GCPs ) are highly expressed in galls . Knockout of either TUBG1 or TUBG2 genes results in inhibition of gall and nematode development , demonstrating that both proteins are important for feeding site formation , while in non-infected plants gene redundancy has been observed [20] . Predominant localization of TUBG2 at the giant cell cortex compared to a more cytoplasmic distribution of TUBG1 suggests that they might exert different functions within these large feeding cells . Double knockout completely abolishes gall development . The reduction of γ-tubulin protein level compromises the MT network integrity . Our previous findings demonstrate the significance of MT and actin filaments during interphase as well as mitosis [7]-[9] for successful nematode infection and reproduction . γ-tubulin is crucial for the organization and function of mitotic spindle and phragmoplast . The overall effect of one or both γ-tubulin and the GCP4 genes knockout on the MT network of giant cells agrees with the hypothesis that γTuRC proteins including γ-tubulins and GCP4 are required for proper functioning of mitotic arrays and MT nucleation . Bowed spindles observed in root tip cells can be caused by abnormal nucleation of MTs . Since unusually shaped and enlarged spindles are normally observed in giant cells , these anomalies might be caused by the unbalance between γ-tubulin and GCP3 concentration in giant cells , ultimately resulting in aneuploid nuclei that failed to divide [5] , [53] . High γ-tubulin and GCP3 concentration may play a role in the formation of disarrayed phragmoplasts observed in giant-feeding cells and might contribute to the misalignment of the cell plate and arrest in giant cell division . The apparent fragmentation and reduction in density of interphase MTs in giant cells could result from defective nucleation . Nucleation , dynamics , and spatial organization of MTs are tightly coordinated processes in plant cells . Proteins secreted by nematodes may induce MT reorganization by altering MT dynamics . Up-regulation of γTuRC proteins in galls provides an excess of MT nucleation sites , which in combination with other factors might control MT dynamics leading to the reorganization of the entire array by site-specific destabilization/stabilization activity . Microtubule response to pathogen invasion depends on the type of plant–microbe interactions [33] and nematodes are the only known pathogens capable of inducing the long-term cytoskeleton rearrangements on their host plants . The local reduction of density of the cytoplasmic MTs may facilitate susceptibility of host plants to nematodes in a similar manner to other systems . For example , microtubule depolymerisation at the infection site has been observed in the soybean and parsley cells attacked by the oomycete Phytophthora sojae [54] , [55] . Stabilization of MTs by taxol blocks gall development , while breakdown permits nematode reproduction [7] . Although what stimulates cytoskeletal responses in nematode feeding cells is still not known , MT and actin rearrangements might be directly or indirectly induced by effectors secreted by the nematodes still to be identified . Here we show that a tight balance between MT nucleation and dynamics is required for the successful nematode infection and γTuRC is essential to exercise this balance . The knockout of the individual components of γTuRCs reduces the efficiency of MT nucleation and consequently affects gall development and inhibits nematode reproduction , while the overexpression of γ-tubulin causes overall stabilisation of the MT network and produces a similar effect on the nematode life cycle . Since upregulation of MT nucleation in developing giant-feeding cells is essential for nematode parasitism , the components of γTuRC can be envisaged as potential targets to design alternative strategies to control pathogen invasion and spread .
T-DNA mutagenized lines of the two γ-tubulin genes of Arabidopsis thaliana were obtained from the Versailles and SALK collections as described by Pastuglia et al . [20] . The CVP11 ( tubg1-1 ) and T628 ( tubg2-2 ) lines have been acquired from a T-DNA–mutagenised population of the ecotype Wassilevskija ( WS ) from Pastuglia et al . [56] . The SALK_004612 ( tubg2-1 ) line was acquired from the ABRC and originates from a T-DNA–mutagenised population of the Columbia-0 ( Col-0 ) ecotype [57] . Surface-sterilized seeds of WS , Col-0 and γ-tubulin knockdown lines were germinated on Murashige and Skoog medium ( Duchefa , Haarlem , the Netherlands ) containing 1% sucrose and 0 . 8% plant cell culture–tested agar ( Sigma-Aldrich ) and 50 mg/mL kanamycin for T-DNA mutagenised lines . After 10 days under growth chamber conditions of 16-h/8-h light/dark cycles at 25°C kanamycin resistant seedlings were transferred to Knop medium which favors root development [58] . Plates were inclined at an angle of 60° to allow the roots to grow along the surface to facilitate harvesting of roots and galls . For nematodes inoculation , around 100 surface-sterilized , freshly hatched second stage juveniles ( J2s ) of Meloidogyne incognita were adjusted on each 14 day old seedling as previously described [9] . For nematode infection tests seedlings were kept in Knop medium during 60 days to allow nematodes to complete their life cycle . At 14 days after inoculation galls were counted and after 60 days egg mass numbers were scored . Total RNA of non-meristematic roots and galls of A . thaliana cv . WS dissected at various time points after nematode inoculation ( 7 , 14 , 21 DAI ) was extracted according to the procedure described by Laroche-Raynal et al . [59] . One microgram of high-quality RNA was reverse-transcribed using the iScript cDNA Synthesis Kit ( Bio-Rad , Marnes-la-Coquette , France ) . Amplification and detection were performed in the Opticon 2 system ( MJ research , Bio-Rad ) . Reaction mixtures were of a final volume of 15 µL , containing 7 . 5 µL qPCR MasterMix Plus For SYBR Green I No Rox ( Eurogentec , Angers , France ) , 0 . 13 µM of each primer and 5 µL of 50-fold diluted cDNA templates . PCR conditions were as follows: 95°C for 15 min , followed by 40 cycles of 95°C for 15 s , 58°C for 30 s and 72°C for 30s . At the end of the programme , a melting curve ( from 65 to 95°C , read every 0 . 5°C ) was determined to ensure that only single products were generated . At5g10790 and At5g62050 were used for normalization of qRT-PCR data . These two genes were previously identified as showing constant expression in response to different biotic stimuli [60] , [61] , [62] . Raw data were treated using the MJ Opticon Monitor Analysis software ( version 3 . 1; Bio-Rad ) . Quantifications were performed with the modified ÄCt method employed by the qBase1 . 3 . 5 software , and expressed as normalized relative quantity . The primer list is given in Table S1 . Three independent quantitative RT-PCR reactions were carried out per sample and three biological replicates were performed . The bars represent mean values from three independent experiments . For morphology observation of uninfected roots and galls of γ-tubulin knockout lines compared to control WS and Col-0 ( tubg1-1 line was made in WS and tubg2-1 in Col-0; 20 ) , samples grown in vitro were fixed in 2% glutaraldehyde in 50 mM PIPES buffer , pH 6 . 9 , and then dehydrated and embedded in Technovit 7100 ( Heraeus Kulzer ) as described by the manufacturer . Embedded roots and gall tissues were sectioned ( 3 µm ) and stained in 0 . 05% toluidine blue and mounted in Depex ( Sigma-Aldrich ) . Microscopic observations were performed using bright-field optics and images were performed with a digital camera ( Axiocam , Zeiss ) . Uninfected roots and feeding sites of roots , inoculated with J2s of M . incognita , of A . thaliana cv . WS , Col-0 , tubg1-1 , tubg2-1 , and tubg1-1 tubg2-2 mutant lines were fixed in 4% formaldehyde in 50mM Pipes buffer ( pH 6 . 9 ) . Dissected galls ( 7 , 14 , and 21 DAI ) were dehydrated and embedded in butyl-methylmethacrylate as described by Kronenberger et al . [63] with some modifications . The immunolocalization procedure was performed essentially as described by de Almeida-Engler et al . [7] . Slides containing sectioned nematode feeding sites were incubated with acetone absolute for 30 min to remove the plastic . Primary and secondary antibodies were diluted 1∶300 and 1∶500 ( v/v ) respectively , in blocking solution ( 1% bovine serum albumin in 50 mM Pipes , pH 6 . 9 ) . Sections were incubated with blocking solution for 30 min , and overnight at 4°C with the primary antibodies . As controls , sections were incubated with pre-immune serum or without primary antibodies . Anti-γ-tubulin and anti-GCP3 were generated respectively as described by Pastuglia et al . [20] and Seltzer et al . [22] . First , incubation with a polyclonal anti-γ-tubulin or anti-GCP3 has been performed and slides were then washed in Pipes buffer for 15 min . Slides were subsequently incubated for 2 h at 37°C with the secondary antibody anti-rabbit Alexa 488 ( green fluorescence ) or Alexa 594 ( red fluorescence ) and washed in Pipes buffer for 15 min . For DNA visualization sections were stained with 4′ , 6-diamidino-2-phenylindole ( DAPI ) ( Sigma-Aldrich ) at 1 µg/ml in Pipes buffer . Slides were mounted in 90% glycerol in ddH2O and were observed with a microscope ( Axioplan 2; Zeiss ) equipped for epifluorescence microscopy and differential interference contrast optics , and images were collected with a digital Axiocam ( Zeiss ) . Root galls of A . thaliana cv . WS and Col-0 were dissected at 14 DAI after inoculation and fixed in a mixture of 1 . 5% glutaraldehyde , 3% paraformaldehyde in 10 mM PBS containing 150 mM NaCl ( pH 7 . 2 ) for 3 h at room temperature . After several washes in PBS buffer fixed galls were incubated in 0 . 5 M NH4Cl for 1 h , dehydrated in graded ethanol series , embedded in acrylic resin LR White ( Sigma ) , and polymerized overnight at 60°C . Ultrathin sections were collected on parlodion-coated nickel grids , treated with 0 . 1 M HCl for 10 min , and washed at least twice with bidistilled water . The grids were pre-incubated in 1% bovin serum albumin ( BSA ) in PBS for 15 min , prior to incubation with pre-immune goat serum ( Sigma ) diluted 1∶10 in BSA-PBS for 1 h . Immuno-labelling was performed with primary antibodies diluted with BSA-PBS . Double labelling was done by treatment of grids either with rabbit polyclonal anti-γ-tubulin ( 1∶500 ) and the monoclonal anti-α-tubulin ( 1∶500 ) , for 90 min at room temperature . The grids were washed 3 times for 5 min in BSA-PBS and incubated for 1 h with secondary antibodies ( 10 nm gold-goat anti-rabbit and 5 nm gold-goat anti-mouse antisera; British BioCell International ) diluted 1∶20 with BSA-PBS . Other grids containing gall sections were incubated in the following immunoreagents: first , grids were incubated with rabbit polyclonal antiserum anti-GCP3 diluted in BSA-PBS . Samples were washed 3 times for 5 min in BSA-PBS and incubated for 1 h with the secondary antibody ( 10 nm gold-goat anti-rabbit ) . Secondly , grids were incubated with the rabbit polyclonal antibody against γ-tubulin ( 1∶300 ) , washed 3 times for 5 min in BSA-PBS and incubated for 1 h with the secondary antibody goat anti-rabbit gold ( 5 nm ) . All grids were rinsed 3 times for 5 min in PBS , rinsed in bidistilled water and stained 5 min in 2% aqueous uranyl acetate and 2 min in 1% lead citrate . Samples were observed under a Philips 400 T electron microscope . Observation of MTs and nuclei in the nematode feeding sites in wild-type and mutant lines was performed in nematode infected A . thaliana seedlings harbouring the MT binding domain of MAP4 fused to -GFP and H2B Histone-YFP ( MBD::GFP- His::YFP ) . More than 50 root meristems and young galls ( 2 to 7 DAI ) were directly observed under the microscope . At least 50 galls at various time points after infection ( 7 to 15 DAI ) were dissected from roots and embedded in 5% agar . Fresh thick sections of 50–100 µm ( 7 to 10 DAI ) or 150–200 µm ( 10 to 14 DAI ) were performed with a HM650V vibrotome Microm ( Walldorf , Germany ) . Whole roots and fresh slices were observed using an inverted confocal microscope ( model LSM510 META; Zeiss ) . YFP and GFP fluorescence were monitored in Lambda mode with a 499 to 550 nm beam path ( 488 nm excitation line ) . The fluorescent dye Syto-82 ( Molecular Probes ) was used at 1 mM final concentration . GFP or YFP and Syto-82 fluorescence were monitored in Lambda mode using the 488 nm argon laser line excitation and spectral detection using 10 nm steps for emission between 499 to 620 nm . All observations were obtained from at least three independent experiments . The complete coding sequences of TUBG1 gene was amplified by PCR , using ecotype cDNAs as the template , the primers pairs attB1F [5′-AA AAA GCA GGC TTC- ( ACC ATG ) - ( 18/25 nt template-specific seq ) -3′] and attB2R [5′-A GAA AGC TGG GTG ( TTA ) - ( template-specific seq ) -3′] with adapter primers attB1F ( 5′-GGGACAAGTTTGTACAAA AAAGCAGGCT-3′ ) and attB2 R ( 5′-GGGGACCACTTTGTACAAGAAAGCTGGGT-3′ ) . These sequences were then inserted into plant expression vectors , using Gateway™ Technology ( Invitrogen ) and the pDONR207 donor vector ( Invitrogen ) . Cloning was carried out in Escherichia coli DH10β cells . Transient expression of γ-tubulin in tobacco leaves was performed in leaves of Nicotiana benthamiana by infiltrating Agrobacterium tumefasciens strain C58C1 harbouring the 35Spro:TUBG1-GFP and 35Spro:GFP-TUBG1 with a syringe and observations were performed 48 hours after infiltration on the confocal microscope ( Zeiss , LSM 510 ) . To generate transgenic plants expressing the γ-tubulin protein the 35Spro:TUBG1-GFP ( TUBG1-GFP ) plasmid was transformed into Agrobacterium tumefasciens strain C58C1 and Arabidospsis thaliana Col-0 were transformed by floral dipping [64] . Plants with γ-tubulin-GFP fluorescence which did not present a visible phenotype and showing less expression were used for localization studies . On the other hand , seedlings with root twisting harbouring γ-tubulin-GFP fluorescence and showing additional γ-tubulin-GFP expression were used for overexpression studies . To test the effect of MT cytoskeleton inhibitors on the γ-tubulin overexpressing line presenting twisted roots , TUBG1-GFP seeds were germinated on MS medium , transferred to the same medium supplemented with oryzalin ( 0 . 5 µM ) or propyzamide ( 5 µM ) and incubated overnight at room temperature . Control experiments were performed using the same medium in the absence of inhibitors . Treated and untreated roots were imaged with a confocal laser scanning microscope ( LSM510 META; Zeiss ) using the tile-scan mode with the 10x/0 . 3NA objective . GFP fluorescence was recorded using 488-nm laser excitation and the 505–530 nm BP emission filter . Arabidopsis Genome Initiative locus identifiers for the genes mentioned in this article are as follows: At3g61650 ( TUBG1 ) , At5g05620 ( TUBG2 ) , At5g06680 ( GCP3 ) and At3g53760 ( GCP4 ) .
|
Among plant pathogens , sedentary endoparasitic nematodes are one of the most damaging pests in global agriculture . Nematodes are greatly resistant due to a broad physiological variability , consequently difficult to fight against . The use of pesticides is highly pollutant to the environment and therefore new strategies must be envisaged . As nematodes induce fragmentation and long-term rearrangements of the plant cytoskeleton during infection , manipulation of cytoskeleton components necessary for parasitism could be used as targets for the development of resistant plants provoking the awareness of biotechnology companies and crop breeders in developing new strategies for the control of pathogen infection . Here we report the first stable γ-tubulin-GFP expressing plant line and provide compelling evidence for the physical interaction between components of the γTuRC , γ-tubulin and γ-tubulin-complex protein 3 ( GCP3 ) as part of free cytoplasmic and microtubules associated complexes . We show here that γTuRC is an essential component of the microtubule nucleation machinery during giant cell development . The reduction of γ-tubulin and GCP4 levels compromises γTuRC functioning and affects microtubule nucleation in giant-feeding cells , delaying their development and affecting nematode reproduction . We conclude that upregulation of microtubule nucleation induced by γTuRC is essential for the nematode parasitism and this process can be targeted in order to protect plants against nematode infection .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biotechnology",
"transgenic",
"plants",
"plant",
"biology",
"botany",
"anatomy",
"and",
"physiology",
"histology",
"plant",
"science",
"plant",
"morphology",
"plant",
"pathology",
"genetically",
"modified",
"organisms",
"biology",
"agriculture",
"physiology",
"plant",
"pathogens",
"molecular",
"cell",
"biology",
"plant",
"biotechnology"
] |
2011
|
Feeding Cells Induced by Phytoparasitic Nematodes Require γ-Tubulin Ring Complex for Microtubule Reorganization
|
B-Lymphotropic Polyomavirus ( LPyV ) serves as a paradigm of virus receptor binding and tropism , and is the closest relative of the recently discovered Human Polyomavirus 9 ( HPyV9 ) . LPyV infection depends on sialic acid on host cells , but the molecular interactions underlying LPyV-receptor binding were unknown . We find by glycan array screening that LPyV specifically recognizes a linear carbohydrate motif that contains α2 , 3-linked sialic acid . High-resolution crystal structures of the LPyV capsid protein VP1 alone and in complex with the trisaccharide ligands 3′-sialyllactose and 3′-sialyl-N-acetyl-lactosamine ( 3SL and 3SLN , respectively ) show essentially identical interactions . Most contacts are contributed by the sialic acid moiety , which is almost entirely buried in a narrow , preformed cleft at the outer surface of the capsid . The recessed nature of the binding site on VP1 and the nature of the observed glycan interactions differ from those of related polyomaviruses and most other sialic acid-binding viruses , which bind sialic acid in shallow , more exposed grooves . Despite their different modes for recognition , the sialic acid binding sites of LPyV and SV40 are half-conserved , hinting at an evolutionary strategy for diversification of binding sites . Our analysis provides a structural basis for the observed specificity of LPyV for linear glycan motifs terminating in α2 , 3-linked sialic acid , and links the different tropisms of known LPyV strains to the receptor binding site . It also serves as a useful template for understanding the ligand-binding properties and serological crossreactivity of HPyV9 .
The B-Lymphotropic Polyomavirus ( LPyV ) was originally isolated from African Green Monkey lymph node cultures [1] and attracted interest because of its narrow tropism for the human B-lymphoblastoid tumor cell line BJA-B . In addition , significant antibody binding to LPyV was observed in human sera , raising the possibility that LPyV , or an LPyV-like human polyomavirus , might be a human oncovirus [2] . Polyomaviruses are a growing family of non-enveloped , icosahedral DNA viruses . Several members of the polyomavirus family , such as Simian Virus 40 ( SV40 ) , can transform cells in culture and cause tumors in animals [3] , [4] . The recently discovered human Merkel Cell Polyomavirus ( MCPyV ) is clearly implicated in a human cancer [5] . However , infections by the human JC and BK Polyomaviruses ( JCPyV and BKPyV , respectively ) remain subclinical in healthy individuals and cause severe acute disease , not cancer , in immunocompromised patients [6] . It is not known whether LPyV is endemic in humans , but a closely related virus , Human Polyomavirus 9 ( HPyV9 ) , was identified in 2011 and detected in human serum , plasma , urine and skin [7] , [8] . Although there are no data yet on the pathogenicity of HPyV9 in the human population , the seroprevalence is 47% in adults [9] . The narrow cell tropism of LPyV has made it a paradigm for studying viral tropism . Attachment of the LPyV major capsid protein VP1 to receptors on host cells is clearly critical for the restricted tropism of the virus [10] . Sialic acid is a crucial component of the LPyV receptor as removal or modification of cellular sialic acids abolished or modulated LPyV cell binding and infection [10] , [11] , [12] , but the identity of the sialylated LPyV receptor and its role in tropism are not known . Sialic acids are a group of acidic monosaccharides that are based on neuraminic acid and decorate eukaryotic cell surfaces . The most prevalent sialic acid in humans is 5-N-acetyl-neuraminic acid ( Neu5Ac ) [13] . Sialic acids occur with different modifications and glycosidic linkages at the peripheral domains of a diversity of carbohydrate sequences of glycoproteins and glycolipids . Initial cell contacts by many viruses involve sialylated glycans [14] . In most cases , the interrelationship between the recognition of specific carbohydrate sequences by viruses and the effects on viral tropism and pathogenesis are only beginning to emerge . Several polyomaviruses use specific sialylated carbohydrates as receptors [15] , [16] , [17] . Structural studies have shown that carbohydrate receptors are bound in shallow grooves on the polyomavirus capsid surface , which is formed by 72 pentamers of the major capsid protein VP1 [17] , [18] , [19] , [20] , [21] . In order to advance an understanding of the receptor-binding properties of LPyV , we expressed and purified its VP1 pentamers and subjected them to screening on a glycan microarray featuring a diverse set of sialylated carbohydrates . We detected specific and restricted binding only to the short trisaccharide probes , 3′-sialyllactose ( 3SL ) and 3′-sialyl-N-acetyllactosamine ( 3SLN ) , and solved crystal structures of both glycans in complex with LPyV VP1 . The structures reveal a preformed , recessed binding site for sialic acid that is different in architecture and location from known sialic acid binding sites of other viruses , and that essentially buries the sialic acid . Due to the high level of sequence similarity to HPyV9 VP1 , we are also able to draw conclusions about the structure and receptor binding properties of this newly discovered human virus as well as its serological cross-reactivity with LPyV .
In order to elucidate the carbohydrate-binding specificity of LPyV in a controlled setting , we recombinantly expressed VP1 pentamers that are unable to assemble into capsids due to truncations of 27 and 66 residues at the N- and C-termini , respectively . These truncations do not affect the overall structure and receptor-binding properties of VP1 pentamers , as demonstrated by structure-function studies of related polyomavirus VP1 pentamers [17] , [20] , [21] , [22] and their comparison with structures of entire virus particles [18] , [23] . The purified pentamers were analyzed on a glycan microarray containing 117 sialylated and 6 non-sialylated lipid-linked oligosaccharide probes , representing sequences occurring on N- and O-linked glycoproteins as well as glycolipids . We detected LPyV VP1 binding signals above background to α2 , 3-sialylated probes bearing sequences related to 3SL and 3SLN ( Fig . 1 , probes 12 and 29 ) . Both 3SL and 3SLN are linear trisaccharides with sequences Neu5Ac-α2 , 3-Gal-β1 , 4-Glc and Neu5Ac-α2 , 3-Gal-β1 , 4-GlcNAc , respectively ( Fig . 1 and Table S1 ) . The two additional probes ( 31 and 33 ) that yielded signals are chemically synthesized derivatives of 3SLN with additional 6-N-acetyl and 6-N-benzoyl functional groups , respectively , at the Gal moiety . There was no binding to a 3SL-derived structure bearing 4-O-acetylated Neu5Ac ( probe 6 ) , nor to the α2 , 6-linked sialyl analogs of 3SL and 3SLN ( probes 77 and 81 ) or the α2 , 8-linked disialyl analog of 3SL ( probe 104 ) . A 3SLN analog with a β1 , 3 linkage between Gal and GlcNAc ( probe 27 ) did not elicit a signal . In addition , ganglioside sequences that are branched at the Gal residue ( e . g . probes 67 and 72 ) were not recognized . Taken together , glycan microarray analysis shows that binding to 3SL/3SLN is specific for defined linkages between each disaccharide , and only select modifications of individual sugar moieties can be tolerated . There were three unusual observations in LPyV binding on the microarrays: First , there were no binding signals detected with longer oligosaccharide sequences with the same α2 , 3-sialyl trisaccharide terminus , such as probes with sialyl-lacto-N-neo-tetraose sequences ( probes 42 and 43 ) and the sialyl-N-glycan ( probe 62 ) . Second , there was stronger binding to the ligand-positive probes at low levels ( 2 fmol per spot ) compared to higher levels ( 5 fmol per spot ) ( Fig . 1 ) . Third , strong LPyV binding was detected only to 3SL and 3SLN oligosaccharide probes that were linked to lipid by reductive amination and thus have ring-opened , reduced monosaccharide cores [24] . The probes with the same oligosaccharide sequences but prepared via oxime ligation without reduction [25] were not bound ( probes 13 and 30 ) , nor were glycosylceramides ( probes 8–11 ) . The reduced monosaccharide cores most likely increase the flexibility of the glycans , thereby facilitating engagement . We observed a related but less pronounced phenomenon with VP1 pentamers of SV40 , which also elicited lower binding signals for receptor oligosaccharides on glycosylceramides than to the same sequences in the neoglycolipids prepared by reductive amination ( Figure S1 ) . In order to provide a structural basis for the observed interactions , we solved crystal structures of unliganded LPyV VP1 as well as complexes of LPyV VP1 with 3SL and 3SLN at resolutions of 1 . 92 , 1 . 48 and 1 . 75 Å , respectively ( Fig . 2 , Table 1 ) . All crystals contained two VP1 pentamers in their asymmetric units , and each model comprises 10 polypeptide chains . LPyV VP1 forms a ring-shaped homopentamer in which the five monomers are arranged around a central five-fold axis and connected by extensive interaction surfaces ( Fig . 2A ) . At the core of each monomer , two antiparallel β-sheets , consisting of β-strands termed B , I , D , G and C , H , E , F , respectively , form a compact β-sandwich . To facilitate discussion of single residues in the symmetric , ring-shaped pentamer , one monomer will serve as reference monomer with no special designation; its clockwise and counterclockwise neighbors will be denoted cw and ccw , respectively . The LPyV VP1 structure contains one bound Ca2+ ion per chain ( Fig . 2A ) that is coordinated by the carboxylate group of E46 and the carbonyl group of S214cw as well as several water molecules . In addition , the carboxylate group of E217cw is close enough to the Ca2+ to engage in an ionic interaction . These residues are highly conserved among all polyomaviruses and form one of the two calcium binding sites in the SV40 virion that are important for capsid stability and regulating assembly [23] . In the virion , the Ca2+ coordination is completed by a glutamate residue of the incoming C-terminal arm from another VP1 pentamer . While the Ca2+ ions in our structures likely come from the crystallization solution containing 0 . 2 M calcium chloride , they nevertheless demonstrate that unassembled VP1 pentamers can weakly bind Ca2+ ions . We observed clear electron density for the oligosaccharide ligands in crystals soaked in 3SL and 3SLN . In contrast , crystals soaked in the same concentration of 6′-sialyllactose ( 6SL ) did not yield any electron density for the glycan ( data not shown ) , confirming the specificity found by glycan array screening . Each LPyV VP1 pentamer contains five oligosaccharide binding sites , which are located at the top of the pentamer ( Fig . 2A–C ) , corresponding to the outer surface of the virion . Only some of these binding sites were occupied with ligand , whereas access to the remaining binding sites was blocked by crystal contacts . Both compounds bound in essentially the same manner to each binding site . The linear 3SL and 3SLN chains can assume a range of possible conformations in solution due to rotational freedom of the glycosidic bonds [26] , [27] . LPyV VP1 binds both compounds in a conformation of the Neu5Ac-α2 , 3-Gal linkage that is preferred in solution ( mean torsion angles 65° , −26° ) [26] , [27] . The terminal Neu5Ac residue , which is best defined by electron density and has low temperature factors ( B-factors ) , inserts deeply into a cleft and engages the protein with multiple contacts ( Fig . 3A , B ) . The adjacent Gal makes fewer contacts and has elevated B-factors . The terminal Glc and GlcNAc residues of 3SL and 3SLN have the weakest electron density and are only visible in the best occupied sites in each structure ( Fig . 2B , C ) . Each carbohydrate binding site lies at the contact between two VP1 monomers and is formed by residues from the BC- , DE- and HI-loops of one monomer as well as the BCcw-loop of the clockwise neighboring monomer ( Fig . 2B , C ) . The long BC-loop can be divided into two substructures , termed BC1- and BC2-loop , that point into different directions . In contrast to most viral sialic acid-binding sites [14] , the LPyV site is not a shallow depression on the protein surface , but a deep cleft that contacts both faces of the Neu5Ac ring ( Fig . 3A , B ) , mostly via van der Waals interactions . Consequently , 80% of the accessible surface area of Neu5Ac ( 317 Å2 of a total of 399 Å2 ) is buried upon binding . The methyl group projects most deeply into the binding site and is surrounded by the side chains of L62 , Y65 , Q130 , H271 and V279 on three different VP1 loops ( Fig . 3A ) . These interactions bury the entire accessible surface ( 72 Å2 ) of the methyl group . In addition , there is a water-mediated hydrogen bond between the Neu5Ac carbonyl group and K75cw on the BC2cw-loop . The Neu5Ac binding pocket is characterized by a high level of surface complementarity for its ligand ( Fig . 3B ) . One face of Neu5Ac packs against the HI-loop of LPyV VP1 and makes van der Waals contacts with H271 and N273 . On the same face , the carboxylate group of Neu5Ac is recognized by a hydrogen bond to the backbone amine of Y274 and water-mediated hydrogen bonds to the backbone amine and side chain of S275 . The other face of Neu5Ac is covered with the hydrophobic part of the K75cw side chain in the BC2cw-loop , forming a lid that lies on top of the partially hydrophobic surface of the sugar ring . In addition to these interactions , residues Q130-G132 contact Neu5Ac from the rear of the binding site . The glycerol chain of Neu5Ac points away from the binding pocket , and its terminus adopts different conformations in different binding sites . However , the glycerol group engages in van der Waals contacts , a water-mediated hydrogen bond from O8 to T277 ( Fig . 3B ) and sometimes a water-mediated hydrogen bond from O9 to S68 . The Gal residue engages in fewer contacts that contribute about 25% of the total buried surface area . Contacts include a hydrogen bond between O2 of Gal and the backbone of S74cw and a water-mediated hydrogen bond between O4 of Gal and the backbone of F73cw , both in the BC2cw-loop . Moreover , there are van der Waals interactions with the Y274 side chain in the HI-loop ( Fig . 2A ) . The terminal Glc and GlcNAc residues are only observed in few binding sites in both complexes . They are within 5 Å of the side chains of F73cw and S74cw . Although the hydrophobic , solvent-exposed side chain of F73cw is not well defined by electron density , it lies close enough to the Glc or GlcNAc residues to allow for weak van der Waals interactions . It is not clear from the structure whether these strengthen binding or cause weak steric hindrance . In order to determine whether conformational changes occur in the protein during ligand binding , we compared the complex structures with the unliganded LPyV VP1 structure . In all chains in which the tip of the BC2cw-loop and adjacent residues are not engaged in crystal contacts , the structures are very similar , indicating that the binding site does not undergo a permanent induced fit movement . In two chains , however , a crystal contact perturbed the native structure of the BC2cw-loop , but not the ligand-bound one . Thus , complex formation might stabilize the receptor binding site , and there might be some flexibility in the unbound structure that allows ligand entry into the site . This hypothesis is supported by the B-factors of receptor-binding residues , which are elevated in the unbound structure compared with the complexes . The LPyV VP1 structures help to rationalize the specific recognition of the trisaccharide motif Neu5Ac-α2 , 3-Gal-β1 , 4-Glc ( NAc ) observed in glycan array screening . This motif can tolerate benzoyl or acetyl substituents at position 6 of the Gal ring , both of which are unlikely to interfere with binding ( Fig . 3B ) . Whereas Neu5Ac in the context of 3SL was readily bound , the analogous α2 , 6-linked oligosaccharide 6SL was not recognized in both glycan array and crystal soaking experiments . Due to the α2 , 6-linkage , the overall shape of the trisaccharide is somewhat kinked and differs from the linear orientation found in α2 , 3-linked compounds . An α2 , 6-linkage would likely lead to loss of the hydrogen bonds involving Gal and clashes with the BC2cw- or HI-loops of LPyV VP1 . Moreover , our structures explain the inability of LPyV VP1 to bind to an α2 , 8-linked disialic acid sequence . The Neu5Ac binding site can only accommodate terminal Neu5Ac , and a second Neu5Ac in a disialic acid sequence would occupy the place of the Gal residue , where it would cause steric clashes . Consistent with the results from glycan array screening , the β1 , 3 analog of 3SL would lead to clashes with protein residues in some conformations . Finally , LPyV VP1 does not bind to branched sequences that carry additional sugar residues attached to the Gal residue at position 4 , such as gangliosides GM1 or GD1a . In both conformations of the α2 , 3-linkage that have been observed in such compounds , steric clashes with the protein would occur if residues were added at that position . There are currently three different sequences of LPyV available in GenBank , which correspond to the K38 , L02 and LPV-76 strains , which are all based on the same isolate ( GenBank NC004763 . 1 ( used here ) , AAA47067 . 1 , AAA47076 . 1 , respectively ) [28] , [29] , [30] . The tropism of the K38 and L02 strains is restricted to proliferating B-lymphocyte lines , while the LPV-76 strain is also able to infect select T-lymphocyte lines [30] . The changes in tropism have been mapped to the VP1 proteins . The K38 and L02 VP1 proteins differ by three point mutations , while LPV-76 VP1 has three additional point mutations . Interestingly , the critical three amino acids linked to changes in tropism ( F73S , S139Y and V/S279N ) all cluster at the carbohydrate binding site ( Fig . 3C ) . Our structure shows that none of the mutations would directly block binding of 3SL or 3SLN . The V/S279N mutation would only influence interactions with the terminal sialic acid , maybe altering specificity for modified sialic acids . The hydrophobic , but entirely solvent accessible side chain of F73 ( Fig . 3C ) can adopt two different conformations in the native structure , one of which approaches the Glc ( NAc ) residue in the oligosaccharide complexes . Mutation to serine could take away this weak contact or relieve steric hindrance and allow binding of a different carbohydrate . Similarly , S139Y might interfere with the binding of longer or branched oligosaccharides . While this residue is not part of the primary sialic acid binding site , it lies directly adjacent to it ( Fig . 3C ) . Thus , our data would suggest that the different tropisms of the three strains are linked to small differences in receptor binding properties . Sialic acid is used by many viral attachment proteins as a tightly bound “hook” to grasp the oligosaccharide , while residues outside the sialic acid binding site modulate binding specificity for sialic acid in different contexts [14] , [17] , [31] . LPyV is the closest homolog of the recently discovered human polyomavirus HPyV9 [7] , [8] . The two VP1 proteins share 87% sequence identity , indicative of a high level of conservation at the structural level . HPyV9 had long been suspected to exist in the human population based on serological reactivity of human sera against LPyV [2] , [32] , and there is significant cross-reactivity for LPyV and HPyV9 in both human and African Green Monkey sera [9] . Surface-exposed LPyV VP1 residues show the same high level of sequence identity with HPyV9 VP1 as residues in the protein interior ( 87% ) . However , divergent residues are distributed unevenly on the LPyV VP1 surface , with most changes occurring on the top surface of VP1 , which would be most accessible to antibodies in the context of the virion ( Fig . 4 ) . The long BC-loop ( residues 55–85 ) , which contributes most to this surface , is especially divergent ( red line in Fig . 4 ) . Surface residues of the BC-loop are only 52% identical with HPyV9 VP1 residues , while BC-loop residues facing towards the interior are 100% identical in sequence . The inner surface of the ring-shaped pentamer ( Fig . 4C ) , which contacts the minor capsid proteins , and its side surface ( Fig . 4A ) , which forms contacts between VP1 pentamers during capsid assembly , are entirely conserved between LPyV and HPyV9 . Most residues in the LPyV VP1 sialic acid binding site are conserved in HPyV9 VP1 ( Fig . 4 ) , and none of the substitutions would sterically interfere with Neu5Ac binding . This suggests that HPyV9 is also capable of interacting with sialylated oligosaccharides , and that the two viruses might even share a similar sialic acid binding mode . Interestingly , two of the tropism-widening substitutions among different LPyV strains are also present in HPyV9 . However , the prediction of carbohydrate binding sites based on homology modeling alone remains challenging , and further studies are necessary to confirm the attachment of HPyV9 to sialylated oligosaccharides on host cells . The interaction of LPyV VP1 with Neu5Ac in a narrow , slot-like binding site is unique among polyomavirus-receptor complexes [17] , [20] , [21] , [22] , and also among the many other viruses that engage sialic acids [14] . There is only one direct hydrogen bond between LPyV VP1 and Neu5Ac , whereas there are typically at least four and up to seven direct hydrogen bonds in other virus-sialic acid complexes [14] . To achieve specificity for Neu5Ac , the LPyV binding site seems instead to rely more on shape complementarity and van der Waals contacts than on a distinct pattern of hydrogen bonds . Despite these differences , the LPyV binding site lies in a region that partially overlaps with the sialic acid binding sites on other polyomaviruses . LPyV engages Neu5Ac in an orientation that resembles that seen in the complex of SV40 VP1 with GM1 , and it is therefore useful to compare the two modes of interaction ( Fig . 5 ) . Interestingly , the two binding sites are “half-conserved” . The HI-loop , which contributes the “back” wall of the sialic acid binding site , is conserved , while the BC1- , BC2cw- and DE-loops feature marked differences , which explain the different orientations of bound Neu5Ac . In the SV40 VP1 complex , the side chain of F75cw is a central hydrophobic contact for the Neu5Ac methyl group . This residue would interfere with the binding of Neu5Ac in the orientation observed in the LPyV VP1 complex . In LPyV VP1 , replacement of F75cw with lysine , a different conformation of the BC2cw-loop and a more distant DE-loop create a recessed surface with an especially deep and narrow pocket that can accommodate Neu5Ac . As a consequence of these changes , the conserved residues of the HI-loop engage in different contacts with the two Neu5Ac orientations ( Fig . 5 ) . Taken together , the comparison highlights how the architecture of the sialic acid binding site , constructed from several loops , can be varied by mutation of some modules while conserving others . With LPyV , there are four known orientations of sialic acid in polyomavirus binding sites . Except for the LPyV-SV40 pair , none of the amino acids that contact them are conserved , but they tend to occupy equivalent positions in sequence and in structure . The observed partial conservation might be a general strategy for evolving binding sites with new properties through functional intermediates that minimize the risk of losing binding altogether .
In this study , we have established the linear , short sequence Neu5Ac-α2 , 3-Gal-β1 , 4-Glc ( NAc ) as a binding motif for the LPyV attachment protein VP1 , solved X-ray structures of LPyV VP1 in complex with two cognate oligosaccharide ligands , and defined contacts in the sialic acid binding site . Neu5Ac clearly serves as the primary point of contact for LPyV , in agreement with previous data showing that LPyV binding and infection were dramatically decreased upon neuraminidase treatment of cells [10] , [33] . Cells treated with modified sialic acid precursors and thus bearing modified sialic acids were found to no longer support LPyV infection if the sialic acids contained long acyl chains on the N-substituent [11] . These longer chains would clash with the side chains of Y65 or V279 in the sialic acid binding site ( Fig . 3B ) . Cells that had incorporated 9-iodo-Neu5Ac and 5-N-fluoroacetyl-Neu5Ac exhibited increased LPyV infection compared with cells carrying unmodified Neu5Ac [12] . In 9-iodo-Neu5Ac , the outermost hydroxyl group of the glycerol chain is replaced with the bigger and more hydrophobic iodine . The iodine could interact favorably with a hydrophobic patch on the LPyV VP1 surface formed mainly by the L62 side chain ( Fig . 3B ) . This patch might also conceivably interact with parts of the naturally occuring 9-O-acetyl Neu5Ac , which was not present in our arrays in the context of 3SL or 3SLN . 5-N-fluoroacetyl-Neu5Ac carries a polar fluorine attached to the N-acetyl group , which could likely be accommodated by the binding site defined in our structure . The natural host of LPyV is the African Green Monkey , in which the predominant sialic acid , 5-N-glycolyl neuraminic acid ( Neu5Gc ) carries an additional hydroxyl group attached to the N-acetyl group . Like 5-N-fluoroacetyl-Neu5Ac , Neu5Gc could also be accommodated by LPyV VP1 , reflecting the host preference of the virus . Glycan array and structural analyses show that LPyV VP1 specifically recognizes linear trisaccharides terminating in α2 , 3-linked sialic acid . It was previously shown that LPyV and SV40 , which binds the branched α2 , 3-sialylated glycolipid GM1 , do not compete for receptors on host cells [10] . This finding can easily be rationalized because the branched GM1 sequence would clash with LPyV VP1 residues . Interestingly , the short binding sequence we describe here contrasts with the longer glycan sequence required by JCPyV [17] , which may be quite restricted in cellular expression . The 3SL/3SLN sequence on the other hand is present on glycoproteins and glycolipids of many cell types . Why then is there a narrow cell tropism of LPyV in human cell lines ? The trisaccharide 3SL corresponds to the glycan portion of the ganglioside GM3 , a possible receptor candidate . However , the Glc residue in GM3 would most likely be buried in the head group layer of the membrane [34] , and modelling suggests that membrane-bound GM3 would clash with LPyV residue F73 when the Neu5Ac is inserted into the binding pocket . It is however possible that GM3 can be engaged in certain contexts , for example when linked to a specific membrane anchor [35] . The 3SL/3SLN trisaccharide sequence could also be part of a longer , yet uncharacterized oligosaccharide chain . Further work is required to determine whether the receptor for LPyV is a glycolipid with a particular ceramide moiety as discussed in [36] or possibly a glycoprotein that displays a particular cluster of sialyl trisaccharides and that perhaps also contributes to tropism . Based on our microarray data , the narrow tropism could also be , at least in part , due to a requirement for a particular mode of the presentation of the short sialyl motif to elicit binding of the VP1 pentamer , similar to what has been observed in other cases [36] . On the array , the glycans have flexible linkages to lipid , and the non-covalent attachment of the probes to the nitrocellulose matrix in the presence of carrier lipids [37] provides them with an element of mobility that enables them to be presented in the required geometry as long as they are not too densely packed . Assuming that the glycan probes are distributed uniformly over the surface of a spot , the distances between one glycan and a neighbouring glycan would be 20–30 Å at 5 fmol/spot , and 40–50 Å at 2 fmol/spot . Interestingly , these distances are in the same range as the distances between two binding sites on a pentamer , which are 30 Å for adjacent and 47 Å for non-adjacent sites . It is therefore at least conceivable that a 5 fmol/spot concentration of glycans does not allow for an effective interaction of more than one glycan with the LPV pentamer , perhaps for steric or entropic reasons . The unusual ligand-binding site of LPyV may therefore reflect a strategy to efficiently engage less densely packed , more accessible ligands at the cell surface .
DNA coding for amino acids 28–301 of LPyV VP1 ( K38 strain , GenBank accession no . NC 004763 . 1 ) was amplified by PCR and cloned into the pET15b expression vector ( Novagen ) in frame with an N-terminal hexahistidine tag ( His-tag ) and a thrombin cleavage site . The protein was overexpressed in E . coli BL21 ( DE3 ) and purified by nickel affinity chromatography . For glycan array screening and crystallization , the protein was further purified by gel filtration on Superdex-200 . For crystallization , the His-tag was cleaved with thrombin before the gel filtration step , leaving the non-native amino acids GSHM at the N-terminus . After gel filtration , the protein was kept in a buffer comprised of 20 mM HEPES pH 7 . 5 , 150 mM NaCl and 10 mM DTT . The microarray was composed of 123 sequence-defined lipid-linked oligosaccharide probes: 117 sialyl-terminating probes and 6 neutral probes as negative controls ( Glycosciences Array Set 30–31 , Table S1 ) . The probes were robotically printed in duplicate on nitrocellulose-coated glass slides at 2 and 5 fmol per spot using a non-contact instrument [38] , [39] . The his-tagged recombinant VP1 protein was pre-complexed with mouse monoclonal anti-poly-histidine ( Ab1 ) and biotinylated anti-mouse IgG antibodies ( Ab2 ) ( both from Sigma ) in a ratio of 4∶2∶1 ( by weight ) . In brief , the LPyV VP1-His tagged protein-antibody pre-complexes were prepared by pre-incubating Ab1 with Ab2 for 15 min at ambient temperature , followed by addition of VP1 and incubation for an additional 15 min on ice . The VP1-antibody complexes were diluted in 5 mM HEPES pH 7 . 4 , 150 mM NaCl , 3% ( w/v ) bovine serum albumin ( Sigma ) and 5 mM CaCl2 , to give a final VP1 concentration of 150 µg/ml in the presence of 2 . 2 mM DTT , and overlaid onto the arrays at 20°C for 2 h . Binding was detected with Alexa Fluor-647-labelled streptavidin ( Molecular Probes ) ; imaging and data analysis was as described [38] , [40] . LPyV VP1 was crystallized by sitting drop vapor diffusion at a concentration of 7 mg/mL against a reservoir comprising 22% ( v/v ) isopropanol , 0 . 1 M sodium acetate pH 4 . 8 and 0 . 2 M calcium chloride . For complex formation , LPyV VP1 was crystallized by hanging drop vapor diffusion and microseeding at a lower protein concentration of 3 . 8–4 . 25 mg/mL against a reservoir that containing a lower isopropanol concentration of 12% ( v/v ) . Crystals were harvested into the respective reservoir solutions and cryoprotected by soaking them for 10 s in reservoir solution containing 25% ( v/v ) ethylene glycol . They were then flash-frozen in liquid nitrogen . For complex formation , crystals were soaked in reservoir solution supplemented with 40 mM 3′-sialyllactose ( Dextra , UK ) or 40 mM 3′-sialyllactosamine ( Carbosynth , UK ) for 45 and 10 min , respectively . They were then cryoprotected in reservoir solution supplemented with 25% ( v/v ) ethylene glycol and 40 mM oligosaccharide , and flash-frozen . Diffraction data were collected at ESRF ( Grenoble , F ) ( beamline ID23-1 ) and at SLS ( Villigen , CH ) ( beamlines X06SA and X06DA ) . Data were processed with xds [41] , and the structure was solved by molecular replacement with Phaser in CCP4 [42] , [43] using the β-sandwich core of the mPyV VP1 pentamer structure ( 1VPS ) as a search model [22] . The crystals belong to space group C2 with two pentamers in their asymmetric unit ( Table 1 ) . After rigid body and simulated annealing coordinate refinement in Phenix [44] , missing parts of the model such as the surface loops appeared in electron density maps and could be built in Coot [45] . Refinement proceeded by alternating rounds of restrained coordinate , isotropic B-factor and TLS refinement in Phenix or Refmac5 [46] , and model building in Coot . The non-crystallographic symmetry relating the ten LPyV VP1 monomers in the asymmetric unit was used as a restraint throughout refinement . In data from soaked crystals , the ligands were located in weighted 2 mFo-DFc and mFo-DFc electron density maps . The carbohydrates were refined using restraints from the CCP4 library , with the exception of the α2 , 3-glycosidic bond , which had to be user-defined . Waters were incorporated using Coot and ARP/wARP . The final models have good stereochemistry and low Rfree values [47] ( Table 1 ) . Residues 28–98 and 106–297 could be modeled for all 10 chains in all structures , with one loop being disordered and additional residues at the C-termini visible in a subset of chains . In addition , the vector-encoded sequence SHM was observed at the N-terminus in all copies . Coordinates and structure factor amplitudes were deposited with the RCSB data bank ( www . rcsb . org ) with entry codes 4MBX ( unliganded LPyV VP1 ) , 4MBY ( complex with 3SL ) and 4MBZ ( complex with 3SLN ) . Figures showing the X-ray structures were prepared with PyMol ( Schrödinger Inc . ) .
|
Viruses must engage specific receptors on host cells in order to initiate infection . The type of receptor and its concentration on cells determine viral spread and tropism , but for many viruses , the receptor and the mode of recognition by the virus are not known . We have characterized the structural requirements for receptor binding of B-lymphotropic polyomavirus ( LPyV ) . This virus was originally isolated from African Green Monkey lymph node cultures and attracted interest because of its narrow tropism for a human tumor cell line . LPyV is also the closest relative of the recently discovered Human Polyomavirus 9 ( HPyV9 ) . We screened the LPyV coat protein VP1 on an carbohydrate microarray and found that it specifically recognizes a linear sugar motif that terminates in α2 , 3-linked sialic acid . We then determined the structures LPyV VP1 bound to these carbohydrates . The protein has a preformed , deeply recessed binding site for sialic acid . The binding site differs in both architecture and mode of recognition from the binding sites of other viruses . LPyV only binds linear carbohydrates that are able to penetrate into the binding slot .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2013
|
Structures of B-Lymphotropic Polyomavirus VP1 in Complex with Oligosaccharide Ligands
|
Since the major outbreak in 2007 in the Yap Island , Zika virus ( ZIKV ) causing dengue-like syndromes has affected multiple islands of the South Pacific region . In May 2015 , the virus was detected in Brazil and then spread through South and Central America . In December 2015 , ZIKV was detected in French Guiana and Martinique . The aim of the study was to evaluate the vector competence of the mosquito spp . Aedes aegypti and Aedes albopictus from the Caribbean ( Martinique , Guadeloupe ) , North America ( southern United States ) , South America ( Brazil , French Guiana ) for the currently circulating Asian genotype of ZIKV isolated from a patient in April 2014 in New Caledonia . Mosquitoes were orally exposed to an Asian genotype of ZIKV ( NC-2014-5132 ) . Upon exposure , engorged mosquitoes were maintained at 28°±1°C , a 16h:8h light:dark cycle and 80% humidity . 25–30 mosquitoes were processed at 4 , 7 and 14 days post-infection ( dpi ) . Mosquito bodies ( thorax and abdomen ) , heads and saliva were analyzed to measure infection , dissemination and transmission , respectively . High infection but lower disseminated infection and transmission rates were observed for both Ae . aegypti and Ae . albopictus . Ae . aegypti populations from Guadeloupe and French Guiana exhibited a higher dissemination of ZIKV than the other Ae . aegypti populations examined . Transmission of ZIKV was observed in both mosquito species at 14 dpi but at a low level . This study suggests that although susceptible to infection , Ae . aegypti and Ae . albopictus were unexpectedly low competent vectors for ZIKV . This may suggest that other factors such as the large naïve population for ZIKV and the high densities of human-biting mosquitoes contribute to the rapid spread of ZIKV during the current outbreak .
Zika virus ( ZIKV; family Flaviviridae , genus Flavivirus ) was first isolated from a sentinel rhesus monkey in the Zika forest of Uganda in 1947 [[1]] . Since then , it has emerged outside of its natural range of distribution , Africa and Asia: Yap Island ( Micronesia ) in 2007 [2] , French Polynesia in 2013 [3] , New Caledonia in 2014 [4] , Easter Island in 2014 [5] , the Cook Islands in 2014 [6] and more recently , northeastern Brazil in May 2015 [7 , 8] , the starting point of a pandemic in the Americas with 26 American countries reporting active ZIKV transmission ( http://www . cdc . gov/zika/geo/active-countries . html ) . Although reports indicate that most infections produce mild signs and symptoms of rash , fever , arthritis or arthralgia , and conjunctivitis , recent infections have been associated with more severe disease outcomes with neurological or auto-immune complications such as Guillain-Barre syndrome [9] and microcephaly ( http://www . cdc . gov/zika/pdfs/possible-association-between-zika-virus-and-microcephaly . pdf ) . This virus has a high potential for geographic expansion into countries where Aedes spp . mosquitoes are present notably Aedes aegypti mosquitoes . The primary vectors of ZIKV in Africa are Aedes mosquitoes with reported viral isolations from Ae . africanus and Ae . apicoargenteus [10] , Ae . luteocephalus [11] , Ae . furcifer and Ae . taylori [12] , and Ae . vittatus [13] . The human-biting mosquito Ae . aegypti is usually considered as a laboratory-competent vector of ZIKV [14] and viral isolations were reported from the species in the wild [13 , 15 , 16] . However , transmission of ZIKV by African Ae . aegypti has been unexpectedly low to null [17] , underlining the importance of genetic delineation of mosquito populations on vector competence [18 , 19] . In addition , Aedes albopictus has also been shown to be an efficient laboratory vector of ZIKV [20] , with viral isolations from field-collected mosquitoes [21] . This positive-sense , single-stranded RNA virus of 10 , 794-nt is composed of three major lineages: East African , West African , or Asian [22] . The Asian genotype is responsible for the current expansion of ZIKV in the Americas [22–24] . As the outcome of transmission depends on the specific pairing of vector and pathogen genotypes [25] , we investigated the vector competence of populations of Ae . aegypti and Ae . albopictus from the Caribbean ( Martinique , Guadeloupe ) , North America ( southern United States ) , South America ( Brazil , French Guiana ) for an Asian genotype of ZIKV .
The Institut Pasteur animal facility has received accreditation from the French Ministry of Agriculture to perform experiments on live animals in compliance with the French and European regulations on care and protection of laboratory animals . This study was approved by the Institutional Animal Care and Use Committee ( IACUC ) at the Institut Pasteur . No specific permits were required for the described field studies in locations which are not protected in any way and did not involve endangered or protected species . Seven populations of mosquitoes ( 5 populations of Ae . aegypti and 2 of Ae . albopictus; ( Table 1 ) from the Caribbean ( Martinique , Guadeloupe ) and continental America ( southern United States , French Guiana , Brazil ) were collected as larvae or using ovitraps . Eggs were hatched in dechlorinated tap water and larvae were reared under controlled conditions of 150–200 larvae per 1 liter and fed with one yeast tablet renewed every 3–4 days . Adults were kept in cages at 28°±1°C with a 16h:8h light:dark cycle , 80% relative humidity , and supplied with a 10% sucrose solution . The F1-F2 generation of mosquitoes was used for infection assays except for Ae . aegypti from Orlando ( > F10 ) and Ae . albopictus from Vero Beach ( F7 ) . The ZIKV strain ( NC-2014-5132 ) was isolated from a patient in April 2014 in New Caledonia . Viral stocks were prepared after five passages of the isolate onto Vero cells maintained at 37°C; cell infection was tracked by observation of cytopathic effect ( CPE ) . Supernatants were collected and the viral titer was estimated by serial 10-fold dilutions on Vero cells expressed in TCID50/mL . The virus stock was divided into 1 mL aliquots and stored at—80°C until use . Partial sequences of the NC-2014-5132 strain showed that it is phylogenetically related to the Asian genotype as are ZIKV strains circulating in the South Pacific region [26] and Brazil [7] . Indeed , based on the NS5 gene sequence , the NC-2014-5132 strain exhibited 99 . 4% identity with ZIKV from Brazil ( Dupont-Rouzeyrol , personal communication ) . Seven day-old females were fed an infectious blood-meal containing 1 . 4 mL of washed rabbit erythrocytes and 700 μL of viral suspension supplemented with a phagostimulant ( ATP ) at a final concentration of 5 mM . For each population , 4–6 boxes of 60 mosquitoes each were exposed to the ZIKV NC-2014-5132 strain . The titer of infectious blood-meals was 107 TCID50/mL . After the infectious blood-meal , engorged females were transferred to small containers and fed with 10% sucrose in a chamber maintained at 28°±1°C , a 16h:8h light:dark cycle and 80% humidity . For each population , batches of 25–30 mosquitoes were analyzed at 4 and 7 days post-infection ( dpi ) . Additionally , Ae . albopictus from Vero-Beach ( VRB ) and Ae . aegypti from Tubiacanga ( TUB ) were examined at 14 dpi . Each mosquito was processed as follows: abdomen and thorax were examined to estimate infection , head for dissemination and collected saliva for transmission . Abdomen and thorax , and head were individually ground in 300 μL of DMEM medium supplemented with 2% fetal bovine serum ( FBS ) . Homogenates were centrifuged at 10 , 000 g for 5 min before titration . Saliva was collected from individual mosquitoes as described in [27] . Briefly , wings and legs of each mosquito were removed and the proboscis was inserted into a 20 μL tip containing 5 μL of FBS . After 45 min , FBS containing saliva was expelled in 45 μL of DMEM medium for titration . Infection rate ( IR ) refers to the proportion of mosquitoes with infected body ( abdomen and thorax ) among tested mosquitoes . Disseminated infection rate ( DIR ) corresponds to the proportion of mosquitoes with infected head among the previously detected infected mosquitoes ( i . e; abdomen/thorax positive ) . Transmission rate ( TR ) represents the proportion of mosquitoes with infectious saliva among mosquitoes with disseminated infection . Transmission efficiency ( TE ) represents the proportion of mosquitoes with infectious saliva among the total number of mosquitoes tested . Body and head homogenates were serially diluted and inoculated onto monolayers of Vero cells in 96-well plates . Cells were incubated for 7 days at 37°C then stained with a solution of crystal violet ( 0 . 2% in 10% formaldehyde and 20% ethanol ) . Presence of viral particles was assessed by detection of CPE . Saliva was titrated on monolayer of Vero cells in 6 well plates incubated 7 days under an agarose overlay . Titers of saliva were expressed as pfu ( plaque-forming unit ) /saliva . All statistical tests were conducted using the STATA software ( StataCorp LP , Texas , USA ) . Rates were compared using Fisher’s exact test and sample distributions with the Kruskal-Wallis test . P-values>0·05 were considered non-significant .
To define whether Ae . aegypti or Ae . albopictus were more likely to sustain a ZIKV outbreak , we analyzed the susceptibility to infection , as well as the ability of the virus to establish disseminated infection at 4 and 7 dpi in the two mosquito species collected from sites where they coexist , Rio de Janeiro ( Brazil ) and Florida ( United States ) ( Fig 1A ) . When examining infection rates ( IR ) ( Fig 1B ) , Ae . aegypti ( TUB and ORL ) were more likely to become infected than Ae . albopictus ( JUR and VRB ) ( p < 0 . 001 ) . Whereas the two Ae . aegypti populations examined behaved similarly ( TUB versus ORL , p > 0 . 05 ) , Ae . albopictus VRB were more infected than Ae . albopictus JUR ( p = 10−3 ) ; infection rates at 4 and 7 dpi were lowest for Ae . albopictus JUR ( N = 7 positive among 30 tested ) . When analyzing dissemination of infected mosquitoes ( Fig 1C ) , disseminated infection rates ( DIR ) were low at 4 and 7 dpi regardless of the mosquito species and the collection site ( p > 0 . 05 ) . Transmission determined by detecting the presence of virus in mosquito saliva was not observed at early dpi ( 4 and 7 ) for any mosquito populations . At late dpi ( 14 ) , IRs , DIRs , TRs , and TEs were examined for Ae . aegypti TUB and Ae . albopictus VRB . Ae . aegypti TUB displayed higher IR and DIR than Ae . albopictus ( IR: Ae . aegypti TUB: 76 . 7% ± 7 . 8 versus Ae . albopictus VRB: 50% ± 9 . 3 , Fig 2A; DIR: Ae . aegypti TUB: 60 . 7% ± 10 . 4 versus Ae . albopictus VRB: 13 . 3% ± 9 . 1 , Fig 2B ) . When examining the saliva of Ae . aegypti TUB and Ae . albopictus VRB at 14 dpi , TRs and viral load in saliva were higher for Ae . albopictus VRB ( TR: 50% ± 50 , Fig 2C; viral load: 134 ± 0 ( mean ± SE ) , data not shown ) compared to Ae . aegypti TUB ( TR: 21 . 4% ± 11 . 4 , Fig 2C; viral load: 18 . 7 ± 10 . 3 , data not shown ) , even if it was not significant ( P = 0 . 383 ) . We should note that the number of mosquitoes examined for transmission was low despite the 30 mosquitoes initially examined . Thus we calculated the transmission efficiency showing that TEs drastically decreased to 3 . 3% ± 3 . 3 for Ae . albopictus VRB and 10% ± 5 . 5% for Ae . aegypti TUB ( Fig 2D ) suggesting that these two species were less competent to ZIKV than expected . Ae . aegypti is present in French Guiana , Guadeloupe and Martinique , but where no Ae . albopictus has yet been reported ( Fig 3A ) . We therefore determined the ability of Ae . aegypti from these territories to become infected and disseminate virus after oral exposure to ZIKV . Infection rates were high and similar regardless of dpi , and the mosquito population ( Fig 3B; p > 0 . 05 ( 0 . 025 at 4 dpi and 0 . 133 at 7 dpi ) ) . However , DIRs were reduced as compared to infection rates ( Fig 3C ) . Viral dissemination rate , however , increased significantly with dpi except for Ae . aegypti MAR . No viral transmission was observed at 4 and 7 dpi . When comparing all five Ae . aegypti populations at 7 dpi , DIRs were significantly different ( P = 0 . 002 ) and two homogeneous groups could be distinguished: Ae . aegypti from Guadeloupe and French Guiana ( P = 0 . 803 ) with higher DIRs compared to Ae . aegypti from Martinique , USA and Brazil ( P = 0 . 609 ) .
Zika virus has recently started to spread outside its natural range of distribution . After the South Pacific islands ( Yap island , French Polynesia , New Caledonia , Cook islands , Easter island; [28]] ) , ZIKV was detected in the South American continent: Brazil in May 2015 [7] , and since , 26 American countries ( http://www . cdc . gov/zika/geo/active-countries . html ) . The first autochthonous cases were recently recorded in Martinique and French Guiana where the mosquito Ae . aegypti was assumed to be the unique vector . Our study showed that the Asian genotype of ZIKV infected and was disseminated by the vectors Ae . aegypti and Ae . albopictus collected in the Caribbean and continental America . Furthermore , we showed that Ae . aegypti from Rio de Janeiro in Brazil and Ae . albopictus from Vero Beach in the United States were able to transmit ZIKV at 14 dpi . Although susceptible to infection , these populations were unexpectedly low competent vectors for ZIKV . After the emergence of chikungunya virus ( CHIKV ) from East Africa [29] followed by its worldwide expansion and establishment in the Americas of the Asian lineage since October 2013 [30] , ZIKV became a second example of emergence of a vector-borne disease threatening a new continent . Both viruses originated in Africa where they circulate in an enzootic cycle involving non-human primates and a wide variety of zoophilic mosquitoes [17 , 31] . Human outbreaks due to CHIKV involve anthropophilic vectors such as Ae . aegypti and to a lesser extent , Ae . albopictus . This latter species has been shown to be capable of transmitting at least 26 arboviruses in the laboratory and its implication as a main vector became a reality with the recent CHIKV pandemic [32] . Ae . albopictus transmits preferentially a CHIKV variant presenting an amino-acid change in an envelope glycoprotein [33 , 34] . This viral variant was selected after passing through the midgut barrier , the first step in mosquito infection [35] . We showed that Ae . albopictus VRB restrained ZIKV dissemination highlighting the importance of barriers such as the midgut in Ae . albopictus mosquitoes . The examination of ZIKV infection in Ae . aegypti TUB underlined the significant role of salivary glands in transmission . Therefore , the specific role of salivary glands on ZIKV transmission by Ae . aegypti and the passage of the virus in this mosquito compartment should be explored more in detail as it has been done with CHIKV in Ae . albopictus [36] . However , the proportion of mosquitoes capable of transmitting ZIKV on the total number of tested mosquitoes , was unexpectedly low suggesting that these two species were poorly competent to ZIKV . We also demonstrated that Ae . albopictus from Florida was at least two times more susceptible to ZIKV infection than Ae . albopictus collected in Rio de Janeiro , underlining differences depending on the mosquito population described under genotype-by-genotype ( G x G ) interactions where the outcome of infection depends on the specific pairing of vector and pathogen genotypes [37] . Additionally , the two Ae . albopictus populations examined from the Americas exhibited lower susceptibilities to ZIKV than Ae . albopictus from tropical Asia ( i . e . Singapore ) [20] . Zika disease can be confused with dengue fever and chikungunya fever , all transmitted by Ae . aegypti and Ae . albopictus . Viremia in patients was lower when infected with ZIKV , i . e . 103−106 RNA copies/mL [23] compared to viremia for dengue virus ( DENV ) ( 106−107 RNA copies/mL; [38] and CHIKV ( 107−109 RNA copies/mL; [39] ) . We thus expected a longer extrinsic incubation period ( EIP ) associated with the lower viremia . EIP corresponds to the time necessary for the virus to reach the mosquito saliva after an infectious blood-meal [40] . An increase of the blood-meal viral titer has been demonstrated to decrease the length of the EIP . For other flaviviruses than ZIKV , yellow fever virus and DENV , viral particles started to be detected in salivary glands of Ae . aegypti at 10 dpi [41] and at 7–9 dpi [42 , 43] , respectively . With ZIKV , we showed an EIP longer than 7 days with a blood-meal at 107 TCID50/mL to Ae . aegypti and Ae . albopictus from the Americas . [41] . Of note , artificial feeding systems usually need higher viral titers to reproduce infection rates observed when mosquitoes fed on viremic hosts [44] . Capable of inducing even higher viremia , CHIKV has been mainly associated with very shorter EIP , e . g . 2 days [27] . Therefore ZIKV does not present the same features as CHIKV in mosquito populations from the Americas with a longer EIP; this longer EIP would allow a broader window for implementation of vector control measures . Surveillance and control measures against ZIKV transmission in the Americas and more specifically , in Brazil , the starting point of the Zika outbreak , mainly use measures implemented for dengue control focused on Ae . aegypti ( http://portalsaude . saude . gov . br/images/pdf/2015/dezembro/09/Microcefalia---Protocolo-de-vigil--ncia-e-resposta---vers--o-1----09dez2015-8h . pdf ) . However , if ZIKV is able to infect and be transmitted by other mosquito species ( e . g . Culex spp . ) , their role in transmission would need to be defined to help design of more adapted vector control strategies aiming to impair the spread of the Zika outbreak in the continent . The recent introduction of ZIKV in the Americas and its rapid spread across the continent and the Caribbean is likely attributable to the globalization of trades and travels and also the ability of local Ae . aegypti and Ae . albopictus to disseminate and then to transmit the Asian genotype of ZIKV . Contrary to the scenario with CHIKV , longer EIPs of ZIKV in the populations examined would allow implementation of more adapted vector control measures leading to improved limitation of this new emerging threat to human health in the Americas . Nevertheless , both Ae . aegypti and Ae . albopictus in the Americas do not appear to be highly efficient vectors of ZIKV , which may be balanced by the large number of susceptible humans and their close contacts with Aedes mosquitoes .
|
Zika virus ( ZIKV ) is an emerging mosquito-borne arbovirus causing dengue-like symptoms . This virus was commonly detected in Africa and Asia . Since its emergence in Yap Island in Micronesia in 2007 , ZIKV reemerged in the South Pacific region in 2013 and ultimately reached the American continent in 2015 . The human biting mosquito Aedes aegypti and the less anthropophilic Aedes albopictus have been incriminated as vectors of ZIKV . Our study showed that American populations of Ae . aegypti and Ae . albopictus were able to become infected and disseminate ZIKV within the mosquito general cavity at early days ( 4 , 7 ) post-infection ( dpi ) . Nevertheless , transmission was unexpectedly low and only detected at 14 dpi . Our findings will help in designing more adapted vector control strategies and limiting the impact of a new emerging threat on human health in the Americas as did the chikungunya in 2014 .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"united",
"states",
"invertebrates",
"medicine",
"and",
"health",
"sciences",
"body",
"fluids",
"chikungunya",
"infection",
"geographical",
"locations",
"saliva",
"animals",
"north",
"america",
"caribbean",
"insect",
"vectors",
"guadeloupe",
"infectious",
"diseases",
"south",
"america",
"epidemiology",
"brazil",
"disease",
"vectors",
"insects",
"arthropoda",
"people",
"and",
"places",
"mosquitoes",
"anatomy",
"physiology",
"biology",
"and",
"life",
"sciences",
"viral",
"diseases",
"martinique",
"organisms"
] |
2016
|
Differential Susceptibilities of Aedes aegypti and Aedes albopictus from the Americas to Zika Virus
|
Nuclear pore complexes ( NPCs ) form gateways for material transfer across the nuclear envelope of eukaryotic cells . Disordered proteins , rich in phenylalanine-glycine repeat motifs ( FG-nups ) , form the central transport channel . Understanding how nups are arranged in the interior of the NPC may explain how NPC functions as a selectivity filter for transport of large molecules and a sieve-like filter for diffusion of small molecules ( < or ) . We employed molecular dynamics to model the structures formed by various assemblies of one kind of nup , namely the 609-aa-long FG domain of Nsp1 ( Nsp1-FG ) . The simulations started from different initial conformations and geometrical arrangements of Nsp1-FGs . In all cases Nsp1-FGs collectively formed brush-like structures with bristles made of bundles of 2–27 nups , however , the bundles being cross-linked through single nups leaving one bundle and joining a nearby one . The degree of cross-linking varies with different initial nup conformations and arrangements . Structural analysis reveals that FG-repeats of the nups not only involve formation of bundle structures , but are abundantly present in cross-linking regions where the epitopes of FG-repeats are highly accessible . Large molecules that are assisted by transport factors ( TFs ) are selectively transported through NPC apparently by binding to FG-nups through populated FG-binding pockets on the TF surface . Therefore , our finding suggests that TFs bind concertedly to multiple FGs in cross-linking regions and break-up the bundles to create wide pores for themselves and their cargoes to pass . In addition , the cross-linking between Nsp1-FG bundles , arising from simulations , is found to set a molecular size limit of < for passive diffusion of molecules . Our simulations suggest that the NPC central channel , near the periphery where tethering of nups is dominant , features brush-like moderately cross-linked bundles , but in the central region , where tethering loses its effect , features a sieve-like structure of bundles and frequent cross-links .
Nuclear pore complexes ( NPCs ) are large protein assemblies embedded in the nuclear membrane that provide the only conduit for exchange of molecules between cytoplasm and nucleoplasm , a process defined as nucleocytoplasmic transport ( NCT ) . While the membrane bound nucleus of a eukaryotic cell protectively envelopes the genetic material , separating it from the cytoplasm , processes such as DNA transcription require access to the genetic material in the interior of the nucleus by a myriad of molecules , including large molecules such as proteins . Two modes exist for the NCT: passive diffusion of molecules , smaller than 40 kDa or 9–10 nm in diameter [1] , and selective transport into and out of the nucleus for larger molecules such as proteins , RNA , ribosomal subunits , the larger molecules being recognized by transport factors ( TFs ) , carrier proteins , which shuttle back and forth ferrying their cargo through the NPC . NPCs are giant complexes having molecular masses ranging from ∼65 MDa ( in yeast ) to ∼125 MDa ( in higher eukaryotes ) that are made of copies of 30 different proteins termed nucleoporins ( nups ) [2]–[5] . Recently , a detailed 3D model for the position and abundance of each nup in the Saccharomyces ( S . ) cerevisiae NPC structure was proposed based on experimental data obtained from molecular , biochemical and structural studies of the NPCs and their components [6] , [7] . In the modeled structure , the scaffold of the NPC is formed by two protein subcomplexes that , through linker proteins , anchor a set of FG-containing nups [6] . The complex has 8-fold symmetry about its central axis and 2-fold symmetry about the equatorial plane such that each nup is repeated 8- , 16- , 32- , or 48-fold . Altogether , the NPC forms a multi-protein complex of nearly ∼450 proteins [4]–[7] . About one-third of all pore proteins constitute the barrier proteins also known as FG-nups ( F and G represent amino acids phenylalanine and glycine , respectively ) . FG-nups are intrinsically disordered , rich in FG-repeat motifs [4] , [5] , [8]–[10] and form the central transport channel of the NPC extending , however , also into the cytoplasmic and nucleoplasmic space . While the FG-nups extending toward cytoplasm and nucleoplasm sides are asymmetric in distribution , the FG-nups ( eg . , Nsp1 ) of the central region have copies distributed symmetrically around the scaffold of the NPC and are critical for bidirectional NCT . The FG-nups are present in different lengths ( a few to several hundred amino acids long ) and exhibit different amino acid composition and repeat motifs , typically FG , GLFG , or FxFG motifs ( x being any amino acid , largely serine ) separated by linker regions of 10–20 hydrophilic amino acids [4] , [9] , [10]; their different spatial localizations render the central channel heterogeneous . These nups play a central role in selective transport of the NPC and this selectivity is attained by bearing specific binding sites ( the FG motifs ) for the TFs that undergo multiple , low-affinity interactions with nups as they permeate their way through the channel [11]–[17] . Given that the central transport channel is filled with heterogeneous and natively unstructured FG-nups that are not susceptible to crystallization for structure determination , the important question how the unstructured FG nups are assembled in the interior to promote efficient and selective gating could not be answered unequivocally . Several models have been proposed so far to explain the structure of the interior of the NPC . According to the virtual gate model [4] , [18] , the tethered , unstructured FG-repeat proteins form an entropic barrier that repels non-specific cargo , preventing it from reaching the interior of the channel , whereas TFs that interact with the FG-repeats of nups in the NPC interior channel have higher probability of entering and permeating the channel . An alternative model , namely the polymer brush model [19] , [20] , is based on the above model suggesting that the FG-nups form extended brush-like polymers with inter-linked bristles that reversibly collapse upon binding of TFs . TFs pass the channel by repetitive binding and unbinding to the nups' FG repeats until they reach the exit of the pore [21] . In contrast , the selective phase/hydrogel model [22] proposes that FG nups in the central channel achieve a sieve-like meshwork through weak inter-repeat FG-FG hydrophobic interactions to form a hydrogel within the central channel [23] . TFs bind to the FG-repeats via low-affinity interactions which transiently melt these FG-FG cross-links of the meshwork , allowing thus only TFs to permeate the channel . Furthermore , non-specific cargo that cannot bind is filtered out [23] , [24] . Another model , the reduction of dimensionality ( ROD ) model , proposes that the TFs act as ferries , carrying cargo and sliding on the surface of phenylalanine-glycine ( FG ) motifs by interacting with the FGs according to a 2D random walk rather than a 3D diffusion [25] . The ‘forest’ model , later modified to the tube gate model [26] , [27] , proposes the interior of the NPC to be a hydrogel and the periphery to be brush-like , featuring two separate zones of traffic with distinct physiochemical properties . In their study , based on the hydrodynamic radius of individual nups localized in different regions of the NPC , the respective authors classify nups as ( a ) ‘shrubs’ that form cohesive domains with collapsed-coil and low charge content; ( b ) ‘extended coils’ that form high-charge content , non-cohesive domains; ( c ) ‘trees’ having features of both ( a ) and ( b ) . Thus , a ‘forest-like’ structure of FG-nups is formed in the NPC . In this model , the collapsed-coil domains are hypothesized to form a transport zone 1 in the central pore and the extended-coil domains form a peripheral zone 2 . The forest model is based on the observations of individual nup conformations , not considering interactions between nups . In a recent study on the role of nups in NPC gating , Tagliazucchi et al [28] deduced a potential of mean force of TF transport from the amino acid ( positive , negative , hydrophobic ) distribution and claimed that this potential can explain the use of TFs for traffic across the nuclear envelope . The claim is based , however , on the assumption that nups are individually random and do not form any structures in the channel interior as suggested by the earlier models of the NPC interior discussed above . The authors also do not attribute an explicit role to TFs interacting with nups through FG motifs . Most models propose an FG network as the main mechanism for pore selectivity and address to a much smaller degree a role for the actual molecular structures formed by the disordered nups of the NPC . Likely , the qualitative differences of the models can be reconciled with the heterogeneity found along entire transport pathways through the NPC that furnishes different environments on the cytoplasmic side , in the interior and on the nucleoplasm side of the NPC . To identify computationally the structure formed by the nups in the interior of the NPC pore , Miao and Schulten [20] carried out molecular dynamics ( MD ) simulations of nup polymers , using coarse-grained MD . Unlike the other approaches arriving at models for NPC gating , their approach included details of individual amino acids and , more importantly , included the interactions between nup polymer chains . In the simulations these chains were end-grafted to a flat surface in a 5×5 array . The MD simulations suggest that the nups have a strong tendency to form bundles of typically 2–6 proteins and that the bundles are inter-linked in a mesh-like manner as single nups cross frequently from one bundle to another one . From this behavior results a brush structure that is made of bundles of interlinked bristles . This structure can explain to some degree the gating of NPC pores: the structure can be passed by small biomolecules readily while larger biomolecules need to alter the structure to pass through . This role , namely of cutting through the mesh-like structure , could be played by transport factors . While the structure model of Miao and Schulten is highly suggestive , the study had obvious shortcomings imposed at the time by limitations in available computer power . One shortcoming is the short length ( 100-aa ) of nups employed that is in contrast to the actual , much longer ( about 600-aa ) lengths of many nups . A second shortcoming is that only a single type of nup tethering was investigated , namely tethering to a flat surface . A third shortcoming is that the simulated ensemble started with all nups initially in a highly ordered , namely fully-extended , straight chainconformation . Enjoying today in the form of petascale computers tremendously more computer power than was available to Miao and Schulten we have performed for the present study microsecond-long simulations on various conformational ensembles of the 609-aa-long FG domain of Nsp1 ( hence termed Nsp1-FG ) , a key nup in the NPC central channel . Three different ensembles were investigated , namely ensembles of untethered Nsp1-FGs , of Nsp1-FGs tethered to a flat surface as in the Miao and Schulten study , and of Nsp1-FGs tethered to a ring-like surface that is qualitatively similar to the geometry of the NPC . Most essential , however , is that the present study explores Nsp1-FG ensembles in different initial protein conformations , namely fully-extended conformations as assumed by Miao and Schulten as well as random polymer-like conformations . Nevertheless , the new MD simulations support the structural model as seen in the simulations of Miao and Schulten , in which 100-aa-long fragments of Nsp1-FG form bundles linked through single proteins crossing between bundles , but reveals more physical characteristics and a high degree of heterogeneity of the model . For example , the new and more extensive simulations demonstrate that the actual cross-linked bundle system depends to a significant degree on the actual tethering of nups' ends as well as on the initial nups conformation . A further key finding is that the FG motifs distributed in the cross-linking regions have highly accessible side groups that furnish potential binding epitopes for TFs to assist in cargo transport .
To elucidate how the NPC poses a barrier for transport of molecules into and out of the nucleus , we thought to characterize through simulations the general structural features in the NPC central channel , arising from the assembly of disordered proteins . We investigated the assembly using only one type of FG-nups , namely Nsp1 . Nsp1 is an FxFG-rich nup present in the central channel and has been investigated as a model protein in previous studies [20] , [24] , [26] . The nup is present in 32 copies in both the upper and lower half of the NPC equatorial plane; its location is ∼30 nm from the center of the NPC [29] . The large number of copies and its spatial location inside the central channel renders Nsp1 critical for bi-directional nucleo-cytoplasmic transport ( NCT ) . The 609-aa-long FG-domain of this protein , namely Nsp1-FG , was described through coarse-grained simulations that investigated the structures formed by different ensembles of Nsp1-FGs . In two first simulations , the proteins were tethered to a gold ring surface ( simulations wild-type_ring and mutant_ring ) . For the initial protein conformations we assumed fully-extended Nsp1-FGs as also adopted in the earlier study by Miao et al [20] . However , rather than a 100-aa-long fragment of FG-domain simulated in their study , we used in the present study the full-length ( 609-aa-long ) FG-domain . The two simulations ( wild-type_ring and mutant_ring ) differ in that the first employs wild-type Nsp1-FG , whereas the second employs a mutant Nsp1-FG , with all its phenylalanines and glycines replaced by alanines . In a third simulation , Nsp1-FGs were tethered to a flat gold surface ( simulation random_array ) . In this case the proteins were assumed initially in a random polymer conformation . In a fourth simulation the effect of end-tethering was investigated by simulating Nsp1-FGs untethered and initially homogeneously distributed in bulk solvent in random polymer conformations ( simulation random_bath ) . All ensembles contain the protein at a concentration that is characteristic for the interior of the NPC channel . The simulations carried out are listed in Table 1 . For simulation wild-type_ring , Nsp1-FGs were grafted to a gold nano-ring that matches the pore dimension of the pore system designed by an experimental group [21] , [29] , engineered to mimic the NPC geometry . The system was simulated for after which time the initially fully stretched Nsp1-FGs formed brush-like bundle structures ( see Figure 1 ) , the same structure as reported in an earlier study of 100-aa-long fragments of FG-domains [20] . Such structures can potentially function as a selective barrier to transport , supporting the virtual-gate model of nuclear pore gating [1] , [20] . Key for the gating mechanism is that the bundles are interconnected via single Nsp1-FG chains cross-linking adjacent bundles ( see Figure 1 ( c ) and Figure 1 ( e ) ) . We analyzed the observed brush-like bundle structures through computation of the bundle-thickness distribution . Bundle thickness is determined in terms of the number of different Nsp1-FG chains present in a bundle . For the brush-like structure as shown in Figure 1 , bundles with few chains are most favored entropically , yet the bundle thickness distribution shows that there arise also thick bundles that consist of more than fourteen chains ( Figure 2 ) . To test which role FG-repeats play in the formation of the structure shown in Figure 1 , we simulated a mutant system replacing all phenylalanines and glycines in Nsp1-FG by alanines ( simulation mutant_ring ) . The assembly structure of mutant Nsp1-FGs in the nanopore model resulting from simulations is very similar to that seen for wild-type ( WT ) Nsp1-FGs . The similarity is also reflected in the bundle distribution shown in Figure 2 that shows a bimodal shape with thin and thick mutant Nsp1-FG bundles arising . The dynamics of Nsp1-FG assembly during the simulation and a detailed 360-degree view of the brush-like structures that resulted from the simulation are provided in Videos S1 and S2 , respectively . The initially fully- extended and straight-chain Nsp1-FGs , simulated in [20] and in the present simulations ( wild-type_ring and mutant_ring ) , are highly ordered at the outset . These simulations may lead to trapped and , thus , unrealistic states that are in quasi-equilibrium , calling into question if the assembly structure shown in Figure 1 is representative for the NPC interior . Therefore , we modeled in simulation random_array end-tethered Nsp1-FGs with a grid spacing of 6 nm , similar to that in the gold ring geometry , but assumed , for the initial state , random rather than straight ( Nsp1-FG ) conformations . The simulation involved again the full-length ( 609-aa ) Nsp1 FG-domain , Nsp1-FG . From the present simulation resulted , nevertheless , a brush-like structure of bundles with the bundle distribution shown in Figure 2 . However , the bundles formed exhibit more cross-links than seen in the earlier simulation [20] ( see Figure 3 ) . A view of the formation , during simulation random_array , of brush-like structures with cross-linked bristles and a 360-degree view of the final conformation are shown in Videos S3 and S4 , respectively . The brush-like structure resulting for ensembles of closely spaced Nsp1-FGs can similarly be characterized by brush height . For the end-tethered Nsp1-FGs in simulations wild-type_ring , mutant_ring and random_array we monitored the average brush height , namely the end-to-end distance of the main-chain Cα beads of the Nsp1-FG chains in the -direction ( Figure 4 ) . The average was taken over all Nsp1-FG chains in the given simulation system . The Nsp1-FG chains in the final structures for both simulations , namely wild-type_ring and mutant_ring , exhibit similar average height of about 80 nm . Interestingly , we observed that in both cases the average height decayed over time in a similar fashion on the time-scales of the simulations . The Nsp1-FG chains in the final structure of the simulation random_array exhibited an average height of 60 nm . In order to investigate how the final structure of closely spaced Nsp1-FGs is affected by end-tethering , we carried out a simulation of initially fully disordered Nsp1-FGs freely floating in a bath ( random_bath ) at a protein density similar to that in random_array . These Nsp1-FGs formed again bundles ( see Figure 5 ) , but in this case with many more links between bundles ( formed by single Nsp1-FGs crossing between bundles ) than seen in case of simulations wild-type_ring , mutant_ring and random_array . The resulting mesh-like structure can be clearly distinguished from the brush-like structures of wild-type_ring , mutant_ring and random_array through the bundle thickness distribution shown in Figure 2 as the bundles arising in simulation random_bath exhibit at most fifteen chains per bundle . Videos S5 and S6 show the dynamics of Nsp1-FGs as seen in simulation random_bath as well as a 360-degree view of the resulting structure , respectively . Given the different structures formed by Nsp1-FG assemblies , as they result from the present simulations and represent likely the interior of the NPC , one wonders if any specific interactions , in particular , hydrophobic FG-FG interactions , favor the structures seen . We determined , therefore , for the different types of amino acids involved in the formation of bundles how often particular amino acids arise in bundles . For simulation wild-type_ring all types of amino acids as well as FG motifs exhibit a similar propensity to be involved in bundle formation , suggesting that FGs are not particularly critical for the formation of these structures ( Figure S1 ) . This is also supported by simulation mutant_ring , in which the FG motif was replaced by alanines , and by an earlier study [20] . Moreover , all amino acids , including both hydrophilic and hydrophobic residues , are equally favorable for the brush-like structures of the wild-type_ring . Actually , no structures arising from simulations wild-type_ring , mutant_ring , random_array or random_bath exhibit particular amino acid preferences . Apparently , formation of the observed bundle structures is not sequence-specific and likely comes about through contributions from both hydrophobic and hydrophilic interactions . However , Dolker et al [30] showed hydrophilic residues in the linker regions to play a key role in the nup aggregation process and structure , whereas the role of aromatic phenylalanine residues is less dominant . To further characterize the structural details of bundles , we performed a so-called reverse CG calculation [31]–[33] on a fragment comprising cross-linked bundles arising in the final structure of simulation wild-type_ring . Such calculation assigns an all-atom ( AA ) model consistent with a CG model . We then employed this AA model as a starting point of a 100-ns equilibration simulation ( simulation fragment_AA ) . These fragments of bundles increased bundle thickness during the equilibration ( see Figure S2 ) , likely due to the absence of end-tethering of proteins resulting from the fragmentation applied . Inter-chain backbone-backbone hydrogen bonds ( HBs ) frequently formed ( ∼30–50% ) for most amino acids within the bundles , and played a critical role for the stability of the bundles ( Figure 6a ) . Secondary structure analysis , on the other hand , revealed no apparent preference ( <5% ) for either α-helix or β-sheet formation for any amino acid . To assess the role of FG and FxFG motifs on nup assembly , we determined the solvent accessible surface area ( SASA ) for individual amino acids; the results are shown in Figure 6b for the FG and FxFG motifs present in a bundle region or in a cross-linking region . The SASA values from both CG and AA simulations show that both FxFGs and FGs adopt a higher degree of solvent accessibility when present in a cross-linking region than compared to SASA values arising for these motifs in a bundle region ( Figure 6b ) ; in this respect FxFG and FG motifs behave very similarly ( Figure 6b ) . In order to determine in how far the structures of Nsp1-FG assemblies described above provide a barrier against the diffusion of molecules through them , we computed the size of passing molecules as described in Methods and illustrated in Figure 7 . Table 2 lists for the structures resulting from simulations wild-type_ring , mutant_ring , random_array , and random_bath the average radius for the cargo molecules that can diffuse through the structures . Large pores are available for diffusive passage when Nsp1-FGs are assembled into bundle structures with few cross-links; in this case molecules with a radius as large as ∼77 Å can pass , whereas mesh-like bundle structures with many cross-links furnish only relatively small pores for molecules to pass through , namely only ones with radii smaller than 43–50 Å . Our results suggest structural properties of closely interacting Nsp1-FG proteins . We contrast these properties with the structural property of an isolated Nsp1-FG protein . The structures exhibited by a single untethered Nsp1-FG protein monitored in a CG MD simulation showed very different features from the structures resulting from the ensembles of closely spaced Nsp1-FG proteins in simulations wild-type_ring , mutant_ring and random_array . An initially fully-extended , untethered single Nsp1-FG is found to coil up into a globule-like structure , with its radius of gyration ( Rg ) decreasing from over 300 Å ( corresponding to the initially fully-extended form ) to 65 Å ( corresponding to the final , coiled form ) as shown in Figure S3 . Similar globule-like structures were also observed in earlier studies of dynamics of individual , end-tethered short fragments of Nsp1-FGs [20] .
The structure of the NPC central channel interior , made up of intrinsically disordered nups with FG-repeats , governs transport through the NPC . The structure formed by the assembly of tethered nups determines the size of small molecules that can passively diffuse through . The structure also interacts , likely through the FG repeats , with the transport factors that carry various cargoes through the NPC . To characterize the structure of the channel interior , we investigated the assembly using one type of FG-nups , namely Nsp1-FG , that is present in the central channel . We studied , using CG MD simulations , the dynamics of Nsp1-FGs starting from three different initial states: tethered , straight-chain conformations; tethered , random-chain conformations; and untethered , random-chain conformations , simulating each system for 1 µs . We observed in all cases the formation of brush-like bundles linked through single Nsp1-FGs crossing between bundles . This assembly structure is very similar to one seen in prior simulations of initially fully-extended , Nsp1-FG fragments ( 100-aa-long ) , that were closely grafted ( 2 . 6 nm apart compared to 6 nm-spacing in the present study ) in an array-like arrangement [20] . On the other hand , a new finding of the current study is that frequency of crossings between bundles and bundle thickness depend on protein length , on the geometry of the simulated volume , on the degree of tethering , and , particularly , on the initial conformations of the proteins . An interesting observation is that FGs may not be the only factor that plays a role in the formation of bundles . In our present mutant study ( simulation mutant_ring ) and that reported in [20] , basically the same structures of brush-like bundles with cross-linking nups are formed as with wild-type Nsp1-FGs , suggesting that factors other than FG motifs could also contribute to the formation of bundle structures . This finding is also supported by our analysis showing that all kinds of amino acids have similar propensity to arise in the bundles formed . This finding is in contrast to a study of cross-linking arising in a hydrogel-like structure suggested for the selective phase model: in such structure low affinity inter-FG hydrophobic interactions , supposedly , are responsible for cross-linking [23] , [24] . We also note that Miller et al [34] showed through EM experiments that another nup , namely Nup153 , is less aggregation-prone when having its FG motifs either deleted or mutated into AG . However , owing to different nups and mutations investigated in their study and ours , we are unable to address the discrepancy between the two studies . The FG motifs not only play a role in bundle formation but also assist in transport of TF through the NPC channel . A recent study showed that the presence of unbound and , thus , freely available FG-repeat motifs helps binding all possible sites of the TF and is required for efficient cargo transport [35] . In line with that study , a portion of the FxFG and FG motifs are found in the present study , through a SASA analysis in the framework of both CG and all-atom simulations , to be highly accessible to solvent in the cross-linking regions between the bundles . Accordingly TFs , dotted with FG binding sites , should bind to the multiple FGs arising in the cross-linking regions and , perhaps , tear the bundles apart in a zipper-like fashion . Our interpretation , based on the structural features observed for tethered , intrinsically disordered FG-nups of the transport channel , agrees with the other views [18]–[21] , [23]–[27] , [36] in that the nups form characteristic quasi-stable , i . e . , slowly varying , structures bearing FG motifs , with which TFs are known to interact . Thus , interaction between TF surfaces and nups is structured and cannot solely be described properly by employing the potential of mean force , as suggested by Tagliazucchi et . al , which does not include the assembly of nups in the interior of the channel [28] . Experiments employing single-molecule studies that deduced spatial density plots for TF-FG-repeat interactions have also shown that the interaction sites are not evenly distributed in the NPC; instead , they form spatial clusters inside and outside the central nuclear pore [37] . Other models have been suggested for the structures formed by nups in the NPC . For example , experiment and simulation suggest that nups aggregate as amyloid like fibers , in which β-sheet rich sheets form through inter-chain backbone-backbone hydrogen bonding and then through stacking of these sheets . Indeed solid-state NMR data are consistent with that such amyloid-like structures in gel-like aggregates formed by Nsp1-FGs mainly through interactions between hydrophilic residues [38] . Likewise , all-atom simulations of ensembles involving the small peptide FSFG showed that the hydrophobic residue phenylalanine can form amyloid-like aggregates [30] , a feature apparently also consistent with a high-resolution EM study of nup aggregates [34] . While the MARTINI force field [39] employed in the present study is unable to model β-sheets , our all-atom simulations based on a reverse CG calculation as described above is capable of yielding β-sheet formation , and indeed resulted in considerable inter-chain backbone-backbone hydrogen bonds in the bundles , but yielded only a small degree of β-sheet structure . Clearly , the structural variety of actual nups aggregates in the NPC still needs to be investigated further through experimental and computational means . A recent model [27] for the structure inside a NPC transport channel proposed that the central region is gel-like and the periphery is brush-like . The authors postulate two zones , a central and a peripheral zone , for trafficking of molecules in and out of the NPC . Similarly , we model the structure of the entire transport channel of the NPC based on our simulation results . In our study , we did not simulate the entire region of the channel interior as this is computationally still too expensive presently . Instead , we simulated systems of Nsp1-FGs starting from different initial states and based on the results suggests a heterogenous model for the NPC central channel: ( i ) at the periphery , i . e . , close to the scaffold , the FG nups are end-tethered to the scaffold through linker proteins; ( ii ) in the middle region , FG nups adopt natively disordered conformations filling up the central volume . The salient features of nups in a real NPC channel arise in our model through end-tethering of nups , protein density , random conformations for initial states , and a full-length real FG-nup found in the NPC , namely the Nsp1-FG domain . Therefore , we interpret our simulation results then to imply that the nups form different structures in different regions of the central channel ( see Figure 8 ) . The region in the periphery of the central channel , away from the center , should be represented by a system of tethered nups as in simulation random_array , where brush-like bundles with less cross-linking arise . The NPC inner diameter is about 30 nm and unstructured nups that are several hundred amino acids long span this volume of an open pore . Since tethering effects should be minimal in the central region due to the large distance from the NPC wall , where nups are actually tethered , we expect that the structure found in this region is similar to that seen in simulation random_bath , exhibiting bundles with a high degree of cross-linking forming a mesh . In this region , we observed the pore size to be 9–10 nm ( Table 2 ) which agrees with the experimentally determined size limit for passive diffusion of small molecules . In the present study we modeled a homogeneous system of nups consisting of only Nsp1-FGs . FG-nups in the real NPC central channel are heterogeneous; FG-nups vary across the channel volume in length , amino acid composition , FG repeat motifs ( FG , FxFG and GLFG ) , composition and spacing of the linker regions . However , both simulation of a short , 100-aa-long fragment of Nsp1-FG in [20] and the current simulation of a full length Nsp1-FG ( 609-aa-long ) led to the same brush-like structure model , suggesting that the interlinked brush-like structure discovered are not sensitive to varying length and type of proteins . Also , nups are known to be intrinsically disordered , interacting with each other through FG-repeats distributed throughout their entire sequences . Such disorder and interactions between FG-repeats were also captured in the present model . Therefore , we suggest that FG nup heterogeneity does not yield qualitatively different nup assembly structures . In addition to the structural features of the NPC , it is necessary for an understanding of the transport mechanism of the NPC to examine the dynamics of assembly and conformational change of the nup proteins . Transport arises on a millisecond timescale [40] , which is beyond the microsecond time scale accessible to large scale MD simulation and performed in the present study . Thus , we are unable to infer information of dynamics of the transport process from the current simulations . Instead , our simulation results should be interpreted as quasi-static pictures of dynamic assembly structures of the nups . The major unsolved problem regarding our still limited knowledge of the NPC and NCT is how specific large cargoes with the assistance of TFs manage to pass the NPC . Clearly the NPC being one of the largest and at the same time one of the most dynamic macromolecular systems in eukaryotic cells still holds great secrets and offers opportunity for great discoveries . Our suggestion here of the principal assembly structure of disordered nups in the NPC , even if completely true , does not imply yet how transport factors can melt the assembly structure for passage . Straightforward MD simulations , even when simplified through coarse-graining , cannot bring about answers as the systems and processes that need to be simulated are much too large and much too slow , respectively . This calamity is actually a bonus as the combination of theory , experiment , and simulation needed is intellectually more rewarding than a straightforward tour de force purely computational strategy . But in pursuing the role of transport factors one needs to be open-minded about the possibility that yet unsuspected mechanisms play a major role .
The FG-repeat domain of wild-type Nsp1 , Nsp1-FG , was built from Saccharomyces ( S . ) cerevisiae Nsp1 sequence 1–609 ( Swiss-Prot P14907 ) by using the 2004 . 03 release of Chemical Computing Group's Molecular Operating Environment ( MOE ) software . The backbone dihedral angles ( phi , psi ) were set to ( , ) so that an unstructured straight Nsp1-FG was obtained . In case of simulation wild-type_ring , we tested this model through comparison to experiments reported in [21] , [29] that have engineered , using nanotechnology , an artificial pore channel employing nuclear pore proteins inside a gold ring . The system dimension and protein density were chosen to imitate volume-wise the interior of the NPC . We reproduced the gold ring dimensions adopted in [29] and distributed over the ring tethering points holding the C-terminal ends of full-length Nsp1-FG domain proteins ( 609-aa-long ) . As shown in Figure 1 , 120 fully-extended wild-type Nsp1-FG chains were grafted to the ring in three concentric rows with 6 nm spacing between adjacent Nsp1-FGs [29] . The C-terminus of the end-tethered Nsp1-FGs was modified by adding five cystine residues that remained fixed to the gold surface throughout the simulation . In [20] , [21] , [42] , these cystine residues formed thiol linkages with the gold substrate . With the C-terminus of each chain attached to the gold-ring and the rest of the chain fully-extended in the direction shown in Figure 1 , the whole system was coarse-grained and solvated with CG water in a box large enough to prevent proteins from interacting with their periodic images . A total of 100 mM NaCl was added to the water box , adjusting the relative concentrations of and to render the whole system neutral . The resulting system simulation wild-type_ring has 15 , 453 , 214 CG beads . For the simulation mutant_ring we constructed the ring system as above and replaced all phenylalanines and glycines of the Nsp1-FGs by alanines . The resulting system described in simulation mutant_ring has 16 , 019 , 434 CG beads . Both systems were simulated for using coarse-grained molecular dynamics simulations as described below . To introduce disorder in Nsp1-FG chains , initial random Nsp1-FG conformations needed as starting points for simulations random_array and random_bath were modeled according to the widely used worm-like chain model [43] . The assignment of the random conformations proceeded in two steps . In a first step , the main-chain Cα beads of Nsp1-FG constituting the protein backbone were modeled as random homo-polymers constructing the backbone from a self-avoiding walk ( SAW ) procedure [44] . In this procedure , the SAW is directed under two local geometric restraints , namely keeping a fixed distance of 3 . 7 Å between neighboring Cα beads and keeping a fixed angle of for three adjacent Cα beads . The stated values are used in the MARTINI force field [39] for polypeptide chains with coil conformations . Any two Cα beads were considered to be in close contact if their distance is shorter than 8 Å . We discarded any conformations of backbone chains with beads in close contact within one chain ( intra ) or between different chains ( inter ) . In the second step , CG side-chains of amino acids were grafted on to the resulting backbone chains . The geometries of the side-chains and of the gold nano-particles were modeled with standard parameters in the MARTINI force field [39] , [45] . The MARTINI CG force field has been extensively optimized in order to correctly model partition between water and non-polar solvent for amino acids including phenylalanine , glycine and alanine that are of interest in the current study . Moreover , multi-site representation of large amino acids is employed in MARTINI in order to realistically model special geometric features , such as aromatic rings in phenylalanine , which are essential for side-chain interactions . In the past few years , the MARTINI force field has been successfully applied in numerous studies of peptide and peptide-lipid interactions , as well as in studies of the assembly of micelles and bilayers around membrane proteins [46] . However , due to lack of treatment of backbone-backbone hydrogen bond interactions , the MARTINI force field is unable to model formation of certain type of structures , such as -sheets . For modeling end-tethered Nsp1-FGs in simulation random_array , we employed again for the construction of a random initial state the worm-like chain model described above and chose the last five Cα beads of cysteine residues 610–614 as the starting points for the SAW . In the experimental systems [21] nups are tethered to the gold substrate by means of thiol bonds to cysteine residues added to the C-terminus . In order to be consistent with the description adopted for the ring-like geometry , the five Cα beads of each Nsp1-FG chain were stretched toward the -direction and placed on a 5×5 grid in the -plane , with a grid spacing of 6 nm . The full-length worm-like chain was then modeled as described above , such that 25 Nsp1-FGs were placed as shown in Figure 3 . The whole system was coarse-grained according to the MARTINI force field [39] , [45] and solvated with CG water . The system was then ionized with 100 mM NaCl , adjusting again the relative concentrations of and to render the whole system neutral . The resulting simulation random_array involves 1 , 097 , 433 CG beads . The system was simulated for as described below . For simulation random_bath , the first three Cα beads of each Nsp1-FG chain , treated as a rigid body , were chosen as starting points . They were randomly placed in simulation boxes with random orientations . The full-length random conformation Nsp1-FGs were then modeled as worm-like chains as described above , with 120 self-avoiding Nsp1-FGs being placed in a box of volume 725 Å×725 Å×725 Å to match the concentration of Nsp1-FGs as in the simulations wild-type_ring and random_array ( see Figure 5 ) . The whole system was coarse-grained according to the MARTINI force field [39] and solvated in a CG water box . 100 mM NaCl was then added to the sytem , adjusting the relative concentrations of and to render the whole system neutral . The resulting simulation random_bath involved 3 , 091 , 910 CG beads . The system random_bath was simulated for as described below . Structures resulting from CG simulations can be “reverse-coarse-grained” to obtain corresponding all-atom ( AA ) structures , where each amino acid can be mapped back to an AA representation by replacing the CG beads by all the atoms it represents . The method is detailed in [31]–[33] . In the present study , CG MD was employed to extend the simulation time scale to microseconds while AA MD was used to refine a segment of the final structure arising in simulation wild-type_ring to investigate the chemical details of the final brush-like structures . For the AA simulation fragment_AA , eleven Nsp1-FG chains , namely , L17 ( 275–507 ) , L18 ( 275–523 ) , L19 ( 293–513 ) , M21 ( 302–544 ) , M22 ( 316–545 ) , M23 ( 317–548 ) , M24 ( 326–546 ) , N26 ( 328–579 ) , N27 ( 336–579 ) , N28 ( 332–570 ) and N29 ( 345–581 ) of CG simulation wild-type_ring were included; the numbers in parenthesis are the residue numbers for amino acids of the respective chains included , having been selected for exhibiting a cross-link . The CG representation of the stated Nsp1-FG bundle segments , after being mapped back to an AA representation , were solvated into an AA water box . A total of 100 mM NaCl was then added to the water box , adjusting the relative concentrations of and to render the system neutral . The resulting box has a volume of 123 Å×198 Å×414 Å and includes 967 , 595 atoms . The resulting AA structure was simulated for 100 ns as described below . Coarse-grained ( CG ) molecular dynamics simulations were performed based on the MARTINI model for proteins [39] in NAMD 2 . 9 [47] . For use of the MARTINI force field we adapted the GROMACS switching function for the LJ potential and a shifting function for the Coulomb potential . Non-bonded interactions were cut off at 12 Å , with shifting throughout the interaction range for electrostatic interactions and beginning at 9 Å for vdW interactions , implementing a smooth cut-off . Simulations were performed using a 10 fs timestep . Pair lists were updated at least once every ten steps , with a 14 Å pair list cut-off . In all cases we performed Langevin dynamics with a damping coefficient of . A constant pressure of 1 atm was maintained with a Nosé-Hoover Langevin piston [48] , using a piston period of 2000 fs and a decay time of 1000 fs . All systems were allowed to equilibrate as follows: first , the system was energy minimized for 5000 steps and molecular dynamics was performed for 2 ns in an NVT ensemble ( T = 300 K ) . The resulting system was then simulated for assuming an NPT ensemble . The all-atom simulations were performed using NAMD 2 . 9 [47]; non-bonded interactions were cut off at 12 Å , with a switching function beginning at 10 Å and implementing a smooth cutoff . The simulations involved multiple timestepping , with a base timestep of 1 fs , short-range interactions calculated every step , and long-range electrostatics every two steps . Electrostatic forces were evaluated through the particle-mesh Ewald method [47] with a grid density of 1 . 0 Å−3 . The AA system was first energy minimized for 5000 steps and then was simulated for 100 ns assuming an NPT ensemble ( T = 300 K ) . Periodic boundary conditions were assumed for all ( CG and all-atom ) simulations . As described in Results , the simulated Nsp1-FGs formed strands that we refer to as bundles . We define a bundle as a linearly arranged cluster of multiple parallel Nsp1-FG chains . In such a cluster , every amino acid in a segment of one Nsp1-FG chain is in contact with at least one amino acid of another Nsp1-FG chain . In order to provide a quantitative characterization of these bundles we applied graph theory to identify bundle segments in a given configuration of the Nsp1-FG chains . In this approach , each amino acid is represented by a node in a graph . If the distance of two amino acids A and B from different chains is shorter than 6 Å , the corresponding nodes in the graph are connected with an edge . In addition , the nodes for amino acids adjacent in sequence to the two amino acids A and B forming an edge are considered to be connected to the nodes for amino acids A and B . Therefore , bundle segments can be identified by examining connected components in a graph as constructed above . For this purpose we employed the breadth-first search algorithm [49] . Thickness of a bundle is determined as the number of different Nsp1-FG chains that belong to the bundle . Another bundle characteristic determined here is the fraction of each type of amino acid involved in a bundle . Nsp1-FG polymers self-assemble , as shown in Results , into network structures . In order to assess how these structures are related to gating in the NPC we determine the size limit of spherical particles being capable geometrically to pass through the structures . This size limit defines the pore size , namely as the largest possible radius of passing particles . Figure 7 illustrates how the pore size was determined by us algorithmically . Starting from one side of the network structure , spheres of various sizes are moved towards the other side . The analysis was applied to the final structures resulting from simulations random_array , random_bath , wild-type_ring , and mutant_ring; in the case of the latter two simulations structures are formed with Nsp1-FGs arranged in a ring-like arrangement . In these cases , the spheres were moved radially from the inside of the Nsp1-FG ring structure to its outside . We implemented the above analysis through an algorithm in which the simulation box with the network structure was mapped into a cubic lattice . We then determined for each grid point , , in the lattice the radius of the largest sphere that could be placed on this grid point without sterical clash with any Nsp1-FG chain; this radius property was calculated as the shortest distance between the grid point and protein beads . A sphere is considered capable of moving in the lattice along a given pathway , if the size of the sphere does not exceed the bottleneck of the pathway , i . e . , the smallest of the grid points on the pathway . There could be multiple pathways for a sphere to permeate from one side to the other side , each pathway having its own bottleneck . To characterize the largest sphere that could permeate the network structures we search for the pathway for which the bottleneck is the widest among all the possible pathways . The pathway search was performed using Dijkstra's algorithm [50] . To illustrate a typical outcome of the pore analysis , we show in Video S7 a 360-degree view of all possible pores identified in the final structure resulting from simulation random_array .
|
Cells of higher life forms separate their genomes from the rest of the cell in a nucleus that surrounds the genome by a nuclear envelope . Hundreds of pores , each a complex made of many proteins , assure traffic into and out of the nucleus through highly selective transport: small biomolecules can pass unhindered , whereas large biomolecules need to associate with proteins called transport factors , to pass . Little is known about how the nuclear pore complexes function , a key impediment to observation being their huge size and the disordered nature of the pore interior . We investigated computationally what kind of structure the nuclear pore proteins ( nups ) form . In the computation we place many nups , each a 600 amino acid-long protein , into arrangements considered representative for the nuclear pore , and simulate the subsequent molecular behavior . We find that the nups form bundles of 2–27 proteins , the bundles being cross-linked when a single nup leaves a bundle and joins an adjacent one . The finding suggests an adaptive molecular mesh arrangement of nups in the nuclear pore and explains how selective transport is accomplished , namely that passage of sufficiently small molecules is unhindered by the cross-linking , but that large molecules need the assistance of transport factors to melt the cross-linking .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"biophysics",
"biology",
"computational",
"biology"
] |
2014
|
Assembly of Nsp1 Nucleoporins Provides Insight into Nuclear Pore Complex Gating
|
Since 2011 , cohorts of schoolchildren in regions bordering Lake Victoria in Kenya and Tanzania have been investigated for morbidity caused by Schistosoma mansoni infection . Despite being neighbouring countries with similar lifestyles and ecological environments , Tanzanian schoolchildren had lower S . mansoni prevalence and intensity and they were taller and heavier , fewer were wasted and anaemic , and more were physical fit compared to their Kenyan peers . The aim of the present study was to evaluate whether diet and school-related markers of socioeconomic status ( SES ) could explain differences in morbidity beyond the effect of infection levels . Parasitological and morbidity data from surveys in 2013–2014 were compared with information on diet and school-related markers of SES collected in 2015 using questionnaires . A total of 490 schoolchildren ( 163 Kenyans and 327 Tanzanians ) aged 9–11 years provided data . A higher proportion of Tanzanian pupils ( 69 . 4% , 95% CI: 64 . 3–74 . 5 ) knew where to wash hands after toilet visits compared to Kenyan pupils ( 48 . 5% , 95% CI: 40 . 9–56 . 1; P<0 . 0005 ) . Similar proportions of children in the two countries ate breakfast , lunch and dinner , but the content of the meals differed . At all three meals , a higher proportion ( 95% CI ) of Tanzanian pupils consumed animal proteins ( mostly fish proteins ) compared to their Kenyan peers ( 35 . 0% ( 28 . 3–41 . 7 ) vs . 0%; P<0 . 0005 at breakfast; 69 . 0% ( 63 . 9–74 . 1 ) vs . 43 . 6% ( 35 . 8–51 . 4 ) ; P<0 . 0005 at lunch; and 67 . 2% ( 62 . 1–72 . 3 ) vs . 53 . 4% ( 45 . 8–61 . 0 ) ; P = 0 . 003 at dinner ) . Multivariable analyses investigating risk factors for important morbidity markers among individuals revealed that after controlling for schistosome and malaria infections , eating animal proteins ( fish ) and knowing where to wash hands after toilet visits were significant predictors for both haemoglobin levels and physical fitness ( measured as VO2 max ) . These results suggest that the differences in morbidity may be affected by factors other than S . mansoni infection alone . Diet and hygiene practice differences were associated with health status of schoolchildren along Lake Victoria in Kenya and Tanzania . Trials Registration numbers: ISRCT 16755535 ( Kenya ) , ISRCT 95819193 ( Tanzania ) .
Schistosomiasis , also known as bilharzia , is an infectious disease caused by parasitic flatworms of the genus Schistosoma . Schistosomiasis is considered as one of the Neglected Tropical Diseases ( NTDs ) and is estimated to affect at least 230 million people annually [1] with a majority in sub-Saharan Africa [2] . In terms of public health impact , schistosomiasis is second only to malaria as the most important parasitic disease in developing countries [3] . Schistosome infections can result in anaemia , stunted growth , malnutrition , impaired physical fitness and numerous other complications [1 , 4] . Although praziquantel has been available for decades and is effective for the treatment of Schistosoma infections , schistosomiasis remains a major health concern . Schistosomiasis control efforts usually focus on reducing the prevalence and intensity of infection by Preventive Chemotherapy ( PC ) . Socioeconomic status ( SES ) has been linked to various health issues , such as nutritional status , disease burden and mortality , as well as accessibility and affordability of health services [5] . As schistosomiasis typically occurs in rural areas where the majority of the population is highly affected by poverty , the impact of schistosomiasis in a given area may be exacerbated by low SES . Collecting socioeconomic information may help identify potential risk factors that contribute to schistosomiasis as well as associated morbidity , and therefore help improve the impact of the national schistosomiasis control programmes . Addressing both schistosomiasis and these other factors would help direct resources to areas most in need . The current investigation was conducted as a part of two cohort studies within a larger multi-country Schistosomiasis Consortium for Operational Research and Evaluation ( SCORE ) research project on gaining and sustaining control of schistosomiasis [6] . The SCORE projects in Kenya and Tanzania were implemented in areas near Lake Victoria , where prevalence of S . mansoni infection was 25% or greater [7 , 8] . The nested cohort studies were conducted to assess S . mansoni infection markers of morbidity over five years and compare the effect of different treatment strategies . Morbidity data collected from these cohorts during the year 3 assessment ( 2013–2014 ) showed that Tanzanian schoolchildren were taller and heavier , fewer were nutritionally wasted and anaemic , and they scored higher in a physical fitness test compared to Kenyan schoolchildren . The Tanzanian pupils also had lower S . mansoni prevalence and intensity despite assumed similar exposure risks and ecological setting . The aim of the present study was to evaluate whether diet and school-related markers of SES could explain differences in morbidity across study sites beyond the effect of infection levels .
Approval for the SCORE Kenya and Tanzania gaining and sustaining control of schistosomiasis and cohort studies was obtained from Institutional Review Boards at the Scientific and Ethical Review Committees of the Kenya Medical Research Institute ( Nairobi , Kenya ) and the Medical Research Coordination Committee ( MRCC ) of the National Institute for Medical Research ( Tanzania ) . Trials Registration numbers: ISRCT 16755535 ( Kenya ) , ISRCT 95819193 ( Tanzania ) . The parasitological and morbidity data ( year 3 cross-sectional data ) were previously collected by the SCORE team in Kenya and Tanzania and are covered by the above mentioned approval , while the questionnaire data were collected from children who separately assented to participate and had written informed consent from parents or legally authorized representatives . The study took place in Nyanza Province ( Bondo District ) of Kenya and Mwanza Region ( Sengerema District ) of Tanzania , along the Lake Victoria shoreline . For the SCORE cohort studies , communities were randomly selected from the two arms with the most intense level of treatment ( annual community-wide treatment over four years ) and the less intense treatment strategy ( biannual school-base treatment ) as part of the larger cross-sectional study ( Fig 1 ) . In each of these two study arms , 4 ( Tanzania ) and 6 ( Kenya ) out of 25 communities were randomly selected in order to achieve a baseline cohort of 800 schoolchildren from each country . These children were 7–8 years of age at the initiation of the intervention study . The children were enrolled for 4 years of intervention succeeded by a 5th year follow-up testing . Morbidity parameters were measured at baseline , and in year 3 and 5 prior to PC with praziquantel . The year 3 cross-sectional data on parasitology and morbidity were collected in 2013 in Kenya and 2014 in Tanzania . The one year difference was due to a one year delay in the larger cross-sectional study in Tanzania and do not reflect different intervention periods . Questionnaire data were collected in Kenya in January and February 2015 and in Tanzania from March to April 2015 . Sample size calculations were not performed specifically for this section of the larger study , but all the cohort pupils available at the time of our visits were enrolled . The methods for collection of cohort data are described in detail elsewhere [9 , 10] , but are briefly presented below . Participants were given stool containers and asked to bring fresh stool specimens to the school on three consecutive days . Specimens were processed using duplicate Kato-Katz thick smears with a 41 . 7 mg template [11] from each specimen in the school ( Tanzania ) or after being transported to the KEMRI/CDC laboratory ( Kisumu , Kenya ) . All slides were examined at the laboratories in Mwanza or Kisumu for S . mansoni eggs . The number of S . mansoni eggs was multiplied by 24 and expressed as eggs per gram of stool ( epg ) . Intensity was reported as the arithmetic mean of epg from the total number of slides per person . Infections with STHs were not investigated in Tanzania as Ascaris lumbricoides and Trichuris trichiura have been recorded to be seldom present [12 , 13] and because eggs of hookworms were not visible because of the time span between preparation and reading . A 5 mL of venous blood sample ( Kenya ) or finger-prick blood sample ( Tanzania ) was collected from each individual as part of the larger study design within each country and haemoglobin ( Hb ) measured using a portable HemoCue photometer ( Ängelholm , Sweden ) . Hb level was reported in g/L and final values used in analysis were adjusted for altitude by subtracting 2 g/L from the raw values for both study sites [14] . Anaemia was defined as Hb values below 115 g/L according to the World Health Organization guidelines [14] . Infection with Plasmodium falciparum was determined by examination of blood smears by experienced microscopists in Kenya and by rapid diagnostic test ( RDT ) ( SD Bioline , Republic of Korea ) in Tanzania . All children were asymptomatic for malaria . Height was measured on barefooted children using a wooden stadiometer . The child stood on the base of the stadiometer with their heels , buttocks , shoulder blades and back of the head touching the vertical backboard and looking straight ahead . When correctly positioned , the ruler was lowered and the height measured in centimetres with one decimal . Weight was measured on a digital scale on barefooted children having removed any excess clothing . Weight was measured in kilograms to one decimal . Height and weight were measured twice by the same examiner and the mean recorded . Z-scores were calculated using the WHO growth reference table [15] . Wasting was defined as a BMI-for-age Z-score of <-2 SD . Physical fitness was assessed using the 20 metre shuttle run fitness test ( 20mSRT ) as described in Bustinduy and colleagues [16] . In brief , during the test , children run continuously between two lines 20 meters apart at increasing speeds , turning when signalled to do so by recorded beeps . A “shuttle” is defined as a run from one line to the other . The running field was prepared in the school compound and runners were separated with at least one meter . Recorders were placed at each end of the field and every recorder was responsible for taking notes of three to five children . The recorder noted the level at which the test subject stopped and how many shuttles the child completed within that level . These numbers are correlated to a maximal oxygen uptake , the VO2max in mL/kg/min as described in Müller and colleagues [17] . The questionnaire was generated by the different research scientists , including social scientists , involved in the SCORE project [18–21] . The questionnaire was developed in English first and translated into Dholuo/Kiswahili language then verbally administered to participants . Prior to data collection , the questionnaires were pilot-tested in groups of children that were not part of the research project to ensure that the tool was feasible to administer and easy to understand for the respondent . The questionnaire contained questions on diet and a range of school-related markers of SES and is attached as supporting information ( S1 Questionnaire ) . Statistical analyses were performed using SPSS version 24 ( IBM , Armonk , NY ) . Summary statistics were calculated and all statistical tests used were two sided , and P<0 . 05 was considered significant . To assess differences between the two countries , Pearson Chi-square was used to assess differences in proportions . The Student’s t-test and one-way ANOVA were used to assess differences between normally distributed means ( two and more than two , respectively ) , while the Mann-Whitney U test was used to compare means that were not normally distributed . To further analyse the potential determinants for morbidity differences seen in the pupils , we performed a more detailed statistical analysis using uni- and multivariable linear regression analyses . The regression analyses were applied with the two important morbidity indicators: ‘Haemoglobin ( Hb ) ’and ‘Maximum oxygen uptake ( VO2 max ) ’ as dependent variables . As too few individuals were nutritionally wasted , we did not perform an analysis with ‘Wasting’ as dependent variable . First , univariable analyses were performed on each of the two dependent variables on the following independent variables: ‘gender’ , ‘age’ , ‘height’ , ‘Schistosoma infection’ , ‘treatment arm’ , ‘malaria infection’ , ‘anaemia’ ( but only for VO2 max ) , ‘transportation to school’ , ‘distance to school’ , ‘health information from teachers’ , ‘knowing where to find a place to wash hands after toilet visits’ , ‘wearing shoes at school’ , ‘having animal proteins at meals’ , and ‘having vegetables at meals’ . The variable ‘having grain at meals’ was omitted as almost all pupils had grain for all meals . As height and weight are highly correlated ( S1 Data ) ; only height was included in the analyses . Age was divided in two groups; 9 years and a combined 10–11 years as only 11 children were 11 years old . Height was divided in two groups with the mean height as cut-off value . The presence of STH infections were recorded in Kenya and the effect of these infections on Hb and VO2max were performed on the Kenyan data only . All variables with P≤0 . 10 ( plus ‘Schistosoma infection’ and ‘malaria infection’ to control for the two infections ) were then included in the subsequent multivariable analyses with Hb and VO2 max as dependent variables using a stepwise strategy ( criteria for enter ≤0 . 05 and for remove ≥0 . 10 ) .
A total of 163 pupils from the Kenyan cohort and a total of 327 pupils from the Tanzanian cohort provided information for the current study ( Table 1 and Table 2 ) . Table 1 shows the year 3 cross-sectional measurements from both countries . Apart from gender distribution , all parameters were significantly different between the two countries . The pupils in Tanzania had lower prevalence and intensity of S . mansoni infections compared to pupils in Kenya . More Tanzanian pupils were diagnosed with malaria but they were tested with the slightly more sensitive RDT , while the Kenyan students were diagnosed by microscopic detection of parasites in their blood . In addition , the Tanzanian pupils were taller and heavier , and fewer were nutritionally wasted compared to their peers in Kenya . Finally , fewer Tanzanian children were anaemic and in the physical fitness test they scored a higher VO2 max . In Kenya , information on STH infections was obtained from 150 individuals . Of those , eight ( 5 . 3% ) were A . lumbricoides and T . trichiura positives and two ( 1 . 3% ) had hookworm infections . The results of the questionnaire are shown in Table 2 . The majority of the pupils , 83 . 4% in the Kenyan cohort and 97 . 2% in the Tanzanian cohort put on shoes when they went to school , and the difference between the two cohorts was statistically significant ( P<0 . 0005 ) . While less than half ( 48 . 5% ) of the Kenyan pupils knew where to wash their hands following a toilet visit , more than two-thirds ( 69 . 4% ) of the Tanzanian pupils were aware ( P<0 . 0005 ) . More Kenyan pupils ( 92 . 0% ) reported that they received health-related information from their teacher compared to Tanzanian pupils ( 54 . 7% , P<0 . 0005 ) . The pupils were asked specifically what they had for each meal the day before , categorised as grain , animal protein and vegetable . Grain ( carbohydrate rich ) was the main source of nutrients in both cohorts for all three meals of the day . However , the intake of animal protein was significantly different between the two cohorts for all three meals as more Tanzanian pupils consumed animal proteins . By contrast , a higher proportion of Kenyan pupils consumed vegetables . Of the animal protein consumed , fish proteins contributed to more than 80% of all meals for both cohorts . Table 3 shows the univariable associations of demographic , anthropometric , parasitological ( including treatment history ) , and diet and school-related markers of SES with mean Hb level ( g/L ) of the schoolchildren in both countries combined . The following variables were significantly associated with Hb: ‘age’ , ‘height’ , ‘know where to wash hands after toilet visit’ and ‘meals with animal protein’ . In the multivariable analysis where all variables with P≤0 . 10 were included and after controlling for schistosome and malaria infections; ‘age’ , ‘height’ , ‘know where to wash hands after toilet visit’ and ‘meals with animal protein’ were retained in the model ( Table 4 ) . The regression coefficient of ‘knowing where to wash hands after toilet visits’ means that those pupils who knew had on average 5 . 33 g/L higher Hb than their peers who did not know . The regression coefficient of ‘animal protein at meals’ corresponds to a 2 . 98 g/L increase in Hb for every step between having no animal proteins , having animal proteins once , twice or three times a day . The regression coefficient of ‘height’ means that pupils with a height of 135 cm or taller had a 4 . 50 g/L higher Hb compared to the pupils lower than 135 cm . Finally , the coefficient of ‘age’ means that the pupils of the age of 10–11 years had a 4 . 34 g/L higher Hb compared to the 9 year old pupils . For separate analysis on the Kenyan data where STH infections were included , none of the three infections were retained in the multivariable analysis . The analyses for VO2 max showed more significant predicators compared to the Hb ( Table 5 ) . The following were found to be significant: ‘gender’ , ‘height’ , ‘Schistosoma infection’ , ‘where to wash hands’ , ‘animal protein at meals’ and ‘animal protein at meals ( grouped ) ’ . In the multivariable linear regression analysis , which included all variables from the univariable analyses with P≤0 . 10 and after controlling for schistosome and malaria infections , the following variables were significant for VO2 max: ‘gender’ , ‘height’ , ‘animal protein at meals ( grouped ) ’ and ‘know where to wash hands after toilet visit’ ( Table 6 ) . The regression coefficient of ‘gender’ means that boys had a 2 . 80 mL/kg/min higher VO2 max compared to girls . The regression coefficient of ‘height’ means that pupils with a height of 135 cm or taller had a 1 . 69 mL/kg/min higher VO2 max compared to those pupils shorter than 135 cm . The regression coefficient of ‘meals with animal protein ( grouped ) ’ means than those pupils having at least one meal with animal protein had a 1 . 42 mL/kg/min higher VO2 max compared to those having no meals with animal protein . Finally , pupils knowing where to wash hands after toilet visits had on average 1 . 04 mL/kg/min higher VO2 max than their peers who did not know . For separate analysis on the Kenyan data where STH infections were included , none of the three infections were retained in the multivariable analysis .
Differences in diet and hygiene practices may explain differences in morbidities commonly associated with schistosomiasis . Schoolchildren in Tanzania more often consumed animal proteins compared to their Kenyan peers and this difference could possibly explain the difference in several morbidity markers between the two populations . Thus , when the two populations were combined and analysed at the individual level , consumption of animal proteins was a significant predictor of both Hb levels and physical fitness . The consumption of animal proteins was associated with increased Hb levels with 3 . 0 g/L for every increase in number of meals with animal proteins per day . In a nation-wide survey in 2004/2005 , the most common cause of anaemia among children in Tanzania was identified to be nutritional anaemia resulting from inadequate dietary intake of nutrients [22] . Furthermore , Mboera and colleagues [22] demonstrated in their study in central Tanzania that anaemia was most prevalent among communities with low prevalence of malaria , suggesting that the anaemia was most likely to be a result of dietary deficiency or caused by other infections than malaria . The consumption of vegetables was not associated with Hb levels despite the fact that these vegetables contain iron . This is probably because this iron is of the non-heme type and that non-heme iron is not as easily absorbed as the heme iron . Furthermore , iron-rich vegetables also contain oxalates and phytates , which impair iron absorption , and boiling decreases the content of iron in vegetables [23] . By contrast , fish contains high amounts of the more easily absorbed heme iron and cooking does not significantly reduce the content of iron in animal products [23] . A total of 69 . 4% of the Tanzanian pupils knew where they could wash their hands after toilet visit compared to 48 . 5% of the Kenyan pupils . The question ‘After toilet visit is there a place to wash your hands’ is meant to reveal whether there are possibilities for hand washing in the school , but it is also reflecting on the pupil’s knowledge about personal hygiene . These two parts cannot be separated based on the answers . Although hand washing after toilet visits is not one of the important tools in the toolbox of schistosomiasis control , it certainly has an effect on other infections such as the STHs A . lumbricoides and T . trichiura and other pathogens causing diarrhoeal diseases . These infections may also have an impact on Hb level . Unfortunately , information on STH infections is only available for the Kenyan cohort , but not for the Tanzanian cohort for reasons described in the method section . Although literature reports that A . lumbricoides and T . trichiura are more common in the Kenyan part compared to the Tanzanian part [24] , the prevalence of the two infections in Kenya in this study was less than 6% . Thus , it is not likely that the difference in Hb levels between the two countries can be explained by a difference in prevalence of either of these infections . Hookworm infections , and especially high intensity infections , have an impact on Hb levels . In the present study , hookworm prevalence was only 1 . 3% in Kenya ( only two individuals ) , while information from Tanzania is lacking . However , a recent study reported hookworm infections in 16% of school and pre-school children in Magu District , which is a neighbouring district to Sengerema District , where the present study took place [13] . It is therefore not plausible that the lower Hb levels in Kenya compared to Tanzania can be explained by a higher prevalence or intensity of hookworm infections in Kenya . Lack of shoes was not a risk factor for Hb levels in this study probably because of the assumed low levels of hookworm infections . Schistosoma mansoni infection was not a risk factor for Hb levels in this study . This is in contrast to the results from the baseline survey in Kenya where heavy S . mansoni infections were predictors of anaemia [9] . In a recent systematic review and meta-analyses on the effect of treatment on mean Hb levels there was no consistent or significant changes between pre- and post-treatment surveys in ten different studies in schoolchildren [25] . However , only one of these studies was on S . mansoni infection and although it documented a considerable impact on Hb and anaemia after two years of treatment through the Ugandan National Control Programme [26] , it is not possible to attribute the improvements to praziquantel treatment alone . This is because the population was also treated with albendazole , which decreased the prevalence of hookworms from 50 . 9% to 10 . 7% and the mean hookworm intensity from 309 epg to 22 epg . Our data showed that infection with Plasmodium had no significant association with Hb levels . Malaria and iron have a complex but important relationship . Plasmodium proliferation requires iron , both during the clinically silent liver stage of growth and in the disease-associated phase of erythrocyte infection [27] . Interestingly , human iron deficiency appears to protect against severe malaria , while iron supplementation may increase risks of infection and disease [27] . This could explain why the Tanzanian pupils had higher prevalence of malaria compared to the Kenyan pupils . However , it is important to note that the diagnostic techniques differed in the two countries; infection with Plasmodium falciparum was determined by examination of blood smears via microscope in Kenya and by rapid diagnostic tests in Tanzania . The rapid diagnostic test is known to be only slightly more sensitive compared to microscopy [28] when microscopy is performed by experienced technicians; still , the prevalence of malaria could be underreported in the Kenyan pupils . Schistosome infection was negatively associated with physical fitness although only in the univariable analysis . In the multivariable analysis other measured parameters seem to be more important for physical fitness resulting in the exclusion of schistosome infection in the final model . Thus , being a boy , being taller than average , having animal proteins at least once a day and knowing where to wash hands after toilet visits were significant predictors of physical fitness . The lack of association between physical fitness and infection in the multivariable analysis is in accordance with other recent studies using the 20mSRT [16 , 17 , 29] and with the baseline results of the two cohorts investigated in this study [9 , 10] . However , in line with the present study , boys had better physical fitness compared to girls in the three studies reporting the associations [16 , 17 , 29] , an association which was lacking at baseline in Tanzania [10] and not investigated in Kenya [9] . As age , height and weight were similar between genders in the present study ( S1 Table ) , this difference is reflecting gender-specific differences and is less related to physical features . Besides the gender differences , Bustinduy et al . [16] found anaemia and growth stunting to be predictors of physical fitness , while only age was predictor in the study of Müller et al . [17] . The role of T . trichiura infections for the physical fitness of school-age children is unclear according to two Chinese studies [30 , 31] . The impact of awareness of personal hygiene and/or content of diet intake on physical fitness has not been assessed previously in the two study areas . Reduced physical fitness is a manifestation of the body's inability to maintain adequate oxygen supply to the tissues and may have many different causes . In developing countries , low physical fitness is often the result of anaemia and under nutrition , which have multifactorial aetiologies such as poor diet and chronic infections . Most important of these infections are malaria , hookworm and schistosomiasis [16] . Having animal protein at meals was a strong predicator for VO2 max uptake in the multivariable linear regression analysis , suggesting that animal protein is a decisive factor for fitness . This is consistent with Bustinduy and colleague’s study , which demonstrated malnutrition parameters as strong predicators of decreased fitness [16] . As eating animal protein also was a strong predictor for Hb level , the physical fitness might be affected in two ways . The animal protein increased the Hb level which again increased the fitness , but at the same time the mere consumption of animal proteins might better satisfy the child’s nutritional needs , resulting in higher fitness . There were limitations in our study , which need to be taken into account during interpretation of the results . Initially , the baseline study enrolled 800 pupils from each cohort , however , through the years , many pupils dropped out and the present study is thus implemented in a sub-group of the baseline cohort . The design of the questionnaire was useful; however , the question ‘After toilet visit is there a place to wash your hands’ could have been more specific so the presence of hand washing facilities could be clearly separated from the pupils’ knowledge on hygiene practices . It would have been valuable if information on prevalence and intensity of STH infections had been available from both study areas . In addition , it was difficult to compare prevalence of malaria infections between the two countries as the two diagnostic tests for malaria have slightly different sensitivities . In conclusion , these results suggest that the differences in morbidity parameters between Kenyan and Tanzanian schoolchildren living near Lake Victoria may be due to factors other than S . mansoni infection alone . Knowing where to wash hands after toilet visits and having a diet rich in fish were associated with higher haemoglobin levels and a better physical fitness . The consequence of these results is that control programmes may improve their interventions by encouraging the communities to provide hand washing facilities in schools , strengthen education on good personal hygiene and promoting a healthy and nutritional diet rich in protein and iron .
|
Millions of school-age children in Kenya and Tanzania are at risk for infection with Schistosoma mansoni , which has an impact on their physical health . A total of 490 schoolchildren ( 163 from Kenya and 327 from Tanzania ) aged 9–11 years living along the shores of Lake Victoria in Kenya and Tanzania provided data on S . mansoni and malaria infections , weight , height , anaemia and physical fitness . The pupils in Tanzania had lower prevalence and intensity of S . mansoni infection compared to pupils in Kenya , but more had malaria parasites in their blood . In addition , Tanzanian pupils were taller and heavier , fewer were anaemic and they scored higher in a physical fitness test . Questionnaire data showed that more of the Tanzanian pupils knew where to wash hands after toilet visits , and more consumed animal protein ( mostly fish protein ) for breakfast , lunch and dinner . Multivariable analyses revealed that eating animal protein and knowing where to wash hands after toilet visits were significant predictors for haemoglobin levels and physical fitness . These results suggest that the differences in morbidity parameters may be affected by factors other than S . mansoni infection alone . Diet and hygiene practices seem to contribute to the health status of schoolchildren along Lake Victoria in Kenya and Tanzania .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] |
[
"schistosoma",
"invertebrates",
"schistosoma",
"mansoni",
"schoolchildren",
"medicine",
"and",
"health",
"sciences",
"education",
"helminths",
"sociology",
"tropical",
"diseases",
"social",
"sciences",
"diet",
"parasitic",
"diseases",
"physical",
"activity",
"animals",
"health",
"care",
"nutrition",
"neglected",
"tropical",
"diseases",
"educational",
"status",
"morbidity",
"public",
"and",
"occupational",
"health",
"schools",
"people",
"and",
"places",
"helminth",
"infections",
"schistosomiasis",
"physical",
"fitness",
"health",
"statistics",
"eukaryota",
"biology",
"and",
"life",
"sciences",
"population",
"groupings",
"malaria",
"organisms"
] |
2018
|
Diet and hygiene practices influence morbidity in schoolchildren living in Schistosomiasis endemic areas along Lake Victoria in Kenya and Tanzania—A cross-sectional study
|
Trachomatous trichiasis significantly reduces vision and health related quality of life ( QoL ) . Although trichiasis surgery is widely performed to treat trichiasis , there is little data on the effect of surgery on QoL . We measured the impact of trichiasis surgery on vision and health related QoL in a longitudinal study from Amhara Region , Ethiopia . We recruited 1000 adult participants with trichiasis ( cases ) and 200 comparison participants , matched to every fifth trichiasis case on age ( +/- two years ) , sex and location . Vision-related quality of life ( VRQoL ) and health-related quality of life ( HRQoL ) were measured using the WHO/PBD-VF20 and WHOQOL-BREF questionnaires respectively , at enrolment and 12 months after enrolment . All trichiasis cases received free standard trichiasis surgery immediately after enrolment . The mean difference in QoL scores between enrolment and follow-up for cases and comparison participants , and the difference-in-differences by baseline trichiasis status was analysed using random effects linear regression , the later adjusted for age , sex and socioeconomic status . At 12-months follow-up , data was collected from 980 ( 98% ) and 198 ( 98% ) trichiasis cases and comparison participants respectively . At this follow-up visit , VRQoL and HRQoL scores of trichiasis cases improved substantially in all subscales and domains by 19 . 1–42 . 0 points ( p<0 . 0001 ) and 4 . 7–17 . 2 points ( p<0 . 0001 ) , respectively . In contrast , among the comparison participants , there was no evidence of improvement in VRQoL and HRQoL domain scores during follow-up . The improvement in VRQoL and HRQoL in cases was independent of the presence of visual acuity improvement at 12 months . Trichiasis surgery substantially improves both VRQoL and HRQoL regardless of visual acuity change . Unprecedented effort is needed to scale-up trichiasis surgical programmes not only to prevent the risk of sight loss but also to improve overall wellbeing and health perception of affected individuals .
Trachoma , an eye disease caused by Chlamydia trachomatis , is the leading infectious cause of blindness worldwide [1] . The infection can lead to progressive conjunctival scarring and subsequently trachomatous trichiasis , the in-turning of eyelashes . Trichiasis in turn can cause constant painful abrasion to the cornea , irreversible corneal opacification and ultimately visual impairment and blindness . Approximately 7 . 3 million people have untreated trichiasis , and 2 . 4 million people are visually impaired from trachoma worldwide [2 , 3] . Trachomatous trichiasis is a painful condition , which can have a major impact on the individual’s general health and wellbeing , even prior to the development of visual impairment [4] . Moreover , it may have major socioeconomic consequences for affected families and communities [4–6] . We have previously reported that trichiasis adversely impacts vision and health related quality of life ( QoL ) , even before visual impairment develops [7] . Other studies have found trichiasis causes considerable functional and physical impairment , social withdrawal and exclusion , inability to work and earn an income [4 , 6 , 8 , 9] . The World Health Organization ( WHO ) recommends corrective eyelid surgery for trichiasis , to reduce the risk of sight loss [10] . Limited evidence suggests that the benefits of surgery may go beyond preventing vision loss and in fact help to restore the physical , social , psychological , environmental and economic wellbeing of individuals through improving vision and reducing pain and discomfort [8 , 11 , 12] . However , detailed empirical data on the impact of surgery on QoL is lacking . One longitudinal study from Ethiopia assessed the effect of trichiasis surgery on physical functioning of 411 trichiasis patients , six months after trichiasis surgery [9] . Another study , conducted in India , assessed HRQoL in 60 trachomatous entropion patients before and 1 month after trichiasis surgery and epilation [13] . No studies have measured the long-term overall effect that trichiasis surgery has on the different QoL domains and overall wellbeing . The WHO has developed and validated several tools for measuring QoL . These include the WHO/PBD-VF20 which is designed to measure vision related quality of life ( VRQoL ) ; and the WHOQOL-BREF which measures general health related quality of life [14 , 15] . We have previously reported a case-control study , which used both of these tools to compare the QoL of individuals with trichiasis to healthy controls , and found substantial differences [7] . Here we report a longitudinal study of these same cases and controls ( hereafter referred to as comparison participants ) , in which we explore the longer-term impact of trichiasis surgery on vision and health-related QoL .
This study was reviewed and approved by the National Health Research Ethics Review Committee of the Ethiopian Ministry of Science and Technology , the London School of Hygiene & Tropical Medicine ( LSHTM ) Ethics Committee , and Emory University Institutional Review Board . Written informed consent in Amharic was obtained prior to enrolment from participants . It was conducted in accordance with the Declaration of Helsinki . If the participant was unable to read and write , the information sheet and consent form were read to them and their consent recorded by thumbprint . Interviews were conducted privately , paper data were archived in a locked cabinet and electronic data were stored on a password-protected computer isolated from the Internet in a secured dedicated study office . Study participants with identified ocular problems were managed as per local protocol . This longitudinal study was nested within a clinical trial of two alternative surgical treatments for trichiasis [16] . We recruited 1000 trichiasis cases into the trial , who were also enrolled into this QoL study . The pre-operative baseline data from this study have been previously reported in detail [7] . Cases were defined as individuals with one or more eyelashes touching the eyeball or with evidence of epilation in either or both eyes in association with tarsal conjunctival scarring . People with trichiasis from other causes , recurrent trichiasis and those aged <18 years were excluded . Trichiasis cases were identified mainly through community-based screening [17] . Recruitment was done in three districts of West Gojam Zone , Amhara Region , Ethiopia between February and May 2014 . This area has one of the highest burdens of trachoma worldwide [18] . We also recruited 200 comparison participants . These were individuals without clinical evidence or a history of trichiasis ( including epilation ) , who came from households without a family member with trichiasis or a history of trichiasis . Comparison participants were individually matched to every fifth trichiasis case by location , sex and age ( +/- two years ) . The research team visited the sub-village ( 30–50 households ) of the trichiasis case that required a matched control . A list of all potentially eligible people living in the sub-village of the case was compiled with the help of the sub-village administrator . One person was randomly selected from this list using a lottery method , given details of the study and invited to participate if eligible . If a selected individual refused or was ineligible , another was randomly selected from the list . When eligible comparison participants were not identified within the sub-village of the index case , recruitment was done in the nearest neighbouring sub-village , using the same procedures . Data from trichiasis cases were collected at health facilities at the time of enrolment into the clinical trial , prior to trichiasis surgery . Data from the comparison participants were collected at their homes . Six trained Amharic speaking interviewers collected data from participants using a standardised questionnaire , including socio-demographic variables ( age , sex , marital and literacy status ) , presence of other health problems in the last month and self-rated socioeconomic status ( SES ) . For self-rated SES , participants were asked to rate the wealth of their households in relation to other households in their village by choosing one of the following options: ( 1 ) very poor , ( 2 ) poor , ( 3 ) average , ( 4 ) wealthy or ( 5 ) very wealthy [5] . Immediately after baseline data collection was completed , all cases received trichiasis surgical management . They were randomised to receive either the bilamellar or the posterior lamellar tarsal rotation , which were being compared in the clinical trial [16] . Both surgical procedures involve an incision through the scarred upper eyelid , parallel to and about 3mm above the lid margin , followed by outward rotation and suturing in the corrected position [26] . Six standardised trichiasis surgeons performed the surgery . Follow-up was conducted approximately one year after enrolment ( minimum 10 and maximum 14 months ) , during the same season as the baseline . For cases , a reminder was sent to attend the 12-month follow-up . Follow-up data were collected for the majority of cases at a health facility . For cases who could not come to the health facility , data were collected during a home visit . Follow-up data were collected on comparison participants at their homes . Participants were interviewed using the same QoL tools at baseline , and clinical data were collected using the same procedures by the same interviewers and a clinical grader . The sample size of 1000 cases and 200 comparison participants has 94% power to detect an effect size of about 0 . 27 ( standardised mean difference ( 3/11 ) ) with a Type I error of 5% . Data were double-entered into Access ( Microsoft ) , cleaned in Epidata 3 . 1 and transferred to Stata 11 ( StataCorp ) for analysis . Analyses were restricted to participants with both baseline and follow-up data .
At baseline 1000 trachomatous trichiasis cases and 200 comparison participants were recruited . At the 12-month follow-up , complete QoL data were collected from 980 ( 98% ) cases and 198 ( 98% ) comparison participants . Cases and comparison participants had a similar age distribution ( mean 47 . 0 vs 45 . 7 years ) , but there were fewer females among the cases than among the comparison participants ( 76 . 4% vs 84 . 3% ) . This difference occurred by chance , as the randomly selected 200 trichiasis cases used to determine the matching characteristics had a higher proportion of females than the full group of 1000 trichiasis cases , Table 1 . After adjusting analyses for age and gender , the trichiasis cases were significantly more likely to be illiterate , widowed or divorced , be from poorer households , and report other health problems in the past month than comparison participants , Table 1 . At baseline almost all comparison participants ( 97% ) had normal vision ( ≥6/18 ) compared with about half ( 52% ) of the cases ( in the operated eye ) . Trichiasis cases also had significantly lower contrast sensitivity score than the comparison participants at baseline ( P<0 . 0001 ) , and poorer visual acuity and contrast sensitivity scores at 12-months ( Table 1 ) . In contrast , at 12 months follow-up there was strong evidence of an improvement in visual acuity ( mean LogMAR change: 0 . 08; 95%CI: 0 . 05 to 0 . 10 ) and contrast sensitivity scores ( mean contrast sensitivity score change: 2 . 41; 95%CI: 1 . 48 to 3 . 35 ) among trichiasis cases . Moreover , among comparison participants there was a small but significant reduction in visual acuity ( mean LogMAR change: -0 . 04; 95%CI: -0 . 07 to -0 . 02 ) and no evidence of a change in contrast sensitivity scores ( mean contrast sensitivity score change: -0 . 14; 95%CI: -0 . 40 to 0 . 13 , p = 0 . 31 ) . At baseline , trichiasis cases had substantially lower VRQoL scores than comparison participants in all four subscales ( p<0 . 0001 ) , Table 2 . One year after trichiasis surgery the mean VRQoL score of cases had improved substantially in all subscales by 19 . 1 to 42 . 3 points ( p<0 . 0001 ) , with large effect sizes in the visual symptom ( 2 . 03 ) , overall eyesight ( 1 . 57 ) and psychosocial ( 0 . 88 ) subscales and a moderate effect size in the general functioning ( 0 . 67 ) subscale . In contrast , there was no evidence of a difference between baseline and follow-up VRQoL scores in all subscales scores in the comparison participants and the effect sizes were very small and negative ( -0 . 08 to -0 . 03 ) , Table 2 . The difference-in-differences analysis showed strong evidence that the improvement in mean VRQoL score from baseline to 12-month follow-up was greater for cases than comparison participants in all sub-scales ( difference-in-differences score: 18 . 9 to 42 . 9 points , p<0 . 0001 ) , Table 2 . Similar results were seen when analyses were stratified by level of visual acuity change at 12-months , Table 3 . VRQoL score among trichiasis cases improved significantly in all subscales , independent of visual changes over 12-months , while the VRQoL scores of comparison participants remained the same ( Table 3 ) . The largest VRQoL change was observed in trichiasis cases with visual improvement , Table 3 . In addition , among trichiasis cases , VRQoL improved in all subscales independent of postoperative recurrent trichiasis , although the improvements in the overall eyesight was significantly less in those with recurrent trichiasis than their counterparts . ( 30 . 8 vs 35 . 7; p = 0 . 04 ) .
For VRQoL , the largest improvement was seen in the visual symptom subscale indicating the major effect that surgery has on relieving pain and discomfort from trichiasis . Trichiasis surgery is also shown to significantly improve visual acuity and contrast sensitivity , which might be related to elimination of the photophobia and tears and restoration of the corneal epithelium as result of the removal of the trichiatic lashes [8 , 11] . Larger gains in VRQoL were seen in those with improved vision . However , substantial improvement in VRQoL also occurred in patients without visual acuity improvement or even in those with deteriorating vision . This strongly supports the view that trichiasis surgery should be performed not only to save vision but also to treat other debilitating symptoms and improve the VRQoL of affected individuals . In contrast , the comparison participants had no improvement in visual acuity , contrast sensitivity or VRQoL 12-months after enrolment . There has been no previous longitudinal study quantifying the effect of trichiasis surgery on VRQoL . However , our findings are consistent with a retrospective qualitative study of 13 women with trichiasis in Niger , in which most participants reported that trichiasis surgery markedly improved their quality of life in association with a complete disappearance of the painful physical symptoms [6] . Interestingly , despite a fundamental difference between cataract and trichiasis in the amount of visual loss , the effect size for the improvement in overall eyesight between our study and a study of the impact of cataract surgery in Kenya was comparable ( 1 . 6 vs 1 . 5 ) and there were large effect sizes ( >0 . 80 ) in the psychosocial subscales in both studies [30] . Among trichiasis cases , there was marked improvement in HRQoL one year after trichiasis surgery , independent of change in vision . The largest improvement was seen in the physical health domain indicating that trichiasis surgery considerably improves work capacity and ability to function without pain . Similar findings have been reported in the Niger qualitative study [6] . In contrast , there was either reduction or no change in the HRQoL of comparison participants one year after enrolment . The results of our study were consistent with the only longitudinal study on HRQoL and trichiasis , conducted in India , which assessed HRQoL in patients before and one month after trichiasis surgery for trachomatous entropion ( n = 41 ) and 15 days after epilation for minor trichiasis cases ( n = 19 ) . In the Indian study , HRQoL significantly improved after treatment of patients with normal and poor vision in the physical health , psychological and environment domains , while there was no significant improvement in the social domain . Consistent with our study , the greatest and least improvements were seen in the physical health ( mean 21 . 3 to 23 . 1 points ) and social domains ( 2 . 0 to 3 . 3 points ) , respectively . The social domain is a composite of social support and personal relationships , which may be less altered by trichiasis surgery . The QoL improvement in our study was larger in the three HRQoL domains except the physical domain , than reported in the Indian study . These two studies have the following differences that might explain the differences in HRQoL gains . Firstly , most of the cases ( 67% ) in the Indian study had entropion without trichiasis while all had trachomatous trichiasis in our study . Secondly , 19 cases with minor trichiasis received epilation ( rather than surgery ) while everyone in our study underwent trichiasis surgery . Thirdly , there was less than one month of follow-up in the Indian study during which the surgical wound healing process might influence QoL results [13] . Longer trichiasis duration and central corneal opacity at baseline predict larger VRQoL gains in all subscales . This suggests that cases with severe disease probably benefit more in terms of improved VRQoL following trichiasis surgery . The results also suggest that bilateral cases benefit more from bilateral surgery to restore their functioning . Older trichiasis cases had greater improvement in the general functioning subscale than younger cases , while younger cases showed greater improvement in the psychosocial scores . Older individuals had more difficulty in general functioning in relation to distance and near vision difficulties at baseline , hence should benefit more from trichiasis surgery in improving their vision and thereby participation in day to day activities . Younger cases may be more likely to be embarrassed , worry about losing eyesight and hesitate to participate in social activities due to their trichiasis than older cases , which may be the reason for younger trichiasis cases to having greater improvement in the psychosocial subscale than older people one year after trichiasis surgery . This is the first large longitudinal study to measure VRQoL and HRQoL change after trichiasis surgery using validated WHO tools . The same interviewers collected data at both baseline and follow-up to ensure questionnaires were administered in a standard way at baseline and follow-up . The study used comparison participants . Absence of significant improvement in QoL among the comparison participants compared to the cases lends weight to the view that the positive QoL change observed is attributable to trichiasis surgery . The study has a number of limitations . The interviewers were not masked in the trichiasis status of the participants and we cannot rule out the possibility of response bias with cases providing a positive answer to satisfy the interviewers regardless of real change or improvement . To limit these , participants were asked to provide honest answers and reassured that their answers would not affect their treatment in any way . In addition , knowledge of trichiasis status by the interviewers might possibly introduced bias in the outcome assessment . Although the improved QoL probably is largely related to vision improvement , less pain and irritation after the surgery , it is difficult to rule out the possibility that the extent of positive change observed among cases might have been related to the effect of just receiving a health intervention rather than solely attributed to the effect of trichiasis surgery . This possibly would explain the 28 . 4 score improvement in the overall eyesight subscale in those with vision deterioration one year after trichiasis surgery . Overall this study demonstrated that trichiasis surgery substantially improves both vision and health related QoL regardless of the visual acuity improvement , suggesting that the effect of trichiasis surgery goes beyond preventing the risk of blindness and improves the overall wellbeing and health perception of affected individuals . Unprecedented effort is needed to scale-up trichiasis surgical programmes and provide prompt surgical intervention to improve overall wellbeing of affected individuals .
|
We previously reported that Trachomatous Trichiasis ( TT ) has a profound impact on vision and heath related quality of life ( QoL ) , even when vision is not impaired . The World Health Organization ( WHO ) recommends corrective eyelid surgery for trichiasis to reduce the risk of vision loss . However , trichiasis surgery may also improve overall wellbeing . There is very limited evidence on the long-term impact of trichiasis surgery on QoL . We measured vision and health-related quality of life of 1000 TT patients before and one year after receiving TT surgery and compared the QoL scores of these with the baseline and 1 year follow-up QoL score of 200 matched individuals who have never had trichiasis or trichiasis surgery . We found strong evidence that surgery substantially improves both vision and heath related QoL of TT case , even when there is no improvement in vision; while there was no evidence of improvement in the QoL of the trichiasis free participants . The results provide clear evidence that the benefit of trichiasis surgery goes beyond preventing the risk of blindness and improves the overall wellbeing and health perception of affected individuals , indicating the need to provide prompt surgical intervention for affected individuals .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"medicine",
"and",
"health",
"sciences",
"social",
"sciences",
"neuroscience",
"surgical",
"and",
"invasive",
"medical",
"procedures",
"visual",
"acuity",
"health",
"care",
"regression",
"analysis",
"optical",
"properties",
"mathematics",
"statistics",
"(mathematics)",
"materials",
"science",
"vision",
"eyes",
"research",
"and",
"analysis",
"methods",
"mathematical",
"and",
"statistical",
"techniques",
"head",
"psychology",
"visual",
"impairments",
"quality",
"of",
"life",
"anatomy",
"linear",
"regression",
"analysis",
"ophthalmology",
"biology",
"and",
"life",
"sciences",
"ocular",
"system",
"sensory",
"perception",
"physical",
"sciences",
"material",
"properties",
"statistical",
"methods",
"opacity"
] |
2016
|
Impact of Trichiasis Surgery on Quality of Life: A Longitudinal Study in Ethiopia
|
The packaging of DNA inside a nucleus shows complex structure stabilized by a host of DNA-bound factors . Both the distribution of these factors and the contacts between different genomic locations of the DNA can now be measured on a genome-wide scale . This has advanced the development of models aimed at predicting the conformation of DNA given only the locations of bound factors—the chromatin folding problem . Here we present a maximum-entropy model that is able to predict a contact map representation of structure given a sequence of bound factors . Non-local effects due to the sequence neighborhood around contacting sites are found to be important for making accurate predictions . Lastly , we show that the model can be used to infer a sequence of bound factors given only a measurement of structure . This opens up the possibility for efficiently predicting sequence regions that may play a role in generating cell-type specific structural differences .
In higher-complexity organisms , the packaged molecule of DNA inside the nucleus of a cell consists not only of the DNA polymer but also a large number of DNA-associated protein and RNA complexes that together form what is known as chromatin . These complexes bind to the DNA by either sequence affinity or by interactions with other bound factors [1] , and their presence is known to be correlated with the 3D conformation of the DNA polymer [2] . They impact the structure of the DNA on many length scales: from the packing of nucleosomes at the smallest scale ( ∼200 basepair ) to the stabilization of loops and genomic domains on much larger scales ( ∼10-100 kilobasepair ) . As a result , they regulate a host of important cell functions from DNA replication to gene expression [3 , 4] . Developing models that can predict how chromatin folds given only the locations of bound factors is of key importance to better understand how they regulate such processes by shaping DNA structure . The average structure of DNA over a population of cells can be measured genome-wide using a high-throughput DNA sequencing method known as Hi-C [5] . Briefly , contacting sites in the genome are cross linked; then the DNA is fragmented and the contacting pairs are sequenced . These sequenced pairs are then used to construct a contact map at a given spatial resolution that gives the number of times any pair of sites along the genome were found to be in contact . Analysis of the contact maps have shown that chromatin can be classified into structural types such as A/B compartments [5 , 6] or topologically associated domains ( TADs ) [7] just from the spatial distribution of contacts . These structural features are known to be strongly correlated with different types of DNA-bound factors . The locations of these bound factors along the genome constitute a form of sequence that helps drive the folding of the DNA—similar to the specific sequence of amino acids that drives a protein to fold . High throughput methods can provide the binding locations of such factors on a genome-wide scale [8] . Interestingly , despite tens to hundreds of different chromatin associated factors , clustering of their binding locations shows that there are only few unique bound states [9–13] ( similar to the grouping of amino acids into just hydrophobic and polar types ) . With the richness of this structural and sequence data for chromatin , predictive models that aim to solve the chromatin folding problem—namely predicting the structure of DNA inside a cell given only the locations of the DNA-bound chromatin factors—are now being developed . Recent modeling efforts have used polymer-based models whose parameters can be tuned to reproduce experimental observations , such as the contact map from Hi-C . One set of approaches tries to find the best 3D polymer structure that is consistent with the constraints imposed by the Hi-C contact map [14–19] . Other methods include bound factors by adding sequence-specific interactions to a given polymer model for the DNA [20–33] . These approaches have been successful in showing how interactions between factors together with topological constraints may be responsible for the observed chromatin structures . Challenges involve continuing to improve the physics of the interacting polymer model using data-driven methods and the time-consuming process of carrying out the polymer simulations . Complementing the polymer simulation methods are purely statistical approaches that aim to predict a contact map instead of a full 3D structure . Sexton et al . [34] developed a statistical model based on site-specific scaling factors that could predict a matrix of expected counts . Other work included sequence information by fitting a pairwise interaction model for the DNA-bound factors that could then predict the probability of contact between pairs of sites given only the sequence at those locations [35] . However it did not include an important effect that is present in polymer simulations , namely the role of neighbors . A growing body of experimental evidence supports the importance of the local sequence neighborhood of bound factors in mediating contacts between pairs of genomic sites [36] . In particular the probability of contact between two sites i and j at a certain genomic distance apart is altered if some site k in the neighborhood of j is attracted to i . Sites k and i will spend a fraction of the time in contact , thus altering the effective polymer distance between i and j ( see S1 Fig ) . Here , our aim is to take such effects into account and predict the probability of two sites being in contact ( i . e . the contact map ) given a local sequence neighborhood . The results presented in the following sections demonstrate how these probabilities can be modelled by a maximum entropy distribution . In addition , by formulating the problem in a Bayesian fashion , we are not only able predict probabilities of contact from an experimentally measured sequence of bound factors , but also predict the probability of a site having a particular bound factor using only the measured local Hi-C contact map
Fig 1 shows a schematic of our model and resulting calculation . First , the genome is discretized into non-overlaping sites of fixed size . A particular chromatin state σk is assigned to to each site k based on the bound chromatin factors there . As mentioned in the introduction , only a few bound states exist and for simplicity we classify each site into only one of two possible sequence states corresponding to active/euchromatic and inactive/heterochromatic DNA , respectively labeled as spin-up ( σk = 1 ) and spin-down ( σk = −1 ) analogous to the physics of ferromagnets ( see Fig 1 ( A ) and Materials and methods for details ) . With respect to structure , Fig 1 ( A ) and 1 ( B ) shows two sites i and j separated by a genomic distance , d = |j − i| , that have a certain chance of forming a contact that we assume is determined by the sequence neighborhood , σ → , situated around them . Hi-C data gives an estimate of this chance by reporting the number of times nij that this pair formed contacts , and hence the likelihood that their sequence neighborhood will form a contact ( Fig 1 ( B ) ) . We denote this probability of contact as P ( c | σ → , d ) which depends on the genomic distance d between sites and sequence neighborhood σ → ( here c indicates that the two sites are in contact; c is therefore one of the values of the random variable z: z = c if contact and z = c ¯ otherwise ) . The neighborhood σ → is a vector containing N sites built from the union of sequence windows situated around the contacting sites i and j and thus depends on d ( see Fig 1 ( B ) and 1 ( C ) , Materials and methods and S2 Fig for further details ) . Using Bayes’ rule we can rewrite P ( c | σ → , d ) as P ( c | σ → , d ) = P ( σ → | c , d ) P ( c | d ) P ( σ → | d ) , ( 1 ) where P ( σ → | c , d ) and P ( σ → | d ) are the probability of observing the given sequence neighborhood σ → given that it has formed a contact and regardless of contact , respectively . P ( c|d ) is the sequence-independent probability of contact at a separation d and can be estimated directly from the Hi-C experiments . In principle , Hi-C data could also be used to directly estimate P ( σ → | c , d ) since the contact map gives the number of times that a given neighborhood was found to form a contact ( see Fig 1C ) . Similarly , P ( σ → | d ) could be estimated from the frequency of occurrence of that sequence neighborhood in the genome . However , because the number of different sequence neighborhoods goes as 2N , even for modest N there is not enough data to make accurate estimates . To overcome this we utilize the principle of maximum entropy to calculate probability distributions , P ( σ → | c , d ) and P ( σ → | d ) , that reproduce a few reliably estimated statistics of σ → extracted from experimental Hi-C data [37 , 38] . The statistics we use to constrain both of the above distributions are the average spin 〈σk〉 at a given position k in the neighborhood ( which represents the average chromatin state at that neighbourhood postion ) and the average spin-spin correlation 〈σkσl〉 between sites k and l in the neighborhood ( which represents the average correlation between the chromatin states at the two neighbourhood positions ) . The averages denoted by 〈⋅〉 are calculated over their respective ensembles of sequence neighbourhoods { σ → } d that surround pairs of sites at a distance of contact d = |j − i| . The ensembles for both distributions can be extracted by scrolling along the sequence of binary states in the genome . When characterizing P ( σ → | c , d ) , each neighbourhood has a weight equal to the number of contacts nij observed between sites i and j; when characterizing P ( σ → | d ) , each neighbourhood has a unit weight for every time it appears in the genome ( as shown in Fig 1 and Materials and methods ) . With the above statistics as constraints , the maximum-entropy distribution can be found using the method of Lagrange multipliers [37–39] and has the form of a Boltzman distribution for the Ising model at kBT = 1 , P ( σ → | · ) = e ∑ k h k σ k + ∑ ∑ l > k J k l σ l σ k Z ( σ → | · ) , ( 2 ) where hk and Jkl are Lagrange multipliers that constitute the fitting parameters of the model . The partition function Z ( σ → | · ) is a normalization constant obtained by summing the numerator of Eq 2 over all possible σ → neighborhoods , Z ( σ → | · ) = ∑ σ → exp ( ∑ k h k σ k + ∑ ∑ l > k J k l σ l σ k ) . In the above , we use “|⋅” to summarize in a single notation two different conditions at each distance of contact , namely P ( σ → | c , d ) and P ( σ → | d ) ( “⋅” substitutes “c , d” or “d” ) . For neighborhoods of size up to N ∼ 23 , exact enumeration can be used to evaluate Eq 2 in minimal time . Beyond that , one would need to estimate this distribution by Monte Carlo sampling [40] . We have thus restricted our neighbourhoods to a maximum size N = 20 so that we can conveniently calculate Eq 2 by enumerating the 2N possible neighbourhoods at a given d ( see Materials and methods ) . At a given genomic distance d the distribution in Eq 2 can be fit to reproduce the experimental statistics 〈σk〉 and 〈σkσl〉 calculated from the sequence ensemble at that distance ( see Materials and methods ) . We now detail how we applied this method to estimate these distributions and ultimately the contact probability , Eq 1 , from real experimental sequence and structural data . As a test of the method , we used the measured structure and sequence data for Drosophila Melanogaster . For structure , we used a Hi-C contact map from Drosophila embryos , generated using sites that were 10 kilobasepair ( kbp ) in size [41] . For sequence , we used our own binary classification ( spin-up or spin-down ) of the measured Drosophila chromatin-associated factors at each 10 kbp site , which was based on the “chromatin colors” classification [11] ( see Materials and methods ) . To avoid overfitting the model , we divided the data from Drosophila chromosomes 2 and 3 into a training set used for fitting and a test set for prediction ( see Materials and methods ) . For genomic distances ranging from d = 10 kbp to 800 kbp in 10 kbp steps , we used the above maximum-entropy method to estimate the two conditional probability distributions , P ( σ → | c , d ) and P ( σ → | d ) . Fig 2 ( A ) –2 ( D ) shows that the fitted distributions successfully predicted the experimental statistics 〈σk〉 and 〈σkσl〉 on the test data . They also captured three-point correlations 〈σkσlσm〉 ( see Fig 2 ( E ) and 2 ( F ) ) despite not having incorporated them into the fit . In Fig 2 ( G ) and 2 ( H ) we also see that the predicted sequence neighborhood probabilities P ( σ → | · ) from the model agreed with their frequencies as seen in the data . Thus these second-order maximum-entropy distributions seemed to be adequate approximations to the true distributions ( first-order maximum-entropy distributions were also tested and failed to reproduce experimental statistics , see S3 Fig and S1 Text ) . Inspection of the associated fit parameters hk and Jlk as a function of genomic distance , d , showed that they naturally clustered into two groups ( through both K-means and Principal Component Analysis ) , with a transition from one to the other occurring at d ≈ 390 kbp ( see S4 Fig and S1 Text ) . Averaging the parameters within each group together we found that 〈hk〉 of the two sites in contact were positive ( especially for the group above transition ) . Positive values for hk favour the spin-up active/euchromatic state and thus , the above finding shows that contacting sites had a tendency to be in the such a state . Interestingly , the sites in between the contacting sites had an inactive/heterochromatic preference ( i . e . negative values for 〈hk〉 ) in the group below the transition , and no chromatin preference in the group above . Therefore at short genomic separations , a heterochromatic region between contacting sites favored contact . We also found that the interaction terms 〈Jlk〉 for the neighbors between contacting sites were ferromagnetic for the group below the transition ( i . e . 〈Jlk〉 > 0 which favors sites l and k to be in the same state ) and were antiferromagnetic for the group above ( i . e Jkl < 0 which favours l and k to be in opposite states ) . Thus at short genomic distances , the sequence between the contacting sites was favored to be homogeneous , whereas at longer distances heterogeneity or a more random sequence seemed to be the case ( see S4 Fig and S1 Text ) . We used the fitted maximum entropy distributions , P ( σ → | c , d ) and P ( σ → | d ) at each distance , d , along with Eq ( 1 ) to make predictions for distance-normalized contact maps P ( c | σ → , d ) / P ( c | d ) on the test data from Drosophila chromosomes 2 and 3 ( red heat maps in Fig 3 ) . Each ( i , j ) pair had an associated σ → from which we could then evaluate Eq 1 using the fitted distributions . ( We normalized by P ( c|d ) so as to remove the strong decay with distance of the probability of making a contact ) . These were then correlated to the normalized experimental Hi-C contact maps nij/〈n ( d ) 〉 of the test data ( blue heat maps in Fig 3 ) , where 〈n ( d ) 〉 is the average number of Hi-C contact counts at a given distance d , that also decays strongly with distance . The correlations with the Drosophila chromosomes were 0 . 39 ( chromosome 2L ) , 0 . 53 ( chromosome 2R ) , 0 . 53 ( chromosome 3L ) , and 0 . 40 ( chromosome 3R ) . These correlations are remarkable given that only a binary model for the sequence was used . The areas of discrepancy in Fig 3 may be due to the fact that the binding factors were measured from different cell lines than the Hi-C ones , so it is possible that in those regions the actual underlying sequence that generated the observed contact counts may be different than what was used in making the prediction . The fitted maximum-entropy model that can predict contact maps from chromatin sequence provides an opportunity to also predict structural changes that might arise due to mutations in the underlying sequence of bound factors . We thus set to identify which genomic locations are expected to disrupt the local structure the most in the event that their chromatin state is flipped . This analysis was performed in genomic regions where the locally-predicted contact probabilities agreed the most with experimental measurements ( correlation between predicted and experimental distance-normalized probabilities of contact c > 0 . 6 ) . The chromatin state , σ , of each of these well-predicted locations was individually inverted and the correlation , c′ , between the newly predicted probabilities and the experimental ones was calculated ( see Materials and methods for details ) . We interpret the change in correlation Δc = c′ − c to be a measure of structural-sensitivity to chromatin sequence mutations for that position . In Fig 4 we show histograms of Δc for genomic locations categorized by genomic feature . Regardless of genomic feature , the vast majority of sites had a negative Δc after their chromatin state was inverted , indicating that the local predicted structure tended to depart from the experimental structure when the sequence state was mutated . In Fig 4A we found that the inactive/heterochromatic ( spin down ) sites were significantly more structurally sensitive than the active/euchromatin ( spin up ) sites ( average Δc was -0 . 11 for spin-down whereas -0 . 04 for spin-up sites ) . In other words , a change from inactive to active DNA tended to produce a greater structural change than the opposite . In Fig 4B we classified genomic sites into three non-overlapping categories: sites where no genes were present , sites that contained gene promoters , and “gene body” sites corresponding to locations occupied by genes but with no promoters . We found that sites with no genes were the most structurally sensitive , followed by “gene body” sites and lastly , the least structurally sensitive were sites containing promoters ( average Δc was -0 . 15 for no genes , -0 . 10 for “gene body” and -0 . 06 for promoter sites ) . These findings highlight that non-coding regions of the genome have the greatest capacity to alter DNA structure , whereas mutations in gene rich regions are less likely to cause significant alterations to structure . Here we show that it is possible to solve the inverse problem , namely , given a Hi-C map of contact counts , {nij} , and a model for the probability of a sequence neighborhood σ → to be in contact , determine the probability of each site k to be in a particular sequence state ( σk = 1 or σk = −1 ) . We denote this probability as P ( σk|{nij} ) where {i , j} is the set of all contacting pairs that contain genomic site k in their sequence neighborhood . By applying Bayes’ rule we can write it as P ( σ k | { n i j } ) = P ( { n i j } | σ k ) P ( σ k ) P ( { n i j } ) , ( 3 ) where P ( σk ) is a prior on the sequence state at site k , and P ( {nij} ) the probability of the data that simply acts as a normalization constant . As we show in S1 Text , Eq 3 can be rewritten as P ( σ k | { n i j } ) = P ( σ k ) 1 - M P ( { n i j } ) ∏ i , j ∑ σ → P ( n i j | σ → , d ) P ( σ → | d ) δ σ k ′ , σ k , ( 4 ) where P ( n i j | σ → , d ) is the probability of observing a particular number of counts between sites i and j given their sequence neighborhood , σ → , k′ is the position of genomic site k in that neighborhood , and M is the the number of {i , j} pairs . We take P ( n i j | σ → , d ) to be a Gaussian distribution N ( λ σ → , d , ζ σ → , d 2 ) with mean equal to the average number of counts for a given sequence σ → , λ σ → , d = λ ( c | σ → , d ) , which is proportional to the probability of contact P ( c | σ → , d ) that can be evaluated from Eq 1 . The variance , ζ σ → , d 2 can be estimated from the Hi-C data ( see S1 Text for more details ) . Using just the observed Hi-C counts nij for the test data and our prior fitted maximum entropy models for P ( c | σ → , d ) and P ( σ → | d ) , we calculated the probability of each site in the test set being spin-up . Fig 5 shows this probability as a function of genomic location along the four test chromosome regions . Applying a threshold to these probabilities ( P ( σk = 1 ) = 0 . 5 ) , we predicted a sequence with a percent agreement with the original test sequence of 78% ( chromosome 2L ) , 72% ( chromosome 2R ) , 77% ( chromosome 3L ) and 77% ( chromosome 3R ) . Thus structure alone can yield an important amount of information about the underlying sequence of bound chromatin factors . We feel that this could have significant impact in determining important regions for regulating structure . In particular , if the maximum entropy model is reliably capturing the essential aspects that connect sequence to structure and if one believes that these aspects are conserved across different cell types , one could use a fitted model from one cell type to predict sequence in another for which only structural information is known .
Our results demonstrate that a model for the probability of contact between two genomic sites given just their neighborhood of bound factors can be estimated using a maximum-entropy method . The resulting model did well at predicting the contact map of a test set using only sequence information . This model connects sequence to structure and thus offers the capability of testing the structural effect of mutating the chromatin sequence . Our analysis highlights that the alteration of DNA conformation by mutating chromatin sequence is particularly strong in sites containing no genes , while only minimal alterations to structure tend to occur when mutating active gene-rich regions . Although we applied our fit model to data from Drosophila embryos , we envision that it could be useful in predicting contact maps from sequence data taken from other cell types from the same species ( e . g . different developmental times or from mutated cell lines ) , or potentially other nearby species where the contact model might expect to hold . We also showed how a fitted model connecting sequence to structure could be used to solve the inverse problem , namely that a sequence of bound factors could be predicted using just measured counts from a Hi-C contact map and a model for the probability of contact given sequence . One could imagine that making such sequence predictions from Hi-C measurements on individual cell lines may be more efficient than carrying out the potentially tens to hundreds of experiments needed to map the locations of all the bound factors . The classification of the genome into chromatin states used to fit our model was based on the occupation patterns of specific binding factors rather than structure . Therefore , we consider the possibility that other classifications may have a greater ability to predict structure from sequence than the one presented . Future work will consider other possible ways of defining sequence states from data for the purpose of predicting structure through this maximum entropy scheme , which may help to further elucidate the biological mechanisms behind chromatin folding . Despite the great variety of DNA-binding chromatin factors identified so far this method is able to reproduce DNA contact maps from binary classifications that group all of the observed marks into only two classes . It is therefore tempting to speculate that at the > 10 Kbp scale , once DNA-bound factors are positioned on the genome , they tend to behave alike when it comes to generating DNA contacts . If correct , this would imply that the utility of such variety of chromatin marks is not entirely related to stabilizing complex DNA structures where a large number of different architectural elements is required . Instead , our findings favor the idea that the observed diversity of chromatin states is primarily related to alternative functions other than creating complex large-scale structures , such as the programming of chromatin states at different developmental times , or genetic regulation . Overall our model has the potential for efficiently making predictions about chromatin structure depending on whether such structural or sequence data are available .
DNA contact maps were obtained from the publicly available Hi-C experiments done by Schuettengruber et al [41] ( GSE61471 ) , performed on 3000-4000 Drosophila melanogaster embryos , 16-18 hours after egg laying . The contact map is an array whose elements are the number of times a particular pair of genomic sites were found to be in contact . The contact map we used defined sites to be of a fixed size ( 10 kilobasepair ( kbp ) ) and were non-overlapping across the genome . The counts , nij , at a particular pair of sites i and j in the contact map were determined by counting up all sequenced pairs from the Hi-C measurement that fell into those sites . We normalized the contact map using the ICE method [6] so that total number of counts along each row across the contact map was the same . This removed potential biases between sites due to the Hi-C protocol . For every 10 kbp site , k , in the genome , we classified the genomic distribution of its DNA-bound factors into two possible chromatin sequence states corresponding to bound states that are associated with euchromatin ( active DNA regions ) , σk = 1 , and heterochromatin ( inactive DNA regions ) , σk = −1 . In order to make these sequence assignments , we used the “chromatin colors” classification of the Drosophila melanogaster genome by Filion et al [11] ( GSE22069 ) which categorizes the distribution of bound factors into five different bound states ( black , blue , yellow , green and red ) based on the DamID binding profiles of 53 different chromatin marks in the embryonic Drosophila melanogaster cell line Kc167 ( 8-12 hours ) . We then grouped the five “chromatin colors” into just a spin-up and spin-down state based on the biological functions of their associated DNA-bound factors: spin-down = black and blue , corresponding to inactive/heterochromatin and spin-up = yellow , green and red that correspond active/euchromatin . The color data gives the coordinates of the regions of the various colours . In order to assign a particular sequence state to each of our 10 kbp sites , we took the dominant color group ( either active , or inactive ) to be the unique sequence state of that site . For the purpose of validating our model , we divided the Drosophila melanogaster autosomal chromosomes , 2 ( comes as 2L and 2R ) and 3 ( comes as 3L and 3R ) into a training set and test set . The training dataset consisted of genomic locations 2L: 1Mbp-17 . 02Mbp , 2R: 1Mbp-15 . 15Mbp , 3L: 1Mbp-18 . 55Mbp , 3R: 1Mbp-21 . 91Mbp . The testing dataset consisted of genomic locations 2L: 17 . 02Mbp-22 . 02Mbp , 2R: 15 . 15Mbp-20 . 15Mbp , 3L: 18 . 55Mbp-23 . 55Mbp , 3R: 21 . 91Mbp-26 . 91Mbp . We constructed sets of sequence neighborhoods { σ → } d for genomic distances between pairs d = |j − i| ranging from d = 1 to d = 80 in units of the site size , 10 kbp . For a pair of sites i and j in the genome separated by d , we took the particular sequence neighborhood , σ → , to be the union of sequence windows centered around each site . Specifically , we took 10 sites centered around site i and 10 sites centered around site j to create a neighborhood size of N = 20 ( see S2 Fig ) where the given sequence neighborhood consists of sites σ → = ( σ i - 4 , … , σ i , … , σ i + 5 , σ j - 5 , … , σ j , … , σ j + 4 ) . For distances , d < 11 , it was not possible to take sequence windows of size 10 sites centered on each site , so for these cases , we defined the neighborhoods as σ → = ( σ i - 4 , … , σ i , … , σ j , … , σ j + 4 ) which had a size N = d + 9 ( S2 Fig ) . For a fixed genomic distance between sites d , we then scrolled through all pairs of sites , {i , j} in both the training and test regions of the genome extracting their corresponding neighborhoods σ → . Since the directionality ( left-right ) of the sequence neighborhood should not influence the contact probability between i and j , we also added the inverted sequence for each σ → to the collection of neighborhoods . Each genomic distance , d , thus had its own unique ensemble of sequence neighborhoods spin-vectors for both the training and test data . The maximum entropy distributions , P ( σ → | c , d ) and P ( σ → | d ) , were constrained to match several statistics of the extracted sequence neighborhood ensembles . The statistics used are 〈σk〉 and 〈σlσk〉 . For the distribution P ( σ → | c , d ) , these statistics were calculated as weighted averages of the extracted ensemble of sequence neighborhoods , where for each neighborhood we used as a weight the number of times , nij , that its contacting sites i and j were observed to be in contact . For P ( σ → | d ) the above statistics were calculated as weighted averages where the weight of each sequence in the ensemble was equal to the number of times it appeared in the genome . The parameters hk and Jlk in P ( σ → | · ) = e ∑ k h k σ k + ∑ ∑ l > k J k l σ l σ k Z ( σ → | · ) ( 5 ) were fit to reproduce the experimental statistics 〈σk〉exp and 〈σkσl〉exp by following the method described by Tkačik et al . [42] that we summarize next . The term “|⋅” simultaneously denotes the two conditions that we characterized at each distance of contact: The probability of observing σ → given contact P ( σ → | c , d ) and the probability of observing σ → regardless of contact P ( σ → | d ) . We therefore fit a set of parameters {hk , Jlk} for each of the two conditions denoted by |⋅ and for each distance of contact d . Briefly , given one of the two conditions above , we used the corresponding collection of spin vectors and their associated weights ( as described in the previous section ) to calculate the experimental statistics 〈σk〉exp and 〈σkσl〉exp . Then we implemented the following iterative scheme: This scheme is guaranteed to converge towards a unique solution regardless of the initial choice of hk and Jlk . First , we identified the sites k for which the local predicted contact map was in good agreement with the experimental contact map . The local contact map consists of only the subset of genomic pairs of sites {i , j} whose contact probability P ( c | σ → , d ) is influenced by the state of site k ( ie . k is one of the spins in the neighborhood σ → of i and j ) . At each genomic site , we calculated the correlation c between the local predicted distance-normalized contacts , P ( c | σ → , d ) / P ( c | d ) , and the local experimental distance-normalized contacts counts , nij/〈n ( d ) 〉 ) , and selected the best correlating locations of the genome ( c > 0 . 6 ) for further analysis . Next , at each of these selected locations the chromatin state was flipped and the correlation c′ between predicted and experimental distance-normalized counts was measured . We then defined Δc = c′ − c as a measurement of structural disruption due to sequence mutation at a given site .
|
The three-dimensional folding of DNA inside the nucleus into specific conformations is necessary for the proper functioning of cells . These structures can be measured by chromosome conformation capture methods ( Hi-C ) that report the number of times that each pair of genomic sites are found in proximal location in a cell population experiment . A number of protein complexes that bind to the DNA have been discovered to be responsible for the stabilization of such conformations . However , identifying the precise relation between the positioning of binding proteins and the resulting structures is still an open problem . Here we present a maximum-entropy method able to predict Hi-C contact probabilities from a sequence of binding factors without the need of performing any polymer simulations . We envision that this method will allow experimentalists to efficiently calculate the expected structural effect of altering the sequence of binding factors . In addition , we also show that our model is capable of solving the inverse problem , namely predicting the underlying sequence of binding factors from a set of observed contact probabilities .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"invertebrates",
"invertebrate",
"genomics",
"animals",
"animal",
"models",
"drosophila",
"melanogaster",
"model",
"organisms",
"mathematics",
"materials",
"science",
"experimental",
"organism",
"systems",
"epigenetics",
"dna",
"dna",
"structure",
"macromolecules",
"structural",
"genomics",
"chromatin",
"drosophila",
"materials",
"by",
"structure",
"research",
"and",
"analysis",
"methods",
"statistical",
"distributions",
"polymers",
"polymer",
"chemistry",
"genomics",
"chromosome",
"biology",
"gene",
"expression",
"chemistry",
"molecular",
"biology",
"animal",
"genomics",
"insects",
"probability",
"theory",
"arthropoda",
"biochemistry",
"eukaryota",
"cell",
"biology",
"nucleic",
"acids",
"genetics",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"computational",
"biology",
"genomics",
"statistics",
"organisms",
"macromolecular",
"structure",
"analysis"
] |
2018
|
A maximum-entropy model for predicting chromatin contacts
|
Ciliopathies are human disorders caused by dysfunction of primary cilia , ubiquitous organelles involved in transduction of environmental signals such as light sensation in photoreceptors . Concentration of signal detection proteins such as opsins in the ciliary membrane is achieved by RabGTPase-regulated polarized vesicle trafficking and by a selective barrier at the ciliary base , the transition zone ( TZ ) . Dysfunction of the TZ protein CC2D2A causes Joubert/Meckel syndromes in humans and loss of ciliary protein localization in animal models , including opsins in retinal photoreceptors . The link between the TZ and upstream vesicle trafficking has been little explored to date . Moreover , the role of the small GTPase Rab8 in opsin-carrier vesicle ( OCV ) trafficking has been recently questioned in a mouse model . Using correlative light and electron microscopy and live imaging in zebrafish photoreceptors , we provide the first live characterization of Rab8-mediated trafficking in photoreceptors in vivo . Our results support a possibly redundant role for both Rab8a/b paralogs in OCV trafficking , based on co-localization of Rab8 and opsins in vesicular structures , and joint movement of Rab8-tagged particles with opsin . We further investigate the role of the TZ protein Cc2d2a in Rab8-mediated trafficking using cc2d2a zebrafish mutants and identify a requirement for Cc2d2a in the latest step of OCV trafficking , namely vesicle fusion . Progressive accumulation of opsin-containing vesicles in the apical portion of photoreceptors lacking Cc2d2a is caused by disorganization of the vesicle fusion machinery at the periciliary membrane with mislocalization and loss of the t-SNAREs SNAP25 and Syntaxin3 and of the exocyst component Exoc4 . We further observe secondary defects on upstream Rab8-trafficking with cytoplasmic accumulation of Rab8 . Taken together , our results support participation of Rab8 in OCV trafficking and identify a novel role for the TZ protein Cc2d2a in fusion of incoming ciliary-directed vesicles , through organization of the vesicle fusion machinery at the periciliary membrane .
Ciliopathies are an expanding group of human disorders caused by primary cilium dysfunction and unified by a wide array of overlapping phenotypes: cystic kidneys , central nervous system malformations and retinal degeneration among others [1–3] . Primary cilia are ubiquitous organelles that consist of a mother centriole-derived basal body ( BB ) and a microtubule-based axoneme ensheathed by a specialized membrane . Primary cilia are involved in transduction of a variety of environmental signals to the cell , including morphogens and light . To serve this purpose , the ciliary membrane is enriched with specific receptors and channels required for signal detection [4] . As cilia are devoid of protein synthesis machinery [5] , ciliary membrane compartmentalization is achieved by highly controlled polarized vesicle trafficking subject to RabGTPase regulation [6] and by a selective barrier at the base of the cilium likely formed by the transition zone ( TZ ) [7] . Mutations in several genes encoding TZ proteins lead to Joubert syndrome ( JBTS; OMIM: 213300 ) , a representative ciliopathy characterized by a very specific cerebellar malformation–the molar tooth sign–and associated in 30% of cases with retinal involvement due to photoreceptor ( PR ) dysfunction [8] . Retinal involvement is common in ciliopathies since the outer segment ( OS ) of PRs , which is the site of the phototransduction cascade , is a highly specialized primary cilium [9] . OSs consist of stacks of membranous disks organized around a microtubule-based axoneme . These membranous disks contain the proteins required for phototransduction , such as the photopigment opsin [10] . Trafficking of opsins towards the OSs is thought to be controlled by the small GTPase Rab8 , as dominant negative GDP-locked Rab8 expression in frog PRs leads to opsin-carrier vesicle ( OCV ) accumulation at the base of the PR OSs [11 , 12] . Further work in frog PRs identified a complex initiated by Arf4 in the trans-Golgi network including ASAP1 , Rab11 , FIP3 , and subsequently Rabin8 and Rab8 . This complex was shown to be involved in opsin delivery to the ciliary compartment [13 , 14] . In addition , Rab11 and Rabin8 , the guanidine exchange factor for Rab8 , were shown to promote ciliary membrane biogenesis in RPE-hTERT cells [15–17] . While two RAB8 paralogs exist in humans ( RAB8A and RAB8B ) , which were both found to be involved in ciliogenesis in RPE cells [18] , it remains undefined which paralog is important in OCV trafficking in PRs . In addition , work on Rab8 knock-out mice questioned the importance of this GTPase in ciliogenesis . Indeed , double rab8a;rab8b knock-out mice revealed mislocalization of apical proteins in intestinal cells but required additional knock-down of the close Rab8-relative Rab10 to develop ciliogenesis defects in fibroblasts [19] . This work suggested that various Rabs can compensate for loss of function of each other . However , a mouse rab8a knock-out displayed no retinal phenotype despite additional expression of dominant negative forms of Rab8b , Rab11a , Rab11b and Rab10 , questioning the role of Rab8 in OCV trafficking in PRs [20] . Once OCVs have reached their target membrane , vesicle fusion is thought to be mediated by the Exocyst and SNARE ( soluble N-ethylmaleimide-sensitive factor attachment protein receptor ) proteins . SNAREs present on the vesicle surface ( v-SNAREs ) and on the target membrane ( t-SNAREs ) pair up and work as catalysts to provide the mechanical force required to mediate membrane fusion [21] . The Exocyst is a multisubunit protein complex implicated in tethering of secretory vesicles to the plasma membrane in several exocytosis processes including ciliogenesis [22 , 23] . The Exocyst localizes at the ciliary base in cell culture [24] and in the ciliary stalk of PRs in frog retina [25] . Moreover one of its components , Sec15 , has been found to interact directly with Rab8 [22] . Interestingly , the Exocyst components Sec8 , Sec10 and Exo84 have been involved in the pathogenicity of various ciliopathies , including Joubert syndrome and the related Meckel syndrome [26–30] . Joubert syndrome is a ciliopathy caused by mutations in one of at least 30 genes , many of which encode TZ proteins interacting with each other in large multi-protein complexes . Dysfunction of CC2D2A , one such TZ protein , is one of the most common causes for JBTS , accounting for ~10% of JBTS families [8] . As for other TZ proteins , its dysfunction leads to loss of ciliary protein localization in mice [31] . Similarly , we have previously reported intracellular opsin mislocalization and massive vesicle accumulation in PRs of the zebrafish cc2d2a uw38 mutant , suggesting the involvement of Cc2d2a in ciliary trafficking and supporting the role of the TZ in ciliary protein content regulation [32 , 33] . While we also observed loss of Rab8 puncta in the absence of Cc2d2a function in this model , the link between the TZ protein Cc2d2a , the small GTPase Rab8 and ciliary-directed trafficking remained unexplored . In this study , we show that Rab8-coated opsin-containing vesicles accumulate progressively at the apical portion of cc2d2a-/- PRs as a result of a vesicle fusion defect . Using correlative light and electron microscopy ( CLEM ) as well as live imaging in zebrafish PRs , we provide a detailed characterization of Rab8-mediated trafficking in this cell type . Our findings support a role for Rab8 in OCV trafficking and demonstrate redundancy between Rab8 paralogs in both rods and cones of wild-type fish . Time-lapse analysis of Rab8-tagged particles in cc2d2a-/- PRs reveals , besides partial mislocalization of Rab8 to the cytoplasm , no substantial difference in movement kinetics for remaining vesicular-associated Rab8 particles compared to wild-type PRs , suggesting that the observed OCV accumulation is not caused by a direct defect in trafficking but rather by deficient vesicle fusion . In support of this model , we observe loss of elements of the vesicle fusion/tethering machinery normally present at the apical membrane of PRs such as t-SNAREs SNAP25 and Syntaxin3 and the Exocyst component Exoc4/Sec8 consequent to Cc2d2a loss-of-function . Together , these results indicate that Cc2d2a plays a crucial role in organization of the vesicular docking sites at the periciliary membrane , allowing OCVs to fuse and deliver their cargo at the base of the OSs .
Our previous studies of the cc2d2auw38 zebrafish null mutant identified massive accumulation of vesicular structures in the apical portion of zebrafish photoreceptors ( PRs ) co-existing with misshapen outer segments ( OSs ) at 5 days post fertilization ( dpf ) [33] . To rule out the possibility that these structures are the result of a degenerative process , we first performed a TUNEL assay to assess PR cell death , which we found to be minimally increased compared to wild-type at 3 dpf only and not increased at 4 dpf ( S1A–S1B” and S1F Fig ) . Furthermore , the morphology of PR nuclei was not affected in cc2d2a-/- retinae , in contrast to what is observed in the degenerating retina of the well-studied ciliary zebrafish mutant ift88-/-/oval [34] ( S1C–S1E Fig ) . We further performed transmission electron microscopy ( TEM ) to analyze the ultrastructure of cc2d2a mutant retinae at earlier time points spanning OS formation . At 60 hpf , wild-type ( wt ) PRs have taken on their typical cell morphology including an apico-basally elongated nucleus , an inner segment containing the forming mitochondrial cluster ( Fig 1A ) and a basal body ( BB ) that has already docked to the apical membrane with extension of the connecting cilium ( Fig 1C ) . At this stage , cc2d2a-/- PRs are morphologically indistinguishable from wt PRs ( Fig 1B ) including in particular docking of the BB to the apical membrane and extension of the connecting cilium ( Fig 1D and S2 Fig ) . However , at later stages we observe that ciliary membrane stacking is impaired in cc2d2a-/- PRs . Nascent OSs are observed in a majority of wt PRs at 72 hpf ( Fig 1E and 1G ) but are mostly absent in cc2d2a-/- retinae where vesicular structures begin to accumulate instead ( Fig 1F and 1H ) . While wt PRs extend OSs of increasing length at 4 dpf ( Fig 1I and 1K ) , vesicular accumulation increases progressively over time in cc2d2a-/- PRs ( Fig 1J and 1L ) becoming massive at 5 dpf , as previously reported [33] . Therefore , we conclude that accumulation of vesicles is progressive from onset of OS formation and not the result of a degenerative process . To ensure that these accumulated membranous structures are opsin-carrier vesicles ( OCVs ) and thus were ciliary targeted , we used a recently developed correlative light and electron microscopy ( CLEM ) method [35 , 36] . This method enables us to overlay immunodetected opsin on ultrathin sections with scanning electron microscopy ( SEM ) images to determine the precise ultrastructural localization of the opsins . Structures of interest such as OSs ( Fig 2A–2A’ ) , accumulated vesicles ( Fig 2B–2B’ , arrows ) or cilia ( Fig 2C–2C’ , arrowhead ) are easily recognizable on the SEM images . Using the 4D2 antibody , which recognizes rhod- and red-green cone opsin , we found opsin signal in OSs of wild-type PRs ( Fig 2A ) . In cc2d2a-/- PRs , opsins localize to aberrant membrane stacks of dysmorphic OSs as well as to the accumulated vesicles at 5 dpf ( Fig 2B and 2C ) . Similar opsin-positive vesicle accumulation was observed already at 3 dpf with onset of aberrant membrane stacking ( S3 Fig ) . In comparison , transport of transmembrane proteins such as Cacna1fa to other cellular compartments than the OS ( synapse ) is unaffected in cc2d2a-/- PRs ( S4 Fig ) . Collectively , our data indicate that OCVs accumulate progressively in cc2d2a-/- PRs as a result of either a ciliary-specific trafficking or a vesicle fusion defect . Our previous study of the zebrafish cc2d2auw38 mutant described mislocalization of the small GTPase Rab8 in PRs lacking Cc2d2a [33] . This suggested that a defect in Rab8-mediated trafficking could underlie the observed vesicle accumulation , given the ascribed roles for Rab8 in regulation of polarized vesicle trafficking to cilia [15 , 16 , 18] and in delivery of opsin-carrier-vesicles ( OCVs ) in PRs [14] . However , recent work has questioned this role for Rab8 paralogs in OCV transport in mouse [20] . To elucidate whether Rab8 participates in opsin transport in zebrafish and to investigate the consequences of Cc2d2a loss-of-function on Rab8-mediated trafficking , we generated transgenic fish lines stably expressing tagged Rab8 in PRs . Zebrafish have 3 Rab8 paralogs: Rab8a , Rab8b and Rab8b-like . Based on synteny and sequence similarity ( 93 . 2% identity at the amino acid level ) , zebrafish Rab8a ( NM_001089562 ) appears to be the true ortholog of human RAB8A ( NM_005370 ) . The chromosomal region around the human RAB8B gene ( NM_016530 , chromosome 15 ) shows conserved synteny with two zebrafish genomic regions , on chromosome 7 ( Rab8b , NM_001099259 ) and on chromosome 25 ( Rab8b-like , XM_021470743 . 1 ) , suggesting that these are zebrafish paralogs resulting from the teleost-specific genome duplication ( S5 Fig ) . Consequently , we now call Rab8b-like Rab8ba and Rab8b Rab8bb ( S6 Fig ) . In any case , all three zebrafish Rab8 paralogs share very high sequence similarity with each other and with the human RAB8A and B paralogs ( S7 Fig ) . To characterize Rab8-directed trafficking in zebrafish PRs and determine whether different Rab8 paralogs could play a role in OCV trafficking , we chose to use the zebrafish Rab8a and Rab8ba sequences . We generated lines expressing N-terminal mCherry-tagged Rab8a in rods ( tg ( rhod:mCherry-rab8a ) ) and cones ( tg ( tacp:mCherry-rab8a ) ) as well as Rab8ba in rods ( tg ( rhod:mCherry-rab8ba ) ) . To determine if the different Rab8 paralogs co-localized with opsin-carrier vesicles , we first performed immunofluorescence using the 4D2 antibody ( recognizing rhodopsin and red-green cone opsin ) on retinal sections of transgenic animals expressing either mCherry-Rab8a or mCherry-Rab8ba . In both cases , we observed co-localization of mCherry-Rab8 with endogenous opsins in both cones and rods ( Fig 3A–3I ) , whereby ~ 50% of opsin puncta co-localized with Rab8a/Rab8ba ( 48 . 8%±27 . 8% for Rab8a and 50 . 2%±21 . 9% for Rab8ba , n > 50 PRs from 11 animals per condition; the remaining opsin-positive puncta could be co-localizing with either endogenous Rab8 paralog or with other Rabs ) . We further transiently co-expressed heat-shock inducible GFP-tagged rhodopsin ( hsp70:rhod-GFP ) [37] in Rab8 transgenic fish to obtain time-lapse videos in live 5 dpf larvae and observed co-movement of Rhodopsin-GFP particles with mCherry-Rab8 particles in vivo ( S1 video ) . Finally , using CLEM , we found that opsin signal is associated with mCherry-Rab8a in vesicular-like structures ( Fig 4 ) . Together , these results confirm that the small GTPase Rab8 is present at the surface of OCVs in both cones and rods and suggest possible redundancy between Rab8 paralogs in the transport of opsins . Our previous work suggested Cc2d2a to be required for the punctate localization of Rab8 , as transiently expressed mCherry-Rab8 was diffusely localized in a majority of cc2d2a-/- cone PRs , while only a subset of mutant PRs maintained punctate Rab8 expression [33] . We now confirmed these observations in transgenic lines stably expressing mCherry-Rab8 . In wild-type PRs , mCherry-Rab8 signal was present as puncta , predominantly in the inner segment ( IS ) , but also to a lesser extent more basally in the PRs , with no clear signal inside the OSs . Immunostaining using an anti-Rab8 antibody confirmed that these mCherry-puncta contain Rab8 ( S8A–S8F Fig ) . In non-transgenic retinae , the signal provided by the anti-Rab8 antibody was less strong than observed with the transgene but showed similar puncta in the IS which were also positive for opsins , indicating that endogenous Rab8 has a similar subcellular localization pattern as observed with the transgene ( S8G–S8I Fig ) . To further evaluate the localization of Rab8 at the ultrastructural level , we performed CLEM on mCherry-Rab8 expressing eyes , which required enhancement of the mCherry signal with an anti-mCherry antibody . In wt PRs , we observed mCherry-positive puncta that localized to membrane-delimited vesicular structures ( Fig 5A–5A’ , arrowheads ) . In cc2d2a-/- PRs , mCherry signal was localized to accumulated vesicles ( Fig 5B–5B’ bracket ) . In addition , we observed a weaker diffuse signal in the cytoplasm of mutant PRs , not delimited by membrane ( Fig 5C–5C’ , arrows ) . Of note , mCherry signal in OSs ( Fig 5A ) became apparent only after antibody-enhancement of the signal , and was absent from non-enhanced transgenic retinae ( S8A and S8D Fig ) or from non-transgenic retinae stained with anti-Rab8 antibody ( S8H–S8I Fig ) , suggesting either non-specific OS antibody signal or very low-grade OS expression of transgenic mCherry-Rab8 due to overexpression and visible only after enhancement . Collectively , these data indicate that Rab8 ( both endogenous and mCherry-tagged ) localizes to puncta representing vesicular structures and that loss of Cc2d2a function leads to partial mislocalization of Rab8 to the cytoplasmic compartment , while Rab8-coated vesicles also remain present in cc2d2a mutant PRs . Taking advantage of the stable transgenic lines expressing mCherry-tagged Rab8 paralogs in rods and cones , we went on to characterize Rab8-trafficking in PRs in vivo in a whole tissue context . For that purpose , we obtained time-lapse videos of live 5 dpf old transgenic larvae stably co-expressing mCherry-Rab8 with GFP-hCentrin , the latter serving as an immobile cellular reference to label the basal body ( BB ) ( S2 video ) . We found Rab8-tagged puncta to display a complex apico-basal shuffling movement , mostly in the inner segment but also spanning the entire length of the PR from the BB to the synapse , with a subset of puncta approaching the BB ( Fig 6A–6D ) . To ensure that this movement was not an artifact driven by the overexpression of Rab8 , we transiently overexpressed mCherry-Rab3aa and analyzed its behavior via live imaging . mCherry-Rab3aa strongly localizes at the PR synapse as predicted [38] , where it shows predominantly localized movement in the synaptic region over the 10 minute duration of the time-lapse ( S3 video and Fig 6E and 6F’ ) , indicating that overexpressed Rabs exhibit the predicted endogenous behavior . Similarly , previous work in zebrafish using overexpressed fluorescently-tagged Rab5 , Rab7 and Rab11 showed distinct localization patterns for the different Rabs consistent with the predicted endosomal compartments for each Rab and a live lipophilic dye uptake assay supported the biological relevance of such overexpression assays [39] . Furthermore , in our assay , despite some increase in vesiculo-tubular structures in a subset of PRs ( S9 Fig ) , overexpression of Rab8 did not substantially affect retinal ultrastructure in the transgenic animals at 5 dpf . We thus pursued to track the movement of Rab8a or Rab8ba-tagged particles over 10 minutes in wild-type cones and rods ( S4–S6 videos ) using the tracking software Ilastik ( Fig 6A’–6B’ ) . mCherry-tagged Rab8 particles demonstrated a complex movement pattern with “shuffling” along the apico-basal axis predominantly in the inner segment of photoreceptors ( S4–S6 videos ) . Interestingly , we observed no significant difference in displacement ( defined as the distance between the first and last coordinates of the recording: on average 890±28 nm; Fig 7A ) , maximum speed ( defined as the largest distance traveled by a particle between two consecutive frames taken at 1 second intervals , on average 1016±98 nm/s; Fig 7B ) or trajectory ( total distance traveled during the entire duration of the recording , on average 21±12 μm; Fig 7D ) between Rab8a and Rab8ba paralogs or between rods and cones . Tracked particles had a cross-sectional area of around 0 . 3 μm on average ( calculated based on the number of constituting pixels , Fig 7E ) , which is consistent with the size of polylobulated vesicular structures seen on CLEM ( arrows in Fig 4A’ ) . Altogether , our data suggest that Rab8a and Rab8ba have a similar behavior in PRs , which could indicate functional redundancy , and that Rab8a-controlled trafficking occurs similarly in rods and cones . To determine whether the vesicle accumulation observed in cc2d2a-/- photoreceptors ( PRs ) is secondary to a defect in Rab8-mediated trafficking , we compared kinetics of mCherry-Rab8a and -Rab8ba particles between wt and cc2d2a-/- rods . Indeed , while mCherry-Rab8 is in part mislocalized diffusely to the cytoplasm in mutant PRs , some mCherry-Rab8 particles are retained in mutant PRs . Using the same video duration and acquisition rate paradigm described above , we observed no major differences in movement kinetics in mutant PRs compared to wt including displacement , speed or particle size ( S7 and S8 videos; quantification in Fig 7A–7F ) . The only variable differing significantly in mutant PRs was the directionality of Rab8ba puncta , which was shifted towards an increased lateral movement in cc2d2a-/- rods ( Fig 7C ) . Given that cell-shape is grossly maintained in cc2d2a-/- PRs , this difference could be explained by increased difficulty of particle movement through accumulated vesicles . However , when quantifying the relative number of particles reaching the periciliary membrane region and approaching the basal body ( BB ) ( Fig 7F ) , we did not observe a significant difference for either of the Rab8 paralogs in mutant vs wt PRs . The absolute number of Rab8 particles coming in proximity with the BB varied widely between different PRs in both wt and mutants , ranging from 0 to 10 in wt and 0 to 28 in mutants . We conclude that loss of Cc2d2a function does not directly affect Rab8-trafficking . Because the massive accumulation of vesicles in cc2d2a-/- PRs occurs only apically and Rab8-trafficking seems unaffected , we hypothesized that the trafficking defect observed in these mutants must happen at late steps of the trafficking process , namely vesicle fusion . Therefore , we focused on the fusion machinery components at the ciliary base . It was previously shown that the t-SNAREs required for the delivery of OCVs in frog and mammalian retina are SNAP25 and Syntaxin3 [25 , 40] . We performed immunostaining to determine the localization of these proteins in zebrafish PRs . In wild-type ( wt ) animals and consistent with previous reports , we found SNAP25 to be localized along the plasma membrane ( Fig 8A–8A’ ) , including at the synapse and at the apical membrane between the inner segment ( IS ) and outer segment ( arrowheads in Fig 8A–8A’ ) . In cc2d2a-/- PRs , however , SNAP25 was mislocalized apically in a membranous compartment highlighted by BODIPY , while its synaptic localization was unaffected , suggesting again a ciliary-specific effect ( Fig 8B–8B’ ) . Moreoever , SNAP25 mislocalization appears to be specific to loss of Cc2d2a , as it was not present in PRs of the ciliary mutant oval/ift88 , defective for intraflagellar transport ( IFT ) and unable to form OSs , in which SNAP25 signal remains clearly present at the apical membrane just apical to the mitochondrial cluster ( Fig 8C–8C’ ) . To determine the precise subcellular localization of SNAP25 in wild-type and mutant PRs , we performed CLEM . In the apical portion of PRs , SNAP25 localized to the apical membrane of the IS , including the periciliary membrane ( Fig 9C–9C’ , arrow ) at the base of the anti-acetylated tubulin-highlighted primary cilium ( Fig 9C–9C’ , arrowhead ) , and to the membrane of calycal processes ( Fig 9A , empty arrowhead ) . In cc2d2a-/- PRs , however , SNAP25 was mislocalized to accumulated vesicles and to dysmorphic OSs by CLEM ( Fig 9D–9D’ and 9E–9E’ respectively ) . Importantly , this SNAP25 mislocalization was observed as early as 3 dpf ( S10 Fig ) , paralleling the accumulation of vesicles in the apical region of PRs ( S11 Fig ) . Taken together , our data suggest that Cc2d2a is required for the correct localization of SNAP25 at the periciliary membrane from the onset of OS formation . We next turned to the second known t-SNARE involved in fusion of OCVs in PRs , Syntaxin3 ( Stx3 ) . We found Syntaxin3 localization to mirror SNAP25 localization at the plasma membrane , including the periciliary membrane , in wild-type PRs at 4 dpf and 6 dpf ( Fig 10A–10A’ and 10C–10C’ , arrowheads ) . In 4 dpf cc2d2a-/- PRs we observed minor mislocalization of Syntaxin3 ( Fig 10B–10B’ ) , which was much less prominent than for SNAP25 and decreased further at 6 dpf ( Fig 10D–10D’ ) . However , at both time points we consistently observed a strong decrease in fluorescence intensity . We confirmed by western blots that both Syntaxin3 and SNAP25 levels were decreased in whole 6 dpf larval eyes ( Fig 10E and 10F and S12 Fig ) . We next focused our analysis on components of the Exocyst , which is a tethering complex downstream of Rab-mediated targeting and upstream of SNARE-mediated fusion . Specifically , we measured levels of the Sec8 homolog Exoc4 , which is a predicted interactor of Rab8 , and found these to be decreased in cc2d2a mutant eyes ( Fig 10E and 10F and S12 Fig ) . Collectively , our data suggest that loss of Cc2d2a leads to mislocalization and/or depletion of various components required for fusion of OCVs at the periciliary membrane .
Primary cilia are devoted to the transduction of a plethora of signals crucial for embryonic development , adult tissue homeostasis and interpretation of environmental stimuli , such as light [41] . Therefore , regulation of ciliary protein content , in particular of transmembrane receptors and channels , is indispensable for primary cilium function [4] . While the ciliary transition zone ( TZ ) is thought to act as a gatekeeper in this process [7] , its link to upstream protein sorting mechanisms and polarized vesicular trafficking has not been elucidated so far . Our study using CC2D2A as a representative TZ protein indicates that CC2D2A is required for the last steps of trafficking , namely vesicle fusion . Furthermore , our work provides novel evidence in support of a role for Rab8 in opsin-carrier vesicle ( OCVs ) trafficking in photoreceptors and suggests that this trafficking is only indirectly affected by dysfunction of the TZ . The vesicle fusion defects observed in cc2d2a-/- PRs are secondary to loss of the vesicle tethering and fusion machinery components SNAP25 , Syntaxin3 and Exoc4 from the periciliary region . These vesicle fusion defects are ciliary-specific , despite the broader localization of the analyzed SNAREs along the PR cell membrane , since delivery of non-ciliary proteins to other highly organized membrane compartments such as the synapse remains unaffected . Furthermore , SNAP25 mislocalization as seen in cc2d2a-/- PRs is not observed in the non-TZ ciliary mutant ift88-/- , which also lacks vesicle accumulation [34] , strongly suggesting that the fusion machinery defects are specific to loss of Cc2d2a/TZ function . We have previously shown CC2D2A to interact with NINL [42] , a centrosomal protein that also interacts with dynein-dynactin motor proteins and MICAL3 , which is a proposed Rab8 effector [43] . Both Rab8 and MICAL3 have catalytic properties that can modify cortical actin cytoskeleton [43–45] , thus facilitating vesicle docking [46] . In addition , Rab8 recruits the Exocyst to initiate vesicle fusion [23 , 47] . Taken together with the data presented in this work , we propose a model whereby Cc2d2a at the TZ may provide a docking point for incoming OCVs through its interaction with NINL and control the localization of the t-SNAREs required for fusion , thereby bringing all components required for vesicle fusion in proximity with each other at the periciliary region ( Fig 11 ) . The mechanisms underlying SNAP25 mislocalization and vesicle fusion machinery depletion ( including SNAP25 , Syntaxin3 and Exoc4 ) in cc2d2a mutant PRs remain to be elucidated . Previous work has shown that SNAREs can be downregulated if their interactors are missing or not appropriately localized [48] . Given the observed mislocalization of SNAP25 , the cell might try to prevent aberrant fusion of OCVs to the wrong membranes by downregulating Syntaxin3 , depleting the Exocyst and reducing the amount of Rab8 engaged on OCVs . SNARE mislocalization was described in the ciliopathy bbs17-/- mouse model , where Syntaxin3 and Stx1bp are mislocalized in the OSs of mutant PRs [49] . The authors propose that Bbs17 is required for retrograde transport of proteins out of the ciliary compartment . In our experiments , accumulation of SNAP25-positive vesicles occurred in cc2d2a-/- PRs even in the absence of formation of OSs , speaking against a role for Cc2d2a in active retrograde transport . One possible explanation for SNAP25 mislocalization could be disrupted phospholipid composition of the periciliary membrane . Indeed , recent work determined that the JBTS-protein INPP5E indirectly controls protein content of the ciliary membrane by regulating ciliary phosphoinositides ( PIPs ) [50] . SNAP25 does not possess a transmembrane domain but is inserted in the target membrane through palmitoylation [51]; thus , its localization might be influenced by the lipid composition of the periciliary membrane . Other proteins regulating intracellular trafficking such as FIP3 or Exocyst subunits contain PIP-binding domains that can be highly specific for certain phospholipids [46 , 52] . Moreover , SNAREs are arranged around cholesterol and PIP2-rich membrane microdomains [53–55] . OCV delivery was shown to depend on PIP2 recognition by FIP3 and to be enhanced by providing docosahexaenoic acid to the PRs [14 , 25] . Interestingly , MICAL3 recognizes membrane-associated proteins bound to specific PIPs as the vesicle “landing site” [43 , 56] . Given that the ciliary localization of INPP5E depends on the TZ [50 , 57] , loss of the TZ protein Cc2d2a could impact the balance of phosphoinositides not only in the ciliary , but also in the periciliary membrane , which could in turn lead to mislocalization of SNAP25 . A large body of work has studied OCV trafficking and opsin delivery to the OSs . OCVs are thought to travel from the Golgi to the OS along microtubules using a dynein motor [10 , 58] . Multiple players thought to be involved in sorting and targeting of OCVs at various steps along this path include Arf4 , ASAP1 , Rab11 , FIP3 and ultimately Rabin8 , which binds and activates Rab8 , priming the OCV for its fusion [14] . However , a recent publication has questioned this model , suggesting that Rab8a and Rab8b paralogs are dispensable for rhodopsin transport: in that study , double rab8a;rab11a knockout mice and rab8a knockout mice with additional expression of dominant negative forms of Rab8b , Rab11a , Rab11b and of the close Rab8-relative Rab10 lack a retinal phenotype [20] . In contrast , our study relying on live imaging of tagged Rab8 and CLEM in photoreceptors supports a role for Rab8 paralogs in rhodopsin transport . We demonstrate presence of Rab8 on vesicular structures containing opsins . Moreover , the movement pattern of Rab8 particles shows apico-basal directionality towards the BB ( as expected with movement along microtubules ) at a speed consistent with cytoplasmic dynein-driven transport [59] and we show Rab8 to move together with opsin in whole live retina . In addition , our work investigated for the first time both rod and cone PRs with respect to Rab8-trafficking , and found no biologically relevant difference between these PR types . Our findings also support redundancy between Rab8 paralogs , which might extend to other Rabs as well . While one of the Rab8 paralogs might dominate in the physiologic situation , the other Rab8 paralogs , and potentially additional Rabs such as the closest Rab8-relatives Rab10 and Rab13 , may be able to take over in case of loss-of-function of Rab8 . Rab13 in particular was not investigated in the work by Ying and collaborators [20] . While Rabs are thought to provide membrane specificity , given the large number of Rab genes ( close to 70 in humans ) and their high degree of sequence conservation , redundancy may be operating in case of disruptions . In this framework , where Rabs act redundantly in trafficking , and SNAREs are distributed along multiple membrane domains , the function of the TZ would be to bring all the components together to the correct place to engage fusion . Association of mutations in Exocyst components with ciliopathies , in particular JBTS [27 , 30] , suggest that the mechanisms uncovered in this work to explain the ciliary dysfunction in cc2d2a-/- PRs might represent a more general mechanism underlying JBTS . Since about half of JBTS proteins are TZ proteins , abnormal ciliary protein composition through a similar trafficking/fusion defect could represent a common pathogenic mechanism underlying JBTS . Previous evidence for a role of TZ proteins in general , and Cc2d2a in particular , in controlling the protein composition of the cilium was provided by a Cc2d2a-knockout mouse model , in which localization of ciliary proteins adenylyl cyclase III , Arl13b , Smoothened and PKD1 was lost from mutant cilia [31] . In a second Cc2d2a mouse knockout , Rab8 signal was lost from the ciliary base , transport vesicles accumulated and ciliogenesis was impaired in cc2d2a-/- mice [60] . While the majority of these findings in mice are consistent with our observations in the zebrafish cc2d2a mutant model , we do not observe ciliogenesis defects in our cc2d2a-/- fish , despite them being true null mutants [33] . Interestingly , only a subset of tissues showed ciliogenesis defects in the first Cc2d2a knockout mouse , suggesting the requirement for Cc2d2a in ciliogenesis to be tissue-specific . Given that vertebrate animal models display very early embryonic lethality in complete absence of cilia , it is unlikely that major ciliogenesis defects would lead to the relatively milder phenotypes of JBTS in humans . Taken together , our findings suggest that dysfunction of the TZ leads to disorganization of the vesicle fusion machinery , causing accumulation of ciliary-bound vesicles , secondary trafficking defects and aberrant protein content of the ciliary membrane . This is consistent with the ascribed gate-keeper function of the TZ , but suggests a more complex and active role , where the TZ influences both sides of the barrier . Further work will be required to elucidate the precise mechanisms through which individual TZ proteins affect the periciliary membrane , in our quest to uncovering the pathogenesis underlying JBTS .
All animal protocols were in compliance with internationally recognized and with Swiss legal ethical guidelines for the use of fish in biomedical research and experiments were approved by the local authorities ( Veterinäramt Zürich Tierhaltungsnummer 150 ) . Zebrafish ( Danio rerio ) were maintained as described [61] . Embryos were raised at 28°C in embryo medium and staged according to development in days post fertilization ( dpf ) [62] . 0 . 003% PTU ( 1-phenyl-2-thiourea ) in embryo medium was used to inhibit melanin synthesis during larval development and facilitate fluorescent microscopy . The cc2d2auw38 , the ift88tz288 and the casper mutants were previously described [32–34 , 63] . Stable zebrafish transgenic lines used in this study included Tg ( rhod:mCherry-rab8ba ) [33] , Tg ( rhod:mCherry-rab8a ) , Tg ( tacp:mCherry-rab8a ) , Tg ( rhod:hCentrin-GFP ) , Tg ( tacp:hCentrin-GFP ) ; the latter four lines were generated in this study ( see next paragraph ) . Constructs injected transiently included rhod:mCherry-rab3aa ( generated in this study ) and hsp70:rhod-GFP ( a gift from J . Malicki [37] ) . RT-PCR was performed using cDNA obtained from whole larvae at 5 dpf to generate a Gateway ( Invitrogen ) p3’ entry clone including the full length coding sequence for Rab3aa ( NM_001003419 . 1 ) . The following primers were used ( attB site sequences are written in lowercase ) : Rab3aa attB2r: 5’-ggggacagctttcttgtacaaagtggcgATGGCTTCAGCGAATGATGC-3’ Rab3aa attB3: 5’-ggggacaactttgtataataaagttgtTTAGCAAGCGCAGTCCTTGT-3’ Gateway ( Invitrogen ) recombination was performed using the Tol2 system [64] . p5’ entry clones included the rod-specific rhodopsin promoter ( gift from the Link lab ) or the cone-specific alpha-transducin promoter ( tacp , [65] ) . mCherry or GFP middle-entry clones and p3’ entry clones containing the human Centrin sequence ( gift from the Link lab ) , the Rab3aa and the Rab8a zebrafish sequence ( NM_001089562; gift from the Beales lab ) were used to generate N-terminal fusions of Rab proteins or Centrin . The resulting constructs were co-injected with Tol2 transposase mRNA as previously described [66] into 1-cell stage embryos . Injected fish were either imaged ( transients ) or raised and further generations outcrossed to casper zebrafish for at least two generations . As gene predictions within GenBank are produced by automated processes which may contain errors , Rab cDNA sequences used in this study were manually annotated . Sequences were identified and annotated using combined information from expressed sequence tags and genome databases ( GeneBank , http://www . ncbi . nlm . nih . gov; Ensembl , http://www . ensembl . org/index . html ) . Human and mouse sequences were used as initial query ( for more details on sequence annotation see [67] ) . The phylogenetic analysis was performed on the Phylogeny . fr platform ( http://www . phylogeny . fr/ ) comprising the following steps [68] . Sequences were aligned using MUSCLE ( v3 . 7 ) [69] configured for highest accuracy ( MUSCLE with default settings ) . Length of sequences varied between 214 and 700 amino acids . After alignment , ambiguous regions ( i . e . containing gaps and/or poorly aligned ) were removed using Gblocks ( v0 . 91b ) [70] . The following parameters were implemented: The minimum length of a block after gap cleaning was set to 5; positions with a gap in less than 50% of the sequences were selected in the final alignment if they were within an appropriate block; all segments with contiguous non-conserved positions bigger than 8 were rejected; minimum number of sequences for a flank position were 55% . After curation , 203 amino acids were chosen for further analysis . The phylogenetic tree was reconstructed using the maximum likelihood method implemented in the PhyML program ( v3 . 0 aLRT ) [71] . The default substitution model was selected assuming an estimated proportion of invariant sites ( of 0 . 000 ) and 4 gamma-distributed rate categories to account for rate heterogeneity across sites . The gamma shape parameter was estimated directly from the data ( gamma = 0 . 728 ) . Reliability for internal branch was assessed using the aLRT test [72] . Graphical representation and edition of the phylogenetic tree were performed with TreeDyn ( v198 . 3 ) and the svg file imported into CorelDraw ( version x5; Corel Corporation Ottawa , Canada ) for final editing . Synteny analysis was done using the synteny database ( http://syntenydb . uoregon . edu/synteny_db/ ) [73] . Parameters were adjusted to a sliding window size of 50 , and several genes in the vicinity of rab8b were used for additional syntenic comparison . The final graph was assembled using a combination of the obtained synteny files and adjusted in size and color using CorelDraw . Several synteny hits in the output files were used as initial queries for a tBLASTx search against the zebrafish or human database ( ncbi nr/nt database ) to verify orthology . Zebrafish larvae were fixed overnight at 4°C in a freshly prepared mixture of 2 . 5% glutaraldehyde and 2% formaldehyde ( FA ) in 0 . 1M sodium cacodylate buffer ( pH 7 . 4 ) . After rinsing in buffer , specimens were washed in 1% osmiumtetroxide and 1% potassiumferrocyanide in 0 . 1 M sodiumcacodylate buffer ( pH 7 . 4 ) , during 2 h at room temperature . After rinsing , tissues were dehydrated through a graded series of ethanol and embedded in epon . Ultrathin sections ( 70nm ) comprising zebrafish eyes were collected on formvar coated grids , subsequently stained with 2% uranyl acetate and Reynold’s lead citrate , and examined with a transmission electron microscope Philips CM-100 . A detailed protocol is available at JoVE [35] . Briefly , 5 dpf old larvae were euthanised in Tricaine ( ethyl 3-aminobenzoate methanesulfonate , Sigma-Aldrich ) and their sectioned heads fixed in 4% formaldehyde/0 . 025% glutaraldehyde in cacodylate buffer overnight at 4°C . Eyes were dissected out in fixative , washed in PBS , embedded in 12% gelatin in 0 . 1 M PBS at 40°C , cooled down and immersed and stored in 2 . 3 M sucrose at 4°C . Prepared samples were frozen in liquid nitrogen and sectioned with a cryo-ultramicrotome ( Ultracut EM FC6 , Leica Microsystems ) . 100 nm ultrathin sections were transferred to a 7 × 7 mm silicon wafer ( Si-Mat Silicon Materials ) and stored at 4°C . Sections were stained with mouse anti-rhodopsin ( 4D2 , gift from R . Molday , University of British Columbia ) 1:100 , rabbit anti-SNAP25 1:250 ( StressGen Biotechnologies , VAP-SV002 ) , chicken anti-mCherry 1:50 ( Abcam ) , mouse anti-acetylated tubulin 1:100 ( Sigma Aldrich , clone 6-11B-1 ) and counterstained when necessary with BODIPY TR methyl ester ( Invitrogen ) 1:100 and DAPI ( 4′ , 6-diamidino-2-phenylindole dihydrochloride ) 1:100 . Alexa Fluor 488-conjugated secondary antibodies ( Life Technologies ) were used at 1:200 . After confocal laser scanning microscopy , samples were postfixed with 0 . 1% glutaraldehyde in PBS , covered with a thin layer of methylcellulose and coated with 10 nm Platinum/Carbon by rotary shadowing at an angle of 8 degrees . SEM images were taken using a Zeiss Supra 50 VP and a Zeiss Auriga 40 SEM . Alignment of light and electron microscopy images was done with the open-source platform Fiji , based on manually inserted landmarks from the nuclear DAPI signal and using the TrakEM2 plugin . At least 3 animals per condition were used for CLEM . Transiently injected Tg ( rhod:mCherry-rab3aa ) and Tg ( hsp70:rhod-GFP ) and stable Tg ( rhod:mCherry-rab8a;rhod:GFP-hCentrin ) , Tg ( rhod:mCherry-rab8ba;rhod:GFP-hCentrin ) and Tg ( tacp:mCherry-rab8a ) 5 dpf zebrafish larvae were anesthetized in Tricaine , embedded in 1 . 8% low melting agarose ( Lonza ) , mounted on glass-bottom Petri dishes and covered with Tricaine-supplemented embryo medium . Time-lapse videos from a single focal plane were obtained with a spinning disk microscope using a 60x 1 . 4 NA objective ( Visitron , CSU-W1 ) . The acquisition lasted for 10 min at a rate of 1 frame per second including both channels ( excitation lasers at 488 nm and 561 nm ) . 13 photoreceptors from 4–7 animals per condition were analyzed . All videos shown are displayed with a 15-fold acceleration ( 15 frames/second ) and edited using Fiji and Adobe Premiere Pro CC . Time-lapse live imaging videos were registered using the MultiStackReg plugin in Fiji . To avoid sampling error carryover , only three to four photoreceptors ( PRs ) were selected per animal for analysis . The analyzed PRs were randomly picked ( random . org ) from the pool of eligible PRs in the videos . The chosen PRs were first isolated using Fiji followed by signal segmentation and automated tracking using the respective pipelines in the open-source software Ilastik ( versions 1 . 1 . 5 , 1 . 1 . 7 and 1 . 2 . 0 [74] ) . Segmentation cues were provided every 5 to 10 frames . Segmented videos were processed for tracking using sigma values of 0 . 1 for Gaussian blurring and a threshold between 0 . 29 and 0 . 31 . Potential objects were only considered when they had a minimum size of 2 pixels ( 1 pixel = 210 nm ) . Optimized tracking parameters were: max . number of objects per merger = 1 , division weight = 10 , transition weight = 3 , appearance cost = 9 , disappearance cost = 47 . Objects were not considered divisible . The selected software outputs were the raw coordinates of the particle barycenter in each frame ( used to calculate the maximum speed and the displacement ( measured as the distance between the first and the last set of coordinates of a particle ) ) . Directionality was expressed as the quotient of the maximum distance spanned in the lateral X axis over the maximum distance spanned in the apico-basal Y axis ( after correcting the coordinates in the reference Cartesian system using Euler angle rotations: xnew = xoriginal * cosθ + yoriginal * sinθ; and ynew = yoriginal * cosθ—xoriginal * sinθ ) . Thus , results <1 represent apico-basally directed movement . Particle size was calculated based on the number of consituting pixels and the surface of individual pixels ( pixel size 210nm x 210nm ) and expressed as the cross-sectional surface area ( surface of each pixel = 0 . 044 μm2 , 10 pixels = 0 . 44 μm2 ) . A particle approaching the BB was considered to reach the periciliary membrane ( determined as a “contact” ) when the barycenter coordinates of the particle were within a 3 pixel radius ( 630nm ) from the barycenter of the BB . Zebrafish larvae were fixed in 4% PFA at room temperature ( RT ) for 30 min , embedded in Neg50 ( Richard-Allan Scientific ) and cryosectioned following standard protocols . Non-specific binding was blocked using PBDT ( PBS , 1% DMSO , 0 . 5% Triton X-100 , 2mg/ml BSA ) with 10% goat serum for 30 minutes at RT before incubation with primary antibodies overnight at 4°C . Primary antibodies were rabbit anti-Syntaxin3 ( 1:400 , Alomone labs , ANR-005 ) , rabbit anti-SNAP25 ( 1:1000 , StressGen Biotechnologies , VAP-SV002 ) , mouse monoclonal anti-Rab8a ( 1:100 , Novus Biologicals , clone 3G1 ) , rabbit anti-Cacna1fa ( 1:5000 , a gift from Michael Taylor , University of Wisconsin [75] ) , rabbit anti-blue opsin ( 1:250 , gift from David Hyde ) and rabbit anti green opsin ( 1:400 , gift from David Hyde , University of Notre Dame ) . Secondary antibodies were Alexa Fluor-conjugated goat anti-rabbit or goat anti-mouse IgG ( 1:400 , Life Technologies ) . BODIPY TR methyl ester ( 1:300 , Invitrogen ) or Vybrant DiO ( 1:200 , ThermoFischer ) was applied for 20 min and nuclei were counterstained with DAPI . Confocal laser scanning microscopy was performed on a Leica HCS LSI or a Leica SP5 microscope . All immunofluorescence experiments were performed at least in duplicate with at least 10 animals per condition . 6 dpf zebrafish larvae were anesthetized in Tricaine . Whole larval eyes were dissected and collected in PBS and transferred to urea buffer ( 65 mM Tris HCl pH 6 . 75 , 8 M urea , 20% glycerine , 5% SDS , 5% β-mercaptoethanol ) containing protease inhibitors ( cOmplete EDTA-free Protease Inhibitor Cocktail , Roche ) . Samples were subsequently lysed using a Sonopuls HD 2070 sonicator , and separated on Mini-PROTEAN TGX 4–15% precast polyacrylamide gels ( Bio-Rad ) . Proteins were transferred to a polyvinylidene difluoride ( PVDF ) membrane ( Invitrogen ) . Nonspecific antibody binding was inhibited by incubation in PBST ( PBS , 0 . 05% Tween-20 ) supplemented with 3% skimmed milk powder . Membranes were probed using antibodies against Syntaxin3 ( rabbit , 1:2000 , Alomone labs , ANR-005 ) , rSec8 ( mouse , 1:1000 , Enzo , clone 14G1 ) , SNAP25 ( rabbit , 1:5000 , StressGen Biotechnologies , VAP-SV002 ) , Rab8 ( mouse , 1:500 , Novus Biologicals , clone 3G1 ) and β-actin ( mouse , 1:1000 , Sigma , A1978 ) . HRP-conjugated goat anti-mouse ( 1:3000 , Merck ) or anti-rabbit ( 1:5000 , Merck ) were used to detect proteins of interest and subsequently visualized by chemiluminescence using luminol/peroxide substrate ( SuperSignal West Dura Extended Duration Substrate , Life Technologies ) and an ImageQuant LAS 4000 imager . All statistical tests were run using GraphPad Prism . Student’s t-tests were used for pairwise comparisons between wild-type ( wt ) and cc2d2a-/- to analyze tracking and western blot results . Western blot results were first normalized to the wt levels . Because parameters relative to periciliary membrane contact do not have a normal distribution , we used a Mann-Whitney U-test instead of a t-test . ANOVA was used to compare kinetic parameters between wt Rab8a cones , wt Rab8a rods and wt Rab8ba rods .
|
Ciliopathies are human disorders caused by dysfunction of primary cilia , ubiquitous organelles involved in transduction of environmental signals to the cells . Concentration and regulation of signal detection proteins in the ciliary membrane is therefore tightly regulated through polarized vesicle trafficking and through a selective barrier at the ciliary base called the transition zone ( TZ ) . Dysfunction of TZ proteins leads to human ciliopathies and to aberrant localization of ciliary proteins in animal models . In this work , we use zebrafish retinal photoreceptors as a model to explore the relationship between the TZ and upstream vesicle trafficking . Relying on modern technologies such as correlative light and electron microscopy and live imaging of fluorescently-tagged proteins , we identify a role for the TZ protein CC2D2A in organizing the components required for vesicle fusion at the periciliary membrane . We also characterize the movement dynamics of vesicles carrying light-detection proteins ( opsins ) towards the ciliary compartment of photoreceptors in vivo and provide novel data in support of the recently questioned involvement of the small GTPase Rab8 in opsin-carrier vesicle trafficking in photoreceptors .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"cell",
"physiology",
"fish",
"medicine",
"and",
"health",
"sciences",
"vesicles",
"ocular",
"anatomy",
"vertebrates",
"social",
"sciences",
"neuroscience",
"animals",
"animal",
"models",
"osteichthyes",
"membrane",
"fusion",
"model",
"organisms",
"experimental",
"organism",
"systems",
"cellular",
"structures",
"and",
"organelles",
"research",
"and",
"analysis",
"methods",
"animal",
"cells",
"sensory",
"receptors",
"cell",
"membranes",
"membrane",
"trafficking",
"zebrafish",
"signal",
"transduction",
"cellular",
"neuroscience",
"psychology",
"eukaryota",
"retina",
"cell",
"biology",
"anatomy",
"cilia",
"neurons",
"photoreceptors",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"afferent",
"neurons",
"sensory",
"perception",
"ocular",
"system",
"organisms"
] |
2017
|
Loss-of-function of the ciliopathy protein Cc2d2a disorganizes the vesicle fusion machinery at the periciliary membrane and indirectly affects Rab8-trafficking in zebrafish photoreceptors
|
Simultaneous spike-counts of neural populations are typically modeled by a Gaussian distribution . On short time scales , however , this distribution is too restrictive to describe and analyze multivariate distributions of discrete spike-counts . We present an alternative that is based on copulas and can account for arbitrary marginal distributions , including Poisson and negative binomial distributions as well as second and higher-order interactions . We describe maximum likelihood-based procedures for fitting copula-based models to spike-count data , and we derive a so-called flashlight transformation which makes it possible to move the tail dependence of an arbitrary copula into an arbitrary orthant of the multivariate probability distribution . Mixtures of copulas that combine different dependence structures and thereby model different driving processes simultaneously are also introduced . First , we apply copula-based models to populations of integrate-and-fire neurons receiving partially correlated input and show that the best fitting copulas provide information about the functional connectivity of coupled neurons which can be extracted using the flashlight transformation . We then apply the new method to data which were recorded from macaque prefrontal cortex using a multi-tetrode array . We find that copula-based distributions with negative binomial marginals provide an appropriate stochastic model for the multivariate spike-count distributions rather than the multivariate Poisson latent variables distribution and the often used multivariate normal distribution . The dependence structure of these distributions provides evidence for common inhibitory input to all recorded stimulus encoding neurons . Finally , we show that copula-based models can be successfully used to evaluate neural codes , e . g . , to characterize stimulus-dependent spike-count distributions with information measures . This demonstrates that copula-based models are not only a versatile class of models for multivariate distributions of spike-counts , but that those models can be exploited to understand functional dependencies .
So far , it is still unknown which statistics are crucial for analysis in order to understand the neural code . One approach is to analyze simultaneous spike-counts of neural populations . Responses from populations of sensory neurons vary even when the same stimulus is presented repeatedly , and the variations between the simultaneous spike-counts are usually correlated ( noise correlations ) at least for neighboring neurons . These noise correlations have been subject of a substantial number of studies ( see [1] for a review ) . For computational reasons , however , these studies typically assume Gaussian noise . Thus , correlated spike rates are generally modeled by multivariate normal distributions with a specific covariance matrix that describes all pairwise linear correlations . For long time intervals or high firing rates , the average number of spikes is sufficiently large for the central limit theorem to apply and the normal distribution is a good approximation for the spike-count distributions . Several experimental findings , however , suggest that processing of sensory information can take place on shorter time scales , involving only tens to hundreds of milliseconds [2] , [3] . In this regime the normal distribution is no longer a valid approximation: Though not widespread for modeling spike-counts , alternative models have been proposed in previous studies that have Poisson distributed marginals and separate parameters for higher order correlations , e . g . the multiple interaction process model [6] and the compound Poisson model [7] . Both models are point processes . In terms of their induced spike-count distributions these models are special cases of the multivariate Poisson latent variables distribution first introduced by Kawamura [8] and presented in a compact matrix notation by Karlis and Meligkotsidou [9] . Similar to the multivariate normal distribution this model has also a couple of shortcomings for spike-count modeling: ( 1 ) Only Poisson-marginals can be modeled . ( 2 ) Negative correlations cannot be represented . ( 3 ) The dependence structure is inflexible: features like tail dependence cannot be modeled . We use and extend a versatile class of models for multivariate discrete distributions that overcome the shortcomings of the afore-mentioned models [10] , [11] . These models are based on the concept of copulas [12] , which allow to combine arbitrary marginal distributions using a rich set of dependence structures . In neuroscience they were also applied to model the distribution of continuous first-spike-latencies [13] . Figure 1 illustrates the copula concept using spike-count data from two real neurons . Figure 1A shows the bivariate empirical distribution and its two marginals . The distribution of the counts depends on the length of the time bin that is used to count the spikes , here . In the case considered , the correlation at low counts is higher than at high counts . This is called lower tail dependence [12] . Figure 1B shows the discretized and rectified multivariate normal distribution . On the other hand , the spike-count probabilities for a copula-based distribution ( Figure 1C ) correspond well to the empirical distribution in Figure 1A . The paper is organized as follows . The next Section “Materials and Methods” contains a description of methodological details regarding the multivariate normal distribution , the multivariate Poisson latent variables distribution , the copula approach for spike-counts and the model fitting procedures . In this section we will also introduce a novel transformation for copula families . The method is innovative and yields a novel result . We will then describe the computational model used to generate synthetic data and the experimental recording and analysis procedures . In the Section “Results” copula-based models will be applied to artificial data generated by integrate-and-fire models of coupled neural populations and to data recorded from macaque prefrontal cortex ( PFC ) during a visual memory task . The appropriateness of the models is also investigated . The paper concludes with a discussion of the strengths and weaknesses of the copula approach for spike-counts .
All procedures were approved by the local authorities ( Regierungspräsidium ) and are in full compliance with the guidelines of the European Community ( EUVD 86/609/EEC ) for the care and use of laboratory animals . The multivariate normal ( MVN ) distribution is characterized by a probability density over continuous variables and its cumulative distribution function ( CDF ) with mean and covariance matrix is given by In order to apply it to spike-count distributions ( which are discrete and non-negative ) it is discretized and rectified ( probability for negative values is set to zero ) . Its CDF is given bywhere denotes the floor operation for the discretization . The probability mass function will have peaks at the zero count rows , due to the rectification of the CDF . It would be desirable to distribute the cut off mass equally to the complete domain . However , this is infeasible for more than three dimensions , because the necessary normalization term is computationally too time-consuming . Note that is no longer the mean of the distribution corresponding to , because the mean is shifted to larger values as is rectified . This shift grows with the dimension . The Poisson latent variables distribution is characterized by a probability mass function ( PMF ) over non-negative integer variables [9] . A random variable with this distribution is composed of latent variables . These latent variables are independent univariate Poisson distributed with rates takes the form , where is a mixture matrix . The PMF of is then given by When we set to we can vary all pairwise and higher order interactions separately using the rates of the latent variables . However , only non-negative correlations can be modeled , because the rates of the latent variables are non-negative . Furthermore , the are marginally Poisson distributed . A copula is a cumulative distribution function ( CDF ) which is defined on the unit hypercube and which has uniform marginals [12] . Formally , a copula is defined as follows: Our goal is to construct multivariate distributions for simultaneously recorded spike-counts that can model a wide range of dependence structures . Copulas make it possible to model multivariate distributions based on two distinct parts: the distributions of the individual elements and the dependence structure . Let us now assume that represents the spike-count of neuron within a given interval . According to Theorem 1 we can then describe the joint cumulative distribution function of the spike counts by choosing a copula from a particular family , and by setting and . are the models of the marginal distributions , i . e . the cumulative distributions of spike-counts of the individual neurons . Often , the Poisson distribution is a good approximation to spike-count variations of single neurons [16] , hence the CDFs of the marginals take the form is the mean spike-count of neuron for a given bin size . A more flexible marginal is the negative binomial distribution , which allows to model spike-count distributions showing overdispersion . Here is the gamma function , is again the mean spike-count of neuron , and is a positive parameter , which controls the degree of overdispersion . The smaller the value of , the greater is the Fano factor , and as approaches infinity , the negative binomial distribution converges to the Poisson distribution . The second part of the model is the copula family . Many different families have been discussed in the literature in the past . Families differ by the number of free parameters and by the class of dependence structures they can represent . The most simplistic copula is the product copula defined as for which independence is attained . We selected a number of useful copula families ( see Table 1 ) . Figure 2 shows their bivariate probability density functions ( PDFs ) . The Clayton family has a so-called lower tail dependence: the correlation between its elements is higher for low values than for high values ( see Figure 2A ) . The scalar parameter controls the strength of dependence . Note that does not only control the strength of pairwise interactions but also the degree of higher order interactions . We define . The Gumbel-Hougaard ( short Gumbel ) family has an upper tail dependence . Here , the region of high correlation is in the upper right corner of the density . Hence , the correlation between its elements is higher for high values than for low values ( see Figure 2B ) . The scalar parameter controls the strength of dependence . The Frank family has no tail dependence . There is no difference between the correlation for low and for high values ( see Figure 2C ) . Again , the scalar parameter controls the strength of dependence and we define . The Ali-Mikhail-Haq ( AMH ) family models are positively ordered , i . e . for it holds for all ( see Figure 2D ) . Again we define . The Farlie-Gumbel-Morgenstern ( FGM ) family has parameters that individually determine the pairwise and higher order interactions . It has parameters less than the Poisson latent variables distribution because the rates of the neurons can be parametrized by the marginals . Non-zero values of the parameter indicate the presence of order interaction . For order interactions are absent . If , for example all for , the corresponding probability distribution includes only parameters of second order , similar to the multivariate normal distribution . The constraints on the parameters , however , constrain the corresponding correlation to be small in terms of their absolute value . We now introduce a novel extension of standard copula models , which is particularly useful for modeling distributions of spike-counts . It is based on the orthant dependence concept . Here , an orthant refers to one of the hypercubes of equal size in the unit hypercube , i . e . a “corner” of the copula distribution . Let us consider a distribution with a so-called lower tail dependence ( see Figure 3A ) , i . e . a distribution , for which the correlation between spike-counts of two neurons is higher for low values than for high values . We now introduce the flashlight transformation which allows to shift the region of high correlation to an arbitrary orthant ( see Figure 3B–D ) . The whole dependence structure between spike-counts is rotated accordingly , but remains unchanged otherwise . The transformation is a function that operates on CDFs . Yet , it rotates the corresponding PDF , not the CDF . The flashlight transformation is specified in the following theorem ( see Text S1 in the supplementary material ) : Once a family of marginal distributions and a family of copulas for describing the dependence structure has been selected , model parameters have to be estimated from the data , i . e . from the empirical distribution . Here we suggest a method which is similar to maximum likelihood estimation . Theorem 1 provides a method to construct multivariate CDFs based on copulas . Therefore , the approach yields a CDF of a multivariate distribution . In order to calculate the likelihood we have to transform the CDF to a probability mass function ( PMF ) . For this purpose we define the sets and , . The probability of a particular set of spike-counts can then be expressed using only the CDF , making use of the so-called inclusion-exclusion principle of Poincaré and Sylvester [19]: ( 2 ) Letdenote the sum of log likelihoods of the marginal distribution , where are the parameters of the chosen family of marginals . Furthermore , letbe the log likelihood of the joint probability mass function , where denotes the parameter of the chosen copula family . The so-called inference for margins ( IFM ) method [20] now proceeds in two steps . First , the marginal likelihoods are maximized separately: Then , the full likelihood is maximized given the estimated marginal parameters: It was shown that the IFM estimator is asymptotically efficient [20] . The estimator is computationally more convenient than the maximum likelihood estimator , because parameter optimization in low dimensional parameter spaces needs less computation time . Depending on whether the copula parameters are constrained , either the Nelder-Mead simplex method for unconstrained nonlinear optimization [21] or the line-search algorithm for constrained nonlinear optimization [22] can be applied to estimate the copula parameters using Eqn 2 as the objective function . For mixtures of copulas , where the values of the latent variables have to be estimated in addition , we suggest to use the expectation-maximization algorithm [23] , [24] . In the expectation step , the weights are updated usingwhere is the PMF of the model based on the copula . In the maximization step the copula parameters are determined for fixed values of by applying the IFM method . Both steps are repeated until parameter values converge . The leaky integrate-and-fire neuron is a simple neuron model that models only subthreshold membrane potentials . The equation for the membrane potential is given bywhere denotes the resting membrane potential , is the total membrane resistance , is the synaptic input current , and is the time constant . The model is completed by a rule which states that whenever reaches a threshold , an action potential is fired and is reset to [25] . In all of our simulations we used , , , , and initialized with . These are typical values that can be found in [25] . Current-based synaptic input for an isolated presynaptic release that occurs at time can be modeled by the so-called -function [25]: The function reaches its peak at time and then decays with time constant . We can model an excitatory synapse by a positive and an inhibitory synapse by a negative . We used for excitatory synapses , for inhibitory synapses , and . Neural activity was recorded from the lateral prefrontal cortex within an area of located on the ventral bank of the principal sulcus of an adult female rhesus monkey ( macaca mulatta ) . Recordings were performed simultaneously from up to adjacent sites with an array of individually movable fiber micro-tetrodes ( manufactured by Thomas Recording ) with an inter-tetrode distance of . Data were sampled at and bandpass filtered between and . Recording positions of individual electrodes were chosen to maximize the recorded activity and the signal quality . The recorded data were processed by a principal component analysis-based spike sorting method . Automatic cluster cutting was manually corrected by subsequent cluster merging if indicated by quantitative criteria such as the ISI-histograms or amplitude stability . Activity was recorded while the monkey performed a visual working memory task . One out of visual stimuli ( fruits and vegetables ) were presented for approximately . After a delay of , during which the monkey had to memorize the sample , a test stimulus ( “test” ) was presented and the monkey had to decide by differential button press whether both stimuli were the same or not . Correct responses were rewarded . Match and non-match trials were randomly presented with equal probability ( ) . The mutual information between spike-counts and stimuli is given by ( 3 ) where is the set of stimuli , is the probability distribution over the stimuli , and is the likelihood of a neural response given a stimulus . For higher dimensions the sum over prohibits an exact computation of , since the number of terms of the sum grows exponentially with . The evaluation of this sum is therefore practically infeasible unless the number of neurons is very small . However , we can estimate the mutual information using Monte Carlo sampling . For each of the stimuli , we can estimate the second sum by drawing samples with probability . The termwill then converge to the second sum in Eqn 3 , as approaches infinity [26] .
Typically the number of samples that can be obtained in electro-physiological experiments is small . Thus , it might appear to be hopeless to estimate a multidimensional model with a detailed dependence structure . However , since our marginal distributions are discrete the copula matters only at a small number of points . In the following , we will demonstrate that it is not always necessary to obtain a great number of samples for a reliable model estimation . For this purpose we selected the Clayton-copula model with negative binomial marginals as a ground truth model which was used to draw samples . We calculated the deviation of the log likelihood of the estimated model from the log likelihood of the ground truth model in percent of the ground truth log likelihood . The correlation strength of the ground truth model was varied by the Clayton parameter . The results are shown in Figure 4 for three different Clayton parameters of the ground truth model . For moderate dependence strengths ( as are typically found in the data ) samples were sufficient for estimations of the log likelihood with an error of less than . One cause for dependence between spike-counts of different neurons are common input populations . Therefore , we investigated network models with different types of common input . We set up two current based leaky integrate-and-fire neurons ( see Section “Materials and Methods” ) and three input populations modeled as Poisson spike generators . The left input population projected only to neuron 1 and the right input population projected only to neuron 2 . The center input population was the common input population , projecting to both neuron 1 and neuron 2 . We investigated all four combinations of excitatory ( E ) and inhibitory ( I ) projections from the common population to the two neurons ( see Figure 5A1–A4 ) . In this network model a lower tail dependence should arise if the projections from the common input projection are mostly inhibitory: each time the common population is active the firing rates of both neurons will decrease simultaneously . Therefore , only low spike-counts should be strongly correlated and the Clayton family should provide a good fit to the responses of such a network . Similarly , two excitatory projections should result in an upper tail dependence and other combinations should become apparent as dependence blobs in other corners of the probability density function of the copula . The flashlight transformation shifts the dependence blob of a given copula with orthant dependence into other orthants of the probability density function and is thus capable of modeling different types of common input populations in a stochastic manner . For two neurons , the lower left corner models an inhibitory input population , the upper right corner models an excitatory input population , and the other corners model a combination of excitatory and inhibitory input populations . The spike trains of the two neurons were binned into intervals . We applied copula-based models with negative binomial marginals to fit the generated data from the four models using the IFM method ( see Section “Model Fitting” ) . Four different copula families were applied: the unmodified bivariate Clayton family and the three remaining flashlight transformations of the Clayton family ( Figure 3 ) . Figure 5C1–C4 shows the log likelihoods of the fits for the corresponding networks as shown in Figure 5A1–A4 . The respective model performed best for the combination of projection types of the common input population it was supposed to model , i . e . Clayton for I-I , Clayton survival for E-E , etc . Hence , by determining the best fitting transformation the most likely combination of input types could be identified . Each of the transformations could be associated with a distinct combination of projection types . To investigate whether the results of the reconstruction depend on the strengths of the synapses we varied between and for excitatory synapses and between and for inhibitory synapses ( data not shown ) . While the relation of the best fitting copula families was constant across all strengths the differences between the curves decreased for decreasing strengths . For it was hard to distinguish between the likelihoods of lower and upper tail dependencies . Therefore , tail dependencies were less pronounced in the spike-counts . In the multi-tetrode data , however , we found significant differences between the likelihoods of the copula families ( see Section “Application of Copula-Based Models to Multi-Tetrode Data” ) . To investigate the impact of the bin size on the reconstruction performance we also binned the data into smaller and larger intervals ( data not shown ) . When the bin size was too small or too large ( and ) the reconstruction did not succeed . In the intermediate range ( , ) , however , the connection types could be reconstructed . This can be explained by the asymptotic distributions of the multivariate spike-counts . According to the central limit theorem the multivariate normal distribution provides a good approximation when the bin size is sufficiently large . Hence , tail dependencies will vanish . On the contrary , when the bin size becomes too small the marginal distributions are essentially Bernoulli distributed and the tail dependencies will vanish as well . Of course , the range of the intermediate bin size depends on the rates of the neurons . The larger the rates the smaller the bin sizes in the intermediate range . For the simulated data the rates were comparable to the data recorded from the prefrontal cortex ( see Section “Multi-Tetrode Recordings” ) . Our copula-based models are capable of modeling different dependence structures with marginals that are tailored to single neuron spike-count distributions . Thus , we expected that the copula-based models would provide a much better fit to data recorded from real neurons than the multivariate normal distribution or the multivariate Poisson latent variables distribution . To test this , we applied copula-based models from different families and with different marginal distributions to data , which has been recorded from macaque prefrontal cortex for each of the twenty presented stimuli and each of the four phases ( pre-stimulus presentation , stimulus presentation , delay , presentation of the test stimulus ) of the visual working memory task . We compared the results to models of the discretized multivariate normal and the Poisson latent variables distribution ( see Section “Materials and Methods” ) We randomly selected count vectors for each task phase and each stimulus as the validation set . We then estimated the model parameters on the remaining count vectors ( training set ) and used the validation set for obtaining an unbiased estimate of the likelihoods of the selected models . We used the IFM-estimator for the copula-based models and the maximum likelihood estimator for the Poisson latent variables distribution . The parameters and of the discretized MVN distribution were estimated by the sample mean and the sample covariance matrix of the spike-counts . This procedure does not correspond to the maximum likelihood estimate of the discretized distribution . We used it , because the maximum likelihood estimator was too expensive to compute for six neurons . The high computational costs come from the estimation of the CDF of the MVN . The rate parameter for the Poisson distribution and negative binomial distribution were estimated via the sample mean . The maximum likelihood estimates for the overdispersion parameter were computed iteratively by Newton's method . Figure 6A summarizes the results for the discretized MVN , the Poisson latent variables distribution , and two copula-based distributions with different marginals , the Poisson distribution , and the negative binomial distribution . The negative binomial distribution provided for all four task phases a significantly better fit than the Poisson distribution , the MVN distribution , and the Poisson latent variables distribution . The likelihood for the copula-based models was significantly greater than for the discretized MVN model ( , paired-sample Student's t test over stimuli ) and the Poisson latent variables model ( ) . Moreover , the likelihood for the negative binomial marginals was even greater than that for the Poisson marginals ( ) . Thus , the copula-based approach provided models that were indeed superior for the data at hand . Moreover , the additional flexibility of the negative binomial marginals improved the fit significantly . We applied different copula families to examine the importance of the dependence structure for the model fit . Figure 6B shows an evaluation of the different copula families with different dependence structures for the best fitting marginal , which was the negative binomial distribution . The model based on the Clayton copula family provided the best fit . The fit was significantly better than for the second best fitting copula family ( ) , the Gumbel family . In spite of having more parameters , the FGM copulas performed worse . However , the FGM model with third order interactions fitted the data significantly better than the model that included only pairwise interactions ( ) . The best fitting copula-based model , the Clayton copula , is characterized by a lower tail dependence . Apart of the Gumbel family , the other families that we applied so far do not model orthant dependencies . To check whether other orthant dependencies would improve the fit , we applied the flashlight transformation and we transformed the Clayton copula tail towards all corners of the six dimensional hyper cube . The results are shown in Figure 7 . The standard Clayton copula with lower tail dependence had the significantly highest value of the log likelihood on the validation set indicating that the empirical spike-count distribution has indeed a lower tail dependence . The second highest peak was reached by the Clayton survival copula . The central peak corresponded to those transformations that were close to the Clayton and the Clayton survival copulas: sectors and ( and decimal ) . Thus , a common lower tail dependence was prominent in the data . We applied mixtures of copulas as described in Section “The Flashlight Transformation and Mixtures of Copulas” to check whether there was indeed a prominent common upper tail dependence beside the lower tail dependence in the data . Therefore , we fixed the Clayton copula ( which models a lower tail dependence ) as the first mixture component and varied the sector of the flashlight transformed Clayton copula for the second mixture component . Figure 7C shows the mean log likelihoods of the mixture models with negative binomial marginals on the same data set used for Figure 7B . All of the mixture models exhibit similar performance . Therefore , the upper tail dependence that we observed for the unmixed model appears to be an artifact of the lower tail dependence . In summary , we could show that the copula-based approach provided a significant improvement in the goodness of fit compared to the discretized and rectified multivariate normal distribution and the Poisson latent variables distribution . Moreover , the dependence structure alone has a significant impact as well . Our model consists of two parts: 1 ) the copula and 2 ) the marginals . We already analyzed the effect of the copula . In this section we describe the investigation of the marginals . In particular , we are interested in understanding how the goodness of fit is influenced by the marginals . For this purpose we compared the log likelihoods of the Clayton-copula model with Poisson , negative binomial , and empirical marginals fitted to the training set of the sample stimulus presentation phase . The model with empirical marginals was a so-called semiparametric distribution consisting of a parametric dependence structure ( the copula family ) and nonparametric marginals . We drew samples from these distributions in order to learn whether the training and validation sets were typical samples from the fitted distributions . For a complex model we expect the likelihood of training samples to be close to the mode of the histogram , while we expect the validation samples to have a much smaller likelihood . Contrary , for a model with small complexity we expect the likelihood of the training samples to be close to the likelihood of the validation samples . When the complexity is too small we expect the likelihoods of the training and the validation samples to be much smaller than the mode of the histogram . In our setting the most complex model is the one with empirical marginals . Histograms of the log likelihoods for copula models with the three different marginals are shown in Figure 8 . For Poisson marginals , the log likelihoods of both the training set and the validation set were much smaller than the log likelihoods of the samples drawn from the fitted distribution . Thus , the Poisson marginals seem to be too simple for a good fit to the data , whereas the negative binomial marginals generalized well in spite of their increased complexity . On the training set the model with the empirical marginals performed best . However , there was a huge discrepancy to the likelihood of the model with empirical marginals on the validation set , whereas the likelihoods of the other two models did not change much . This result can be explained by overfitting . The empirical marginals matched the marginals of the training set perfectly . The empirical marginals of the training set , however , were noisy representations of the true marginals , because of the limited sample size . Hence , a perfect fit is not beneficial when it comes to novel data . In contrast to that , the likelihoods of the models with Poisson and negative binomial marginals were almost equal to the respective likelihoods on the training set . Thus , these models did not suffer from overfitting . In order to relate these findings to the number of samples in our training set we can compare the number of samples to the estimated number of required samples for the toy example in Section “Model Fitting” . Figure 6 shows that the log likelihood for the Clayton-copula model deviated from the second best family by . In Section “Model Fitting” we showed that for this model samples were sufficient for good estimations of the log likelihood . For the delay phase and for the test stimulus phase , the number of samples varied between and per stimulus . Therefore , the number of samples was sufficient for these phases . Taken together with the histogram analysis , we found that the model complexity was appropriate for the available amount of data at hand . We will now show that the copula-based models can be used to measure the short-term information about a stimulus that is encoded by the spike-count dependence structure of the recorded neurons . The first step is to estimate the total information of the spike-count responses . We applied the best fitting copula model , the Clayton-copula model with negative binomial marginals , to estimate the mutual information between stimuli and responses via Monte Carlo sampling ( see Section “Materials and Methods” ) . Figure 9A shows the estimated mutual information for each of the four task phases . The mutual information was greater during the sample stimulus interval and the test stimulus interval than during the delay interval . Therefore , a stimulus presentation evoked a spike-count response which instantly encoded information about the stimulus . In the test stimulus phase the dotted line is above the dashed line , so the spike-counts coded more information about the sample stimulus that was previously presented than about the test stimulus . Figure 9B shows the information estimate , normalized to the mutual information that is shown in Figure 9A . The dependence structure carried between and of the mutual information . During the test stimulus interval the dependence structure encoded almost twice as much information about the test stimulus as about the sample stimulus that was previously presented .
We developed a framework for analyzing the noise dependence of spike-counts and used synthetic data from a model of leaky integrate-and-fire neurons to derive interpretations for different dependence structures . Applying the framework to our data from the macaque prefrontal cortex we found that: ( 1 ) copula-based models with negative binomial marginals rather than the multivariate normal distribution or the Poisson latent variables distribution are appropriate models of spike-count data for short time intervals; ( 2 ) the dependence structure encodes between and of the mutual information about the presented stimuli; ( 3 ) the amount of data required for a good likelihood estimation is present in our data set; and ( 4 ) a lower tail dependence between all neurons is present in the data and can be explained by common inhibitory input . The copula approach has many advantages compared to previous models . Recently , the Ising model gained a lot of attention in neuroscience [4] , [27] . This model is a maximum entropy model of binary variables called spins that have only pairwise interactions [28] . The model is applied to the neuroscience setting by binning spike trains into very short time intervals such that at most one spike falls into each bin . The spin for that bin then indicates whether or not a spike was present . Using this model pairwise interactions between simultaneously recorded neurons can be modeled [4] . The Ising model is a special case of a more general class of nested maximum entropy models [29] . Other models in this class can be used to model higher order interactions between neurons . Nevertheless , an independence assumption for subsequent bins is necessary due to the limited number of samples present in typical neuroscience settings . Therefore , the marginal spike-counts of individual neurons will be binomial distributed . The variance of this distribution is always smaller than its mean which is a severe disadvantage of this model class . The copula approach on the other hand can model arbitrary marginals . Another class of models are doubly stochastic models where some parameters of the data distribution are themselves random variables . The doubly stochastic Poisson point process presented by Krumin and Shoham belongs to this class [30] . For such models the marginal distributions change whenever the dependence is modified . It is thus very hard to disentangle the effects of the dependence structure from the effects of the marginals . In contrast to the multivariate normal distribution and the multivariate Poisson latent variables distribution the copula approach can be used to model arbitrary marginal distributions that are appropriate for the data at hand . The marginal distributions can therefore be discrete without any mass on the negative axis and with variance greater than the mean . We compared the fits of negative binomial marginals to Poisson and empirical marginals and found that only the negative binomial marginals provided a reasonable fit to the data . Contrary to the Poisson marginals , the negative binomial marginals were complex enough such that likelihoods of samples from the model were consistent with the likelihood of the data . Moreover , the negative binomial marginals did not suffer from overfitting as did the empirical marginals . We conclude that the negative binomial marginals are appropriate to describe the spike-counts recorded from the prefrontal cortex . The dependence structure of the copula approach is flexible . Higher order interactions can be parametrized separately if desired . Furthermore , in contrast to the multivariate Poisson latent variables distribution , negative correlations can be modeled as well . Another advantage of the copula approach is that it is modular in the sense that the copula family used for the data analysis can be easily exchanged by another family . Many different copula families exist , each representing and parameterizing different properties of the dependence structures . Thus , it is easy to test for different properties of a distribution . Specific examples are the Clayton and Gumbel families . These families have lower and upper tail dependencies , respectively . Lower and upper tail dependencies can arise from common input populations with inhibitory and excitatory projections , respectively . By deriving the flashlight transformation we could construct additional families that account for combinations of inhibition and excitation . When applying the flashlight transformation to the data from the prefrontal cortex , we found that the unmodified Clayton family provided the best fit to the test data . Therefore , a common lower tail dependence to all neurons is present in the data . One explanation is a common input population whose projections are mostly inhibitory to all the analyzed neurons . Two types of common inhibitory sources are possible: ( 1 ) A local source of inhibitory input such as common interneurons . ( 2 ) Another area projecting to the prefrontal cortex . It was found that interneurons have a reach of no more than a few hundred micrometers whereas the inter-tetrode distance was . Thus , it is unlikely that a population of common interneurons inhibits all the stimulus specific neurons that we recorded from . Another area , therefore , is more likely to be the source of the common inhibitory input . One possibility could be the ventral tegmental area ( VTA ) . In the rat cortex it was found that the VTA exerts a direct inhibitory influence on the PFC . In a study of recorded PFC neurons were inhibited as a result of VTA stimulation [31] . Moreover , the VTA is thought to be a central component of the reward system [32] which is essential for a memory task . Our analysis provides evidence for such an influence based on the spike-count statistics . The second best fit was achieved by the Clayton survival family . One explanation for this result is provided by an upper tail dependence between all neurons in addition to the stronger lower tail dependence . We applied mixtures of copulas to elucidate this issue and found that a mixture of the Clayton and Clayton survival family did not provide the best fit out of all mixtures of the Clayton family with a Clayton flashlight transformation . At first sight it is puzzling that the upper tail dependence seems to disappear when mixed with the lower tail dependence . However , the Clayton copula and the Clayton survival copula have their dependence along the same line in the six dimensional space that is spanned by the neuronal spike-counts , though predominantly at different ends of this line . Hence , the Clayton survival family can capture some of the dependence that is inherent to the Clayton family . We conclude that the prominence of the upper tail dependence that was observed for the unmixed model is an artifact of the lower tail dependence component . The results show that important properties of dependence structures such as tail dependencies arise very naturally in simple input scenarios , and that the copula approach can be used to construct generative models that are capable of capturing these aspects of this underlying connectivity . In principle , copula-based models can be used to guide reconstructions of functional connectivity , but this topic is outside the scope of this study . If the reader is interested in detailed reconstruction of functional connectivity we recommend the studies in [33]–[35] as a starting point . We could show that there is important information represented in the dependence structure which has been ignored in studies reporting only the correlation coefficient . Based on the flashlight transformation we could derive novel copula families with interesting interpretations for neuroscience: the statistical dependence gives insight into possible connections of the underlying network . Other copula families might be applicable to investigate different properties of the network . We could also show that the Gaussian distribution is not an appropriate approximation of the spike-count distribution of short time intervals . Yet , many studies applied this approximation in their investigations . Therefore , these studies should be reassessed with respect to their validity for short-term coding . We also compared the copula-based approach to the multivariate Poisson latent variables distribution . In terms of spike-counts this model corresponds to previous point process models that account for higher order correlations . The copula-based approach overcomes a number of shortcomings of this distribution , namely the Poisson marginals , the restriction to non-negative correlations and the inflexible dependence structure . We could show that the improvement in the goodness-of-fit is significant . Taken together , the copula-based approach allows us to model and analyze spike-count dependencies in much more detail than previously applied models . A drawback is the small number of neurons to which the approach can be applied so far . The approach is computationally too demanding for higher numbers of neurons because the model fitting complexity is exponential in the number of neurons . Approximate inference methods might provide a solution to the computational problem . However , another problem is the number of samples available in typical electro-physiological experiments . We could show that samples are sufficient for six dimensional data with moderate dependence strengths . Nevertheless , the amount of required data increases dramatically for increasing dimensions , i . e . for the number of neurons . A combination with dimensionality reduction techniques might provide a solution to this problem .
|
The brain has an enormous number of neurons that do not work alone but in an ensemble . Yet , mostly individual neurons were measured in the past and therefore models were restricted to independent neurons . With the advent of new multi-electrode techniques , however , it becomes possible to measure a great number of neurons simultaneously . As a result , models of how populations of neurons co-vary are becoming increasingly important . Here , we describe such a framework based on so-called copulas . Copulas allow to separate the neural variation structure of the population from the variability of the individual neurons . Contrary to standard models , versatile dependence structures can be described using this approach . We explore what additional information is provided by the detailed dependence . For simulated neurons , we show that the variation structure of the population allows inference of the underlying connectivity structure of the neurons . The power of the approach is demonstrated on a memory experiment in macaque monkey . We show that our framework describes the measurements better than the standard models and identify possible network connections of the measured neurons .
|
[
"Abstract",
"Introduction",
"Materials",
"and",
"Methods",
"Results",
"Discussion"
] |
[
"neuroscience/theoretical",
"neuroscience",
"neuroscience/sensory",
"systems"
] |
2009
|
Analyzing Short-Term Noise Dependencies of Spike-Counts in Macaque Prefrontal Cortex Using Copulas and the Flashlight Transformation
|
Recent genome-wide analyses have uncovered a high accumulation of RNA polymerase II ( Pol II ) at the 5′ end of genes . This elevated Pol II presence at promoters , referred to here as Poll II poising , is mainly ( but not exclusively ) attributed to temporal pausing of transcription during early elongation which , in turn , has been proposed to be a regulatory step for processes that need to be activated “on demand” . Yet , the full genome-wide regulatory role of Pol II poising is yet to be delineated . To elucidate the role of Pol II poising in B cell activation , we compared Pol II profiles in resting and activated B cells . We found that while Pol II poised genes generally overlap functionally among different B cell states and correspond to the functional groups previously identified for other cell types , non-poised genes are B cell state specific . Focusing on the changes in transcription activity upon B cell activation , we found that the majority of such changes were from poised to non-poised state . The genes showing this type of transition were functionally enriched in translation , RNA processing and mRNA metabolic process . Interestingly , we also observed a transition from non-poised to poised state . Within this set of genes we identified several Immediate Early Genes ( IEG ) , which were highly expressed in resting B cell and shifted from non-poised to poised state after B cell activation . Thus Pol II poising does not only mark genes for rapid expression in the future , but it is also associated with genes that are silenced after a burst of their expression . Finally , we performed comparative analysis of the presence of G4 motifs in the context of poised versus non-poised but active genes . Interestingly we observed a differential enrichment of these motifs upstream versus downstream of TSS depending on poising status . The enrichment of G4 sequence motifs upstream of TSS of non-poised active genes suggests a potential role of quadruplexes in expression regulation .
Transcription of protein-coding genes by RNA polymerase II ( Pol II ) is a complex , multistep process [1–5] . Potentially each of the transcription steps such as polymerase II recruitment , pre-initiation complex ( PIC ) assembly , open complex formation , promoter escape , pausing , elongation , and termination provides opportunity for a regulatory action [6] . Until recently it has been assumed that the assembly of the pre-initiation complex and Pol II recruitment are the main regulatory steps [7] . Recent genome-wide chromatin immunoprecipitation ( ChIP ) studies in human [8–10] and Drosophila melanogaster [11 , 12] cells have shown a high accumulation of Pol II at the 5′ end of genes . Previously , several genes , including HSP70 and Myc [13–16] , have been known to harbor promoter proximal paused polymerase whose release regulates Pol II entrance into the productive elongation step; this mode of regulation was assumed to be rare . In those studies , Pol II pausing has been defined formally as an event in which the forward movement of elongation-competent transcription complexes is temporarily stopped owing to template sequence , regulatory factors , or both [1] . Genome-wide studies demonstrated that in addition to Pol II pausing , in some cell types Poll II accumulates at the promoters due to different reasons , some of which could also be regulatory . In particular , Maxwell et al . demonstrated that during starvation in C . elegans , in addition to pausing occurring at active stress-response genes , an inactive ‘‘docked” Pol II accumulates upstream of inactive growth genes [17] . Kouzine et al . showed that promoter melting is another key regulatory step of gene expression in resting B cells [18] and it also leads to accumulation of Pol II in the promoter region . Thus promoter-proximal accumulation of Pol II surveyed , for example , by a ChIP-seq experiment , is clearly not sufficient to determine the precise transcriptional status of Pol II . Bona fide pausing can be observed by experiments such as permanganate sensitivity assays [3 , 11 , 12] , scRNA-seq [19] [20] , GRO-seq [21] , or PRO-seq [22] . To make a distinction between bone fide pausing and accumulation of Pol II in Chip-seq experiments , accumulation of Pol II at promoter , independent of its status , is often referred to as polymerase poising [11 , 23 , 24] . Following this practice , here we consider Pol II to be poised if its density at the promoter is significantly higher than in the gene body independent of a specific Pol II status . We caution readers that the term “poised” has also been used in literature in other contexts , specifically to denote Pol II phosphorylated on Ser-5 residues ( RNAPIIS5P ) [25 , 26] or to denote genes with promoters that comprise a bivalent chromatin domain containing a histone modification associated with transcriptional activation , histone H3 trimethylated at lysine 4 ( H3K4me3 ) , along with another associated with transcriptional repression , H3K27me3 ( for a review , see [27] ) . Pol II poising allows for pre-recruitment of Pol II ahead of gene expression . In particular , it is now broadly accepted that Pol II poising facilitates a rapid response to stimuli [12 , 28–33] . In agreement , promoter-proximal Pol II poising was shown to be prevalent at genes involved in response to stimuli , immune response , and development [11 , 12 , 34–37] . An important group of genes susceptible to fast induction are Immediate Early Genes ( IEG ) –genes that are activated within minutes from stimuli . Pol II poising has been described for several mammalian IEGs including JunB , cFos , and cMyc [36 , 38 , 39] . Analyzing rat neurons , Saha et al . found that several IEGs , such as Arc ( also known as activity-regulated gene 3 . 1 ) , are poised for near-instantaneous transcription by Pol II poising [40] . Similarly , immediate mediators of the inflammatory response were shown to be poised for gene activation through RNA polymerase II stalling [34] . Pol II poising provides an opportunity not only for fast but also for synchronized response to stimuli [41] . However the exact mechanism regulating such synchronization is yet to be elucidated . In particular , not all poised genes are induced in response to stimuli [42–44] thus such synchronization would have to be conditioned on additional factors . In addition , studies of Pol II poising during Drosophila development indicated that de novo recruitment of poised Pol II does not occur in a tissue-specific manner , necessitating additional tissue-specific regulation [23] . Given the role Pol II poising is proposed to play in the regulation of gene expression , comparative analysis of cells in different stages provides an important tool for understanding this mode of regulation . Recently , such a comparative GRO-Seq based analysis of mouse embryonic stem cells ( ESCs ) and mouse embryonic fibroblasts ( MEFs ) [45] has suggested that the transition of Pol II from the poised to the productive elongation stage of transcription is a major regulated step during early differentiation in mouse cells . In Drosophila melanogaster , Gaertner et al . showed that Pol II poised status changes during development consistent with the view that poising prepares genes for future expression [23] . While Pol II poising has been proposed as an important regulator of response to stimuli , no analysis of Pol II poising in one of the most informative settings—mouse resting ( RESTB ) and activated B cells ( ACTB ) —has been done . Here , we close this gap by performing a comparative analysis of Pol II poising in these cells . Using a classification of Pol II profiles into three groups , we compared functional enrichment of genes in these classes across cell states . Analysis of resting and activated B cells allowed us to identify and investigate groups of genes whose Pol II profile changes in response to B cell activation and thus to provide novel insight into the role of Pol II poising in gene regulation .
It has been broadly recognized that the distribution of Pol II across gene is not uniform and is often characterized by an overall larger accumulation of Pol II in the promoter region than in the gene body . To examine the Pol II distribution in the analyzed cells , we processed reads from ChIP-seq data of IgG and Pol II as described in Material and Methods section . We considered Pol II to be present at the promoter ( Pol II+ ) if the number of Pol II reads at the promoter was significantly higher ( p<0 . 001 , Fisher's exact test ) than the number of IgG control reads in the same region . Unless otherwise stated , we focused on the Pol II+ genes . We use Pol II− to denote genes not in the Pol II+ group . Following the well-established practice [11] , Poising Index ( PI ) is defined as the ratio of Pol II density in the promoter to Pol II density in the gene body , as described in Material and Methods section . The distributions of Pol II density in the promoter and gene body regions , and PI values for resting and activated B cells are summarized in Fig 1 . The PI distributions differ between two B cell states with the lower PI values in activated B cells ( p<1 . 3e-178 , Mann-Whitney U test and Cohen's effect size d = -0 . 41 ) . We further classified genes with Pol II+ promoters based on the relative Pol II density in the promoter and gene body regions . Specifically , if Pol II presence in the promoter is significantly greater than in the gene body , we call these genes Pol II poised genes; otherwise we call them Pol II non-poised genes . The classification was done using Fisher's exact test where we assessed a null hypothesis that the Pol II density in the promoter and gene body is equal ( see Material and Methods ) . In the end , we defined three major classes of genes based on Pol II activity across gene: Fig 2 shows the number of genes in each class across different cell states . Both cell states have a similar number of Pol II+ promoter genes ( 51–54% overall ) : 9710 genes in activated and 9290 genes in resting B cells . However , ACTB have higher number of class NP genes ( 14% overall and 26% among Pol II+ promoter genes ) . We next analyzed the functional enrichment of genes in class NP and P in both cell states . Interestingly , considering all genes as a background ( Table 1 ) , class P genes are enriched for DNA repair , apoptosis , cell cycle , cellular macromolecule catabolic process , translation , and transcription in both resting and activated B cells . These classes generally correspond to genes that function in response to extracellular or intracellular stimuli [45] . Importantly , the enriched functional categories we identified for B cells largely overlap with categories identified using GRO-seq in mouse embryonic stem cells and mouse embryonic fibroblasts when similarly all genes were taken as the background [45] . This suggests that functional enrichment of poised genes is not only common across different cell stages but also across different cell types . This is consistent with recent results showing that during Drosophila development de novo recruitment of poised Pol II does not occur in a tissue-specific manner [23] . In contrast , class NP genes in ACTB ( Table 2 ) are enriched with regulation of lymphocyte activation—a processes that is specific to lymphocyte cells . Thus while Pol II poised genes correspond to the functional groups previously associated with poised genes in other cell types , non-poised genes are B cell state specific . Additionally , class NP genes in ACTB are enriched in translation and transcription . This is consistent with the previous observation that there is an overall 10 fold transcriptional amplification in ACTB relative to RESTB [18] . Gene Ontology ( GO ) enrichment analysis of class NP genes in RESTB does not show enrichment in any specific GO processes , however , as discussed in detail later , we found that the top genes in this class includes Immediate Early Genes—genes that are relevant for the cell activation process . To focus more closely on B cell active genes , we additionally performed enrichment analysis using only Pol II+ genes as the background . Poised RESTB genes remained enriched in translation and metabolic process ( Table 3 ) . Non-poised genes in ACTB remained to be enriched in B cell specific processes and both RESTB and ACTB were enriched in immune response ( Table 4 ) . The fact that Pol II poised genes functionally overlapped between different cell states suggests that the tendency of certain groups of genes to be poised is largely context independent , while non-poised and transcriptionally active genes include functional groups that are cell state specific . To gain insight into the role of Pol II poising in B cell activation , we asked whether there are any functionally coherent groups of genes that change Pol II profile class after B cell activation . If poising prepares genes for a rapid and simultaneous expression , we might observe an enrichment of certain functional groups within the genes moving from Pol II profile class P to class NP upon B cell activation . GO enrichment analysis of genes that changed Pol II profile from class P to NP ( with the background of all Pol II+ promoter genes in both RESTB or ACTB ) showed that these genes are enriched for cellular processes related to cell cycle and transcription such as translation , RNA processing , and mRNA metabolic process . On the other hand , Pol II+ promoter genes non-poised in RESTB ( class NP ) but poised in ACTB ( class P ) are not enriched for any specific cellular processes ( Fig 3 ) . This observation remains true when Pol II profile change is measured by at least two-fold increase of the PI value rather than by switching profile classes . We next examined how the change in gene transcriptional activity upon B cell activation relates to the changes in Pol II densities , poising index and expression . The genes that transitioned from poised to non-poised state upon B cell activation where characterized , on average , by increased Pol II density in the gene body and reduced Pol II density at promoters ( Fig 4a ) . In addition , genes with increased density of Pol II in the gene body were characterized by lower poising index and higher mRNA expression ( Fig 4b ) . Given these observations one might expect that there is a correlation between changes in poising index and changes in gene mRNA expression , but there was no such correlation . However , we observed that the genes that transitioned from poised to non-poised state upon B cell activation experienced higher expression increase than the genes that continued to be poised ( p<4 . 2e-7 , Mann-Whitney-Wilcoxon test ) ; although the effect size was negligible . Thus we separately examined changes in Pol II density at promoters and gene bodies by asking whether the increase in gene expression upon B cell activation correlates more strongly either with increased Pol II promoter proximal density or with its increased density in gene bodies . We calculated the Spearman and partial Spearman correlations between changes in gene expression and changes in Pol II promoter/body density during B cell activation ( Fig 4c ) . The increase in gene expression upon activation correlated with the increase of Pol II density in both promoter and gene body regions; the correlation was stronger for the increase of Pol II in the gene body even after correcting for the Pol II density change in the promoter . The correlation between the change in gene expression and the change in Pol II promoter density became weak ( although statistically significant ) after correcting for correlation of mRNA expression change with Pol II change in gene body ( see Spearman partial correlation in Fig 4c ) . Thus for many genes reduction in poising index value during the transition from RESTB to ACTB can be associated with an increase of Pol II density in the gene body and an increase in gene expression . This is consistent with the enrichment of these genes for cellular processes related to cell cycle and transcription such as translation , RNA processing , and mRNA metabolic process . Indeed activated B cells are much larger and transcriptionally more active than resting B cells [18] . Yet , Fig 4a also indicates that poised to non-poised transition can be a result of reducing of Pol II presence at promoter . Since increased gene expression correlates with increased Pol II promoter density then the increased expression is not the only explanation for reduced poising index . This analysis reveals that while transition from poised to non-poised state is often associated with increased Pol II body density and increased gene expression , change in poising index is not a simple function of expression change . In resting cells , a majority of genes with Pol II presence at the promoter that are non-poised ( class NP ) are characterized by low Pol II presence at both the promoter and the gene body . As we showed above , the non-poised genes in resting cells that transition to poised state in activated cells were not enriched in any GO categories . We were interested to see whether such transition can also occurs for genes that are transcribed at significant levels in resting B cells . Sorting them based on Pol II density in the gene body , we found that the top ten of these genes contain known immediate early genes such as Btg1 , Fos , Jun , Egr1 , Bcl6 , and Zfp36 . As an illustration , Fig 5 shows the Pol II profile of Jun and Fos in RESTB and ACTB . We then used measurements of mRNA expression level in resting B cells and in cells activated for 30 minutes , 3 , 24 , and 72 hours to trace expression dynamics of these genes . We found that their expression increases even further when measured 30 minutes after the B cell activation . Thus despite being highly expressed in resting B cells , these genes have not achieved their full induction level yet . Their expression then gradually drops when measured at 3 and 24 hours after B cell activation and after that remains stable ( Fig 6 ) . ( We note that our measurements of Pol II in ACTB were done 72h after activation . ) Many IEGs have previously been shown to be poised prior to the burst of their expression [39 , 46 , 47] . For example , we noted that an immediate early gene , Egr2 , is poised and expressed at a very low level in resting B cells and experiences an expression burst 30 minutes after activation . It is possible that the IEGs that we found to be highly expressed in RESTB have been spontaneously activated , which prevented us from observing them in their poised state . Interestingly , our analysis reveals that when the cells are fully activated , the expression of these genes decreases while Pol II accumulates at promoters . Thus after expression burst these IEGs return to poised state . These results suggest that Pol II poising is not just a regulatory step that prepares genes for sprint . We found that a number of early response genes , including Jun , Fos and Egr1 , are highly expressed and non-poised in resting B cells but are poised and have relatively low expression in activated B cells , suggesting that Pol II poising also accompanies attenuation of expression of previously active genes . Previous studies identified a correlation between the presence of G4 motifs and promoter-proximal poising [48] . Given the observation that for B cell Pol II poising status is cell state dependent , we asked if there is any difference in such promoter proximal sequence motifs between genes with different poising dynamics . G-quadruplexes are non-canonical conformations of DNA molecules . A necessary , but not sufficient , condition for their formation is a particular sequence motif consisting of four runs of G tracks . In this study quadruplex motif occurrences were predicted using QuadParser [49] as described in Material and Methods . Quadruplex motifs are abundant in the mouse genome and are present in the 2kb upstream regions of promoters of up to 50% of the genes . We found that for both RESTB and ACTB cells quadruplexes were enriched downstream of TSS of genes with poised Pol II . Such enrichment was observed before for human cancer NCI-60 cell line and primary T cells [48] . Interestingly we also observed that , as opposed to the poised genes , the non-poised genes in both B cell stages were enriched in these motifs upstream of TSS ( Fig 7 ) . These results suggest that , in the context of Pol II poising , the role of G4 motifs upstream and downstream of TSS is likely to be quite different . Indeed , formation of a quadruplex on template strand can obstruct Pol II movement . In addition , runs of Gs downstream of TSS can be involved in R-loops formation , which was also proposed to facilitate Pol II poising [48] , [50 , 51] . Such R-loops are found to be even more stable if RNA-DNA quadruplexes are involved [52–54] . The enrichment of G4 motifs in downstream region is strongly associated with CpG islands . In contrast the enrichment of G4 motifs in upstream region cannot be explained by the presence of CpG islands ( S2 Fig ) . Given that the biases with respect to the presence of G4 sequence motifs where similar in both ACTB and RESTB , we asked whether the genes that , similarly to the early response genes , have higher relative expression level shortly after B cell activation as compared to fully activated B cell show any particular bias for the presence of these motifs . We divided the genes into three equal classes: high ( HI ) , medium , and low ( LO ) , depending on the ratio of gene expression in B cells activated for 30 minutes to the expression after 72h of activation . In particular , the genes in the high ratio ( HI ) group are relatively more actively expressed in the cells shortly after the B cell activation ( 30m ) than in the fully activated cells ( 72h ) , as for example the early response genes . Comparing the distribution of G4 motifs in the HI and LO groups we found that the HI group was strongly enriched in G4 motifs upstream of TSS; the enrichment was higher for G4 motifs on the non-template strand than on template strand ( Fig 7 ) suggesting the importance of these motifs for rapid gene expression . Indeed , focusing specifically on early response genes identified above , we found that all of them have G4 motifs upstream of TSS but not downstream . Overall , our analysis suggests that destabilization of the DNA duplex that accompanies formation of the quadruplexes upstream of TSS can facilitate Pol II engagement and in this way might facilitate rapid progression to elongation for genes that are highly expressed immediately after activation .
Polymerase poising has been defined as a significant enrichment of accumulation of Pol II near the promoter relative to the gene body . This accumulation can be attributed to several mechanisms , including pausing of transcription during early elongation and docking . Such Pol II poising reflects pre-recruitment of Pol II ahead of gene expression . To understand the role of poising for B cell activation , we performed a comparative analysis of Pol II poising in resting and activated B cells . We found that on the genome-wide scale poised genes are consistently enriched for DNA repair , apoptosis , cell cycle , cellular macromolecule catabolic process , translation , and transcription . These functional terms were similar to the previously identified for genes exhibiting Pol II poising in embryonic stem cells and mouse embryonic fibroblasts suggesting that these groups of genes are common across cell types [45] . Also consistently with this view , the genes that change from poised to non-poised as the result of B cell activation have higher relative expression . Focusing on B cell active genes , we additionally performed enrichment analysis using only Pol II + genes as the background . Poised RESTB genes were still enriched in translation and metabolic process . Non-poised genes in ACTB remained to be enriched in lymphocyte activation and both RESTB and ACTB were enriched in immune response—all B cell specific processes . It has been proposed that poising provides a mechanism for synchronized response [1] . Indeed , cellular processes related to cell cycle and transcription such as translation , RNA processing , and mRNA metabolic processing consistently change Pol II profiles from poised to non-poised upon cell activation . Our analysis also supports the contribution of poising to regulation of early response genes . In addition , it provides further evidence to the observation made in the context of Drosophila development that the presence of poised polymerase does not necessarily equate to direct regulation through pause release to productive elongation [55] . Our results indicate that that Pol II poising is not just a regulatory step that prepares genes for sprint . We found that a number of early response genes , including Jun , Fos and Egr1 , are highly expressed and non-poised in resting B cells but are poised and have relatively low expression in activated B cells , suggesting that Pol II poising is also associated with previously active genes . A closer analysis of group of genes allowed for identification of interesting relationships between distributions of G4 motifs in poised and non-poised genes . Genes with poised Pol II in both resting and activated B cell where enriched in these motives downstream of TSS while non-poised genes showed upstream enrichment suggesting that the role of G4 motifs in Pol II dynamics might be context dependent . It has been proposed that the runs of Gs downstream of TSS can facilitate Pol II poising by R-loops formation or obstructing Pol II progression [48] . Our results are consistent with such hypothesis . The enrichment in G4 motifs upstream of TSS for non-poised genes has not been observed before . We propose that formation of the quadruplexes upstream of TSS can facilitate Pol II engagement by destabilizing the DNA duplex . Specifically , they could expedite a recruitment of Pol II to the promoter region and accelerate its progression to elongation . Thus , taken together , our analysis of resting and activated B cells allowed us to provide novel insight into the dynamics of Pol II poising .
CD43 negative B cells were isolated from the spleens of 6 to 8 week old C57BL/6J mice ( The Jackson Laboratory ) by immunomagnetic depletion ( Miltenyi Biotech ) . Activation of the purified B cells was as follows: RPMI-1640 containing FCS , Pen/Strep , glutamine , nonessential amino acids , sodium pyruvate , 2-β-mercapto-ethanol , and HEPES for: 30m , 3h , 24h or 72h in the presence of 25 μg/ml of LPS ( E . Coli 0111:B4; Sigma ) , 5 ng/ml of IL-4 ( BioSource ) and purified rat anti-mouse CD180 ( BD Pharmingen ) . The cells were incubated at 37 degrees in 5% CO2 . At the appropriate aforementioned time points the activated B cells were spun down at 1500 RPM for 5min and resuspended in 1mL of Trizol ( Life Technologies ) and processed for RNA isolation . Resting B cells were also spun down at 1500 RPM processed in the same manner . Class switching to IgG1 was verified by FACS analysis at 72h . We downloaded Pol II binding and IgG control data ( ChIP-seq ) , and mRNA sequencing data ( RNA-seq ) for resting and activated B cells from the Gene Expression Omnibus under accession number GSE24178 and from the Short Read Archive under accession number SRA072844 . For each cell line , all replicates were merged for joint analysis . The Chip-seq data with read length from 36 bps to 50 bps were aligned to the mouse reference genome mm9 using Bowtie 2 ( version 2 . 1 . 0 ) [56] allowing no more than 2 mismatches and no gaps . We disregarded reads that have multiple best match aligned loci on the reference genome . We used the spliced read aligner TopHat ( version 1 . 31 ) [57] to map all RNA-seq reads to the mouse genome ( mm9 ) . In order to estimate mRNA expression , we used Cufflinks ( version 2 . 1 . 1 ) [58] with the complete set of mouse RefSeq transcripts . We used NCBI’s DAVID software [59] to perform GO analysis . To summarize and remove the similar GO terms , we used REVIGO [60] with the similarity parameter set to medium . We downloaded RefSeq gene coordinates for mouse mm9 assembly from UCSC Genome Browser . Only protein coding genes were considered in our analysis i . e . we kept only transcripts with tag NM . We also filtered out genes that were shorter than 2kb . For each cluster of overlapping genes , we only kept the gene producing the longest transcript and disregard the rest . After these filtering steps , we obtained 18 , 211 protein coding transcripts . For each gene , we defined its promoter region as a genome segment from 100 bp upstream to 500 bp downstream of its transcription start site . The body region of a gene starts from 1kb downstream of its transcription start site to its transcription termination site . For each gene , the promoter density of Pol II was calculated as the number of reads aligned to its promoter region normalized to reads per kilobase per million reads mapped ( RPKM ) . Similarly , the gene body density of Pol II was computed using RPKM values . The poising index was computed as the ratio between promoter density and gene body density [59] . In our analysis , we were interested in genes with significant Pol II binding in their promoters . These Pol II+ promoter genes have a significantly higher number of Pol II reads than the average number of IgG reads in the promoter region . Given that the total number of reads is large and the lengths of the promoter/body regions are small compared to the length of the genome , we assume that the number of reads in a segment in the IgG control experiment follows a Poisson distribution . For each gene , we evaluated the null hypothesis that its Pol II promoter density is equal to average IgG promoter density using a one sample Poisson test , with p-values from multiple tests being adjusted using Benjamini and Hochberg's procedure [61] . If p-value< 0 . 01 , the gene has Pol II+ promoter; otherwise it has Pol II- promoter . Among Pol II+ promoter genes , we further checked whether they have poised Pol II in their promoters by assessing whether the Pol II promoter density is significantly greater than gene body density . First , we estimated the average IgG promoter and gene body density from the control experiments . Similarly , we calculated the average Pol II promoter and gene body density . For each gene , we formed 2x2 contingency with the rows corresponding to the Pol II and IgG experiments and the columns corresponding to promoter and gene body density . Using Fisher's exact test as in [45] , we then assessed the null hypothesis that the Pol II in the promoter and gene body is equal . The p-values were corrected using Benjamini and Hochberg's procedure [61] with the significance threshold of 0 . 001 . Non-B DNA forming sequences were identified within 3 . 0 kb region upstream of TSS or within 500bp region downstream of TSS . Regions with propensity to form quadruplex were predicted using QuadParser program [49] with at least 3 G bases in each of four runs of G repeat and gap size between 1 and 7 nucleotides; both strands were searched . Overlapping regions were merged into a single region . For each gene , we computed the numbers of quadruplex forming regions; and for each gene group as in Fig 7 and S2 Fig , we calculated median and median absolute deviation of these numbers . Statistical differences between groups were computed using Mann-Whitney-Wilcoxon test .
|
Accumulation of RNA polymerase II ( Pol II ) in the promoter proximal region has been proposed to be important for rapid gene activation , but the full regulatory role of Pol II dynamics is yet to be delineated . Defining polymerase poising as a significant enrichment of accumulation of Pol II near the promoter , and comparing Pol II profiles in resting and activated B cells , we found that Pol II poised genes generally overlap functionally among different B cell states and correspond to the functional groups identified for poised genes in other cell types . In contrast , non-poised genes are B cell state specific . Interestingly , the transition from poised to non-poised state was not associated exclusively with B cell activation . We found quite a few genes expressed in resting B cells that transitioned from non-poised to poised state after cell activation including several Immediate Early Genes ( IEGs ) —genes which are known to be activated transiently and rapidly in response to a wide variety of cellular stimuli . In addition , we observed an enrichment of G4 sequence motifs upstream of TSS of non-poised but active genes , including IEG genes . We propose that formation of quadruplexes upstream of TSS can facilitate Pol II engagement by destabilizing the DNA duplex . Taken together , our analysis of resting and activated B cells allowed us to provide novel insight into the dynamics of Pol II poising .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"blood",
"cells",
"sequencing",
"techniques",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"gene",
"regulation",
"metabolic",
"processes",
"immediate",
"early",
"genes",
"immunology",
"dna",
"transcription",
"gene",
"types",
"sequence",
"motif",
"analysis",
"molecular",
"biology",
"techniques",
"research",
"and",
"analysis",
"methods",
"sequence",
"analysis",
"white",
"blood",
"cells",
"animal",
"cells",
"gene",
"expression",
"molecular",
"biology",
"antibody-producing",
"cells",
"biochemistry",
"cell",
"biology",
"b",
"cells",
"protein",
"translation",
"genetics",
"biology",
"and",
"life",
"sciences",
"cellular",
"types",
"metabolism"
] |
2016
|
Ups and Downs of Poised RNA Polymerase II in B-Cells
|
A novel microduplication syndrome involving various-sized contiguous duplications in 17p13 . 3 has recently been described , suggesting that increased copy number of genes in 17p13 . 3 , particularly PAFAH1B1 , is associated with clinical features including facial dysmorphism , developmental delay , and autism spectrum disorder . We have previously shown that patient-derived cell lines from individuals with haploinsufficiency of RPA1 , a gene within 17p13 . 3 , exhibit an impaired ATR-dependent DNA damage response ( DDR ) . Here , we show that cell lines from patients with duplications specifically incorporating RPA1 exhibit a different although characteristic spectrum of DDR defects including abnormal S phase distribution , attenuated DNA double strand break ( DSB ) -induced RAD51 chromatin retention , elevated genomic instability , and increased sensitivity to DNA damaging agents . Using controlled conditional over-expression of RPA1 in a human model cell system , we also see attenuated DSB-induced RAD51 chromatin retention . Furthermore , we find that transient over-expression of RPA1 can impact on homologous recombination ( HR ) pathways following DSB formation , favouring engagement in aberrant forms of recombination and repair . Our data identifies unanticipated defects in the DDR associated with duplications in 17p13 . 3 in humans involving modest RPA1 over-expression .
Variously sized contiguous deletions within 17p13 . 3-pter are associated with complex clinical features in humans including structural brain abnormalities ( lissencephaly , agyria , microcephaly ) , growth retardation and developmental delay [1] . Multiple pathogenomic studies have identified haploinsufficiency of genes including PAFAH1B1 ( LIS1 ) and YWHAE ( 14-3-3ε ) as being particularly relevant in this context [2]–[5] . Previously , we have shown that patients with haploinsufficiency of RPA1 exhibit defective ATR-dependent DDR including failure of the G2-M cell cycle checkpoint suggesting RPA1 is sensitive to copy number variation [6] . Defective ATR-dependent G2-M arrest is associated with human conditions characterised by severe microcephaly ( e . g . Seckel syndrome , Microcephalic primordial dwarfism type II , MCPH1-dependent Primary microcephaly , Nijmegen breakage syndrome ) [7] . RPA1 ( RPA1: RPA-70KD ) encodes the largest subunit of the Replication Protein A complex , a heterotrimeric complex ( RPA1-2-3: RPA-70KD-RPA-32KD-RPA14KD respectively ) with single stranded DNA binding capability that appears to be involved in multiple DNA transactions . It functions to prevent unregulated nuclease digestion and/or hairpin formation as well as orchestrating the sequential assembly and disassembly of various DNA processing factors during DNA replication , repair and recombination [8]–[10] . With respect to the DDR , the DNA single stranded binding function of RPA1–3 plays a fundamental role in the recruitment of ATR to sites of DNA damage , for example stalled replication forks , via a direct interaction with ATR's binding partner , ATRIP [11] . Furthermore , through interactions with RAD51 and RAD52 , RPA1–3 also plays an essential role in homology directed recombinational repair , likely facilitating RAD51 nucleofilament formation allowing strand invasion and homology searching [12]–[16] . Recently , distinct , variously sized , non-recurrent duplications within 17p13 . 3 have been identified in several individuals defining a novel genomic disorder . In two of these the duplication included RPA1 [17] . Consistent with other genomic disorders , the clinical duplication phenotype appears to be less severe compared to deletions within 17p13 . 3 . Nevertheless , subtle over-expression of ‘normal’ genes within 17p13 . 3 is associated with profound clinical consequences [17]–[19] . Interestingly , over-expression of RPA1 has been implicated in genomic instability in other systems . For example , a quantitative over-expression screen in the budding yeast Saccharomyces cerevisiae found that over-expression of RFA1 , the S . cerevisiae equivalent of mammalian RPA1 , was associated with delayed cell cycle progression through G2-M , impaired chromosomal spindle attachment and activation of the DDR [20] . Furthermore , ectopic over-expression of individual RPA1–3 subunits in the human colorectal carcinoma cell line HCT116 promoted endoreduplication and aneuploidy [21] . Whether RPA1–3 over-expression functionally contributes to any cellular or clinical phenotype associated with genomic disorders has not been investigated . Since we had previously observed specific DDR-defects associated with reduced RPA1 expression in cell lines derived from individuals with variously sized contiguous deletions at 17p13 . 3-pter , we sought to determine if increased levels of RPA1 are associated with identical and/or related DDR-defects [6] . Herein , we show that cell lines derived from patients with 17p13 . 3 duplications that encompass RPA1 exhibit modest RPA1 over-expression , abnormal S phase distribution , attenuated DSB-induced RAD51 chromatin retention and enhanced sensitivity to killing by camptothecin , consistent with compromised homologous recombination ( HR ) . Using various model and reporter systems we demonstrate that subtle over-expression of RPA1 is indeed associated with altered HR-mediated DNA double strand break repair .
Two of the 17p13 . 3 duplication cases recently described by Bi et al involve genomic duplication of RPA1 , amongst other genes [17] . A schematic representation of the various CNVs in this region in several cell lines used in this study is shown in Figure 1A . The cell lines involving RPA1 duplication , BAB2668 and BAB2719 , are shown in red ( Figure 1A ) . We examined RPA1 expression by western blotting following careful titration of whole cell extracts to ascertain the extent of over-expression at the protein level using EBV-transformed lymphoblastoid cells ( LBLs ) from both RPA1-duplication cases , compared to another LBL with a 17p13 . 3 duplication that does not involve RPA1 ( BAB2721 ) , as reported by Bi et al [17] , and to an LBL ( BAB2751; Case 6 [17] ) exhibiting a novel 17p13 . 3 genomic deletion involving haploinsufficiency of RPA1 ( Figure 1B ) . The left-hand panel of Figure 1B shows that RPA1 protein is modestly over-expressed in whole cells extracts from BAB2668 ( Case 4 [17] ) and BAB2719 ( Case7 [17] ) compared to BAB2721 , an LBL from a patient who does not exhibit RPA1 duplication at the genomic level . LBLs from patient BAB2751 ( Del; deleted for one copy of RPA1 ) associated with genomic haploinsufficiency of RPA1 show modestly reduced RPA1 expression at the protein level . Interestingly , modest over-expression of RPA2 was also evident in whole cell extracts from BAB2719 ( Dup ) and BAB2668 ( Dup ) ( Figure 1B middle panels ) . This suggests that the 17p13 . 3 duplications involving RPA1 and resulting in RPA1 over-expression in these cells likely also results in elevated levels of the RPA complex since RPA2 levels appear modestly elevated in these cells ( Figure 1B and 1C ) . Quantification of three separate experiments relative to MCM2 is shown in Figure 1C . Similar data from the other LBLs as described in Figure 1A is shown in Figure S1 . An important role of RPA1–3 in the ATR-dependent DDR is the recruitment of ATR-ATRIP to single stranded DNA ( ssDNA ) generated at the DNA damage site , thereby initiating ATR-dependent signalling [11] , [22] . One aim of this process is the activation of cell cycle checkpoint arrest , particularly at the G2-M transition . Previously , we have shown that Miller-Dieker Syndrome ( MDS ) and severe Isolated Lissencephaly Sequence ( ILS+ ) patient-derived LBLs with RPA1 haploinsufficiency fail to activate the ATR-dependent G2-M checkpoint [6] . Furthermore , we showed that this cellular phenotype was RPA1-dependent since it could be complemented by ectopic expression of RPA1 following transfection [6] . Interestingly , precedent exists whereby over-expression of a DDR-component is actually associated with a functional defect in the DDR [23]–[28] . Nevertheless , we did not observe a defective ATR-dependent G2-M cell cycle checkpoint arrest in LBLs derived from individuals with increased RPA1 levels associated with RPA1 duplication ( Figure 2A ) . This was in contrast to LBLs exhibiting RPA1 haploinsufficiency with reduced RPA1 expression ( BAB2751; Del; deleted for one copy of RPA1 . Figure 2A and [6] ) . Therefore , modest over-expression of RPA1 in the context of 17p13 . 3 duplication is not associated with the same ATR-dependent DDR-defect as that of RPA1 haploinsufficiency . Since ectopic over-expression of RPA1 has previously been shown to induce other forms of genomic instability including aneuploidy , we examined spindle assembly checkpoint ( SAC ) proficiency following prolonged exposure to the spindle poison nocodazole in RPA1-duplicated patient-derived LBLs [21] . Following 24 hrs treatment with 1 . 5 µM nocodazole , cells with a functional SAC exhibit an increased 4N population without any progression to >4N , as demonstrated by the propidium iodide staining flow cytometry profiles shown in Figure 2B ( Unt; untreated . Noc; nocodazole treated ) . Quantification of the 4N population , with or without 24 hrs treatment with nocodazole , demonstrates that BAB2752 ( WT; wild-type RPA1 copy number ) , BAB2668 ( Dup; RPA1 duplication ) and BAB2719 ( Dup ) all exhibit a similar arrest at 4N following nocodazole ( Figure 2C ) . No increase in >4N was seen in either of the RPA1-duplicaiton containing LBLs , BAB2668 or BAB2719 ( Figure 2B ) . Hence , we observed a normal nocodazole-induced arrest at mitosis with 4N DNA content suggestive of a proficient SAC in this context . Ectopic over-expression of RPA1 is associated with endoreduplication in certain cell lines [21] . Since RPA1–3 complex is a fundamental component of normal DNA replication we examined S phase in one of our patient-derived LBLs with RPA1 duplication ( BAB2668 ) using bromodeoxyuridine ( BrdU ) pulse-labelling-coupled two-dimensional flow cytometry ( Figure 3A ) . No evidence for spontaneous endoreduplication was found ( data not shown ) . Whilst we did not observe a difference in the overall amount of BrdU incorporated between patient-derived LBL BAB2668 ( Dup; with RPA1 duplication ) compared to those with normal ( BAB2752;WT ) or haploinsufficient RPA1 copy number ( BAB2751;Del ) ( Figure 3A left hand graph ) , the distribution or pattern of BrdU labelling was specifically and reproducibly altered in BAB2668 LBLs with RPA1 duplication ( Figure 3A middle flow cytometry panels ) . The BrdU positive cells within the boxed area represent those that have DNA content between 2N and 4N ( mid S phase ) but have not incorporated BrdU efficiently . These cells ( i . e . mid S phase yet low BrdU incorporation ) are approximately 3–4 fold more abundant in BAB2668 ( Dup; RPA1 duplicated ) compared to BAB2752 ( WT; RPA1 copy no ) or BAB2751 ( Del; haploinsufficient RPA1 ) ( Figure 3A right hand graph ) . This is suggestive of a stochastic problem in S phase progression or DNA replication in unperturbed asynchronously growing LBLs with RPA1 duplication . We next examined the ability of our patient-derived LBLs to recover DNA replication following prolonged treatment with the DNA polymerase inhibitor aphidicolin ( APH ) . APH treatment for 24 hrs efficiently reduced total BrdU incorporation as expected for all cell lines ( Figure 3B upper panels ) . When this APH-induced DNA replication block was removed we found that RPA1-duplication associated LBL BAB2668 ( Dup ) failed to progress as efficiently as the control LBL BAB2752 ( WT ) through S phase , as judged by the distribution of BrdU positive cells in late S phase as indicated by the boxed area in Figure 3B ( lower panels and graph ) . This is consistent with a constitutional problem in the ability to efficiently complete DNA replication , in this case following recovery from replicative stress , in these patient-derived cells . Interestingly , RPA1 haploinsufficiency ( BAB2751 ) conferred a similar phenotype ( Figure 3B ) . Since the RPA1–3 complex is also an important functional component of HR , we sought to examine HR in our patient-derived LBLs . Furthermore , defective HR has previously been shown to result in impaired S phase progression , a phenotype suggested by our BrdU incorporation data ( Figure 3 ) [29] , [30] . Following IR treatment , we found modestly increased chromatin binding of RPA1 , RPA2 and RAD51 in LBLs with normal RPA1 copy number ( BAB2705; WT ) , in contrast to chromatin extracts from RPA1-duplicated BAB2668 LBLs ( Dup ) ( Figure 4A and 4B ) . In fact , BAB2668 ( Dup ) exhibited increased endogenous levels of chromatin bound RPA1 , RPA2 and RAD51 , even in undamaged cells , in contrast to the WT LBLs , but this level did not change following IR ( Figure 4A and 4B ) . Protein quantifications standardised to histone H2B loading and normalised to the un-irradiated BAB2705 ( WT ) for each of RPA1 , RPA2 and RAD51 from three separate experiments are shown in Figure 4B . The attenuated IR-induced RAD51 chromatin retention observed in BAB2668 ( Dup ) indicates a potential problem in the ability to induce RAD51-dependent HR in these cells following IR-induced DSB formation . Since the patient-derived LBLs with increased RPA1 copy number are also duplicated for other genes , it was important to examine whether indeed RPA1 is the protein conferring this phenotype or whether it is in fact the consequence of combined increased copy number of several genes . To demonstrate that this cellular phenotype was specifically associated with RPA1 over-expression , we constructed a conditional , isopropyl β-D-1-thiogalactopyranoside ( IPTG ) -inducible-RPA1 model system in the human glioblastoma line T98G based on the pTUNE vector ( Origene ) . Interestingly , we found that transient ectopic over-expression of RPA1 from a high expression level CMV-promoter-containing pcDNA3 . 1 mammalian expression vector was consistently associated with overt toxicity ( associated with detectable activated caspase 3; data not shown ) in multiple commonly employed human tumour lines ( e . g . HeLa , MG63 , A549 ) suggesting that strong over-expression of RPA1 is not tolerated . We found significant leakiness associated with the T98G-RPA1 system ( pTUNE-RPA1 ) as RPA1 and RPA2 levels appeared increased even in the absence of IPTG ( Figure 4C ) . Nevertheless , IPTG treatment did induce a modest elevated expression of RPA1 in this context ( Figure 4C Unt; untreated . IPTG; IPTG treated ) . Interestingly , this was also associated with elevated RPA2 expression ( Figure 4C ) , similar to what was found in the patient-derived LBLs ( Figure 1B and 1C ) . We found that the modest IPTG-induced over-expression of RPA1 was associated with attenuated IR-induced chromatin retention of both RPA1 and RAD51 ( Figure 4D ) . Attenuated IR-induced RAD51 chromatin retention is also a feature of the patient-derived LBLs ( Figure 4A and 4B ) . Furthermore , and similar to LBLs exhibiting RPA1 duplication ( BAB2668; Figure 4A ) , we observed more chromatin associated RPA1 even in undamaged cells following induction with IPTG ( Figure 4D ) . Collectively these data suggest that modest over-expression of RPA1 is associated with attenuated RAD51 chromatin recruitment following IR treatment and that this cellular phenotype is observed in LBLs from patients with duplications in 17p13 . 3 involving RPA1 ( Figure 4A ) . To gain direct , independent insight into the consequences of subtle RPA1 over-expression on DSB-induced HR , we exploited I-Sce I restriction enzyme-induced HR using the established model DRneo reporter system in Chinese Hamster Ovary ( CHO ) cells following transient over-expression of human native untagged RPA1 . Unlike other commonly used model HR reporter systems ( e . g . DR-GFP ) , we opted to use this set-up specifically because the DRneo system enables the collective assessment of alternative forms of recombination alongside gene conversion ( GC ) , such as single strand annealing ( SSA ) ( Figure 5A ) [31]–[33] . This is a heterologous system , although similar approaches have been used successfully before to study HR-mediated DSB-repair ( [30] and references therein ) . Unfortunately , since the available RPA1 antibodies fail to cross-react with hamster RPA1 , it was difficult to determine the precise extent of RPA1 over-expression following transient transfection . Nevertheless , transient expression of the human protein under these conditions appeared only modestly greater than endogenous RPA1 expression in T98G cells ( Figure 5B ) . Interestingly , RPA1 expression in this context did not appear to grossly affect GC , although the limitations of such as heterologous system should be kept in mind ( Figure 5C; white bars ) . Nevertheless , a reduced expression-induced RPA-dependent phenotype has been shown to give a different outcome using a similar system arguing against a simple dominant-negative effect here [16] . However , unexpectedly , we observed an approximately 2-fold increase in total levels of HR ( i . e . all forms of GC+SSA; black bars ) following I-Sce I-induced DSB formation ( Figure 5C and 5D ) . This implies that RPA1 over-expression following I-Sce I-induced DSB results in increased forms of recombination such as SSA and/or GC with crossing-over which can be regarded as an aberrant or less favourable forms of recombination since they are associated with loss of genetic material [29] . Interestingly , RPA has recently been shown to be required for SSA in Xenopus [34] . Furthermore , increased RAD51 expression is also associated with increased genomic instability using a similar HR reporter system suggesting that over-expression of functional components of repair pathways likely to be involved in repairing such DSB's can adversely affect repair [35] . These data are consistent with attenuated IR-induced RAD51 chromatin recruitment observed in the RPA1-duplication associated patient-derived LBLs ( Figure 4A ) and the T98G-RPA1 system ( pTUNE-RPA1; Figure 4D ) . Collectively , these results suggest that modest increased expression of RPA1 can influence HR sub-pathway choice . Our findings with the model HR reporter system suggest that increased RPA expression was associated with increased recombination leading to the hypothesis that increased RPA expression could be associated with increased genome instability . Similarly , since RAD51 over-expression can also induce significant genomic instability [23]–[28] , we examined mitotic spreads of our RPA1-over-expression model cell line ( pTUNE-RPA1 ) for evidence of elevated genomic instability . Strikingly , significant levels of chromosome aberrations , fusions/derivatised chromosomes in particular , were observed in these cells ( Figure 6A and 6B ) . Such abnormalities would be consistent with aberrant cross-over and/or ligation events . These aberrations were seen even without induction with IPTG which is a further indication of some inherent leakiness in this system and is consistent with elevated chromatin bound RPA1 seen here ( Figure 4D ) . Despite the limitations of this artificial cell system , these data do demonstrate that subtle over-expression of RPA1 can induce significant levels of genomic instability , specifically elevated levels of derivatised chromosomes . To examine whether the patient-derived LBLs with RPA1 over-expression exhibited a similar phenotype we examined mitotic spreads for chromosomal abnormalities in LBLs from BAB2719 ( Dup; RPA1 duplication ) compared to BAB2752 ( WT; wild-type normal RPA1 copy number ) . We also observed elevated levels of chromosomal aberrations , specifically an over-representation of chromosomal fusions , in LBLs associated with RPA1 over-expression ( Figure 7A and 7B ) . These aberrations were increased following IR-treatment further suggestive of an inability of these cells to properly repair DSBs ( Figure 7B ) . Defective and/or aberrant DNA damage-induced HR is also associated with DNA damaging-induced genomic instability . Consistent with this we found increased levels of hydroxyurea ( HU ) -induced micronuclei ( MN ) formation in LBLs associated with RPA1-duplication ( Dup; BAB2668 , BAB2719 ) indicative of increased DNA breakage following replication fork stalling in these cells ( Figure 7C ) . Furthermore , compromised HR is also specifically associated with sensitivity to killing by topoisomerase I inhibitors such as camptothecin ( CPT ) [36] . Consistent with an underlying problem with HR associated with RPA1-duplication in our patient-derived LBLs , we also found that these lines were sensitive to apoptosis induction following CPT treatment , as judged by increased levels of sub-G1 cells by propidium iodide flow cytometry ( Figure 7D ) . Interestingly , for both of these cellular phenotypes we observed a similar response in BAB2751 ( RPA1 haploinsufficient ) cells to that of lines over-expressing RPA1 ( Dup; BAB2668 and BAB2719 ) . This suggests that manipulation of RPA1 levels ( increase or decrease ) results in increased genomic instability following DNA damage . In summary , we found that duplications involving RPA1 are associated with modest over-expression of RPA1 and also RPA2 at the protein level , impaired S phase distribution and spontaneously elevated levels of chromatin bound RPA1 , RPA2 and RAD51 , along with attenuated IR-induced RAD51 chromatin retention following DSB's suggestive of compromised HR . Using the DRneo model HR-reporter system we observed a hyper-recombinogenic phenotype consistent with a shift towards a less genomically preferable form of HR following modest RPA1 over-expression . We also found increased levels of complex rearrangements especially after DSB-induction in patient derived LBLs with RPA1 duplication . Furthermore , these patient derived cells exhibit other evidence of underlying problems in the DDR such as sensitivity to CPT and elevated HU-induced micronuclei formation .
Variously sized contiguous gene deletions at 17p13 . 3 are associated with severe neurodevelopmental phenotypes including microcephaly and neuronal migration deficits [1] . Recently , duplications within 17p13 . 3 have been identified in several patients exhibiting a milder though distinct phenotype that also incorporates aspects of autism spectrum disorder [17]–[19] . Much attention has focused on characterising the consequence of CNV of PAFAH1B1/LIS1 in this respect [17]–[19] , [37] . Previously , we have shown that LBLs from some ILS+ individuals and from MDS patients , all of whom exhibit haploinsufficiency of RPA1 , a gene telomeric to PAFAH1B1/LIS1 , exhibit impaired ATR-dependent DDR [6] . Here , we find that for the reciprocal situation , that is , in LBLs from patients associated with duplication of RPA1 , we observed a distinct DDR abnormality impacting upon HR . The RPA1–3 complex is a fundamental functional component of many DNA processes involving the generation of single stranded DNA [8] . RPA1–3 complex is essential for several DNA repair pathways ( e . g nucleotide excision repair , mismatch repair , base excision repair ) , for DNA replication and recombination events [12]–[16] . Therefore , a plausible assumption would be that a significant reduction in RPA expression/function results in embryonic lethality . Attempts to create knockout mice for RPA1 have not been reported . Nevertheless , mice bearing a semi-dominant heterozygous mis-sense mutation in Rpa1 ( Rpa1L230P ) exhibit gross genomic rearrangements and are highly cancer prone ( Rpa1L230P homozygosity is cell lethal ) [38] . Hence , precedent exists for altered RPA1 , and likely , consequently RPA complex function , impacting on genomic stability at the organismal level . Furthermore , forced over-expression of RPA1 can cause genomic instability , at least in cancer cell lines [21] . There are several instances whereby over-expression of various DDR and/or cell cycle components disrupts or adversely affects the fundamental cellular processes/pathways in which they are functional components . For example , over-expression of CDC25A phosphatase is thought to be an important contributor to uncontrolled cell cycle progression from G2 into M frequently observed in certain malignancies [24] , [25] . Over-expression of RAD51 and RAD52 has been found to reduce DSB-induced HR in mammalian cells [28] . Indeed , over-expression of separase or the SAC component MAD2 results in aneuploidy and malignancy in mice , consistent with defective SAC activity [26] , [27] . Our findings suggest that a modest over-expression of RPA1 in LBLs derived from individuals with duplications in 17p13 . 3 involving RPA1 results in an abnormal distribution of cells in S phase , adversely impacts on HR and is associated with elevated chromosomal instability and sensitivity to DNA damaging agents . RPA1 is thought to be of particular importance for RPA heterotrimeric function since it can bind DNA independently of the other subunits and contains the greatest surface area available to mediate protein-protein interaction [39] . Interestingly , we found that RPA2 also appeared to be over-expressed in our patient-derived LBLs associated with RPA1 duplication , potentially suggesting elevated levels of RPA complex in this context . This could have adverse implications for coordinating subsequent DNA processing pathways . For example , during HR , a ‘handover’ between RPA1–3 complex coated ssDNA and RAD51 must occur to allow RAD51 nucleofilament formation for strand invasion . As RPA1 can bind RAD51 directly , an excess of chromatin bound RPA complex could interfere with the timing , coordination and/or efficiency of this ‘handover’ ( Figure 8 ) . A direct consequence of this could be either uncontrolled elevated or reduced overall HR capacity and/or a preference for other forms of recombinational repair , aside from gene conversion ( GC ) . Data from our patient-derived LBLs show spontaneously elevated RAD51 on chromatin but attenuated IR-induced recruitment . Furthermore , our data generated following transient RPA1 over-expression in the DRneo system indicates a hyper-recombinogenic phenotype ( i . e . total HR increases whilst levels of GC without crossing-over remain fairly constant ) . Interestingly , RPA has recently been shown to be required for SSA , at least in Xenopus [34] . One possible interpretation of the DRneo system-derived data is a shift towards an elevated level of SSA and/or GC with crossing-over , both of which involved loss of genetic material . The elevated levels of derivatised chromosomes observed in mitotic spreads from the RPA1-duplication associated LBLs and in the pTUNE-RPA1 system cells are consistent with aberrant cross-over and/or ligation events . A hyper-recombinogenic phenotype can have serious consequences for genome stability . For example , elevated ‘mutagenic’ HR has been implicated as pathophysiological contributor to disease progression in haematological malignancies such as Chronic Myelogenous Leukaemia and Multiple Myeloma [40]–[42] . The complex clinical spectrum of 17p13 . 3 microduplication syndrome is undoubtedly a consequence of the combined increased copy number of several genes within 17p13 . 3 , although some candidates may have greater impacts than others [17]–[19] . The specific pathological connection between increased RPA1 expression and the clinical features of those respective patients is unclear . Nevertheless , the cellular phenotypes described here , including impaired S phase and suboptimal HR , could adversely influence apparently unrelated biological pathways by affecting gene expression . For example , components of the DNA replication machinery have been shown to influence epigenetic control of gene silencing [43] . Furthermore , suboptimal/aberrant HR could conceivably ultimately alter the genomic architecture resulting in unanticipated cis and/trans effects on the expression of other genes . Interestingly , RPA1 has been implicated in such ‘allelic phasing’ together with TP53 with respect to carcinogenesis [44] . Congenitally elevated genomic instability is often associated with cancer predisposition , although this has not been noted in either 17p13 . 3 duplication syndrome patients associated with RPA1-duplication [17] . Obviously there are too few patients to make any definitive conclusions , although the cellular defects presented here may warrant consideration in this respect . Clearly , further work is required to untangle the clinical consequences of increased RPA1 expression . In conclusion , we have found that LBLs derived from patients with duplications in 17p13 . 3 specifically incorporating RPA1 exhibit a modest over-expression of RPA1 and RPA2 which is associated with attenuated S phase transit , attenuated IR-induced RAD51 chromatin recruitment , elevated chromosomal instability , increased HU-induced MN formation and sensitivity to killing by CPT . All of these phenotypes are consistent with an inefficient HR pathway . Furthermore , using various model cell systems we showed that modest conditional over-expression of RPA1 alone impacts on IR-induced RAD51 chromatin retention and I-SceI-induced HR in a reporter construct , the latter phenotype indicative of a hyper-recombinogenic shift towards alternative forms of recombination coincident with elevated chromosomal fusions . Collectively , our findings highlight a novel association between impaired DDR and CNV resulting in copy number gain of RPA1 .
EBV-transformed patient-derived lymphoblastoid cell lines ( LBLs ) were cultured in RPMI with 15% FCS , L-Gln and antibiotics ( Pen-Strep ) at 5% CO2 . T98G glioblastoma cells were maintained in MEM supplemented with 10% FCS , pyruvate and non-essential amino acids . Chinese Hamster Ovary cell lines ( CHOs ) were cultured in 10% DMEM , L-Gln and antibiotics ( Pen-Strep ) at 5% CO2 . Urea extraction: Cells were lysed in 150 µl urea buffer ( 9 M urea , 50 mM Tris-HCl at pH 7 . 5 and 10 mM 2-mercaptoethanol ) , followed by 15 s sonication , 30% amplitude using a micro-tip ( SIGMA-Aldrich ) . The supernatant was quantified by Bradford Assay . Detergent lysis: Cell pellets were incubated for 1 hr on ice in buffer containing 50 mM Tris-HCl pH 7 . 5 , 150 mM NaCl , 2 mM EDTA , 2 mM EGTA , 25 mM NaF , 25 mM β-glycerolphosphate , 0 . 1 mM Na-orthovanadate , 0 . 2% Triton X-100 , 0 . 3% IGEPAL and protease inhibitor cocktail tablets as indicated by manufacturer ( Roche ) . The supernatant was quantified by Bradford Assay . Cells were harvested 24 hr after 10 Gy gamma irradiation . Gamma irradiation was performed using a 137Cs γ-ray source at a dose rate of 8 Gy/min . Cells were lysed in detergent lysis buffer ( above ) for 1 hr on ice followed by 15 min in high-salt IP buffer ( an extra 500 mM NaCl added to the regular IP buffer ) . The cell pellet was re-suspended in urea buffer ( see WCE above ) and sonicated for 15 s . The supernatant was quantified by Bradford Assay . Western blots were developed using ECL ( Pierce ) in a luminescent image analyser , Image Quant LAS 4000 ( GE Healthcare ) . This analyser ensures all bands are in the linear range ( during the developing any saturated bands are highlighted so that the exposure can be decreased ) . Image Quant TL 7 . 01 quantification software was used to quantify the band intensities . Alternatively , following ECL , Western blots were developed using film and the scanned images quantified with Image J software . Custom assembled pTUNE-RPA1 was obtained from Origene and stable T98G clones were obtained following transfection with MetafectenePro ( Biontex Laboratories GmbH ) and selection in G418 ( 1 mg/ml ) . For inductions , cells were treated with 500 µM IPTG for 3 hrs . Anti-RPA1 ( Ab-1 #NA13 ) and anti-RPA2 ( Ab-2 #NA18 ) antibodies were from Calbiochem . Anti-RAD51 ( H-92 ) was from Santa Cruz . Anti-H3 was from Cell Signaling and anti-H2B was from Millipore . Anti-BrdU-FITC conjugated antibody ( 347583 ) was from Becton Dickinson . UV irradiation was carried out using a UV-C source ( 0 . 6 J/m2/s ) . Cells were irradiated with 5 J/m2 UV-C in PBS and immediately seeded into complete medium supplemented with 1 . 5 µM nocodazole for 24 hr . Cells were pelleted , swollen with 75 mM KCl for 10 min before fixing in Carnoy's solution ( methanol: glacial acetic acid 3∶1 ) , before counterstaining with 4′-6-Diamidino-2-phenylindole ( DAPI ) . Cells were Cytospun ( Shandon ) onto poly-L-lysine coated slides and mounted with Vectashield ( Vector Labs ) . Slides were scored using a Zeiss AxioPlan microscope . Exponentially growing LBLs were treated with 1 . 5 µM nocodazole for 24 hrs then fixed in 70% ice cold ethanol prior to propidium iodide staining and analysis by flow cytometry . LBLs were fixed in ice-cold 70% ethanol for 24 h and re-suspended in PBS containing 0 . 5% Tween-20 , 10 µg/ml propidium iodide and 500 µg/ml RNase A . Data were collected using a Becton Dickinson FACS Calibur machine and were analysed with CellQuest software . For BrdU incorporation cells were labelled with 50 µM BrdU for 15 min . Incorporated BrdU was detected using FITC-conjugated anti-BrdU antibody ( Becton-Dickson ) . ERCC1 . 17 DRneo CHO cells were grown in 10% DMEM supplemented with 3 µg/ml Blasticidin-S and 0 . 05 mM hygromycin B [33] . Assay: 5×105 cells were seeded in 6 cm plates . Next day cells were co-transfected with 2 µg RPA1 and 2 µg I-SceI ( or CMV control ) using MetafectenePro , according to the manufacturer's protocol . 24 hr after transfection cells were put into selection . For the recombination frequencies , 5×104 cells per 10 cm plate were seeded with 1 mg/ml G418 and/or 0 . 5 mg/ml hygromycin-B . 103 cells were seeded to determine the cloning efficiency . Plates were incubated for 7 days , after which they were stained with methylene blue . Exponentially growing LBLs were treated with 10 µM camptothecin ( CPT ) and incubated for 72 hrs , fixed ( ice cold 70% ethanol ) , stained with propidium iodide and sub-G1 cells quantified by flow cytometry . Cells were treated with 1 mM HU for 4 hrs before incubation for 24 hrs in 5 µg/ml cytochalasin B for bi-nucelate formation . Micronuclei were scored in bi-nucleated cells by immunofluorescence microscopy ( Zeiss AxioPlan ) following swelling in KCl ( 75 mM 10 mins ) , fixation ( Carnoy's fix; 3∶1 methanol∶acetic acid . 10 mins ) and staining with DAPI and acridine orange ( 2 µg/ml ) . The pTUNE-RPA1 T98G cells , and wild type T98G cells , were induced with 500 µM IPTG for 3 hr before adding 0 . 2 µg/ml colcemid for 4 hr prior to harvesting . Cells were swollen ( 75 mM KCl 10 mins ) and then fixed in Carnoy's fixative ( 10 mins ) prior to being dropped onto slides from approx 50 cm above . The slides were air dried and Giemsa stained according to the manufacturer's ( Sigma ) protocol . Images were captured on a Zeiss AxioPlan microscope . Chromosomes spreads were scored blinded according to the following criteria; fusions between different chromosomes , breaks , branched structures and ‘other’ ( a terminal fusion within a single chromosome ) . The results were represented as aberrations per 100 chromosomes , rather than per metaphase due to the aneuploid nature of T98G . LBLs were treated with 2 Gy ionising radiation ( IR ) and allowed to recover for 24 hrs . The IR treated ( IR ) and untreated control ( Unt ) cells were treated with 0 . 2 µg/ml colcemid for 4 hr prior to harvesting . Cells were swollen ( 75 mM KCl 10 mins ) , fixed ( Carnoy's ) , Giemsa stained and analysed as above .
|
The widespread use of genomic array technology has lead to the identification of a plethora of novel human genomic disorders . These complex conditions occur as a consequence of structural genomic alterations ( deletions , amplifications , complex rearrangements ) . Understanding the specific consequences of such alterations on gene expression and unanticipated impacts on biochemical pathways represents an important challenge to help untangle the clinical basis of these conditions and ultimately aid in their management . Here , we demonstrate that individuals with specific duplications of 17p13 . 3 incorporating RPA1 exhibit modest over-expression of RPA1 . Unexpectedly , this is associated with elevated levels of genomic instability and sensitivity to DNA damage . RPA1 is a component of the Replication Protein A heterotrimer , a complex that plays fundamental roles in DNA replication , repair , and recombination . Reduced RPA1 levels are associated with impaired DNA damage checkpoint activation , but the cellular impacts of over-expression of this subunit have not previously been described in the context of a genomic disorder . Using model cell and reporter systems , we show that modestly elevated levels of RPA1 can adversely impact on DNA double-strand break–induced homologous recombination resulting in elevated levels of chromosome fusions . This data highlights an unanticipated consequence of copy number variation on genomic stability .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biochemistry",
"nucleic",
"acids",
"dna",
"biology",
"dna",
"repair"
] |
2011
|
Increased RPA1 Gene Dosage Affects Genomic Stability Potentially Contributing to 17p13.3 Duplication Syndrome
|
The major histocompatibility complex ( MHC ) of immunity genes has been reported to influence mate choice in vertebrates , and a recent study presented genetic evidence for this effect in humans . Specifically , greater dissimilarity at the MHC locus was reported for European-American mates ( parents in HapMap Phase 2 trios ) than for non-mates . Here we show that the results depend on a few extreme data points , are not robust to conservative changes in the analysis procedure , and cannot be reproduced in an equivalent but independent set of European-American mates . Although some evidence suggests an avoidance of extreme MHC similarity between mates , rather than a preference for dissimilarity , limited sample sizes preclude a rigorous investigation . In summary , fine-scale molecular-genetic data do not conclusively support the hypothesis that mate selection in humans is influenced by the MHC locus .
The MHC locus contains genes central to acquired immunity , as well as numerous olfactory receptors [1] . It is reported to influence mate selection in a number of vertebrates , and is thought to act through the sense of smell to favor genetic dissimilarity between parents and thus heterozygosity in offspring [2] . Evidence for these effects in humans includes the high degree of MHC polymorphism [1] , MHC-dependent female sexual interest [3] and preferences for male body odors [4]–[6] , and a depletion of matching five-locus HLA haplotypes in Hutterite couples [7] . Among relevant experiments in model organisms , MHC class I peptides were shown to induce pregnancy blocking in mice [8] . However , this block required the vomeronasal organ , which is not known to function in humans , and human pheromones have not been clearly identified [9] . Despite the advent of MHC-based matchmaking services [10] , not all of the available evidence consistently supports a preference for MHC-dissimilar mates in humans [11] . Although women indicated a preference for the odors of MHC-dissimilar men when knowingly [4] or unknowingly [5] rating potential partners , women who were single [6] or taking oral contraceptives [4] , [6] preferred the odors of MHC-similar men . Women also ranked the faces of MHC-similar men as more attractive [12] . In addition , the relationship of odor preference in these controlled settings to the selection of actual mates in practice is unclear . HapMap genotypes of father-mother-child trios afford the chance to test for an association of MHC similarity with bona fide mate selection rather than a stated preference . A recent study by Chaix et al . sought signs of mate selection in the genotypes of HapMap Phase 2 ( Hap2 ) parents by comparing the genetic relatedness of mated and unmated opposite-sex couples , both at the MHC locus and overall ( using all common autosomal variants ) [13] . Yoruban mates ( N = 27 couples ) were reported to be slightly more similar than expected ( nominal two-sided P<0 . 001 ) overall , but no significant difference in MHC relatedness was detected between mates and non-mates . By contrast , European-American mates ( N = 28 couples ) did not differ significantly from non-mates in autosomal relatedness , but were less similar at the MHC locus than were non-mates ( P = 0 . 015 ) . The latter result was interpreted as supporting a role for the MHC locus in mate choice , and outliers were excluded as an explanation [13] . However , a visual comparison of mate and non-mate pairs ( Figure 1 and Figure 2 ) suggests a weak effect that may derive from a few extreme pairs . Furthermore , adjusting the significance threshold for the fact that multiple hypotheses were tested ( two sets of SNPs in each of two populations ) would have rendered the results insignificant ( see Results ) . The availability of HapMap Phase 3 ( Hap3 ) data now permits a test of these findings using independent samples from the same population .
The findings reported previously in Hap2 couples [13] were replicated as follows . Phased genotypes ( release 21 ) for Yorubans and Europeans were obtained from the HapMap web site ( http://www . hapmap . org ) . There were 30 mated couples in each cohort ( see Text S1 for a description of samples and data sets ) . Only autosomal SNPs were retained . The same couples were excluded as in the previous study to avoid the presence of close relatives ( Text S2 ) , based on independent calculations of relatedness coefficients [14] . In each population , SNPs with a minor allele frequency ( MAF ) <5% ( based on remaining samples ) were excluded . For every pair of individuals a and b , the identity coefficient , Qa , b , was calculated as the fraction of identical alleles , but with two heterozygotes considered 50% identical ( het-het = 50% ) . While Chaix et al . referred to Q as the proportion of identical variants , a comparison of an individual with themselves ( self-self ) using the het-het = 50% score yields values of Q that are variable and lower than unity ( see Supporting Figure 1 in Text S3 ) ; we therefore refer to Q as an identity coefficient . The mean Q for all pairs within the cohort ( after excluding relatives ) , Qmean ( cohort ) , was then derived , and the relatedness coefficient for a and b was calculated as Ra , b = ( Qa , b−Qmean ( cohort ) ) / ( 1−Qmean ( cohort ) ) . The mean R ( Rmean ) for couples , Rmean ( mates ) , was calculated; its significance was then assessed by randomly pairing males and females 1 , 000 times and recording the frequency P with which the absolute value of Rmean from a random trial was equal to or greater than that observed for real couples . Following the previous report , actual couples were allowed in random trials; however , we found that excluding them had little impact on the results ( data not shown; also see description of Z score below ) . Q , R and P values were also calculated based on SNPs at the MHC locus , defined as spanning positions 29 , 700 , 000 to 33 , 300 , 000 on chromosome 6 for NCBI build 35 [13] . Results are shown in Table S1 . The methods were modified in various ways , both to explore the robustness of results to changes in the procedure , and to establish the applicability of the analysis to Hap3 couples; details are provided in Text S3 and results are shown in Table S2 and Table S3 . First , in order to facilitate permutations of excluded couples due to relatives ( see below ) , we calculated MAFs before excluding couples , rather than after , and determined that this change had a negligible effect on results ( Table S2 and Table S3 ) . Second , in addition to P values , the difference in relatedness for mates and non-mates was quantified as a Z score , namely ( Rmean ( mates ) −Rmean ( non-mates ) ) /Rsd ( non-mates ) , where Rmean ( non-mates ) and Rsd ( non-mates ) are the mean and standard deviation , respectively , of the relatedness of non-mate pairs . Because we wished to assess our results using the methods of the previous report [13] , we compared means rather than medians , and based the Z score on standard deviations rather than median deviations . We note that P values are based on random trials in which real mate pairs were allowed , for consistency with the previous report [13] , while Z scores compare mates with the non-mate distribution; however , this distinction has a negligible impact on results ( data not shown ) . We observed inconsistent P values derived from different sets of 1 , 000 random trials ( Table 1 ) . This result was to be expected based on the theoretical standard error , calculated as ( . 05* . 95/1000 ) 1/2 [15] , or 14% of α , for empirical P values near α = 0 . 05 . We therefore increased the number of random trials to 100 , 000 , which reduced the standard error in empirical P values by a factor of 10 . Because the results varied according to which couples were excluded , we opted to analyze all possible permutations and to calculate aggregate R , P and Z values . To enable this analysis , Qmean ( cohort ) and all pairwise R values were first calculated with all samples included in order to detect relatives ( see Text S2 for details ) . Subsequently , for each permutation of excluded couples , Qmean and R values were recalculated based on remaining couples , followed by random trials . Unphased genotypes were obtained for Hap2 ( release 24 ) and Hap3 ( release 2 ) populations with parent-child trios . Mate-pair relatedness analyses were only conducted for Europeans and Yorubans due to insufficient numbers of couples in the other Hap 3 population samples ( N≤16 following exclusions to avoid the presence of close relatives ) . Uncalled alleles , which are corrected in phased genotypes [16] but present in unphased genotypes ( see Text S1 ) , were skipped in all calculations . Because Hap3 SNP positions are reported for NCBI build 36 , we converted the reported MHC coordinates ( specified above ) using the liftOver program and corresponding chain file ( both obtained from the UCSC Genome Bioinformatics web site , http://genome . ucsc . edu ) but found them unchanged . MAFs were based on all available samples , including children and unmated individuals , although this was found to have a negligible impact on the number of remaining SNPs ( not shown ) . The minimum MAF was lowered to 1% to conform more closely to the standard definition of common variant , yielding identity coefficients consistent with those based on MAF ≥5% ( Supporting Figure 1 in Text S3 ) . Effectively , the minimum number of individuals with minor alleles was 2 for Hap2 and 4 for Hap3 . Children and unmated samples were excluded from calculations of mean population-wide identity coefficients ( Qmean ( cohort ) ) , and therefore did not influence comparisons of relatedness ( R ) in mate and non-mate pairs . Concordance of results with phased genotypes was verified for Hap2 ( Table S2 and Table S3 ) . In Hap3 , following the calculation of MAFs , the analysis procedure ( detection of relatives , recalculation of Qmean , etc . ) was conducted separately for the subset of Hap3 samples also present in Hap2 , for all Hap3 samples , and for samples present in Hap3 only . Because the latter subset of samples allows an independent test of findings reported previously in Hap2 , a one-tailed test of significance was conducted in this case for the previously reported hypothesis of interest: 1 ) relatedness is lower than expected for MHC SNPs in European mates , or 2 ) relatedness is higher than expected for autosomal SNPs in Yoruban mates . Chaix et al . calculated the recombination rate and mean relatedness between mates Rmean ( mates ) for 3 . 6 Mbp segments throughout the genome [13] . To replicate this analysis and apply it to non-mates as well as the Phase 3 population , we obtained recombination rates from the HapMap web site ( release 21 for NCBI genome build 35 , and release 22 for NCBI build 36 ) , and the centromere coordinates for the respective genome builds from the UCSC Genome Bioinformatics web site . Following Chaix et al . , the genome was divided into segments of 3 . 6 Mbp tiled every 300 Kbp , segments with fewer than 1000 SNPs or overlapping centromeres were excluded , and Rmean ( mates ) was calculated for each segment based on all SNPs falling within that segment . Additional steps , which may differ from the methods of Chaix et al . , were as follows . First , all mated individuals were included in the analyses but pairs of relatives were not included in the mean population-wide identity coefficient , Qmean ( cohort ) . Second , the recombination rate for each segment ( in cM/Mbp ) was calculated from the difference in recombination ( cM ) between the two data points closest to each end of the segment , with the denominator fixed at 3 . 6 Mbp . Third , we also conducted this analysis for male-female non-mate pairs; in this case , pairs of relatives were excluded from Qmean ( cohort ) and from the mean relatedness for non-mate pairs , Rmean ( non-mates ) . Lastly , segments overlapping the MHC locus were excluded . Chaix et al . tested four specific locus/population combinations ( MHC SNPs in Europeans , MHC SNPs in Yorubans , all autosomal SNPs in Europeans , and all autosomal SNPs in Yorubans ) , reporting that two of these revealed significant phenomena and two did not; however , no adjustment was made for multiple hypothesis testing [13] . Because there is no clear strategy for modeling dependence between these hypotheses , we adopted the Dunn-Šidák method , which treats hypotheses as mutually independent . The case for treating hypotheses related to the two sets of SNPs as independent of one another is strengthened by scatter plots showing that identity at the MHC locus is poorly correlated with identity overall ( Figure 1 ) . Chaix et al . reported a P value of 0 . 015 as significant [13] , thus we used a nominal significance threshold of α = 0 . 05 . For k hypotheses tested , the corrected threshold according to the Dunn-Šidák method [17] is α' = 1- ( 1-α ) 1/k , so that α' = 0 . 0253 for k = 2 and α' = 0 . 0127 for k = 4 . We note that all P values reported in this study are two-sided unless specified otherwise . They are also nominal ( uncorrected for multiple testing ) , so that any correction would further reduce significance .
We successfully replicated each of the specific results of Chaix et al . [13] with phased Hap2 genotypes ( Table S1 ) . However , we found that MHC dissimilarity of European mate pairs was weak ( Z = −0 . 41 ) and was diminished in significance by small changes in the analysis ( Table 1 and Table S2 ) . Significance decreased ( from P = 0 . 014 to P = 0 . 022 ) when the number of random trials was increased ( from 1 , 000 to 100 , 000 ) , and was lost ( P = 0 . 052 ) upon the subsequent exclusion of the single most MHC-dissimilar couple ( Supporting Figure 6 in Text S4 ) . Similarly , the use of median identity and relatedness values instead of means , which should improve robustness to extreme pairs , led to an insignificant result ( P = 0 . 288 ) . In addition , results varied depending on which couples were excluded to avoid the presence of related individuals ( 0 . 019≤P≤0 . 034 ) . Chaix et al . reported results based on an identity measure that assigns a score of 50% to two heterozygous genotypes ( het-het = 50% ) , but noted that results were similar with het-het = 100% [13] . Opting for the latter measure in part because it has the intuitive behavior of yielding self-self coefficients of unity ( see Text S3 ) , we found results to be consistently weaker , and insignificant ( P = 0 . 0546 and P = 0 . 0592 ) in two of the four possible permutations of excluded couples . Subsequent analyses were based on unphased genotypes , SNPs with MAF ≥1% , het-het = 100% , 100 , 000 random trials , and aggregate values of relatedness and significance calculated for all possible ways of excluding samples such that pairs of related invididuals are eliminated ( see Methods ) . In Hap2 couples , these methods produced results in agreement with those reported previously ( Table 1 , Table S2 and Table S3 ) . Our conclusions—that reported mate-pair relatedness effects in Europeans and Yorubans are strongly dependent on extreme pairs—were also confirmed using methods that adhered as closely as possible to the original report ( not shown but see Figure S1 ) . After replicating previously reported results , we sought to test the hypothesis of MHC-dependent mate selection in an independent sample from the same population . Of the 50 couples genotyped in Hap3 , 26 “Hap2∩3” couples were also present in Hap2 while 24 “Hap3-only” couples were unique to Hap3 ( see Supporting Table 1 in Text S1 ) . This allowed us to assess differences between Hap2 and Hap3 data using Hap2∩3 couples , and also to attempt an independent replication of the reported phenomena using Hap3-only couples . First , we verified that Hap2 and Hap3 genotypes were concordant for each Hap2∩3 sample ( Text S3 ) and yielded similar results for MHC dissimilarity ( Table 1 , Table S2 , and Text S5 ) , despite the smaller and partially distinct set of SNPs genotyped in Hap3 . Second , we determined that the Hap2∩3 and Hap3-only samples were drawn from the same population ( Supporting Figure 4 in Text S3 ) . Despite the loss of significance for MHC dissimilarity in the 24 Hap2∩3 couples ( Z = −0 . 34 , P≅0 . 07 ) , the consistency of results obtained with Hap2 and Hap3 genotypes suggested that Hap3-only couples represented a valid test of previously reported findings . The 24 independent European mated Hap3-only couples showed negligible and insignificant dissimilarity at the MHC locus relative to random pairs ( Z = −0 . 08 , one-tailed P = 0 . 351; Figure 1 , Table 1 ) . The absence of significance in Hap3 was corroborated with phased genotypes ( 21 couples; P = 0 . 497 , Z = −0 . 01 ) and further confirmed using the original ( het-het = 50% ) identity score ( not shown ) . A test of the entire Hap3 cohort ( thus including most Hap2 samples ) also yielded an insignificant result ( Z = −0 . 24 , P = 0 . 14; Table 1 ) . Rare instances of very high MHC identity were observed among unrelated non-mate pairs in Hap2 Europeans but not among couples ( Figure 1 and Figure 2 ) , suggesting the possibility of a bias against extremely high MHC similarity in mate pairs ( or their offspring; see Discussion ) , rather than a preference toward dissimilarity . In both the Hap2 and Hap3-only subsets , we observed a possible depletion of high MHC similarity in mate pairs relative to non-mate pairs ( Figure 2 ) . This potential excess of high MHC similarity in European non-mate pairs is not seen in other populations ( Figure S2 ) or in autosomal similarity ( not shown ) . However , sample sizes are too small to permit a rigorous test of this hypothesis . We also re-examined the previous report of excess autosomal relatedness among mated pairs in the Yoruban population [13] . We first replicated previous results in Hap2 couples , and obtained equivalent results using modified methods described above ( Table S1 and Table 2 ) . In addition , based on Hap2 genotypes , the effect remained significant for the 24 Hap2 couples that were also genotyped in Hap3 ( “Hap2∩3”; P = 0 . 01 , Z = 0 . 56; Table 2 ) . However , an examination of these same couples using Hap3 genotypes did not confirm the previous finding ( P = 0 . 13 , Z = 0 . 33 ) , even when only common SNPs were considered ( Table 2 and Table S3; see Text S6 for a discussion of this discrepancy ) . Finally , we examined the 26 independent couples present in Hap3 but not Hap2 ( “Hap3-only” ) and found them to confirm the previous finding of excess similarity among mate pairs ( Table 2 ) . We observed that both the Hap2 and Hap3-only samples contained a small number of mate pairs with unusually high similarity ( Figure 3 ) , suggesting that they may be relatives . The distribution of mate-pair identity coefficients for all Hap3 samples has a shoulder that suggests an underlying mixture of two types of couples ( Figure 3 ) . Thus , an enrichment for autosomal similarity previously reported in Yoruban mate pairs [13] is confirmed , but may be driven by a subset of the couples . In Hap2 Europeans , Chaix et al . found that mean relatedness between mates at the MHC locus was lower than or equal to relatedness between mates at 99 . 6% of similarly-sized genomic segments , or 99 . 9% of segments with the same or lower recombination rate as the MHC locus [13] . Thus , the conclusion that relatedness at the MHC locus was extreme relative to other genomic loci is well supported . However , this observation could be explained by phenomena other than mate selection , e . g . , by elevated positive selection leading to heightened diversity at the MHC locus . This explanation is supported by the high degree of polymorphism observed at the MHC locus [1] . An observation of relatedness between non-mates at the MHC locus that is also systematically lower than other genomic loci would argue against a mate selection explanation . We therefore performed a similar analysis for non-mate pairs , which was not presented in the previous report [13] . We first replicated the reported analysis as closely as possible ( see Methods and Text S7 ) , then applied it to opposite-sex non-mate pairs in Hap2 and to mates and non-mates in Hap3 . In Hap2 Europeans , although we obtained slightly higher numbers of segments than reported with lower relatedness than the MHC locus , we found that the MHC locus is only slightly less extreme in a genome-wide analysis in non-mates than in mates: mean relatedness for non-mates at the MHC locus was lower than or equal to relatedness at 96 . 9% of all segments ( compared with 99 . 0% for mate pairs ) and 94 . 3% of segments with equal or lower recombination rate ( as opposed to 97 . 1% for mate pairs; Supporting Table 4 in Text S7 ) . In the combined Hap3 European population , which showed a slight but insignificant difference in mean MHC relatedness between mates and random couples ( Z = −0 . 24 , P = 0 . 143; Table 1 ) , only 1 . 6% of segments in mates but 91 . 0% in non-mates had lower mean relatedness than the MHC locus . Rather than suggesting that the MHC locus is unique , these results appear to simply reflect the large standard deviations observed for MHC relatedness in both mates and non-mates ( Table 1 ) . In Hap3 Yorubans , the MHC locus is extreme in opposite directions in mates and non-mates: mean relatedness is lower at 94 . 6% of all loci than at the MHC locus in mates , but lower in only 1 . 6% of loci in non-mates . Given that there is no evidence in Yorubans of a significant difference in MHC relatedness between mates and non-mates , these results may be explained by the previous finding [13] ( confirmed here ) that Yorubans exhibit a broad mate-dependent shift in relatedness across autosomal loci . Standards vary on when , whether and how to correct significance tests when multiple hypotheses are examined . Chaix et al . tested four hypotheses regarding the difference between observed and expected relatedness in mate pairs , namely regarding autosomal and MHC relatedness in Yorubans and in Europeans , but did not report any corrections for multiple hypotheses [13] . We examined the effect of correcting previously reported nominal P values for multiple hypothesis testing . First , given that results were presented separately for each population [13] and that each finding was of interest on its own , one might reasonably consider that each population represents an independent hypothesis test ( of excess MHC dissimilarity in mates relative to random couples ) and therefore warrants correction . For two hypotheses , the corrected significance threshold is α' = 0 . 0253 ( see Methods ) , so that the previously-reported MHC relatedness in Hap2 European mates retains significance ( P = 0 . 015≤α' ) . Analyses of autosomal SNPs were introduced in the previous report as negative controls; however , the excess autosomal relatedness in Yoruban mates was presented as a significant result per se [13] . Therefore , another correction may have been warranted for the two sets of SNPs assayed ( MHC and autosomal ) . For four hypotheses , the corrected significance threshold is α' = 0 . 0127 , so that MHC relatedness in Hap2 European mates would no longer differ significantly from expectation ( P = 0 . 015≥α' ) , whereas overall relatedness in Yoruban couples would ( P<0 . 001 ) .
We found that the previously reported MHC dissimilarity among Hap2 European-American mate pairs [13] is apparent but not robustly supported by the underlying genotypic evidence . In addition , the effect essentially disappears in Hap3 for a similar number of independent couples from the same population , and is weak and insignificant for the combined Hap3 cohort . We cannot explain the discrepancy based on differences in SNPs assayed or by the imputation of missing alleles in phased data , given that Hap2 and Hap3 genotypes yield concordant results for the same couples ( Hap2∩3 ) , as do phased and unphased genotypes . In addition , Hap2 and Hap3 samples appear to be drawn from the same population ( Supporting Figure 4 in Text S3 ) , suggesting an explanation other than population structure . The fact that the MHC dissimilarity in Hap2 couples becomes marginal upon minor modifications in the methods and included samples suggests that the result was weaker than reported and did not represent a significant difference between mate pairs and non-mate pairs . This conclusion is supported by the observation that a stringent correction for multiple-hypothesis testing renders the original finding of MHC dissimilarity insignificant , even given the previously-reported nominal ( uncorrected ) P value . It is certainly true that multiple testing corrections are a matter of ongoing debate and diverse preference , and overly conservative approaches can lead to a substantial loss of power . However , even stricter approaches to multiple testing can be argued . For example , the authors might also have corrected for the several instances wherein they tested two alternative methods and chose one . Although our analysis of HapMap genotypes does not support a broad and significant dependence of mate selection on MHC , a weak effect is apparent in Hap2 even when the most extreme couples are excluded ( Supporting Figure 6 in Text S4 ) . In addition , the apparent depletion of very high MHC identity coefficients among mates ( Figure S2 ) hints that mate selection may disfavor extreme MHC similarity . Unfortunately , too few samples are currently available to pursue this hypothesis . Following the previous report [13] , we considered the entire MHC locus as a single unit . Because this genomic region contains many genes and exhibits variable rates of recombination [1] , it is possible that an examination of MHC genotypes at a finer scale would reveal a correlation with mate selection . Given the need to adjust for multiple hypothesis testing , and the likelihood ( based on HapMap samples ) that any effects would be subtle , a rigorous investigation of this question will require many more samples . If mates were found to differ from non-mates in MHC relatedness , in these or other populations , we note that this phenomenon need not stem from mate selection alone , particularly if only couples with children are considered . If offspring with certain MHC allele combinations survive preferentially , exclusion of mated couples without children could yield a non-random MHC similarity distribution amongst the remaining couples . This idea is supported by the increase in MHC heterozygosity of mouse embryos following viral infection of the parents [18] . The reported preference of Hap2 Yoruban individuals for mates more similar to themselves overall [13] was not robust to changing the source of genotype data . However , it was corroborated in an independent set of Hap3 couples . As with MHC dissimilarity in Europeans , the mate-dependent autosomal similarity effect detected in Yorubans appears to be driven by a small subset of pairs . Larger studies will be required to shed further light on these hypotheses .
|
There is evidence in numerous species that genes involved in immunity influence mate choice . Factors like body odor may subconsciously favor partners with different immunity alleles , to avoid inbreeding and/or endow offspring with broad resistance to pathogens . A previous study , based on HapMap genotypes , reported that European-American mates were extremely dissimilar from each other in immunity alleles compared to non-mates . Upon re-examining the results and methods , and visually comparing mates and non-mates , we found that this effect was weak , strongly dependent on extreme pairs and on arbitrary choices in methodology , and not significant after correcting for the multiple hypotheses tested . More importantly , examination of new couples from the same population did not support this hypothesis . Rare instances of very high MHC similarity among non-mates suggest that mates may avoid extreme similarity , rather than favoring dissimilarity . However , too few samples are readily available to test this prospect rigorously . We conclude that HapMap samples do not conclusively support the hypothesis that MHC genotypes exert an influence on mate choice . The same previous study reported that Yorubans appear to prefer mates who are more genetically similar to themselves overall . Our analyses suggest that the effect is driven by a subset of the sample .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"computational",
"biology/population",
"genetics",
"genetics",
"and",
"genomics/genetics",
"of",
"the",
"immune",
"system",
"evolutionary",
"biology/human",
"evolution",
"immunology/genetics",
"of",
"the",
"immune",
"system",
"genetics",
"and",
"genomics/population",
"genetics"
] |
2010
|
Absence of Evidence for MHC–Dependent Mate Selection within HapMap Populations
|
Despite a broad cell-type tropism , cytomegalovirus ( CMV ) is an evidentially pulmonary pathogen . Predilection for the lungs is of medical relevance in immunocompromised recipients of hematopoietic cell transplantation , in whom interstitial CMV pneumonia is a frequent and , if left untreated , fatal clinical manifestation of human CMV infection . A conceivable contribution of CMV to airway diseases of other etiology is an issue that so far attracted little medical attention . As the route of primary CMV infection upon host-to-host transmission in early childhood involves airway mucosa , coincidence of CMV airway infection and exposure to airborne environmental antigens is almost unavoidable . For investigating possible consequences of such a coincidence , we established a mouse model of airway co-exposure to CMV and ovalbumin ( OVA ) representing a protein antigen of an inherently low allergenic potential . Accordingly , intratracheal OVA exposure alone failed to sensitize for allergic airway disease ( AAD ) upon OVA aerosol challenge . In contrast , airway infection at the time of OVA sensitization predisposed for AAD that was characterized by airway inflammation , IgE secretion , thickening of airway epithelia , and goblet cell hyperplasia . This AAD histopathology was associated with a T helper type 2 ( Th2 ) transcription profile in the lungs , including IL-4 , IL-5 , IL-9 , and IL-25 , known inducers of Th2-driven AAD . These symptoms were all prevented by a pre-challenge depletion of CD4+ T cells , but not of CD8+ T cells . As to the underlying mechanism , murine CMV activated migratory CD11b+ as well as CD103+ conventional dendritic cells ( cDCs ) , which have been associated with Th2 cytokine-driven AAD and with antigen cross-presentation , respectively . This resulted in an enhanced OVA uptake and recruitment of the OVA-laden cDCs selectively to the draining tracheal lymph nodes for antigen presentation . We thus propose that CMV , through activation of migratory cDCs in the airway mucosa , can enhance the allergenic potential of otherwise poorly allergenic environmental protein antigens .
It is common knowledge in the allergy field that viral respiratory infections and allergic airway diseases ( AADs ) , for instance allergic asthma , can interdepend [1 , 2] . In principle , mutual interference may aggravate or dampen either of these medical entities . While few publications reported on protective effects of viral infections on the development of asthma [3–5] , the majority of studies revealed an exacerbation of asthma by viral respiratory diseases ( for reviews see [2 , 6–11] ) . On the other hand , damage of the asthmatic lung epithelium can increase the susceptibility to viral infections [12] . Allergic asthma and respiratory viral infections both can affect the physical and functional integrity of the airway epithelium and can thereby destroy its barrier function ( reviewed in [11] ) . This in turn facilitates the penetration of allergens as well as the invasion of pathogens [13] . In addition , expression of specific immune response genes in lung epithelial cells can be modulated by allergens and by viruses [14] , and there exists evidence to propose a cytokine-based interference between allergic reactions and the antiviral immune response ( for an overview see [7] ) . Overall , the pathophysiological interactions between respiratory virus infections and allergic airway diseases are manifold and rather complicated , and are not yet fully understood . In addition to the more typical respiratory viruses , such as respiratory syncytial virus ( RSV ) and rhinoviruses [7] that are characterized by a strict tropism for airway mucosa , cytomegalovirus ( CMV ) , specifically murine CMV ( mCMV ) , was reported to unexpectedly modulate asthma in an experimental mouse model [15 , 16] . CMVs are strictly host species-specific , double-stranded DNA viruses of the beta-subfamily of the herpesvirus family . Co-speciation with their specific hosts during eons of co-evolution has led to an intricate virus-host adaptation that is reflected by a set of “private” genes for each CMV species that is not shared between CMVs of different host species , but “biological convergence” has led to a comparable pathobiology of different CMV-host pairs [17] . Human CMV ( hCMV ) is known for its clinical relevance in congenital CMV infection of the fetus , resulting in birth defects , and in the immunocompromised host , in particular in recipients of hematopoietic cell transplantation ( HCT ) or solid organ transplantation ( SOT ) , in whom it can cause multiple organ failure by infection of a broad range of cell types [18] . This results in tissue-destructive viral histopathology . A link to the airways as a prominent site of CMV pathogenesis is provided by the fact that interstitial CMV pneumonia , associated with infection of interstitial fibroblasts , lung vascular endothelial cells , and pneumocytes , is the lead organ manifestation of hCMV in HCT recipients ( for clinical overviews , see [19 , 20] ) . The mouse model , employing mCMV , has proven its validity for predictions and “proof of concept” in the immunotherapy of clinical CMV ( reviewed in [21–23] ) . Importantly , all cornerstones of hCMV pathogenesis in HCT patients were reproduced with mCMV in the mouse model of experimental HCT , including interstitial pneumonia [24–29] , and lung involvement was also shown in mice infected as neonates [30–32] or as adults [33 , 34] . In all these instances , CMVs reached the lungs “from within” by intra-host dissemination , and not via the airways . However , although CMVs are not to be viewed as typical respiratory viruses , natural host-to-host transmission through saliva involves infection of airway mucosa , so that airway infection “from without” is epidemiologically relevant . Again , this airway mucosa route of hCMV infection was reproduced in the murine model already in early days of CMV research [35] , and revisited more recently [36–38] . Thus , the murine model has proven its validity also with respect to exogenous airway infection . All in all , there exists reasonable evidence to conclude that respiratory tract infection by host-to-host transmission of CMVs and exposure to inhaled environmental allergens share the target site and thus can meet with the potential consequence of a mutual enhancing and/or inhibiting interference . Clinical observations indeed suggest that CMV infections impact the course of airway diseases of other etiology [39 , 40] , although the underlying mechanisms were at that time not yet fully understood . Employing the mouse model , the present study aimed at investigating the epidemiologically realistic possibility of sensitization for AAD by airway co-exposure to CMV and an airborne environmental antigen of an otherwise low intrinsic allergenic potential .
Established murine models of AADs , including asthma , are divided into two phases: a sensitization phase , in which the antigen/allergen is usually applied systemically , alone or in combination with an adjuvant , and a challenge phase , in which the antigen/allergen is locally administered to the lungs . At least for studying a potential role of CMV in the sensitization for AAD , this standard protocol with systemic sensitization has no medical correlate and thus required an adaptation to the more realistic scenario of co-exposure to CMV and inhaled environmental antigen in airway mucosa at the time of virus host-to-host transmission . To model this situation , mice were sensitized by intratracheal administration of purified model antigen OVA in absence or presence of a simultaneous intratracheal infection with mCMV , followed two weeks later by three consecutive inhalative challenge exposures to OVA aerosol [41–43] . The experimental regimen is illustrated in Fig 1A . The four experimental groups , on which most conclusions rely , included OVA challenge without prior OVA sensitization , as well as OVA challenge after prior OVA sensitization , both in absence or presence of an mCMV airway infection at the time of OVA sensitization ( Table 1 ) . The sensitization phase is characterized by an uptake of the antigen/allergen by professional antigen-presenting cells ( APCs ) , in particular dendritic cells ( DCs ) , their migration to the draining regional lymph node/s and priming of an adaptive T-cell and B-cell response . Depending on the activation status of the DCs , tolerance or sensitization are induced . Upon successful sensitization , memory cells develop and can react by a recall response to a recurring antigen/allergen exposure . This can result in AAD , which is characterized by an influx of inflammatory cells , such as lymphocytes , macrophages , and granulocytes into the lungs , associated with the release of a large array of inflammatory mediators . Histopathological criteria for AAD are the thickening of airway epithelia and hyperplasia of mucus-producing goblet cells . Inflammatory influx into the airways and lung tissue are signs of AAD . To investigate the influence of mCMV on OVA-specific inflammatory processes in the lungs , immunocompetent C57BL/6 mice were sensitized by intratracheal infection with mCMV combined with intratracheal administration of OVA on day 0 . For provoking AAD , repetitive OVA-challenge was performed on days 14 , 15 , and 16 ( Fig 1A ) . At the time of the first episode of OVA-challenge ( day 14 ) as well as at the time of read-out ( day 18 ) , virus replicated locally in the lungs , but not in the spleen ( Fig 1B ) . This organ selectivity is in accordance with findings in a related model of mCMV airway involvement after infection via the intranasal route [36] and is explained by an immune response that prevents further intra-host virus dissemination . At 48 hrs after the last challenge , cytospin preparations of the broncho-alveolar lavage ( BAL ) were analyzed for inflammatory cells ( Fig 1C ) , and HE-stained lung sections were scored for an inflammatory cell influx ( Fig 1D ) . Analyzing the cellular composition of the BAL revealed that only those mice that were mCMV-infected and OVA-sensitized showed elevated overall cell numbers ( Fig 1C , left panel ) upon OVA challenge ( group mCMV/OVA//OVA ) . In particular , numbers of lymphocytes and macrophages were increased in absolute terms . Notably , eosinophilia was not observed and also the number of neutrophilic granulocytes was hardly affected . Airway infection with mCMV in absence of OVA sensitization ( group mCMV/—//OVA ) also induced somewhat higher numbers of lung-infiltrating lymphocytes compared to uninfected groups with either only OVA challenge or OVA sensitization and challenge ( groups—/—//OVA and—/OVA//OVA , respectively ) , which is explained by OVA-independent , infection-associated and mast cell-assisted lung infiltration [34] . The quantitatively most significant increase in total cell counts ( Tcc; Fig 1C , left panel ) as well as in the percentage of lymphocytes ( Fig 1C , right panel ) , however , was observed upon OVA challenge when mice were OVA-sensitized in the presence of infection ( group mCMV/OVA//OVA compared to all other groups ) . The relative increase in the number of BAL lymphocytes was associated with a relative decrease in the number of alveolar macrophages ( Fig 1C , right panel ) . These findings from cell quantification in the BAL were consistent with corresponding histological images of lung tissue sections , illustrating the most pronounced inflammatory cell influx after OVA challenge in the group of mice sensitized by OVA in the presence of airway infection by mCMV ( Fig 1D ) . Notably , OVA sensitization and challenge in the group—/OVA//OVA was not associated with an increased cell infiltration of the lungs , as indicated by an inflammation score that was found to be almost identical to the score in the—/—//OVA group of mice with no preceding OVA sensitization ( Fig 1D , right panel ) . In accordance with the cell quantifications , mCMV infection in the OVA-unsensitized control group mCMV/—//OVA led to a slightly increased inflammation score but far below the score of the OVA-specific infiltration in the group mCMV/OVA//OVA . As all experimental groups included an OVA challenge , it is important to point out that an inflammatory cell influx was not observed in absence of OVA challenge ( S1 Fig ) . Notably , the enhancing effect of mCMV on OVA sensitization was found not to depend on replicative virus ( S2 Fig ) . In conclusion , purified OVA has little-to-no allergenic potential in terms of inducing an inflammatory response , unless sensitization to OVA occurs in the presence of airway exposure to mCMV . As mCMV airway infection in the OVA sensitization phase was found to enhance an OVA-specific airway inflammation upon OVA challenge ( see above ) , we also analyzed the influence of mCMV infection on the OVA-specific B-cell response ( Fig 2 ) . To this end , OVA-specific serum immunoglobulins were measured at 48 hrs after the last challenge . Intriguingly , OVA sensitization and challenge in experimental group—/OVA//OVA failed to induce OVA-specific IgE , IgG1 , IgG2b and IgG2c antibodies , neither did mCMV airway infection in absence of OVA sensitization . Again , only a combination of mCMV airway infection with OVA sensitization and challenge in group mCMV/OVA//OVA resulted in significant titers of OVA-specific antibodies . Importantly , as antibody production and immunoglobulin class switch are CD4+ T helper cell-dependent , these results imply that sufficient help was provided only when CD4+ T cells were primed by OVA sensitization under conditions of concomitant infection . Remodeling of the airways by increased numbers of mucus-secreting goblet cells , that is goblet cell hyperplasia , represents a histopathological hallmark defining AAD more stringently than inflammatory cell influx alone , in particular when studied in the presence of infection that by itself contributes to inflammation . Histological images of lung tissue sections document thickening of the bronchial epithelium and enhanced numbers of PAS-stained , mucus-producing goblet cells upon OVA challenge only when OVA sensitization had taken place in the presence of mCMV airway infection ( Fig 3A , lower right panel ) . This visual impression , documented by representatively selected images of tissue sections , is statistically substantiated by histometrical quantitation of the thickness of airway epithelia ( Fig 3B ) and by counting of goblet cells ( Fig 3C ) . It is of interest to note that the comparison between control group—/—//OVA and infected group mCMV/—//OVA did not reveal a significant difference in these parameters of AAD ( Fig 3B and 3C ) . Thus , whereas mCMV infection by itself is associated with cellular infiltration of the lungs ( see above ) , it does not elicit goblet cell hyperplasia , a finding that is most important as it clearly distinguishes viral histopathology in the lungs from OVA-specific AAD . Since CD8+ T cells are long known as the predominant effector cell type that controls acute pulmonary mCMV infection in related experimental mouse models ( [24 , 26 , 27 , 28 , 34] , reviewed in [23 , 44] ) , it was an obvious question if CD8+ T cells with an effector cell phenotype dominate T-cell infiltrates in the lungs also in the here discussed AAD model ( Fig 4 ) . At a glance , CD8+ T cells quantitatively dominated over CD4+ T cells in pulmonary T-cell infiltrates in overall lung tissue ( Fig 4A and 4B ) and in BAL that represents cells present in extravasal airway epithelia including alveolar epithelium of the lungs ( Fig 4B; S3 Fig ) . The highest percentage and absolute number of CD8+ T cells and , accordingly , the highest CD8:CD4 T-cell ratio , was observed in the group mCMV/OVA//OVA . A direct comparison of this group with the likewise infected group mCMV/—//OVA , not sensitized to OVA , suggests an OVA sensitization-specific component of the response , which is more evident in lung tissue than it is in BAL . In the CD8+ T-cell population , numbers of CD62L-KLRG1+ short-lived effector cells ( SLECs ) [45–48] , which include terminally differentiated cells destined to death [49] , were elevated in both infected groups compared to uninfected groups in lungs and BAL . A slightly elevated number of CD8+ SLECs in the infected and OVA-sensitized group mCMV/OVA//OVA may indicate a minor OVA-specific component in an overall primarily virus-specific CD8+ SLEC population . Percentages of CD4+ SLECs were low throughout and were not notably elevated in the infected groups , neither in lungs nor in BAL ( Fig 4A and 4B; S3 Fig ) , although slightly elevated absolute numbers in the group mCMV/OVA//OVA ( Fig 4B ) might indicate a minor OVA sensitization-specific component as well . Pulmonary infiltrates in all groups included an unexpectedly high proportion of TCRβ+CD4-CD8- cells amongst all TCRβ+ cells , likely representing NKT cells [50 , 51] . As their proportion was similar in all groups regardless of infection or OVA sensitization , and as their presence after challenge evidently did not correlate at all with AAD ( recall Fig 3 ) , we decided not to pursue this otherwise interesting phenomenon further in the context of the here discussed AAD model . Despite the low numbers of CD4+ SLECs , viral epitope-specific cells capable of secreting IFN-γ upon short-term stimulation with a panel of known antigenic peptides of mCMV presented by the MHC class-II ( MHCII ) molecule I-Ab [52 , 53] were detectable in the two infected groups ( Fig 5A , left panel ) , including group mCMV/—//OVA in which AAD did not develop . So , virus-specific IFN-γ-secreting CD4+ T cells are apparently not notably involved in AAD . Surprisingly , only baseline levels of CD4+ T cells recognized an immunodominant I-Ab-presented OVA peptide [54] , with no enhancement observed in the AAD group mCMV/OVA//OVA ( Fig 5B , left panel ) . As IFN-γ is a lead cytokine of T helper type-1 ( Th1 ) cells , we also quantitated CD4+ T cells expressing the lead cytokine IL-4 of Th2 cells after short-term stimulation with the I-Ab-presented peptides . Although cells secreting IL-4 were easily detectable in the AAD group mCMV/OVA//OVA after polyclonal stimulation via antibody ligation of the signaling molecule CD3ε , numbers of Th2 cells specific for the tested panel of viral peptides as well as the OVA peptide were baseline throughout ( S4 Fig ) . Thus , viral epitope-specific CD4+ T cells in the infected groups were almost exclusively Th1 cells , whereas the frequencies of OVA-specific Th1 and Th2 cells were all below the detection limit of this assay . As expected , IFN-γ-secreting CD8+ T cells specific for a panel of known antigenic peptides of mCMV presented by the MHC class-I ( MHCI ) molecules Kb and Db [55] were detected in the two infected groups ( Fig 5A , right panel ) . As with the virus-specific CD4+ T cells discussed above , absence of AAD in group mCMV/—//OVA implies that virus-specific CD8+ T cells are also not critically involved in AAD . At first glance , absence of viral epitope-specific CD8+ T cells in the OVA-sensitized but uninfected groups may seem to be trivial , but it actually gives us the important information that there exists no antigenic mimicry between the Kb-presented immunodominant OVA epitope SIINFEKL and any of the Kb- or Db-presented viral epitopes that could lead to cross-reactivity of CD8+ T cells defining heterologous immunity [56 , 57] . A low frequency of SIINFEKL-specific CD8+ T cells was detected in group—/OVA//OVA , and significantly enhanced in the AAD group mCMV/OVA//OVA ( Fig 5B , right panel ) . As antigen processing of exogenously administered OVA and presentation of OVA-derived antigenic peptide by an MHCI molecule involves OVA uptake and cross-presentation by DCs ( for reviews , see [58 , 59] ) , the enhanced SIINFEKL-specific response in group mCMV/OVA//OVA was the first indication for viral activation of antigen cross-presenting DCs in this AAD model . Altogether , at this stage of investigation , it was tempting to conclude that CD4+ T cells are unlikely to contribute decisively to AAD observed in group mCMV/OVA//OVA , whereas an OVA-specific CD8+ T-cell response appeared to correlate with AAD . To unequivocally identify the effector cells in AAD , mice of the AAD group mCMV/OVA//OVA were either left untreated to result in AAD , or were depleted of T-cell subsets in all combinations shortly before the first OVA challenge ( Fig 6A ) . The result was as clear as surprising: goblet cell hyperplasia characterizing AAD in the undepleted mice ( Fig 6B , upper image ) was prevented by double depletion as well as by CD4 depletion alone , but not by CD8 depletion alone ( Fig 6B ) . These findings , visualized by representatively selected histopathological images , were statistically substantiated by measuring the thickness of airway epithelium and by counting the number of goblet cells ( Fig 6C ) . For both parameters of AAD , a significant reduction was achieved by double depletion and by CD4 depletion , but not by CD8 depletion alone . A trend to a slightly more efficient reduction by depletion of both T-cell subsets , as compared to CD4 depletion alone , did not reach statistical significance for goblet cell hyperplasia . Thus , although CD8+ T cells by far dominate the T-cell response in quantity , AAD histopathology is caused by the minority population of CD4+ T cells . In other AAD models , specifically in murine asthma , airway goblet cell hyperplasia was found to be directly induced by Th2-derived IL-9 , whereas IL-4 , IL-5 , and IL-13 independently induce goblet cell hyperplasia , though indirectly via neutrophilic granulocytes [60–62] . IL-25 has also been implicated in AAD , operating indirectly by stimulating the production of IL-4 , IL-5 , and IL-13 ( reviewed in [63] ) . Commemorating the unverifiably low frequency of IL-4-secreting peptide-specific Th2 cells in the lungs of mice of the relevant AAD group mCMV/OVA//OVA ( S4 Fig ) , we decided to perform a cytokine transcription profiling in the lungs , as it is less dependent on the time point of analysis , is more sensitive because of amplification by RT-PCR , and covers more candidates than a protein array . Specifically , cytokines with a short half-life delivered by very few producer cells over a short distance to target cells in tissue are difficult to detect as proteins in situ . Furthermore , at least for T cells there exist no example and no rationale to give reasons for a speculation that gene expression would not lead to the respective cytokine . A comparison of transcriptional activity in lungs of mCMV/OVA-sensitized mice with no OVA challenge ( no AAD ) and after OVA challenge ( AAD ) revealed a broad challenge-dependent induction of cytokine gene expression with a predominantly Th2 profile ( Fig 7; S5 Fig ) , which includes the known AAD/goblet cell hyperplasia-inducing cytokines IL-4 , IL-5 , IL-9 , and IL-25 , with the notable exception of IL-13 . The source of most of the induced cytokines was clearly CD4+ T cells , as CD4 depletion prior to OVA challenge often not only precluded an induction but even led to a reduction to below the pre-challenge level . Notably , cytokine-receptor pairs showed the known inverse regulation in that upregulation of the cytokine leads to downregulation of its receptor ( as an example , see IL-4/IL4rα ) and downregulation of the cytokine leads to upregulation of its receptor ( as an example , see IL-13/IL13rα1 ) . This is a further argument to conclude that the respective cytokines were indeed produced in the lungs . Enhanced presence of Th2 cells in AAD lungs was finally confirmed by amplifying IL-5 production in lung cells , derived from lungs of all four experimental groups ( Table 1 ) , by restimulation with full-length OVA protein in cell culture , which covers all epitopes . Notably , OVA-specific IL-5 production , measured as difference between OVA restimulated cultures and non-restimulated control cultures , was significantly enhanced in the AAD group mCMV/OVA//OVA as compared to the remaining three groups in which AAD did not develop ( Fig 8A ) . Absolute IL-5 levels in cell cultures of lung cells of the relevant AAD group mCMV/OVA//OVA demonstrate the OVA-specificity of this recall response ( Fig 8B ) . This finding finally proved the presence of OVA-specific , IL-5-producing Th2 cells in the AAD lungs . The question remained how mCMV airway infection facilitates OVA sensitization of Th2 cells . Sensitization by an airborne environmental antigen requires antigen uptake by activated DCs in the airway mucosa , migration of antigen-laden DCs to draining regional lymph nodes , specifically the tracheal lymph nodes ( tLN ) , and antigen presentation to naïve Th2 cells . To track the antigen , mice were sensitized with a fluorescent ( Alexa Fluor 647 ) derivate of OVA , administered intratracheally alone or in combination with mCMV ( Fig 9A ) . As soon as 1 day later , labeled OVA had reached the tLN , where it localized almost exclusively to MHCII+CD11c+ DCs . Notably , the number of OVA-laden DCs in the tLN was markedly elevated after airway co-exposure to mCMV , so that one role of mCMV apparently is to facilitate antigen uptake by DCs . More efficient uptake of labeled OVA was associated with activation of DCs ( Fig 9B and 9C ) , as reflected by a significant upregulation of co-stimulatory molecules , such as the CD28 ligands CD80 ( B7-1 ) and CD86 ( B7-2 ) [64] , which are components of the immunological synapse critical for the priming of naïve T cells [65 , 66] . In contrast , cell surface expression of CD40 and the chemokine receptor CCR7/CD197 on OVA-laden DCs was not enhanced by airway co-exposure to mCMV . So far , data have shown the presence of activated , OVA-laden MHCII+CD11c+ DCs in the airway-draining tLN after combined airway exposure to OVA and mCMV . A variety of distinct DC subsets localize to the respiratory tract , where they survey the respiratory mucosa and parenchyma for foreign antigens , including pathogens . Upon receiving an activation stimulus and uptake of antigen/allergen , several subsets of airway-resident DCs are able to migrate to the draining tLNs , where they process and present antigens for sensitization of naïve T cells or for restimulation of memory T cells [67–69] . For identifying the DC subset ( s ) involved in the sensitization for AAD , we set out to first distinguish between a DC population composed of conventional DCs ( cDCs ) and monocyte-derived DCs ( MoDCs ) on the one hand , and plasmacytoid DCs ( pDCs ) on the other . By analyzing the expression of B220 in MHCII+CD11c+ DCs present in tLNs after combined airway exposure to fluorescent OVA and mCMV ( Fig 10 ) , B220low cDCs/MoDCs and the B220high pDCs were found to have taken up the fluorescent OVA ( Fig 10A ) . OVA-laden DCs localized selectively to the draining tLN , and not to the non-draining sub-mandibular LN ( smLN ) ( Fig 10B ) . Corresponding to the conditions for AAD , significant numbers of OVA-laden DCs were found in the tLNs only after combined airway exposure to OVA and mCMV . Notably , although OVA-laden cDCs/MoDCs and pDCs both migrated to the tLNs , only cDCs/MoDCs migrated in response to mCMV in a virus dose-dependent manner , and predominated in number ( Fig 10C ) . This finding led us to focus on cDCs/MoDCs . OVA-laden cDCs and MoDCs were distinguished by low and high expression of Ly6c , respectively , and Ly6clow cDCs were subclassified based on mutually exclusive expression of CD11b and CD103 ( Fig 11A ) . Although Ly6chigh MoDCs were found to be somewhat elevated in response to mCMV , their relative and absolute numbers are very low compared to cDCs ( Fig 11B ) , so that their contribution to sensitization for AAD is supposedly minor . Upon OVA sensitization alone , relative and absolute numbers of OVA-laden CD11b+ cDCs in tLNs exceed those of CD103+ cDCs . Notably , under conditions of co-exposure to mCMV , both subsets equal in significantly increased absolute numbers . Altogether , the data provide a reasonable argument to conclude that both subsets of migratory cDCs are activated by mCMV in the airway mucosa , take up OVA , and migrate to the draining tLNs for antigen presentation .
Host-to-host transmission of hCMV mostly takes place in early childhood during intimate social contacts of the child with family members or peers who shed the virus with saliva . Local mucosal infection can spread to airway mucosa , where virus and inhaled environmental antigens can meet at airway mucosa-resident DCs capable to migrate upon activation . It was the aim of our study to model this epidemiologically realistic situation for investigating a putative contribution of primary CMV airway infection to sensitization by antigens/allergens that predispose the child for AADs , including asthma . The question is difficult to address by clinical investigation , because primary CMV airway infection in early childhood is usually not diagnosed , so that a link between CMV and AAD years later in life is not obvious to be considered . The here presented data from a mouse model predict that CMV airway infection at the time of exposure to inhaled antigens/allergens can indeed enhance allergenic sensitization that predisposes for AAD . Importantly , the model shows that even a protein antigen that has low-to-no allergenic potential on its own can sensitize for AAD when CMV activates airway mucosa-resident DCs for more efficient antigen uptake . Such a mechanism is medically relevant as it significantly broadens the spectrum of potential environmental inducers of AAD . The data have consistently shown that histopathological key parameters defining AAD , namely airway inflammation , thickening of airway epithelium , and the lead symptom goblet cell hyperplasia depend on airway co-exposure to a potential allergen and CMV , whereas either exposure alone fails . IgE secretion and IL-5-driven airway eosinophilia have also been associated with AAD ( for a review see [63] ) . While we detected an enhanced IgE secretion in the AAD group mCMV/OVA//OVA , eosinophila was not observed despite IL-5 upregulation . This finding is in accordance with work in the mouse model , describing that mCMV infection actually decreases airway eosinophilia [15] . In this context it is important to call attention to a clinical asthma phenotype in humans , known as non-eosinophilic asthma ( NEA ) . NEA is characterized by airway inflammation with absence of eosinophils and represents severe asthma cases , as its most relevant clinical trait is its poor response to standard asthma treatments , especially to inhaled corticosteroids [70] . At the present state , we would not go so far to claim to have here described a murine model for NEA , but the clinical example clearly shows that absence of eosinophilia in our AAD model is certainly no valid argument against the predictive value of this mouse model for AAD entities that exist in humans . The grade of AAD observed here was not full-blown asthma , as respiratory function was not yet critically affected in the AAD group mCMV/OVA//OVA . As a perspective , however , we propose that we observed here a pre-asthmatic stage that may develop into clinical asthma when the recall response is further amplified by additional rounds of challenge exposure . This remains to be tested in future work . As to the mechanism , the data delineate a causal chain in which CMV activates DCs residing in the airway mucosa , as indicated by expression of CD80 and CD86 . This activation is associated with a more efficient uptake of airborne antigens . OVA-laden DCs , in particular migratory CD11b+ as well as CD103+ cDCs , then specifically migrate to the draining lymph nodes , the tLNs , for presenting OVA peptides to prime naïve T cells . While CD11b+ cDCs mediate MHCII-restricted Th2 priming to respiratory allergens/antigens and have been associated with the pathogenesis of asthma [71] , CD103+ cDCs represent the antigen cross-presenting cDC subset in airway mucosa [72] . Accordingly , also in our model , OVA-laden CD11b+ cDCs are likely the inducers of the observed Th2-driven AAD , whereas the activation of CD103+ cDCs explains the MHCI-restricted priming of CD8+ T cells specific for the OVA-derived antigenic peptide SIINFEKL . It is established knowledge in the allergy field that Th2-derived IL-9 and IL-4/IL-5 induce goblet cell hyperplasia either directly or indirectly via neutrophilic granulocytes , respectively [60–62] . In addition , IL-25 contributes to AAD by its stimulating effect on IL-4 and IL-5 [63] . The conclusion that viral activation of CD11b+ cDCs in airway mucosa is the critical triggering incident in sensitization for AAD is supported by the previous finding that administration of activated ex vivo OVA-pulsed DCs to the airways primes for a Th2 response in the lungs and predisposes for pulmonary goblet cell hyperplasia [73] . A missing piece in the chain of evidence is the demonstration of a quantifiable number of Th2 cells that recognize a single , though immunodominant , OVA epitope ( OVA323-339 ) in the lungs . This low frequency corresponds to the generally very low response of CD4+ T cells in the lungs of mice of the AAD group mCMV/OVA//OVA . In this context it is informative to note that in an unrelated mouse model of pulmonary infection with a recombinant influenza virus expressing epitope OVA323-339 , the frequency of the OVA epitope-specific Th1 cells in the lungs was found to be 10-fold lower than the frequency of CD8+ T cells specific for MHCI epitope NP264-272 of influenza virus nucleoprotein [74] , even though replication of the recombinant virus is expected to result in a particularly efficient priming to OVA . The quantitatively low OVA-specific Th2 response in the lungs of the AAD group mCMV/OVA//OVA contrasts with the production of isotype-switched OVA-specific serum antibodies also observed only in the AAD group , which clearly indicates sufficient activation of the CD4+ T-helper cell/B-cell axis , though this is likely to take place in the B-cell areas of lymphoid tissues rather than in lung tissue . Nonetheless , OVA-specificity of AAD induction is beyond question since AAD required both OVA sensitization and OVA challenge . In addition , an involvement of Th2 cells that reside in the lungs became evident from a CD4+ T cell-dependent Th2 cytokine transcription profile in the lungs of mCMV/OVA//OVA mice . This included enhanced gene expression of interleukins IL4 , IL-5 , IL-9 , and IL-25 , all known to induce goblet cell hyperplasia , the lead histopathological sign of tissue remodeling that defines AAD . The presence of Th2 cells in AAD lungs became evident from the recall response of in vivo primed pulmonary Th2 cells by restimulation with full-length OVA protein in cell culture , which led to significant levels of IL-5 specifically in the AAD group mCMV/OVA//OVA . We thus conclude that the in situ frequency of OVA peptide-specific Th2 cells is very low in the lungs and that minute numbers of Th2 cells suffice in lung tissue for inducing AAD by local delivery of AAD-driving interleukins IL-4 , IL-5 , IL-9 , and IL-25 . IFN-γ-secreting viral epitope-specific Th1 cells as well as CD8+ T cells are apparently not critically involved in AAD , as their respective frequencies were about the same in AAD group mCMV/OVA//OVA and non-AAD group mCMV/—//OVA ( Fig 5A ) . Notably , UV-inactivated mCMV also induced AAD . In this context it is of interest that non-infectious but enveloped and therefore entry-competent subviral particles of hCMV , the so-called dense bodies , were found to stimulate maturation and activation of immature human DCs [75] . So , apparently , signaling associated with the viral entry process is sufficient to activate DCs . Conformance of mCMV with hCMV in this respect strengthens the validity of the model in predicting activation of DCs in airway mucosa by hCMV . The molecular mechanism by which mCMV activates DCs has been extensively addressed in literature , although studies on mCMV interactions with DC subsets so far focused primarily on pDCs . Despite the fact that OVA-laden cDCs were identified here as the DCs that migrate to draining tLNs for allergenic sensitization , activation by mCMV applies to most DC subsets , including pDCs , as we have found OVA-uptake and CD80/CD86 expression by essentially all MHCII+CD11c+ DCs . For pDCs it is known that they are activated by mCMV through TLR7 and TLR9 sensing and signaling , leading to the production of type I interferons . Furthermore , all subsets of splenic DCs were found to be activated by mCMV through MyD88 adaptor-dependent TLR9 signaling to produce IL-12 ( for reviews , see [76–78] ) . CD8α+cDCs and CD11b+cDCs can be infected by mCMV in vivo , whereas pDCs are not infectable and nonetheless sense mCMV ( reviewed in [76] ) . All this collectively indicates that productive infection of DCs is not required for mCMV-mediated DC activation that represents the initializing event in the allergenic sensitization phase resulting in AAD . Pathophysiological interdependence of seemingly unrelated disease entities , resulting in co-morbidity , is an important , though rarely experimentally addressed , issue in medicine . In a sense , our model is a co-morbidity model in that it links CMV infection to AAD . Notably , while we have shown that viral activation of CD11b+ cDCs in airway mucosa promotes Th2-driven AAD , a mirror image was recently presented by linking AAD/asthma to influenza virus infection [79] . In that work , prior airway exposure to an inhaled potent allergen activated the CD11b+ cDCs in the airway mucosa and their migration into the draining LNs , which unexpectedly resulted in a more efficient priming of protective antiviral CD8+ T cells , a function usually attributed in influenza virus infection to the antigen cross-presenting CD103+ cDCs [80] . All in all , our data provide reasonable evidence to propose that activation of migratory CD11b+ cDCs by mCMV in the airway mucosa to enhanced antigen uptake , migration into the draining tLN , and presentation of allergen epitope ( s ) to Th2 cells elicits AAD . The medically most important aspect , in our view , is the finding that CMV infection , as predicted by the here presented model , can “convert” a harmless , inhaled protein antigen to an allergen . As a perspective , future work will aim to expand the model for defining conditions under which CMV airway infection could possibly contribute to the development of full-blown asthma , and to address the question if CMV might be involved in the pathogenesis of non-eosinophilic human forms of asthma , NEA .
Female C57BL/6 mice were purchased from Harlan Laboratories and housed in the translational animal research center ( TARC ) of the University Medical Center of the Johannes Gutenberg-University Mainz for at least 1 week under specified-pathogen-free ( SPF ) conditions . Mice were used at the age of 8-to-12-weeks . Animal experiments were approved according to German federal law §8 Abs . 1 TierSchG by the ethics committee of the Landesuntersuchungsamt Rheinland-Pfalz , permission numbers 177-07/G09-1-004 and 177-07/G 14-1-015 . Intratracheal ( i . t . ) applications were performed essentially as described [81 , 82] . In brief , mice were anaesthetized by intraperitoneal ( i . p . ) injection of ketamin/rompun ( Ketamin-ratiopharm , Ratiopharm , Ulm , Germany; Rompun 2% , Bayer , Leverkusen , Germany ) and attached on a restraining device . The tongue was carefully pulled aside and with the aid of a cold light positioned at the throat , a crystal tip was placed into the trachea . First , mCMV ( cell culture-propagated and purified strain Smith , ATCC VR-194/1981 , [83] ) was instilled in a volume of 20 μl diluted in PBS . As a number of virological and immunological parameters proved 106 PFU to be the optimal infection dose for i . t . application , experiments were performed with 106 PFU of mCMV if not indicated otherwise . After infection , 20 μl of either OVA-Alexa Fluor 647 conjugate ( 4mg/ml in PBS; catalog no . O34784 , Invitrogen , Germany ) or endotoxin-free OVA ( EndoGrade Ovalbumin , 4mg/ml in PBS; catalog no . 321001 , Hyglos , Bernried , Germany ) was administered . PBS served as control . Experimental groups consisted of at least 4 mice . To induce antigen sensitization via the airways , OVA and mCMV were administered i . t . on day 0 as described above . Subsequently , allergic airway inflammation was provoked by repetitive challenge exposures to OVA ( albumin from chicken egg white; catalog no . A5503 , Sigma ) on days 14 , 15 , and 16 . For this , an OVA solution ( 1% in PBS ) was freshly prepared for every use and nebulized for 20 min with an ultrasonic nebulizer ( NE-U17 , Omron , Hoofdorp , The Netherlands ) . After the last challenge , the following parameters were determined for groups of usually 4–5 mice tested either individually ( i-iv ) or as a pool ( v-vii ) : ( i ) OVA-specific serum immunoglobulins and IL-5 by ELISA , ( ii ) cellular composition of the broncho-alveolar lavage ( BAL ) by cytospins , ( iii ) inflammation of the lungs by histology , ( iv ) virus titers , ( v ) epitope-specificity of pulmonary CD4+ and CD8+ T-cell infiltrates by ELISPOT assay , ( vi ) phenotype of T-cell infiltrates in lung tissue and BAL by cytofluorometry , and ( vii ) gene expression analysis . Blood samples were collected in tubes , using a 1-ml syringe , and stored at room temperature until sera preparation . BAL was performed as described below and its cellular composition was analyzed following fixation and staining of the cytospins by using a Microscopy Hemacolor-Set ( Merck , Darmstadt , Germany ) . Percentages and absolute numbers were calculated for each cell type . Lung parenchyma was prepared as described below for cytofluorometric analysis of T-cell subsets and for determination of CD4+ and CD8+ T-cell frequencies by ELISPOT assay . For histological analysis , lungs were fixed and prepared as described below . To obtain cells for cytofluorometric analysis , the left lung lobe was ligated on the afferent bronchus , removed , and stored in PBS on ice until further preparation . Mice were treated essentially as described in the sensitization/challenge protocol described above . One day before the first challenge ( d13 ) , T-cell depletion was performed by i . v . injection of purified monoclonal antibodies directed against CD4 ( clone: YTS191 . 1; 0 . 8mg/mouse ) or CD8 ( clone: YTS169 . 4; 1 . 3mg/mouse ) . Lungs were fixed by inflation ( 1 ml ) and immersion in 4% formalin , followed by embedding in paraffin . Tissue sections were prepared and stained with Hematoxylin/Eosin ( HE ) to analyse tissue inflammation , or with a combined Periodic Acid Schiff ( PAS ) /HE staining to identify mucus-producing goblet cells . Airway inflammation was scored semi-quantitatively on HE slides . For this , five randomly selected areas were scored by two experienced observers blinded to experimental groups . Inflammation was scored on a scale from 0 to 4 . The score is based on the following inflammatory conditions: {0} , all airways and vessels are free of inflammatory infiltrates . {1} , some infiltrates are detectable around airways and vessels , and are up to two layers in thickness . {2} , the majority of airways and vessels show infiltrates with a thickness of up to two layers , only occasionally more layers can be found . {3} , the majority of airways and vessels show inflammatory infiltrates . Many layers are thicker than two layers , ranging from 3–8 layers . {4} , all airways and vessels are surrounded by thick layers of inflammatory cells . PAS-positive goblet cells were quantified per millimeter of basal membrane on at least three different representative airways on PAS-stained slides . Airway thickness was measured on three independent HE stained slides , with at least three different representative airways per slide , so that at least 9 airways were evaluated . The measurements were performed by using an imaging software ( analySIS , Soft Imaging Systems , Stuttgart , Germany ) . Infectious virus ( expressed as plaque forming units , PFU ) was quantitated for whole organ homogenates by a virus plaque assay performed on monolayers of mouse embryo fibroblasts , making use of increasing the sensitivity by the method of centrifugal enhancement of infectivity ( [83] and references therein ) . Custom peptide synthesis to a purity of > 80% was performed by JPT Peptide Technologies ( Berlin , Germany ) . Synthetic peptides were used for exogenous loading of C57BL/6 ( H-2b ) spleen cells and of EL-4 ( H-2b ) cells with panels of MHCII- and MHCI-presented peptides of mCMV or OVA , respectively . Peptide-laden cells were used as stimulator cells in the ELISPOT assay , as described below . Peptides recognized by CD4+ T cells: m18 ( NERAKSPAAMTAEDE ) [52] , M25 ( LYETPISATAMVIDI ) [53] , m139 ( TRPYRYPRVCDASLS ) [52] , m141 ( LVVFSDPNADAATSV ) [52] , and m142 ( RYLTAAAVTAVLQDF ) [53] , and OVA323-339 ( ISQAVHAAHAEINEAGR ) [54] . Peptides recognized by CD8+ T cells: M38 ( SSPPMFRVP ) , M45 ( HGIRNASFI ) , M57 ( SCLEFWQRV ) , IE3 ( RALEYKNL ) , m139 ( TVYGFCLL ) , m141 ( VIDAFSRL ) , and m164 ( GTTDFLWM ) [55] , and OVA257-264 ( SIINFEKL ) [85] . Frequencies of virus- and OVA-specific CD8+ T cells and CD4+ Th1 cells were determined by an IFN-γ-based ELISPOT assay as described ( [86 , 87] and references therein ) . In brief , graded numbers of immunomagnetically purified CD4+ and CD8+ T cells from lung tissue of 4–5 mice per group were stimulated for IFN-γ secretion in triplicate assay cultures . C57BL/6 spleen cells and EL-4 lymphoma cells were used as APCs exogenously loaded with synthetic peptides at loading concentrations of 10−6 M ( CD4+ T-cell epitopes ) and 10−7 M ( CD8+ T-cell epitopes ) , respectively . Frequencies of IL-4 producing CD4+ Th2 cells were determined by using the IFN-γ/IL-4 double-color ELISPOT assay , according to the manufacturer’s protocol ( CTL , Shaker Hights , OH ) . The total number of CD4+ Th2 cells capable of producing IL-4 upon polyclonal stimulation was determined by the CD3ε-redirected ELISPOT assay ( [83] and references therein ) , using 145-2C11 hybridoma cells as stimulator cells that produce agonistic monoclonal antibody directed against the CD3ε molecule of the T cell-receptor-CD3 signaling complex . Frequencies of IFN-γ and IL-4 secreting , spot-forming cells and the corresponding 95% confidence intervals were calculated by intercept-free linear regression analysis [86 , 87] . Spots were counted automatically based on standardized criteria using ImmunoSpot S4 Pro Analyzer ( CTL , Shaker Hights ) and CTL ImmunoSpot software V5 . 1 . 36 Professional DC . Total lung RNA was purified from lung tissue from a pool of 5 mice for each group by the RNeasy Micro Kit ( Qiagen , Hilden , Germany ) with on-column DNase I treatment according to manufacturer’s instructions ( Qiagen ) . 1μg of high-quality total RNA was reverse transcribed using the First Strand Synthesis Kit ( Qiagen ) to generate the template for the RT2 Profiler PCR Array Mouse Th1&Th2 Responses ( Qiagen ) . The PCR reaction was run on an ABI 7500 Real Time PCR System ( ThermoFisher Scientific , Darmstadt , Germany ) and subsequently analyzed with the SDS Software 1 . 4 . 25 following the provided instructions ( Qiagen ) . The CT cut-off was set to cycle 35 . The web-based Data Analysis Center provided by the manufacturer was used to generate comparative heat maps and scatter plots , and fold-change was calculated by determining the ratio of mRNA levels to control values using the ΔΔCt method . All data were normalized to an average of four housekeeping genes , Gusb , Hsp90ab1 , Gapdh , and Actb . A list of analyzed Th1&Th2 related genes is provided online by the manufacturer . For comparing data from more than two experimental groups , ANOVA was performed with additional Bonferroni correction to determine significant differences between the experimental groups . P values between two groups of interest were calculated by using the unpaired two-tailed Student’s t test with Welch’s correction of unequal variances . Differences between data sets are considered statistically significant ( * ) for P ≤ 0 . 05 , very significant ( ** ) for P ≤ 0 . 01 , and highly significant ( *** ) for P ≤ 0 . 001 . All analyses were performed with Graphpad Prism 6 . 04 ( GraphPad Software , San Diego , CA ) .
|
From an epidemiological perspective , natural host-to-host transmission of human CMV mostly occurs in early childhood through saliva from virus-shedding intimate contact persons , such as family members or peers . The oronasal route of transmission involves also the airway mucosa and airway-draining lymph nodes . Almost unavoidably , CMV airway infection coincides with airway exposure to environmental antigens , which include potent classical allergens but also protein antigens that have low-to-no allergenic potential on their own . Ovalbumin ( OVA ) , when administered as a purified protein , can serve as a well-studied model antigen for only poorly allergenic environmental antigens . In a murine model of airway exposure to OVA for allergenic sensitization , we have addressed the question if a simultaneous airway infection with murine CMV ( mCMV ) can promote allergic airway disease ( AAD ) elicited by challenge exposure to OVA . As anticipated by the model design , exposure to OVA alone did not sensitize for an allergic response to challenge exposure , nor did mCMV infection alone . Notably , based on viral activation of antigen uptake by DCs , both combined sensitized for a type 2 CD4+ T helper ( Th2 ) cell-driven AAD histopathology . This is a novel aspect in CMV pathobiology with the medical relevance that CMV airway infection enlarges the spectrum of environmental allergens .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"blood",
"cells",
"t",
"helper",
"cells",
"medicine",
"and",
"health",
"sciences",
"immune",
"cells",
"enzyme-linked",
"immunoassays",
"pathology",
"and",
"laboratory",
"medicine",
"respiratory",
"infections",
"immunology",
"cloning",
"pulmonology",
"animal",
"models",
"model",
"organisms",
"signs",
"and",
"symptoms",
"experimental",
"organism",
"systems",
"cytotoxic",
"t",
"cells",
"molecular",
"biology",
"techniques",
"immunologic",
"techniques",
"research",
"and",
"analysis",
"methods",
"white",
"blood",
"cells",
"inflammation",
"animal",
"cells",
"animal",
"studies",
"t",
"cells",
"immunoassays",
"mouse",
"models",
"molecular",
"biology",
"immune",
"response",
"diagnostic",
"medicine",
"cell",
"biology",
"biology",
"and",
"life",
"sciences",
"cellular",
"types"
] |
2019
|
Coincident airway exposure to low-potency allergen and cytomegalovirus sensitizes for allergic airway disease by viral activation of migratory dendritic cells
|
Changes in higher order chromatin organisation have been linked to transcriptional regulation; however , little is known about how such organisation alters during embryonic development or how it is regulated by extrinsic signals . Here we analyse changes in chromatin organisation as neural differentiation progresses , exploiting the clear spatial separation of the temporal events of differentiation along the elongating body axis of the mouse embryo . Combining fluorescence in situ hybridisation with super-resolution structured illumination microscopy , we show that chromatin around key differentiation gene loci Pax6 and Irx3 undergoes both decompaction and displacement towards the nuclear centre coincident with transcriptional onset . Conversely , down-regulation of Fgf8 as neural differentiation commences correlates with a more peripheral nuclear position of this locus . During normal neural differentiation , fibroblast growth factor ( FGF ) signalling is repressed by retinoic acid , and this vitamin A derivative is further required for transcription of neural genes . We show here that exposure to retinoic acid or inhibition of FGF signalling promotes precocious decompaction and central nuclear positioning of differentiation gene loci . Using the Raldh2 mutant as a model for retinoid deficiency , we further find that such changes in higher order chromatin organisation are dependent on retinoid signalling . In this retinoid deficient condition , FGF signalling persists ectopically in the elongating body , and importantly , we find that inhibiting FGF receptor ( FGFR ) signalling in Raldh2−/− embryos does not rescue differentiation gene transcription , but does elicit both chromatin decompaction and nuclear position change . These findings demonstrate that regulation of higher order chromatin organisation during differentiation in the embryo can be uncoupled from the machinery that promotes transcription and , for the first time , identify FGF as an extrinsic signal that can direct chromatin compaction and nuclear organisation of gene loci .
Differentiation is directed by extrinsic signals that regulate expression of transcription factors , which determine cell fates . A further critical level of regulation is provided by so-called higher order chromatin organisation , which includes changes in local chromatin compaction and nuclear position of gene loci . Such changes have been documented in in vitro differentiation assays , but this level of organisation has not been analysed as extensively during embryonic development and the role of signalling pathways in modulating chromatin and nuclear organisation in the developing embryo remains unexplored . During vertebrate embryonic development , induction of the future brain is followed by the progressive generation of neural tissue as the body axis elongates and this provides a unique opportunity to investigate steps leading to the onset of neural differentiation . New neural tissue arises from the stem zone/caudal lateral epiblast ( adjacent to the primitive streak ) , which includes resident axial stem cells [1] , [2] ( Figure 1A ) . As cells leave this regressing region they either ingress to form paraxial mesoderm or remain in the epiblast and commence neural differentiation . Stem zone cells are highly proliferative and are maintained by FGF and Wnt signalling [3] , [4] . This is attenuated by retinoid signals synthesised in the forming somites [3] , [5] , [6] ( Figure 1A ) . Retinoic acid ( RA ) promotes neural differentiation in at least two steps; first repressing FGF/Wnt signalling and then promoting expression , in the forming neural tube , of key genes characteristic of neural progenitors , such as Sox1 , Sox3 and Pax6 [3] , [7]–[9] . Importantly , FGF signalling also counteracts retinoid signalling , repressing expression of Raldh2 which encodes retinaldehyde dehydrogenase 2 - an RA synthesising enzyme - in the presomitic mesoderm and RA receptor beta ( RARb ) in the forming neuroepithelium as well as promoting expression of Cyp26a1 , encoding a RA catabolising enzyme reviewed in [1] ( summarised in Figure 1A ) . The underlying molecular mechanisms through which this opposing signalling switch controls differentiation onset in the embryonic body axis are not well understood , but might involve changes in gene expression determined by altered chromatin structure around target gene loci . One way in which chromatin compaction is locally regulated is by the action of polycomb repression complexes ( PRC ) ; and there is some evidence implicating FGF signalling in the regulation of polycomb complex component genes . Polycomb group ( PcG ) proteins are key regulators of cell growth and differentiation genes [10]–[12] and are found in two broad classes of complex; PRC2 , which mediates the histone modification H3K27me3 associated with transcriptional repression through the activity of the Ezh1/2 histone methyltransferase and PRC1 , which mediates local chromatin compaction [13] . In the zebrafish , the epiboly/tailbud phenotype of Ph2α morphants ( homologue of the PRC1 component polyhomeotic ) is similar to that of Fgf8 morphants , and Ph2α acts downstream of FGF signalling , which is necessary , although not sufficient for Ph2α expression [14] . Mice mutant for Fgf8 or for PcG genes ( Eed , Ezh2 or Ring1b ) also share a common gastrulation failure phenotype , with some reported proliferation defects [15]–[19] , suggesting conservation of a relationship between FGF signalling and polycomb function in the early embryo . Retinoic acid can signal directly to chromatin via liganded retinoic acid receptor – retinoid X receptor ( RAR-RXRs ) heterodimers and their sequence specific binding to retinoic acid response elements ( RAREs ) and this is known to attenuate binding of PRC2 components and to decrease H3K27me3 enrichment at these sites [20]–[22] . These observations suggest that in some contexts FGF may promote , while retinoid signalling represses , the action of polycomb complexes . Furthermore , as activation of polycomb target loci , such as the Hox gene clusters , is accompanied by visible unfolding of the compact state [23] , [24] , such signals might alter chromatin compaction at differentiation gene loci . Importantly , changes in chromatin compaction and local organisation are not simply a consequence of transcription; experimental translocation of a 3′ Hoxb1 transgene to the 5′ end of the Hoxd cluster elicited such chromatin changes in a cellular context in which Hoxb1 is not transcribed [25] . This phenomenon shows that alteration of chromatin organisation can prefigure gene transcription . A further important manifestation of higher order chromatin organisation that frequently correlates with transcription is the position of a locus with respect to the nuclear periphery , which can be a repressive environment . Although recent studies have shown that artificial tethering to the nuclear periphery need not necessarily lead to gene silencing [26]–[28] , many loci do exhibit a change in distance to the nuclear edge , and association with the nuclear lamina [29] which correlates with their potential for transcription . As extrinsic signals orchestrate development by directing gene transcription , it is likely that this involves regulation of such higher order organisation , however , it is not known whether particular signalling pathways direct such mechanisms nor whether they can elicit changes in chromatin organisation independently of transcriptional regulation . To assess changes in higher-order chromatin organisation during the progressive generation of neural tissue in the elongating body axis of the mouse embryo , we used fluorescence in situ hybridisation ( FISH ) combined with super-resolution structured illumination microscopy ( SIM ) . We analysed changes in higher-order chromatin organisation at the loci of exemplar neural progenitor genes Pax6 and Irx3 as differentiation takes place and compared this with the Fgf8 locus , which is transcriptionally downregulated as cells leave the stem zone ( Figures 1B–B‴ 1C–C‴ ) . As retinoid signalling is required for transcription of differentiation genes ( including that of Pax6 ) we analysed chromatin organisation around loci in the Raldh2−/− mutant embryo , which is unable to synthesize RA in the elongating embryonic body axis [30] . In retinoid deficient embryos caudal FGF signalling expands rostrally from the stem zone [3] , [6] and by blocking FGFR signalling in Raldh2 mutants we dissected the consequences of FGF loss in a context in which many differentiation genes fail to be transcribed . Our data demonstrate , for the first time , that FGF signalling acts upstream of mechanisms that direct higher-order chromatin organisation around differentiation gene loci and further reveal that such mechanisms can be uncoupled from the machinery that mediates transcription of such genes .
To determine if the onset of Pax6 transcription in the E8 . 5 mouse embryo involves a change in chromatin compaction , fosmid probes separated by 65 kb and specific for sequences flanking Pax6 ( Table S1 ) were used for 3D FISH on wildtype CD1 mouse embryos ( Figure 2A ) . Images were captured using SIM and inter-probe distances , ( d ) in µm , were measured in transverse sections of the stem zone and preneural tube ( which lack Pax6 transcription ) and in the neural tube , where Pax6 is now transcribed ( excluding Pax6 negative cells in dorsal and ventral most positions ) ( Figures 1B–B′ , 2A , B ) . Fosmids were also used to measure changes around a control locus , alpha-globin ( Hba-a1 ) , which is not transcribed in the embryo at this stage [31] ( Figures 2C , D ) . Chromatin compaction was assessed by a comparison of d2 values for each data set , as this is the value that scales linearly with genomic separation [32] and that has been used previously to identify differences in chromatin compaction between cells at different stages of differentiation [24] and between wildtype and mutant cells [13] ( Figure 2E ) ( see Methods for data set collection ) . There was no statistical difference between d2 values for stem zone and preneural tube ( p>0 . 05 ) , but a clear increase in inter-probe distances across Pax6 in neural tube in comparison with measurements in either stem zone or preneural tube nuclei ( p<0 . 05 , Figures 2B , E , Tables S2 , S3 ) . Additionally , in recently formed somites , which lie adjacent to the neural tube and which do not and will not express Pax6 , chromatin across the Pax6 locus is as compact as it is in the stem zone and preneural tube , and significantly more compact than in neural tube ( Figures 2B , E ) . In contrast , inter-probe distances around a control gene locus ( alpha-globin , Hba-a1 ) were not significantly different between stem zone , neural tube and somite data sets ( p>0 . 05 , Figures 2D , F ) . This controls for any overall change in chromatin condensation at the onset of differentiation . These data therefore indicate that chromatin is compact around the Pax6 locus in cells that do not and will not express this gene ( somites ) and those that will later come to express it ( stem zone and preneural tube ) , and that it specifically decompacts coincident with onset of Pax6 transcription in the neural tube . Analysis of genome-wide histone modification data sets in mouse ES cells and derived neural progenitors [11] further reveals that the Pax6 locus is subject to H3K27me3 enrichment in ES cells and is relieved of this mark upon neural differentiation , suggesting that this locus is a target of polycomb mediated repression ( Figure S1 ) . We further assessed chromatin organisation across the locus of the gene Fgf8 , which is expressed in the stem zone and downregulated as neural differentiation commences . Fgf8 is marked by H3K27me3 and H3K4me3 in mES cells and upon neural differentiation H3K4me3 is lost and H3K27me3 retained [11] ( Figure S1 ) . Fgf8 is a smaller gene than Pax6 and FISH signals from fosmids flanking this locus ( 100 kb separation ) were barely resolved in any tissue assessed ( Figure S2 ) . These findings suggest that polycomb group proteins regulate Fgf8 expression , but in a manner that does not involve visibly detectable chromatin compaction . Neighbouring genes Npm3 and Mgea5 show similar patterns of gene expression as Fgf8 ( Figure S3 ) , but not of histone modifications ( Figure S1 ) . This suggests that PRC-mediated chromatin compaction around the Fgf8 locus may be too subtle , or masked by the chromatin environments of neighbouring genes , to be detected by FISH . To investigate the potential relationship between the position of a gene within the nucleus and its transcriptional activity along the embryonic body axis , we analysed the proximity of FISH signals for Pax6 and or Fgf8 ( Figure 3A ) to the nuclear periphery as defined by DAPI staining . The Pax6 locus is closer to the nuclear periphery in the stem zone than in the neural tube ( p<0 . 05; Figures 3B , B′ ) , whereas the converse is the case for Fgf8 ( Figures 3C , C′ ) . The relative nuclear position of the control Hba-a1 locus was the same in the stem zone and neural tube ( Figures 3D , D′ ) . These data show that nuclear position correlates well with transcription of Pax6 and Fgf8 in the normal embryo . Pax6 is not expressed in the neural tube of mouse embryos lacking the RA synthesising enzyme retinaldehyde dehydrogenase 2 ( Raldh2−/− ) [9] , [30] ( Figures 4A , B ) . To determine whether this is also accompanied by failure to undergo changes in higher order chromatin organisation , FISH with probe pairs across the Pax6 locus was carried out on E8 . 5 Raldh2−/− embryos . There was no difference in chromatin compaction ( d2 ) between stem zone and neural tube ( p>0 . 05; Figure 5C ) indicating that chromatin decompaction , normally observed across the Pax6 locus in the wildtype neural tube ( Figures 2A , B ) , does not take place in this retinoid deficient condition . Indeed , the distribution of inter-probe distances in Raldh2−/− neural tube nuclei was similar to that found in stem zone of wildtype mice ( p>0 . 05; Figure 4C ) . The Pax6 locus also remained compact in somites of wildtype and mutant animals ( Figures 4 C , D , E ) . The absence of retinoid signalling also resulted in a failure of Pax6 to reposition away from the nuclear periphery in the neural tube compared to stem zone ( Figure 4F ) . Moreover , Pax6 is more peripherally located in the Raldh2−/− neural tube than in the wildtype neural tube ( p<0 . 05; Figures 4D , E and F ) . These data show that , for the Pax6 locus , neither chromatin decompaction nor a shift away from the nuclear periphery take place in the retinoid deficient neural tube in which Pax6 is not transcribed . To determine whether exposure to retinoic acid leads to decompaction and a more central nuclear position of the Pax6 locus we treated explanted caudal regions with retinoic acid or vehicle DMSO control for 10 h ( Figure 5A ) . Explants were then processed either for in situ hybridisation to monitor Pax6 transcription or for FISH to assess local chromatin organisation ( Figure 5B ) . This confirmed that retinoic acid induces Pax6 expression ( Figures 5B , B′ ) and demonstrated that this correlates with the decompaction and more central nuclear location of this locus ( Figures 5C , D; p<0 . 05 and p<0 . 05 , respectively ) . FGF signalling ectopically persists in the preneural tube of retinoid deficient quail embryos [3] and in the neural tube of Raldh2−/− mouse embryos [5] , [6] . As FGF signalling represses onset of expression of neural differentiation genes , including Pax6 , in the elongating body axis [3] , [33] , it is possible that failure to express Pax6 in the Raldh2 mutant is due to an excess of FGF signalling . To determine whether FGF signalling represses differentiation onset via a mechanism that involves regulation of higher order chromatin organisation , FGF signalling was blocked with the FGFR inhibitor PD173074 [34] . Explanted whole E8 wildtype embryos were cultured in vitro exposed to either DMSO vehicle control or PD173074 for 7 h and then processed for FISH , or analysed for expression of the FGFR pathway target Sprouty2 ( Spry2 ) [35] and for Pax6 . The repression of Spry2 ( DMSO n = 0/5; PD173074 n = 5/5 , Figure S4A ) confirmed the effective blocking of FGFR signalling ( and see [36] ) . Inhibition of FGFR signalling in the elongating neural axis also leads to precocious onset of Pax6 expression , which is then detected more caudally in the preneural tube in the chick embryo [33] . Consistent with this , Pax6 transcripts were detected in the preneural tube of PD173074 treated mouse embryos ( DMSO n = 0/4; PD173074 n = 3/4 , Figures 6A , B ) . Analysis of chromatin compaction across the Pax6 locus by FISH ( Figure 6C ) revealed that , unlike the situation in untreated ( Figure 2 ) and control ( DMSO treated ) embryos where Pax6 chromatin was more compact ( smaller inter-probe distances ) in stem zone and somites than in neural tube , this difference was abolished in PD173074-treated embryos . In these conditions the chromatin across Pax6 appears to decompact in stem zone and the somites to the level normally seen in the neural tube ( p<0 . 05; Figures 6C , D , E ) . This indicates that FGFR signalling normally promotes chromatin compaction around Pax6 in caudal regions and somites ( see Discussion ) . Blocking FGFR signalling also promoted a shift in Pax6 localisation towards the nuclear centre in stem zone and somitic nuclei in comparison with DMSO control ( p<0 . 05 and p<0 . 05 respectively; Figures 5D , E , and F and S7A ) . No significant change in compaction or nuclear position in control and PD173074 treated embryos was seen at the control Hba-a1 locus , indicating that changes in chromatin organisation around the Pax6 locus do not reflect a general consequence of FGFR inhibition ( Figures S4B , C ) . Together these data indicate that FGF signalling acts upstream of mechanisms that regulate chromatin compaction and nuclear position at Pax6 . To extend this analysis we assessed chromatin organisation around the locus of an additional neural progenitor marker gene , Irx3 . Like Pax6 , onset of Irx3 transcription takes place in the neural tube of the elongating body and is initially broadly expressed across the dorso-ventral axis [37] . Blocking FGFR signalling led to a caudal expansion of the Irx3 expression domain ( n = 0/5 DMSO treated and n = 4/6 PD173074 treated embryos ) ( Figures 7A , B ) . Also like Pax6 , the Irx3 locus is a PRC target in ES cells [11] ( Figure S1 ) . Using fosmids flanking Irx3 ( Figure 7C ) FISH analysis confirmed that this region of chromatin decompacts and relocates towards the nuclear centre in the neural tube coincident with its transcription ( Figures 7D , E ) . Furthermore , blocking FGFR signalling led to decompaction and a more central nuclear position of the Irx3 locus in stem zone nuclei and also somites ( Figures 7D , E , S7B ) . These data demonstrate that FGF signalling consistently acts upstream of chromatin re-organisation at differentiation gene loci . To determine whether excess of FGF signalling is responsible for the lack of Pax6 expression in the absence of retinoid signalling , Raldh2 mutant embryos , and their wild type littermates produced from heterozygous Raldh2+/− crosses , were cultured with PD173074 or DMSO . Strikingly , blocking FGFR signalling did not rescue Pax6 expression in the neural tube of Raldh2−/− embryos ( Pax6 mRNA was detected in neural tube of 0/4 mutant embryos and 0/4 PD173074 treated mutant embryos; Figures 8A , B ) . Attenuation of FGFR signalling in this condition is therefore not sufficient for onset of Pax6 expression and is consistent with a further requirement for retinoid signalling to promote neural differentiation [3] . This finding does , however , raise the possibility , that blocking FGFR signalling in retinoid deficient conditions still promotes initial steps in the differentiation process upstream of Pax6 transcription and this perhaps includes chromatin re-organisation . Indeed , FISH revealed that the Pax6 region decompacts in the stem zone and in the neural tube of PD173074 treated Raldh2−/− embryos compared to the control DMSO-treated mutant embryos ( p<0 . 05 for both comparisons; Figures 8C , D , E ) . Blocking FGFR signalling in wildtype or in Raldh2 mutant embryos also decompacts chromatin in somites , despite the absence of Pax6 expression in this tissue ( p<0 . 05 for both comparisons; Figures 8C , D , E ) ( see below ) . Similarly , blocking FGFR signalling in this context also induced a shift towards the nuclear centre of the Pax6 locus in both stem zone and neural tube; and this relative position is similar to that seen in the neural tube of wildtype or DMSO-treated embryos ( p>0 . 05 , Figure 8F ) . In this context , FGFR signalling therefore acts upstream of mechanisms that direct both local chromatin compaction and nuclear position and that can be uncoupled from the activity of retinoid mediated transcription factor complexes that are required to promote expression of neural differentiation genes such as Pax6 . Although we could not use FISH to measure chromatin compaction at the Fgf8 locus , this approach can be used to assess nuclear position . Consistent with rostral expansion of Fgf8 transcription in such mutants [6] ( Figures 9A , A″ ) , the Fgf8 locus fails to locate towards the nuclear periphery in the Raldh2 −/− neural tube , ( p<0 . 05 in comparison with wildtype , Figures 9B , S5 ) . Here , nuclear position therefore correlates with changing Fgf8 expression and these findings indicate that retinoid signalling is upstream of mechanism ( s ) that directs nuclear position of the Fgf8 locus . It is further possible that the location of the Fgf8 locus is influenced by FGF signalling itself , as transcription of Fgf genes can be maintained by positive auto-regulatory feedback loops e . g . [38] , [39] . To address this possibility we blocked FGFR signalling in wildtype and Raldh2−/− mutants ( Figures 9A″ , A‴ , B′ , B″ ) . In both conditions this led to a more peripheral localisation of the Fgf8 locus in the stem zone ( where this gene is expressed ) in comparison with wildtype and Raldh2−/− DMSO controls . Strikingly , in the Raldh2 −/− mutant neural tube blocking FGFR signalling also rescued the failure to shift to the nuclear periphery observed in untreated mutants ( Figures 9B–B″ , S5 ) . Fgf8 transcripts are still detected in PD173074 exposed embryos ( Figures 9A″ , A‴′ ) and this may reflect the known stability of Fgf8 mRNA [40] , ( although some intronic Fgf8 transcripts were detected in the stem zone of PD173074 treated embryos by whole mount in situ hybridisation ( Figure S6 ) , indicating that not all active Fgf8 transcription is lost ) . Overall , these findings demonstrate that FGFR signalling regulates nuclear position of the Fgf8 locus and that it is responsible for the persistent central location of this gene in the retinoid deficient neural tube .
We found that chromatin decompaction around Pax6 and Irx3 correlate with their transcriptional activation in the newly generated neural axis . The decompaction observed at these loci is reminiscent of that seen upon activation of Hox loci during embryonic development [23] , [24] , where chromatin compaction of the silent loci has been shown to be mediated by the PRC1 polycomb complex [13] . Decompaction around Pax6 and Irx3 in the neural tube nuclei corresponds to the first transcription of these genes during development and is consistent with polycomb regulation as indicated by the presence of H3K27me3 [11] and also the PRC1 protein Ring1b at the Pax6 locus in ES cells [41] . H3K27me3 is also associated with the Fgf8 locus in ES cells [11] . As we detect no increase in chromatin compaction when Fgf8 is transcriptionally downregulated , it is possible that in the embryo this does not involve polycomb mediated repression . However , Fgf8 , but not neighbouring genes ( NPM3 and Mgea5 ) , has associated H3K27me3 in ES cells ( Figure S2 ) and we cannot exclude that compaction local only to Fgf8 is not detected by our FISH assay . We found that retinoid signalling is sufficient and necessary for chromatin decompaction around the Pax6 locus . Pax6 is not a direct target of the RAR/RXR transcriptional complex , but has an important cross-regulatory relationship with the proneural gene Ngn2 which is a direct target [8] , [42]–[44] . However , RA signalling has been linked directly to the loss of PRC2 binding and H3K27me3 due to its promotion of MSK1/2 mediated phosphorylation of H3S28 in an embryonic carcinoma cell assay [45] . This modification is adjacent to H3K27 and correlates with the loss of PcG binding and loss of repression at a subset of PRC target genes . Nevertheless , RA is just one of several extrinsic signals that can promote MSK1/2 activity in vitro and mice null for both MSK1 and 2 are viable and fertile [46] , suggesting that regulation of MSK1/2 is not a key endogenous mechanism for removal of polycomb-mediated repression in development . Instead , our data indicate that the requirement for retinoid signalling in the embryo is for removal of FGF signalling , which is in turn responsible for compaction around Pax6 and Irx3 . As well as undergoing chromatin decompaction , Pax6 and Irx3 also relocates to a more central position in the nucleus in the neural tube . For Pax6 , this does not occur in the absence of retinoid signalling . Fgf8 shows the converse pattern of nuclear movements , but in the apparent absence of chromatin compaction changes , suggesting that nuclear position and local chromatin organisation are regulated by distinct molecular mechanisms . The relocation of Fgf8 toward the edge of the nucleus in the neural tube is blocked in Raldh2 mutants which correlates with its ectopic transcription [6] . Fgf8 has upstream RAREs indicating that it may be directly repressed by RA signalling [47] , [48] and our finding here suggests that this may be linked to nuclear positioning mechanisms . Importantly , although nuclear position of the Fgf8 locus correlated well with transcription of this gene , we found no cellular context ( including wildtype , FGFR or retinoid signalling deficient conditions ) in which chromatin decompaction ( observed around Pax6 and Irx3 ) took place without a concomitant shift towards the nuclear centre . This suggests that de-compaction may be contingent upon a more central nuclear position . Association with nuclear lamins correlates well with location of genomic regions to the nuclear periphery and so-called Lamin associated domains ( LADs ) generally have low levels of transcription . Genome-wide maps of Lamin B1 association in mouse ES cells show that many neural differentiation genes are located in LADs [29] . However , neither the Pax6 nor Irx3 loci , nor the whole Fgf8 region is a LAD in either ES cells or ES-derived neural progenitors ( Figure S8 ) suggesting that at least in this in vitro context nuclear re-positioning of these loci is unlikely to be mediated by altered Lamin B1 association . However , while Lamin B genes appear not to be required for differentiation in ES cells , Lamin B null mice do exhibit profound neural defects [49] , [50] . These include both neural progenitor proliferation and nuclear lamina integrity , and so a role for LAD mediated regulation of nuclear positioning of neural differentiation genes in the embryo cannot be ruled out [49] , [50] . Our discovery that blocking FGFR signalling in the Raldh2 mutant , where many differentiation genes fail to be transcribed , restores chromatin decompaction and a more central nuclear position of the Pax6 locus suggests that RA acts first to inhibit FGFR signalling and that FGF is upstream of molecular mechanism ( s ) that direct higher order chromatin organisation at differentiation gene loci . This result is further supported in wildtype embryos in which FGFR signalling is blocked , as here Pax6 transcription extends caudally , being precociously expressed in the preneural tube , but we detect Pax6 decompaction and its more central nuclear position in stem zone cells , which have yet to express Pax6 . Importantly , these experiments uncouple regulation of higher order chromatin organisation from gene expression itself and indicate that these large-scale chromatin changes take place as an initial step in the differentiation process . Although we assess these changes by detailed investigation around two exemplar neural differentiation genes , FGF signalling in this context represses expression of many such genes , including Sox1 and Sox3 ( further regulators of the neural progenitor cell state ) and prevents the onset of ventral patterning and neuron production in the newly generated spinal cord [3] , [7] . It is therefore likely that FGF signalling ( directly or indirectly ) regulates a general mechanism ( s ) that determines chromatin organisation at such differentiation genes , many of which are known PRC2 targets in ES cells . Intriguingly , blocking FGFR signalling also led to decompaction and a more central nuclear position of Pax6 and Irx3 loci in somites , where this gene is never expressed . These somites will have formed ( 1 somite every 2 hours ) during the 7 h period of exposure to FGFR inhibitor , at the start of which these cells would have been experiencing FGF signalling in the presomitic mesoderm . The reorganisation of Pax6 and Irx3 loci in this context may thus reflect the finding that high level FGF signalling is required for mesoderm induction , while reduction elicits neural differentiation , as observed in Fgfr1 mutant mice , reviewed in [1] . Sudden loss of FGFR signalling in the early presomitic mesoderm might therefore elicit initial steps in neural differentiation . Importantly , we show that blocking FGFR signalling does not lead to global chromatin reorganisation , as inter-probe distances and the fractional radius for control Hba-a1 locus and inter-probe distance for the region of the Fgf8 locus remain unchanged in all tissues examined . The Fgf8 locus does , however , alter its nuclear position in response to changes in FGFR signalling . When FGFR signalling is blocked in either wildtype or Raldh2 mutant embryos the Fgf8 locus remains close to the nuclear periphery in all tissues examined , including the stem zone where this gene is normally expressed and in the Raldh2 −/− mutant , where this inhibition of FGF signalling rescues the ectopic centralised location of Fgf8 locus . Although location at the nuclear periphery generally correlates with gene repression , we do detect some intronic Fgf8 transcripts in PD173074 treated embryos , indicating that in the timeframe of this experiment the peripheralisation of the Fgf8 locus does not simply correlate with loss of transcription . This may reflect an initial heterogeneous response to the loss of FGF signalling across the stem zone cell population , however , active transcription and peripheral locus position it is not incompatible with transcription [27] , [28] . Overall then , FGF is upstream of mechanisms in the stem zone that lead to Pax6 and Irx3 compaction and peripheral location , and that promote a central position of Fgf8 within the nucleus . This shows that in this context FGF signalling influences multiple distinct molecular mechanisms , which regulate chromatin compaction and promote movement towards or away from the nuclear centre in a locus specific manner . Attenuation of FGF signalling in human embryonic stem ( hES ) cells and mouse epiblast stem cells leads to loss of self-renewal [51]–[53] . Furthermore , as observed in the elongating embryonic neural axis [33] and in mouse ES cells that have experienced a period of endogenous FGF/Erk [7] , inhibition of FGF/Erk signalling in hES cells induces rapid expression of Pax6 [53] . The attenuation of FGF signalling in stem cells of epiblast origin and in multipotent epiblast cells located in the stem zone/caudal lateral epiblast therefore serves as a common trigger for onset of differentiation and it is likely that conserved molecular mechanisms that include relief from polycomb mediated repression at differentiation genes underlie this initial step . Key future tasks are to determine how FGF signalling regulates local chromatin compaction and orchestrates nuclear positioning to constrain cell differentiation .
Wildtype CD1 embryos were collected at E8 . 5 , dissected , fixed and processed for in situ hybridization ( ISH ) or for FISH as described below . Heterozygous Raldh2 mutant CD1 mice [30] were crossed to generate litters at E8-8 . 5 containing Raldh2−/− , Raldh2+/− and wildtype embryos . These were either dissected , genotyped as described previously [30] , fixed and processed for ISH or FISH ( see below ) , or E8 embryos within yolk sacs were collected in warmed ( 37°C ) culture medium ( rat serum , tyrode solution; 1∶1 ) containing control DMSO ( 0 . 5 µl/1 ml culture medium ) or FGFR inhibitor PD173074 ( Calbiochem ) at 50 µM . Embryos were then cultured for 7 hours in a water-saturated roller-tube incubator at 37°C in 5% CO2 , 20% O2 . These were then dissected , genotyped , fixed and processed for FISH . For treatment with retinoic acid wild type CD1 E8-8 . 5 embryos were dissected to give explants pairs of the caudal embryo ( Figure 5A ) with one explant treated with 250 nM RA and the other DMSO vehicle control cultured in collagen as previously described [3] for 10 h . Explants were then fixed in 4% PFA and processed for ISH or FISH . For FISH analysis nuclei in sections taken from the central third of each explant were measured ( 5 explant pairs , >30 nuclei per explant measured ) for inter-probe distance and fractional radius . Initial analyses compared differences between treated and untreated explants taken from the same embryo and these were all significantly different ( Table S4 ) . We therefore pooled all treated and all untreated explant data ( Figures 5C , E ) . All procedures using animals were performed in accordance with UK and French legislation and guidance on animal use in bioscience research . Standard procedures were used to carry out in situ hybridisation in whole embryos to detect mRNAs for Pax6 , Irx3 , Fgf8 , Spry2 , Npm3 and Mgea5/OGA ( primers used to clone Irx3 , Npm3 and Mgea5/OGA can be found in Figure S3 ) . A subset of these were embedded and cryo-sectioned to visualise mRNA localisation at a cellular level . Intronic Fgf8 was detected using a probe for the region between exons 5 and 6 of the mouse Fgf8 gene ( a kind gift from Olivier Pourquie , [40] ) . Mouse embryos stored in 100% MeOH were cleared in xylene , embedded in wax , sectioned at 7 microns and dried down on thin TESPA-coated 50×22 #1 . 5 coverslips ( Scientific Laboratory Supplies Ltd ) suitable for OMX microscopy . The protocol for FISH on mouse tissue sections was then adapted from [54] . Coverslips with sections were heated to 65°C ( 20 min ) , washed ×4 in xylene ( 10 min ) and re-hydrated through an ethanol series to dH20 . Coverslips were then microwaved for 20 min in 0 . 1 M citrate buffer , pH6 . 0 , cooled in buffer ( 20 min ) washed and stored in dH20 prior to pre-hybridisation steps and denaturation as previously described [54] . Fosmids pairs separated by inter-genomic distance of 60–120 kb were selected from the WIBR-1 Mouse Fosmid Library ( Whitehead Institute/MIT Center for Genomic Research ) and sequences confirmed by targeted PCR ( Table S1 , Figure S9 ) . These were then labelled with either digoxigenin-11-dUTP or biotin-16-dUTP by nick transcription . Approximately 150 ng probe along with 15 µg mouse Cot1 DNA ( Invitrogen ) and 5 µg sonicated salmon sperm DNA ( sssDNA ) were used per coverslip , denatured and hybridised to coverslips [54] . After overnight incubation and washing , digoxigenin labelled probes were detected with anti-dig FITC ( 1∶20 , Roche ) and amplified with anti-sheep Alexa Fluor 488 ( 1∶100 , Molecular Probes ) ; biotin labelled probes with biotinylated anti-avidin ( 1∶100 ) and Alexa streptavidin 594 ( 1∶500 , Molecular Probes ) . Nuclei were counterstained with DAPI and coverslips mounted onto slides with 25 µl of Slowfade Gold ( Molecular Probes ) . Samples were imaged on a Deltavision 3D OMX Structured Illumination Microscope ( Applied Precision ) using a protocol adapted after [55] . Regions of interest ( ROIs ) were identified using a Deltavision microscope , mapped using Softworx ( Applied Precision ) and acquired with a UPlanSApochromat 100× 1 . 4 NA oil-immersion objective lens ( Olympus ) and back-illuminated Cascade II 512×512 EMCCD camera ( Photometrics ) on the OMX version 2 system ( Applied Precision ) equipped with 405 , 488 , and 593 solid-state lasers . Samples were illuminated by a coherent scrambled laser light source that had passed through a diffraction grating to generate the structured illumination . Potential photo-bleaching was minimised by using lowest possible laser power and exposure times ( 50 and 250 ms ) . Raw images were processed and reconstructed using the Softworx structured illumination reconstruction tool ( Applied Precision ) [56] . The 405 , 488 and 593 channels were then aligned in x and y , using predetermined shifts which were measured using a target lens and 100-nm Tetraspeck fluorescent beads ( Invitrogen ) in the Softworx alignment tool ( Applied Precision ) . For analysis of chromatin compaction and nuclear position , measurements were made in images of >50 nuclei per region in each of 3 different embryos per condition . Stem zone was defined as epiblast cells adjacent and just caudal to the node ( ∼5 sections per embryo ) , preneural tube as neuroepithelium rostral to the node underlain by notochord and presomitic mesoderm , neural tube as neuroepithelium flanked by 2 or 3 most recently formed somites , and these adjacent somites were also used to represent somitic tissue . As nuclei in tissues are not as spherical as in cultured cells it was not possible to apply standard nuclear segmentation tools to define nuclear position . Instead sections in which a fosmid signal and nuclear edge were in sharp focus were used to measure the shortest distance from the probe centre to the periphery . The broadest distance across the nucleus was also measured as an indication of nuclear diameter and this was halved and data presented as a proportion the nuclear radius ( fractional radius ) . Super resolution images were uploaded into an OMERO server ( Open Microscopy Environment ) and ROIs containing hybridisation signals for both dig and biotin-labelled probes were identified by manual inspection in OMERO-insight . ROIs typically extended over several z-sections to accommodate the whole volume of the signals . These ROIs were analysed by a custom script developed in MATLAB ( Michael Porter , University of Dundee ) . This script first segments the objects defined by each probe from the background using Otsu thresholding and then calculates the xyz coordinates the centroid in each object . The centroids of these two objects and the distance between them , d ( µm ) , were then output to a spread-sheet . The inter-probe distance was then squared because in interphase nuclei the mean physical distance squared between two points is linearly related to the known genomic distance [32] . Within each nucleus , the line measurement tool was used to determine the distance of the edge of the nucleus from the hybridisation signal of the biotin-labelled probe in sections in which it was in sharp focus and this was then averaged . The radius of the nucleus , also measured with the line measurement tool , was then divided by this distance . This gave the distance of the gene locus from the nuclear periphery as a proportion of nuclear size . Box plots in figures show distribution of data . Top and bottom whiskers show highest and lowest data points respectively . Top and bottom lines of box represent 3rd and 1st inter-quartiles and the middle line represents the median . Non-parametric Mann-Whitney U test was used for analyses as data were not normally distributed . For comparison between explant pairs , a paired-sample Wilcoxon signed-rank test was used ( Table S4 ) .
|
Changes in the position of genes within the nucleus and in their local organisation frequently correlate with whether or not genes are turned on . However , little is known about how such nuclear organisation is controlled and whether this can be separated from the mechanisms that promote transcription . We show here that central nuclear position and chromatin de-compaction correlate with onset of expression at key neural differentiation gene loci in the mouse embryo . Conversely , the locus of a gene that is down-regulated as neural differentiation commences exhibits a shift towards the nuclear periphery as this takes place . Importantly , we show that signalling through the fibroblast growth factor ( FGF ) pathway regulates changes at this level of nuclear organisation . FGF represses differentiation gene transcription and keeps differentiation gene loci compact and at the nuclear periphery . By blocking FGF signalling in a retinoid deficient embryo in which differentiation genes are not expressed , we further show that control of nuclear organisation by FGF is not just a consequence of gene transcription . These findings are the first to demonstrate that such higher order nuclear organisation is regulated in the developing embryo , that this takes place downstream of FGF signaling , and can be uncoupled from the machinery of gene transcription .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"developmental",
"biology",
"gene",
"expression",
"genetics",
"epigenetics",
"biology",
"molecular",
"cell",
"biology"
] |
2013
|
FGF Signalling Regulates Chromatin Organisation during Neural Differentiation via Mechanisms that Can Be Uncoupled from Transcription
|
Fusarium graminearum , the causal agent of Fusarium head blight in cereal crops , produces sexual progeny ( ascospore ) as an important overwintering and dissemination strategy for completing the disease cycle . This homothallic ascomycetous species does not require a partner for sexual mating; instead , it carries two opposite mating-type ( MAT ) loci in a single nucleus to control sexual development . To gain a comprehensive understanding of the regulation of sexual development in F . graminearum , we used in-depth and high-throughput analyses to examine the target genes controlled transcriptionally by two-linked MAT loci ( MAT1-1 , MAT1-2 ) . We hybridized a genome-wide microarray with total RNAs from F . graminearum mutants that lacked each MAT locus individually or together , and overexpressed MAT1-2-1 , as well as their wild-type progenitor , at an early stage of sexual development . A comparison of the gene expression levels revealed a total of 1 , 245 differentially expressed genes ( DEGs ) among all of the mutants examined . Among these , genes involved in metabolism , cell wall organization , cellular response to stimuli , cell adhesion , fertilization , development , chromatin silencing , and signal transduction , were significantly enriched . Protein binding microarray analysis revealed the presence of putative core DNA binding sequences ( ATTAAT or ATTGTT ) for the HMG ( high mobility group ) -box motif in the MAT1-2-1 protein . Targeted deletion of 106 DEGs revealed 25 genes that were specifically required for sexual development , most of which were regulated transcriptionally by both the MAT1-1 and MAT1-2 loci . Taken together with the expression patterns of key target genes , we propose a regulatory pathway for MAT-mediated sexual development , in which both MAT loci may be activated by several environmental cues via chromatin remodeling and/or signaling pathways , and then control the expression of at least 1 , 245 target genes during sexual development via regulatory cascades and/or networks involving several downstream transcription factors and a putative RNA interference pathway .
Fusarium graminearum , a homothallic ( self-fertile ) ascomycetous fungus , causes serious diseases ( e . g . , Fusarium head blight ) in major cereal crops , and produces several mycotoxins in diseased cereals [1] . Recently , this species was defined as a member of the F . graminearum species complex , which consists of more than 16 phylogenetically distinct species found worldwide [2–7] . To complete the recurrent cycle of cereal diseases , F . graminearum produces sexual progeny ( ascospores ) on cereal debris as overwintering propagules [8] . The sexual reproduction of F . graminearum is controlled by master regulators called mating-type ( MAT ) loci [9 , 10] . Unlike their heterothallic relatives , F . graminearum carries two closely linked MAT loci ( MAT1-1 , MAT1-2 ) . A single nucleus contains individual MAT genes in a structural organization ( MAT1-1-1 , MAT1-1-2 , MAT1-1-3 at the MAT1-1 locus; MAT1-2-1 at the MAT1-2 locus ) similar to that of other Sordariomycetes ( e . g . , Neurospora crassa , Podospora anserina , Sordaria macrospora ) [9 , 10] . All of the MAT genes encode transcription factors that carry conserved DNA-binding motifs called an alpha box ( MAT1-1-1 ) , an HMG-box domain ( MAT1-1-3 , MAT1-2-1 ) , and a PHP domain ( MAT1-1-2 ) [9 , 10] . The importance of individual MAT transcripts and MAT loci for sexual development has been intensively studied in F . graminearum , but their functional requirement is not conserved among other fungal species . All of the four individual MAT genes at the MAT loci are essential for sexual development in F . graminearum [11–14] , whereas SmtA-1 and SmtA-3 ( comparable to MAT1-1-1 and MAT1-1-3 , respectively ) are dispensable for fruiting body ( perithecium ) formation in homothallic S . macrospora [15] . In heterothallic species , MAT1-1-2 is essential for perithecium formation in P . anserina , but has a redundant function together with MAT1-1-3 in N . crassa [15] . The phenotypic changes caused by MAT deletions and gene expression patterns in F . graminearum strongly suggest that MAT genes are involved in both the early and late stages of sexual development [13 , 14] . In contrast , the prominent roles of MAT genes in heterothallic species are to maintain the sexual identity of cells that express the opposite MAT gene for mating ( i . e . , controlling sexual compatibility ) and to regulate pheromone-mediated signaling pathways . Recently , an additional transcript ( MAT1-2-3 ) with no DNA-binding motif was identified in the MAT1-2 locus [16] , but its role ( s ) in sexual development is not essential in F . graminearum [14] . MAT transcriptional factors may control the transcriptional expression of downstream genes that are necessary for sexual development in filamentous fungi . Several transcriptional profiling analyses have been performed to identify MAT loci target genes that are differentially expressed in fungal strains lacking MAT genes during sexual development [17–20] . However , the function and related regulatory pathways of MAT-target genes have not been sufficiently elucidated to allow a comprehensive understanding of sexual development under the control of the MAT loci; only homology-based functional categorization and limited information regarding gene function ( such as pheromone/receptor genes ) are available . Very recently , putative target genes of a fungal mating-type gene ( MAT1-1-1 ) carrying a DNA binding alpha box domain were identified by a genome-wide search using chromatin immunoprecipitation combined with next-generation sequencing ( ChIP-seq ) in Penicillium chrysogenum , but only a limited number of target genes were functionally characterized [21] . In F . graminearum , genome-wide transcriptional analyses during perithecium development have been also performed using microarrays [22] and RNA-sequencing ( RNA-seq ) technology [23] , but the functions of most of the highly expressed genes remain unclear . Despite intensive investigation of MAT genes in filamentous fungi , many questions regarding sexual developmental processes regulated by MAT genes remain unanswered . Various cellular and developmental events occur during sexual reproduction in ascomycetes: ascogonium formation , fertilization , nuclear migration and proliferation in ascogonium , nuclear recognition and fusion in dikaryotic hyphae , meiosis , and ascus/ascospore formation . However , little is known about the specific roles of MAT genes in these sexual stages , particularly those after fertilization , although pheromone/receptor-mediated fertilization under control of MAT is well-established in heterothallic species . In homothallic species , the function of MAT loci that are present within a single nucleus are less known compared to those in heterothallic species; even the mechanism by which MAT controls the mating process remains unclear . Homothallic F . graminearum is an ideal species for exploring these unanswered questions for several reasons described below . The presence of both MAT1-1 and MAT1-2 loci in a single nucleus provides a good model system for investigating the roles of both loci after fertilization ( e . g . , nuclear fusion , meiosis , perithecium maturation ) , which requires two parental strains of the opposite mating types in heterothallic species . The capacity of F . graminearum to outcross and self-cross [11] suggests that it has gene regulatory mechanisms for sexual development that are identical to those of heterothallic ascomycetes . Unlike S . macrospora , F . graminearum , requires all of the transcripts at both MAT loci for sexual development [14] , which makes the effects of MAT deletions on the expression and function of MAT target genes more evident . In addition , F . graminearum can be molecularly manipulated to allow high-throughput gene deletions [24] and genetic analyses [11] . Finally , the production of sexual progeny is ecologically important for disease development by F . graminearum , because its sexual cycle predominates in the field , which makes the current study significant both practically and fundamentally . To explore the regulatory mechanisms controlled by MAT genes in F . graminearum , we performed a large-scale study of the target genes of two MAT loci using several strategies including genome-wide transcriptional profiling in various genetic backgrounds , protein binding microarray analysis , in-depth quantitative real-time PCR , and high-throughput gene deletions . The results of this study combined with previous reports provide an insight that allows a comprehensive understanding of the sexual developmental processes under the control of the MAT loci in F . graminearum .
For the microarrays , we used four transgenic strains that were derived from the self-fertile wild-type ( WT ) strain ( Z3643 ) . Three strains , designated ΔMAT1-1 , ΔMAT1-2 , and ΔMAT1-1;ΔMAT1-2 , contained different deletions of the two MAT loci ( MAT1-1 , MAT1-2 ) , and one strain ( OM2 ) overexpressed the MAT1-2-1 allele ( for details , see S1 Text and S1–S3 Figs ) . To identify genes that were regulated by the MAT loci during sexual development , genome-wide microarray analysis was performed using total RNA extracted from mycelia and/or perithecial initials of three MAT-deletion strains , OM2 , and their WT progenitor ( Z3643 ) . Analysis of the transcriptional profiles revealed a total of 1 , 245 genes that were differentially regulated by ≥ 2-fold in all of the transgenic strains compared to Z3643 . Among these , 1 , 106 ( 647 downregulated , 459 upregulated ) were differentially regulated in the three MAT-deletion strains ( ΔMAT1-1 , ΔMAT1-2 , ΔMAT1-1;ΔMAT1-2 ) , and 187 ( 177 downregulated , 10 upregulated ) were in OM2 ( Fig 1 , S1 Table ) . All of the DEGs identified in the three MAT-deletion strains could be categorized into 14 groups according to their expression patterns in each MAT-deletion background ( Fig 1 , S1 Table ) . Of the 647 genes that were downregulated compared to WT , 522 ( 80 . 7% ) were in either or both ΔMAT1-1 and ΔMAT1-2 , but not in ΔMAT1-1;ΔMAT1-2 . Among these , 337 were downregulated in both ΔMAT1-1 and ΔMAT1-2 , but were not significantly changed in ΔMAT1-1;ΔMAT1-2 ( designated DDN , where the first D means downregulated in ΔMAT1-1 , the second D is for the downregulation in ΔMAT1-2 , and N means no change in ΔMAT1-1;ΔMAT1-2 ) , 117 were downregulated in only ΔMAT1-1 ( DNN ) , and 68 were downregulated in only ΔMAT1-2 ( NDN ) . The remaining 125 genes were downregulated in ΔMAT1-1;ΔMAT1-2 , regardless of the differential regulation in either ΔMAT1-1 or ΔMAT1-2 , among which 98 ( 78 . 4% ) were DDD-type ( Fig 1 ) . Among the upregulated genes , most ( 97 . 2% ) were present in either or both ΔMAT1-1 and ΔMAT1-2 , but not in ΔMAT1-1;ΔMAT1-2; 220 , 105 , and 121 genes were UUN- , UNN- , and NUN-type , respectively ( Fig 1 ) . Most of the DEGs identified in OM2 were downregulated , among which 15 were also downregulated and 30 were also upregulated in the MAT-deletion strains ( S1 Table ) . Four of the ten genes upregulated in OM2 were also downregulated in the MAT-deletion strains ( S1 Table ) . MAT1-2-1 was NDD-type and was upregulated in OM2 , and two MAT1-1 transcripts ( MAT1-1-2 , MAT1-1-3 ) were the DND-type , and were unchanged in OM2 . In addition , 729 DEGs ( 325 downregulated , 404 upregulated ) were identified in the two MAT deletion strains ( ΔMAT1-1 and ΔMAT1-2 ) compared to the MAT null strain ( ΔMAT1-1;ΔMAT1-2 ) ( S4 Fig , S2 Table ) . In total , 87 of the 101 NNU-type genes ( unchanged in ΔMAT1-1 or ΔMAT1-2 , but upregulated in WT compared to ΔMAT1-1;ΔMAT1-2 ) overlapped with the DDD-type genes that were identified in comparison with WT ( S2 Table ) . The GO analysis revealed that several Biological Process categories were enriched among the DDN- , DNN- , and NDN-type genes , including various types of metabolism , and developmental processes . Among the genes ( UNN , UUN , NUN ) upregulated in the MAT-deletion strains were enriched the categories of carbohydrate metabolism , developmental processes involved in sporulation , cellular response to chemical stimuli , and cell wall organization . Most of the cellular components categories enriched among these groups were fungal cell wall , plasma membranes , and extracellular ( S3 Table ) . The DDD-type genes were poorly matched to the GO-terms; only those involved in developmental processes ( e . g . , regulation of cell morphogenesis and response to stimuli ) , and organic hydroxyl compound metabolism ( including the polyketide biosynthesis for perithecial pigment ) categories were enriched in this group ( S3 Table ) . Among the genes that were downregulated in OM2 , the categories of metabolism for lipid and nitrogen compounds were enriched . ( S3 Table ) . In addition to the genes enriched for GO terms , 21 genes that might be involved in the cellular processes ( e . g . , cell fusion , nuclear fusion , cell division , chromosome partitioning ) required for sexual development were identified among the DEGs ( S4 Table ) ; the expression of most of these was unchanged in ΔMAT1-1;ΔMAT1-2 ( XXN ) . A total of 58 DEGs identified in this study were analyzed using quantitative real-time PCR in MAT-deletion strains at the perithecial induction stage to confirm that their differential expression was caused by each MAT deletion . The expression patterns of 42 of the 58 genes compared in the three MAT-deletion strains and WT were consistent with the microarray results ( S5 Table ) . The other 16 genes were also differentially expressed , but their patterns in one of three MAT-deletion strains were not consistent with the qPCR data . Among these , eight genes that were identified as DDN-type in the microarrays , exhibited expression that was downregulated almost two-fold in ΔMAT1-1;ΔMAT1-2 compared to WT , and were confirmed as DDD-type using qPCR . Previous studies confirmed that an additional eight genes were downregulated in either ΔMAT1-2 or both ΔMAT1-1 and ΔMAT1-2 strains by Northern blot analysis [17 , 25] . A total of 50 transcription factor ( TF ) genes ( 6 . 9% of the total TFs in the F . graminearum genome ) [24] were identified among the DEGs in the current study ( Fig 2 ) . TF genes with a specific and essential function for sexual development in F . graminearum , which were identified based on phenotypic changes after gene deletions [24] , were only enriched among gene groups that were downregulated in ΔMAT1-1;ΔMAT1-2 ( DND , NDD , NND , DDD ) . All of the sexual development-specific TFs , other than the MAT genes themselves , were DDD-type . The TFs belonging to other gene groups , whose expression levels were unchanged in ΔMAT1-1;ΔMAT1-2 ( DDN , DNN , NDN ) , were either dispensable for sexual development ( ten TFs ) , involved in pleiotropic phenotypes ( i . e . involved in sexual development and other traits; four TFs ) , or involved in traits other than sexual development ( two TFs; Fig 2 ) . In contrast , none of the TFs that were specific to sexual development were identified among the 20 TFs upregulated in the MAT-deletion strains; only 1 TF gene deletion proved lethal . Among the four TFs downregulated in OM2 , only one TF ( FGSG_00404 ) was sexual-specific in the Z3639 strain in a previous study [24] , but was not in Z3643 in the current study . The remaining TFs were dispensable or were involved in sexual development along with other trait ( zearalenone production ) ( FGSG_07368 ) ( Fig 2 ) . Based on the phenotypic changes by gene deletions [23] , the F . graminearum locus IDs ( FGSG_ ) on a gray background indicate the genes specific in function to sexual development , IDs in bold and underline are for those involved in sexual development and other traits , IDs in bold are for those are involved in the traits other than sexual development , and underlined IDs are for those probably lethal . We focused on the expression profiling of genes involved in secondary metabolism since it has been known that secondary metabolism and sexual development are linked in filamentous fungi . To identify SM genes among the DEGs , they were compared against members of the 67 tentative SM gene clusters in F . graminearum [26] . Gene member ( s ) belonging to 22 SM clusters were identified in the DEGs from the MAT-deletion strains or OM2 ( S6 Table ) ; however , only 6 SM clusters included DEGs that encoded key ( signature ) enzymes . Among these , members of two polyketide synthase ( PKS ) gene clusters were downregulated in the MAT-deletion strains: PKS3 ( along with four additional genes ) , which is responsible for the biosynthesis of dark perithecial pigment , and PKS7 ( with one tailoring gene ) , whose chemical product has not yet been identified; these were DDD- and DDN-type , respectively . Two non-ribosomal peptide synthetase genes ( NPS10 , a NPS-like gene ) for unknown metabolites were also identified . In addition to these key enzyme genes , those that encoded either tailoring enzymes or transporters belonging to other PKS clusters ( PKS2 , PKS14 , PKS15 , PKS17 for unknown polyketide compounds ) , an NPS1 cluster for a siderophore ( malonichrome ) [27] , and a butenolide cluster were identified among the DEGs ( S6 Table ) . A total of 169 DEGs ( 13 . 6% ) identified in this study overlapped with the 2 , 064 genes identified previously as sexual development-specific in the F . graminearum PH-1 strain , and whose transcripts were only detected during perithecium formation [22] . The genes downregulated in ΔMAT1-1;ΔMAT1-2 ( i . e . NND- , NDD- , DND- , DDD-type ) overlapped at a higher frequency ( 43 of 125; 34 . 4% ) than those downregulated in both or either ΔMAT1-1 and ΔMAT1-2 ( DDN , DNN , NDN ) ( 76 of 522; 14 . 6% ) ( S7 Table ) . The DEGs identified in this study were also compared to those from the F . graminearum strains lacking FgVelB or GzGPA1 , both of which are self-sterile [28 , 29] . More than half of the genes ( 378 of 647; 58 . 4% ) that were downregulated in the MAT-deletion strains overlapped with those in F . graminearum ΔFgVelB ( Fig 3 , S8 Table ) . In addition , 121 genes ( 18 . 7% ) downregulated in the MAT-deletion strains overlapped with those in the ΔGzGPA1 strain , among which 100 ( 82 . 6% ) were also DEGs in the ΔFgVelB strain ( Fig 3 , S8 Table ) . Interestingly three MAT genes [MAT1-1-3 ( FGSG_08890 ) , MAT1-1-2 ( FGSG_08891 ) , and MAT1-2-1 ( FGSG_08893 ) ] , and one MAT gene [MAT1-1-3 ( FGSG_08890 ) ] were downregulated in the ΔFgVelB [28] and ΔGzGPA1 [29] strains , respectively . Surprisingly , only a small number of DEGs overlapped with DEGs in S . macrospora strains lacking MAT genes . Specifically , 19 DEGs overlapped with 311 genes regulated exclusively in ΔSmtA-2 ( ΔMAT1-1-2 ) , 26 DEGs corresponded to 520 genes regulated in both ΔSmtA-1 and ΔSmtA-2 [15] , and 6 DEGs overlapped with 80 genes from ΔSmta-1 ( ΔMAT1-2 ) [19] ( S9 Table ) . To assess how the MAT loci regulate the expression of genes encoding pheromones ( GzPPG1 , GzPPG2 ) and their cognate receptors ( GzPRE1 , GzPRE2 ) during the early stage of perithecial induction ( 3 days after the removal of the aerial mycelia on carrot agar ) , qPCR was used to compare the transcript levels of each gene in the fungal strains used in microarray analysis , as well as in those lacking individual genes ( MAT1-1-1 , MAT1-1-2 , MAT1-1-3 ) in the MAT1-1 locus ( Table 1 ) . Because northern blotting previously confirmed that all four genes were only expressed in the WT strain during sexual development [30] , we used the transcript level of each gene in WT , or the weakest expression level in GzPRE1 as references to evaluate the effects of MAT deletion or overexpression ( Table 1 ) . The expression of GzPPG1 was significantly reduced in all of the MAT-deletion strains examined compared to WT , with the exception of ΔMAT1-1-2 . Because of the relatively high abundance of the GzPPG1 transcript compared to other genes in WT , this suggests that GzPPG1 is highly expressed only in the WT and ΔMAT1-1-2 strains , but not in a MAT-locus-specific manner . In contrast , GzPPG2 expression was reduced in the ΔMAT1-2 and MAT-null strains , but increased dramatically in the strain lacking the entire MAT1-1 locus ( ΔMAT1-1 ) , and that lacking only the MAT1-1-1 gene at the MAT1-1 locus . This suggests that GzPPG2 is only expressed in the WT and ΔMAT1-1 strains , and therefore exhibits a MAT1-2-locus-specific expression pattern , consistent with our previous study [30] . GzPRE1 was downregulated in both the ΔMAT1-1 and MAT-null strains , but was expressed at comparable levels in the ΔMAT1-2 and WT strains , suggesting a MAT1-1-locus-specific expression . However , the upregulation of GzPRE1 in strains lacking individual MAT1-1 transcripts was surprising , although most of the upregulated transcript were expressed at levels that were weaker than or similar to GzPRE2 in WT . In contrast , GzPRE2 was upregulated in all of the MAT-deletion strains except for ΔMAT1-2 , suggesting that GzPRE2 was constitutively expressed in all of the strains examined . The effects of ΔMAT1-1-1 on the expression of genes encoding pheromones and their receptors were more dramatic than those of other MAT1-1 gene deletions ( ΔMAT1-1-2 , ΔMAT1-1-3 ) , with the exception of GzPRE1 , suggesting that MAT1-1-1 is the major regulator of the pheromone/receptor system among the three transcripts at the MAT1-1 locus . In addition , the expression of GzPPG2 , GzPRE1 , and GzPRE2 was upregulated in the OM2 strain ( Table 1 ) . Recently , similar gene expression data for pheromone/receptor genes in the MAT-deletion strains were reported by Zheng et al [13] . However , those data cannot be directly compared with those in the current study because they were obtained using RNA samples from aerial hyphae on carrot agar before perithecial induction [13] . To determine the functional requirement of the DEGs identified in the current study during sexual development , we selected 106 DEGs based on their expression patterns and putative functional roles . Then , we deleted each DEG from the F . graminearum Z3643 genome using a targeted gene replacement strategy . Including the 32 DEGs that overlapped with those previously identified as being functionally required for or transcriptionally specific for sexual development and/or other traits ( e . g . , hyphal growth , toxin production , virulence ) in F . graminearum , the results of the functional analysis of a total of 127 DEGs were reported here ( S10 Table , S11 Table , Table 2 ) . Based on the phenotypes of the gene deletion strains , 40 genes were responsible for phenotypic changes . Among these , 37 were involved in sexual development alone or together with other traits , and the remaining three were required for phenotypes other than sexual development ( S10 Table ) . Of the 37 genes involved in sexual development , 25 genes were specific to sexual development ( Table 2 ) . The phenotypic changes caused by the deletion of these genes were restricted to only sexual developmental processes ranging from the formation of perithecium initials to ascospore discharge; no changes in other traits such as hyphal growth and pigmentation , conidiation , mycotoxin production , and/or virulence were observed . The transgenic strains in which five genes had been individually deleted ( FGSG_00404 , 04480 , 05239 , 13708 , and 03916 ) produced no perithecium initials on carrot agar , and those in which each of nine genes were deleted ( FGSG_01366 , 08320 , 11826 , 13162 , 10742 , 08890 [MAT1-1-3] , 08891 [MAT1-1-2] , 08892 [MAT1-1-1] , 08893 [MAT1-2-1] ) produced barren perithecia that were smaller in size and/or number than WT and contained no asci/ascospores ( S11 Table , Table 2 , S5 Fig ) . By contrast , the remaining 11 genes were not absolutely required for the production of fertile perithecia , but instead , were specifically involved in sexual development . The mutants in which six genes ( FGSG_03673 , 05151 , 07578 , 06059 , 11962 , and 02655 [GzPRE2] ) were deleted individually produced lower numbers of mature perithecia , whereas those lacking FGSG_06966 and FGSG_01862 produced larger perithecia , or showed delayed perithecia formation , respectively , compared to WT . The deletion mutants of FGSG_00348 , FGSG_02052 , and FGSG_09182 ( PKS3 ) produced perithecia that looked similar to those in WT , but exhibited defects at different stages of perithecia maturation ( Table 2 , S5 Fig ) . The targeted deletion of FGSG_00348 ( designated FgSMS-2 ) from Z3643 , which exhibited sequence similarity to a gene encoding an Argonaute protein known to participate in the RNA interference ( RNAi ) pathway in Drosophila melanogaster [31] and N . crassa [32] , caused no dramatic changes in major traits such as hyphal growth , conidiation , pigmentation , virulence , and perithecia formation in F . graminearum . Unlike the perithecia produced in WT , those in ΔFgSMS-2 produced no cirrhi ( ascospores oozing from the perithecia ) at the ostiole 10 days after perithecial induction , and contained fewer numbers of asci that were formed at least 2 days later than WT ( Fig 4 ) , as previously reported in the Z3639 strain [33] . The germination rate of ascospores was not significantly different from WT . Interestingly , outcrossing the ΔFgSMS-2 strain as a male to the ΔMAT1-2 strain as a female produced incomplete tetrads that mainly carried four ascospores rather than eight , in which the GFP marker did not segregate equally ( Fig 5 ) . Furthermore , the outcross between the ΔFgSMS-2 strain ( female ) and ΔMAT1-2 strain ( male ) produced asci similar to those in the self-cross of ΔFgSMS-2 strain ( Fig 5 ) . qPCR confirmed that FgSMS-2 was specifically expressed at a later stage ( 7 days after perithecial induction ) of sexual development , and was regulated transcriptionally by both MAT loci ( S6 Fig ) . In addition , the expression of three DEGs ( FGSG_02877 and belonging to UUN , and 05906 to DDN ) was upregulated during sexual development in the ΔFgSMS-2 strain compared to the WT strain , suggesting that FgSMS-2 is involved in the degradation of the mRNAs of these DEGs ( S6 Fig ) . The deletion of another Argonaute-like gene ( FGSG_08752 ) in the F . graminearum genome , which was not differentially regulated by the MAT loci , had no effect on sexual development; even the double deletion of FgSMS-2 and FGSG_08752 yielded an identical phenotype as the ΔFgSMS-2 strain ( Fig 4 ) . Unlike ΔFgSMS-2 , the strain lacking FGSG_02052 produced cirrhi at least 2 days earlier than WT , but it produced as many normal-looking ascospores as WT ( Fig 6 ) . Among the 25 genes with sexual development-specific functions , ten ( 40%; FGSG_04480 , 00404 , 01366 , 11826 , 05151 , 06966 , 08890 , 08891 , 08892 , 08893 ) , including four MAT genes , encode transcription factors , two ( FGSG_08320 and 09182 ) encode SM gene cluster members , and the others are involved in metabolism ( FGSG_13708 , FGSG_03673 , FGSG_07578 , FGSG_06059 ) , chromatin silencing ( FGSG_13162 ) , cell adhesion ( FGSG_03916 ) , signaling ( FGSG_05239 ) , cytoskeleton dynamics ( FGSG_01862 ) , RNA inference ( FGSG_00348 ) , or unknown functions ( FGSG_11962 and FGSG_02052 ) ( Table 2 ) . Interestingly , 21 ( 84% ) of the 25 sexual-specific genes were downregulated in ΔMAT1-1;ΔMAT1-2 . Only 1 gene ( FGSG_05239 ) among the 31 DDN-type genes examined had a sexual-specific function ( Table 2 ) . In addition , 12 genes ( FGSG_00532 , 02572 , 06039 , 06228 , 07368 , 07546 [MYT2] , 07869 , 08795 , 09019 , 09834 , 09896 [GzICL1] , 10825] were responsible for pleiotropic phenotypes , including defects in sexual development . Three genes , which were not essential for sexual development , were involved in hyphal growth ( FGSG_04946 ) or virulence toward the host plant ( FGSG_05906 [FGL1] , FGSG_10396] ) ( Table 2 ) . A total of 37 DEGs identified in the current study were analyzed in the WT strain grown on carrot agar using qPCR to determine the time course of the transcriptional profiles during both the vegetative and sexual stages . Most DEGs examined in the current study ( 33 of 38 ) were confirmed as sexual development-specific at the transcriptional level , because they were expressed at higher or lower ( for FGSG_05906 only ) levels in the WT Z3643 strain under perithecial induction conditions compared to vegetative conditions ( S12 Table ) . Among these , four genes ( FGSG_ 05246 , FGSG_05847 , FGSG_06549 , FGSG_01763 ) could be assumed to be involved in the early stage of perithecium formation , because their expression peaked 3 days after perithecial induction . In contrast , the remaining 28 genes are probably specific to a later stage of sexual development ( S12 Table ) . However , the functional assignment of these genes relies on only a correlation , and needs further confirmation . In our previous study [28] , 42 genes that were identified as DEGs in the current study were confirmed to be sexual development-specific in the WT Z3639 strain . Among these , 50% ( 21 genes ) and 28 . 6% were DDD- and DDN-type , respectively ( S8 Table ) . When the DEGs in the current study were searched against the transcriptome data obtained from the fungal PH-1 strain at 6 developmental stages during perithecium formation , data revealed that 1 , 152 DEGs were expressed at these time points [23] ( S13 Table ) . In particular , the transcript accumulation of 72% ( 87/121 ) of the XXD-type genes ( including DDD , DND , NDD , NND ) peaked 96 h after perithecial induction , whereas the expression of 70% of the XXN-type ( DDN , DNN , NDN ) genes peaked at earlier time points ( 2 , 24 , 48 , or 72 h; S13 Table ) . However , the upregulated genes were not enriched at a specific stage . In addition , 45% ( 54/120 ) of the genes that were downregulated in OM2 exhibited the highest transcript accumulation at 72 h ( Table 3 ) . We used PBM technology [34 , 35] to identify a putative binding site for the MAT1-2-1 protein ( S7 and S8 Figs ) . Two PBMs ( Q9-PBM , FgPBM ) were hybridized to the DNA-binding HMG motif of MAT1-2-1 , which was fused to DsRed fluorescent protein , and then expressed in E . coli . A comparison of the putative consensus binding sequences identified using both PBM methods indicated that ATTGTT could be the core binding sequence for the HMG domain of MAT1-2-1 ( Fig 7 ) , which is complementary to the core-binding element ( AACAAT ) of the mammalian sex-determining region Y ( SRY ) or SRY-related HMG box gene ( SOX ) [36–38] . Electrophoretic mobility shift assay ( EMSA ) revealed that the quadruple sequences of the identified motifs ( ATTAAT and ATTGTT ) had binding activities for the MAT1-2-1 HMG box domain ( S9A Fig ) . The promoter regions of three genes ( FGSG_04946 , FGSG_08467 and FGSG_06480 ) could bind to MAT1-2-1 , but their binding activities were relatively weak ( S9B Fig ) . Experimental details and discussion from the PBM and EMSA analyses were described in S2 Text . To assess the possible regulatory relationship among the sexual development-specific MAT downstream regulator genes ( four transcription factors and one RNAi regulator ) , we used qPCR to examine their expression patterns in fungal strains in which each gene was deleted during sexual development ( 3 and 6 days after perithecial induction ) ( Table 4 ) . Using a fold-change threshold of 3 . 0 , because most genes were expressed at relatively low levels ( based on previous transcriptome data [23] ) , a map of the regulatory interactions among the genes was constructed , as previously performed [39] . Because the core binding sequence of MAT1-2-1 was found in only FGSG_01366 , the transcription factor carrying the HMG-box motif , we assumed that FGSG_01366 is the first putative target of MAT1-2-1 in the network ( Fig 8 ) . However , the deletion of FGSG_06966 , and FGSG_11826 had a significant effect on the expression of FGSG_01366 and other regulatory genes , suggesting that interregulatory networks operate among these genes . Interestingly FGSG_00348 , which encodes an Argonaute-like protein was downregulated in strains in which all four transcription factor genes had been individually deleted , suggesting that it was a downstream target of these genes . In addition , none of the target gene deletions had a significant effect on the transcription of MAT1-1-1 and MAT1-2-1 , which confirms that these regulatory genes are downstream of the MAT loci in F . graminearum .
The most significant achievement in this study is that it provided a comprehensive investigation of the putative target genes of the MAT loci during sexual development in self-fertile F . graminearum . Genome-wide transcriptional profiling in various MAT genetic backgrounds , and subsequent in-depth and high-throughput analyses allowed us to explore the regulatory networks and function of MAT-target genes during the early sexual developmental process when the major regulators ( MAT1-1-1 , MAT1-2-1 ) of each MAT locus are expressed at their peak levels [14] . Similar to other filamentous ascomycetes , F . graminearum undergoes various cellular processes during each stage of sexual development including mating , cell fusion , nuclear division , fusion , meiosis , ascus/ascospore development , and perithecium maturation . These aspects of sexual development could include pheromone-mediated membrane function , signal transduction , cytoskeleton dynamics , secretory pathways , cell cycle , cell adhesion , apoptosis , and differentiation [40] . The GO terms associated with genes that are differentially expressed in strains carrying a single MAT gene or no MAT gene were significantly enriched for terms related to sexual development processes . In particular , the terms of metabolism including cell wall organization , developmental processes involved in reproduction , and the cellular response to chemical stimulus were enriched among the DDN- and UUN- type DEGs , which are the most frequent groups . Similarly , those described above as well as related to signaling , cellular homeostasis , and cell cycle were enriched among the DEGs that were found in only ΔMAT1-1 or ΔMAT1-2 ( DNN- , UNN- , NDN- , and NUN-types ) . Unlike the DEGs described above , those from the MAT-null strain ( mostly DDD-type ) were orthologous to the genes associated with the GO terms that were more specific to sexual development ( S3 Table ) . These genes , most of which were also MAT loci-specific , play important roles in various stages of sexual development ranging from mating to fruiting body formation , which was further confirmed using high-throughput gene deletions . The high frequency of genes specific to sexual development ( 86 . 5%; 33 of 38 according to qPCR ) among the DEGs suggests that most MAT-target genes are involved in sexual development . A comparison with the study by Hallen et al . [22] identified 169 additional DEGs that were specifically induced during sexual development , which further supports the transcriptional specificity of MAT-target genes . It is unclear if or how mating identity is maintained by the MAT-specific pheromone/receptor system in homothallic F . graminearum , although it was suggested that the single ( opposite ) MAT-mediated intercellular recognition might occur by differential expression of both MAT loci ( S10 Fig ) [41] . However , in the current study , qPCR of various MAT-deletion strains suggested MAT1-2-locus-specific GzPPG2 expression and MAT1-1-locus-specific GzPRE1 expression ( Fig 9 ) . In the homothallic S . macrospora , the pheromone genes ppg1 and ppg2 were downregulated in the ΔSmtA-1 ( ΔMAT1-1-1 ) strain , whereas only ppg2 was downregulated in the ΔSmta-1 strain ( ΔMAT1-2-1 ) [15] . However , it is unknown if the expression pattern of ppg in S . macrospora is MAT-specific , because no studies have been performed in strains lacking the MAT loci . Despite the dispensability of pheromones and their receptors for the formation of fertile perithecia in F . graminearum [30] , the qPCR results in the current study predict a role for this system during sexual developmental processes ( e . g . , pheromone-mediated fertilization or inter-nuclear recognition ) in F . graminearum ( Fig 9 ) . Once perithecia formation has been induced , WT cells , which carry and express both MAT1-1;MAT1-2 [14] , are capable of producing two pheromone precursors ( GzPPG1 , GzPPG2 ) and their cognate receptors ( GzPRE1 , GzPRE2 ) . However , during the vegetative growth stage , WT cells carrying but not expressing both MAT loci [14] , which are comparable to MAT-null cells during the sexual stage , produced no pheromones or receptors , except for GzPRE2 . ( Table 1 , Fig 9A ) . Cells carrying only the MAT1-1 locus ( the ΔMAT1-2 strain in this study ) , which could be present among WT cells as well by differential expression of MAT loci ( S10 Fig ) , if occurs , produce only two pheromone receptors ( GzPRE1 and GzPRE2 ) . In contrast , those cells carrying only the MAT1-2 locus can produce the pheromone GzPPG2 as well as the receptor GzPRE2 ( Fig 9A ) . If the exclusive expression of one MAT locus occurs within the parental cells ( e . g . , the conidia as the male element and the ascogonium as the female element ) [40] , we could hypothesize that an interaction between two cells is required for fertilization: a cell with only MAT1-1 ( designated MAT1-1 ) with one with MAT1-2 ( MAT1-2 ) , MAT1-1 with a WT cell ( MAT1-1;MAT1-2 ) , and MAT1-2 with MAT1-1;MAT1-2 ( Fig 9B ) . In the first case , which usually occurs in heterothallic species , the interaction between the GzPPG2 pheromone released from MAT1-2 and its cognate receptor GzPRE1 in MAT1-1 might lead to the fertilization . In the other two cases , the interactions between GzPPG1/GzPRE2 and GzPPG2/GzPRE1 would lead to fertilization . Since GzPPG1 is expressed in the germinating conidia of F . graminearum [42] , it is possible that the MAT1-2 and WT strains could act as the male parent in the fertilization model ( Fig 9B ) . The possible fertilization in a heterothallic manner , as proposed in the first combination ( MAT1-1 vs . MAT1-2 ) , was previously demonstrated by outcrosses between the ΔMAT1-1;MAT1-2 and MAT1-1;ΔMAT1-2 strains of F . graminearum , which produced fertile perithecia . However , their numbers and fertility were much lower than in the WT strain [11] . A recent study showed that outcrossing ΔMAT1-1;MAT1-2 as the male parent with MAT1-1;ΔMAT1-2 as the female produced partially normal perithecia , whereas an outcross with the reverse parental roles ( i . e . , ΔMAT1-1;MAT1-2 female × MAT1-1;ΔMAT1-2 male ) produced only small , sterile perithecia [13] . The reduced fertility of perithecia in these outcrosses could be attributed to the low level of expression of GzPRE1 in MAT1-1 strain . The failure in the latter outcross could be ascribed to the lack of ( or poor ) ability of the MAT1-1 strain to produce pheromones as the male parent . The other two combinations for fertilization ( Fig 9B ) could also yield a successful outcross between WT ( male ) and either the ΔMAT1-1;MAT1-2 or MAT1-1;ΔMAT1-2 strain ( female ) , as described previously [11] . The retained sexual ability of the F . graminearum strain lacking all of the pheromone and receptor genes suggests that F . graminearum might mate randomly . In addition , the current results reveal the possibility that pheromone-mediated fertilization and/or internucleus recognition occur after random mating in F . graminearum . The similar effect of ΔMAT1-2 and ΔMAT1-1;ΔMAT1-2 on the expression of two pheromone genes ( i . e . , downregulation of both GzPPG1 and GzPPG2 ) could be attributed to the downregulation of MAT1-1 genes in the ΔMAT1-2 strain . However , the effect of ΔMAT1-1 on the expression of GzPPG2 ( i . e . , upregulation ) was opposite that of ΔMAT1-2 or ΔMAT1-1;ΔMAT1-2 , suggesting that the effect of ΔMAT1-1 on the expression of MAT1-2 might not be as great as the effect of ΔMAT1-2 on MAT1-1 expression . The phenotypic characterization of transgenic strains lacking the selected DEGs provides clear evidence for their functional requirement and specificity for sexual development , as well as the potential stage of their roles during sexual development . Among the 25 genes that were specific to sexual development , 14 ( 56% ) affected the perithecium initial or protoperithecium formation , because the gene deletion strains produced barren protoperithecium ( that did not develop further into fertile perithecium ) , or no perithecium initials ( Table 2 ) . This strongly suggests that the MAT loci regulate the genes that are necessary for the cellular events involved in the developmental stages that occur before nuclear fusion and/or meiosis that may subsequently facilitate the formation of mature perithecium and asci/ascospores ( S10 Fig ) . Based on sequence homology , the cellular functions of these sexual-specific genes could be proposed as follows . Two G-protein coupled receptors ( the pheromone receptor GzPRE2 and FGSG_05239 ) might be involved in mating and subsequent signal transduction . Ten TFs , including four MAT genes , might regulate many downstream genes that are necessary for protoperithecium formation . An additional two TFs ( FGSG_06228 , FGSG_03597 ) , which encode a regulator of G-protein signaling ( RGS ) protein domain , were upregulated by MAT deletions ( UNN- and UUN-type , respectively ) . Both genes had relatively high sequence similarity to the RGS gene in Aspergillus nidulans , designated FlbA ( with 61% and 43% amino acid identities , respectively ) , which is required for the control of mycelial proliferation and asexual sporulation [43] . FGSG_03597 is also downregulated in OM2 . These expression patterns suggest that the MAT1-1 locus might suppress the RGS genes during early sexual development to block the signal transduction pathways required for asexual sporulation . The suppression of asexual sporulation throughout sexual development has been demonstrated in F . graminearum; conidiation in the self-sterile ΔFgVelB strain was derepressed even during perithecial formation [28] . It is likely that at least one ( FGSG_06228 ) of these genes is the functional homolog of A . nidulans FlbA because the targeted deletion of FGSG_06228 had a significant effect ( i . e . an ~11-fold reduction ) on conidiation in F . graminearum Z3643 , which was different from the phenotypic changes in the Z3639 genetic background [24] . In addition , ΔFGSG_06228 exhibited pleiotropic changes in other traits such as perithecium formation , virulence , and mycotoxin production ( S11 Table ) . Importantly , this study is the first to show inactivation of the FlbA-like gene by the MAT loci during sexual development in filamentous ascomycetes , although the regulatory networks for conidiation in which FlbA plays an important role have been studied in Aspergillus [44] . Since the MAT genes were downregulated in the ΔFgVelB strain , it is possible that FgVelB , a component of the FgVeA complex , controls asexual sporulation by activating the MAT loci . Consistent with this notion , FgVelB controls the expression of MAT genes in F . graminearum [28] . FGSG_07368 , a TF that represses ZEA biosynthesis [24] , is involved in both ZEA production and perithecium formation because its deletion strain produced ~three-fold higher amounts of ZEA , but formed lower numbers of fertile perithecia compared to WT . This gene is downregulated in the OM2 strain , suggesting that the MAT1-2-1 protein represses ZEA biosynthesis during the perithecia induction stage by suppressing the expression of FGSG_07368 . Because the negative effect of ZEA overproduction on sexual development was confirmed in F . graminearum [45] , this MAT-induced regulation pattern suggests that the MAT loci acts as a master regulator to both activate and repress the expression of many genes during sexual development . Seven genes ( FGSG_08320 , 13708 , 06059 , 07578 , 07869 , 03673 , 09182 ) encode enzymes that might participate in the metabolic pathways that are required for protoperithecium formation . In particular , FGSG_09182 ( PKS3 ) and FGSG_08320 are required for perithecial pigment [46] and an unidentified secondary metabolite , respectively . Four genes ( FGSG_03916 , 10742 , 05246 , 01862 ) might be involved in various cellular events such as cell adhesion ( FGSG_03916 ) , cellular fusion ( FGSG_10742 ) , paraphyses senescence ( FGSG_05246 ) , and cytoskeletal organization ( FGSG_01862 ) . FGSG_13162 , which encodes a protein that controls the histone promoter , might be involved in chromatin silencing and could be required for the regulation of gene expression during protoperithecium formation . Sexual-specific genes are more required for the early stage of sexual development ( e . g . , the initial perithecium formation ) than for postfertilization or post-meiotic events . This suggests that the MAT genes , particularly MAT1-1-1 and MAT1-2-1 , whose transcripts accumulate at the highest levels during the early stage [13 , 14] , control the genes that are involved in this stage more than those that regulate the later stages ( e . g . , nuclear fusion , meiosis , ascus maturation ) . However , it could also be attributed to the time point ( 48 h after perithecium induction ) at which total RNA samples were extracted for microarray analysis . Nevertheless , the possibility that the MAT loci regulate both the early and late stages of sexual development is supported by the phenotypes of transgenic strains that lack each of three remaining genes ( FGSG_00348 , 00532 , 02052 ) . Specifically , these strains could produce mature perithecium , although the number of perithecia produced was lower ( FGSG_00348 ) or higher ( FGSG_00532 ) than WT . This suggests that the MAT genes also control the expression of the genes that are required for post-fertilization processes . In particular , the deduced product of FGSG_00348 , designated FgSMS-2 , contains the PAZ- and Piwi domains that are also found in SMS-2 from N . crassa and AGO2 ( Argonaute ) from D . melanogaster . These proteins play an important role in the RNA silencing known as meiotic silencing by unpaired DNA ( MSUD ) and siRNA-directed RNA interference , respectively [32 , 47]; they are the active part of the RNA-induced silencing complex and required for cleaving the target mRNA . The formation of asci carrying an incomplete set of ascospores or abnormally shaped ascospores in the ΔFgSMS-2 strain suggests that the MAT loci regulate the meiotic events that lead to asci/ascospore development by activating the expression of FgSMS-2 , which might control the RNAi pathway in F . graminearum . Another SMS-2 ortholog ( FGSG_0872 ) is present in the F . graminearum genome , which is similar to the N . crassa gene Qde-2 , which regulates another RNA silencing pathway ( quelling ) during vegetative growth . However , the lack of transcriptional control of FGSG_0872 by MAT and the phenotypic changes caused by gene deletion suggest that FgSMS-2 is the only Argonaute-like gene that controls meiosis and the subsequent developmental pathways in F . graminearum . The upregulation of three DEGs caused by ΔFgSMS-2 during perithecial formation ( S6 Fig ) suggests that FgSMS-2 is involved in the degradation of the mRNA of these genes in F . graminearum , which is comparable to the role of Argonaute proteins . However , further investigations are necessary for confirmation , such as exploring and characterizing small RNAs . Interestingly , 17 of the 21 sexual-specific genes ( excluding 4 MAT genes ) were downregulated in the MAT-null strain as well as ΔMAT1-1 and ΔMAT1-2 . The four remaining genes included FGSG_03916 , which is repressed during sexual development , and FGSG_02655 ( GzPRE2 ) and FGSG_00404 , which are both differentially regulated only in OM2 . These data suggest that all of the genes with sexual-specific functions are regulated by both the MAT1-1 and MAT1-2 loci . This also indicates that the presence of both MAT1-1 and MAT1-2 proteins within a separate or common nucleus is necessary for controlling most stages of sexual development , from ascogonium formation to perithecium maturation ( S10 Fig ) . In addition , the high frequency ( 72% ) of enrichment of XXD-type DEGs at a later stage ( 96 h after perithecial induction ) , which is dramatically different from the XXN-type DEGs enriched at the earlier stages ( Table 3 ) , strongly suggests that both MAT1-1 and MAT1-2 are required to control stages of development later than fertilization . However , the functional assignment of these DEGs into particular growth stages is based on only developmental time course gene expression , and awaits further confirmation . Taken together , the results of the current and previous studies provide insights toward a comprehensive understanding of the MAT-mediated regulatory pathways that control sexual development in F . graminearum ( Fig 10 ) . The environmental cues for sexual development elucidated to date in F . graminearum are light , nutrient conditions ( probably nitrogen starvation ) , and mycelial autophagy [40] . In contrast to A . nidulans , F . graminearum only produces fertile perithecia on carrot agar in the presence of light . The molecular characterization of F . graminearum orthologs ( FgWC-1 [FGSG_07941] and FgWC-2 [FGSG_00710] ) of the photoreceptors encoding wc-1 and wc-2 in N . crassa revealed that these two genes repress the expression of MAT genes during the sexual development ( within 3−5 days of perithecial induction ) [48] . Similarly , FgLaeA , a component of the FgVeA complex that is involved in chromatin remodeling , is also likely to repress MAT gene expression , because the MAT genes were upregulated in the ΔFgLaeA strain [49] . However , other members of the FgVeA complex , FgVeA and FgVelB , activate MAT gene expression because the MAT genes were downregulated in the ΔFgVelB strain , and the ΔFgVeA strain was self-sterile [28] . Taken together , it is possible to speculate that the expression of the MAT genes could be activated or repressed by chromatin remodeling via FgWC-1/FgWC-2 and FgVeA complexes . However , this speculation awaits further investigation . The downregulation of MAT genes in the ΔGzGPA1 strain [33] suggests that they can also be positively regulated by a G-protein-mediated signal transduction pathway , which is probably induced via nutrient sensing . The deletion of most of the genes in the G-protein signaling pathway , including the MAP kinase and cAMP cascades , caused dramatic changes in several traits , including sexual development [50–54] . This suggests that the MAT loci might be under the control of these signaling pathways . In contrast , most of the genes involved in these signaling pathways ( S14 Table ) were not differentially regulated in the MAT-deletion strains examined in this study . This suggests that MAT does not regulate these pathways . Once activated under perithecial induction conditions [14] , MAT genes control the expression of the pheromone/receptor system in a MAT-specific manner ( particularly for GzPPG2 and its cognate receptor GzPRE1 ) , suggesting that MAT plays a role in pheromone-mediated fertilization . However , it is unclear if the maintenance of cell identity for mating , which might be mediated by the MAT-specific pheromone/receptor system and/or by a single MAT locus , occurs in F . graminearum , although the differential expression of each MAT locus within a single nucleus has been suggested in this regard . It is still possible that the pheromone/receptor system plays an important role in the stages after fertilization , such as during inter-nuclear recognition . Based on the genome and promoter microarray results in the current study , it is possible that the regulation of sexual development under the control of the MAT loci might involve cascades or networks that are downstream of TFs . Focusing on the TFs that are specific to sexual development at both the gene expression and functional levels allows a model of regulation of these TFs by the MAT loci to be proposed ( Fig 10 ) . Although inter-regulatory networks between these TFs have also been proposed , it is possible that these major TFs control the expression of the genes that are involved in various molecular processes related to sexual development , conidiation , and zearalenone production . In this regulatory circuit , the HMG-box motif containing TF FGSG_01366 is the main downstream regulator of the MAT loci . Interestingly , FGSG_01366 is an ortholog of PaHMG5 in P . anserina , which plays a crucial role in a network containing several HMG-box factors that regulate mating-type genes and their target genes [39] . However , the regulatory roles of these HMG-box TFs for sexual development differ between the two species . FGSG_01366 is a downstream target of the MAT loci in F . graminearum , whereas PaHMG5 is an upstream regulator of mating-type genes ( FMR1 and FPR1 , which are comparable to MAT1-1-1 and MAT1-2-1 , respectively ) and the pheromone/receptor system [39] . The expression of the F . graminearum ortholog ( FGSG_05151 ) of the other HMG box gene ( PaHMG8 ) , a downstream target of PaHMG5 , in P . anserina was also downregulated in the MAT-null background ( NND-type ) . This suggests that FGSG_05151 is a member of the regulatory circuit of TFs that are under the control of the MAT loci ( Fig 10 ) , although the gene deletion phenotype was not as dramatic as that for FGSG_01366 ( Table 2 ) . The other regulatory mechanism under the control of the MAT loci might be the RNAi pathway involved in meiosis and/or ascus development . Although RNAi-mediated MSUD was not demonstrated in F . graminearum , the functional requirement of the Argonaute-like protein FgSMS-2 for ascus maturation and ascospore morphology and upregulation of several DEGs by ΔFgSMS-2 suggests that RNAi may have a regulatory role during meiosis in F . graminearum . The time course expression pattern and deletion of selected DEGs suggest that major MAT genes ( MAT1-1-1 , MAT1-2-1 ) , which are highly induced at an early stage , regulate the genes involved in the later as well as early stages of sexual development ( e . g . , meiosis , ascospore formation , discharge ) . This suggests that MAT loci play crucial roles throughout sexual development in F . graminearum . In summary , both MAT loci are activated by several environmental cues via chromatin remodeling and/or signaling pathways , and then control the expression of at least 1 , 245 target genes during the early stage of sexual development via regulatory cascades and/or networks involving several downstream TFs and RNAi . The regulatory effects of the MAT loci on these target genes could be directly achieved by the binding of MAT1-2-1 protein to core sequences within the promoter regions of some target genes , or indirectly via downstream TFs .
The F . graminearum WT strains PH-1 and Z3643 [55] , which belong to lineage seven of the F . graminearum species complex [5 , 17] , are self-fertile . T43ΔM1-3 [25] , T43ΔM2-2 [17] , and T43ΔDM1M2 ( S1 Fig ) are self-sterile transgenic mutants derived from Z3643 , and lack MAT1-1 , MAT1-2 , and both MAT1-1 and MAT1-2 loci , respectively . The ΔMAT1-1-1 , ΔMAT1-1-2 , and ΔMAT1-1-3 strains were previously generated by deleting the MAT1-1-1 , MAT1-1-2 , and MAT1-1-3 transcripts at the MAT1-1 locus of Z3643 , respectively; they are all self-sterile [14] . OM2 ( S2 Fig ) is a MAT1-2-1-overexpressing strain that was generated by insertion of MAT1-2-1 into the T43ΔM2-2 strain under the control of the F . fujikuroi EF1A promoter . The WT and transgenic strains were stored in 20% glycerol at –70°C . Sexual development was induced on carrot agar , as previously described [11 , 56] . For genomic DNA extraction , each strain was grown in 50 mL CM [56] at 25°C for 72 h on a rotary shaker ( 150 rpm ) . Conidiation was induced in CMC liquid medium [57] . Fungal genomic DNA was prepared as previously described [56 , 58] , and total RNA was extracted from mycelia and/or protoperithecia that had formed on carrot agar using an Easy-Spin Total RNA Extraction kit ( Intron Biotech , Seongnam , Korea ) according to the manufacturer’s instructions . All of the PCR primers used in this study ( S15 Table ) were synthesized by Bioneer Corporation ( Chungwon , Korea ) . DNA gel blots were prepared [59] and hybridized using biotinylated DNA probes that were prepared using the BioPrime DNA labeling system ( Invitrogen , Carlsbad , CA , USA ) , and were developed using the BrightStar BioDetect Kit ( Ambion , Austin , TX , USA ) . Other general procedures for nucleic acid manipulations were performed as previously described [59] . qPCR was performed using SYBR Green Super Mix ( Bio-Rad ) with first-strand cDNA synthesized from total RNA [33 , 60] . The amplification efficiency of all genes was determined as previously described [60] . Gene expression was measured in three biological replicates from each time point . EF1A ( FGSG_08811 ) was used as an endogenous control for data normalization [60] . Fungal strains were grown in 20 mL agmatine-amended liquid medium [61] for trichothecene production and SG liquid medium [62] for zearalenone production , as previously described . Luciferase activity was then measured in cell lysates from the strains using GloMax 96 Microplate Luminometer ( Promega ) as previously described [63] . The virulence of the fungal strains was determined on wheat heads as previously described [49] . Total RNAs were isolated from carrot agar cultures of the F . graminearum Z3643 WT strain carrying both MAT1-1 and MAT1-2 , three MAT-deletion strains ( ΔMAT1-1 [carrying only MAT1-2] , ΔMAT1-2 [carrying only MAT1-1] , and ΔMAT1-1;ΔMAT1-2 ) , and the OM2 strain , which had been grown for 2 days after sexual induction . Double strand cDNAs were synthesized as previously described [33] . Microarray analysis was conducted at GreenGene Biotech ( Yongin , Korea ) using the F . graminearum microarray that was manufactured at NimbleGen ( Madison , WI , USA ) , as previously described [33] . This array includes 13 , 382 transcripts from F . graminearum sequencing assembly three . Experiments were repeated three times with total RNA samples that were independently prepared . Microarrays were scanned using Genepix 4000 B ( Axon Instruments , Toronto , Canada ) , and the signals were analyzed with Nimblescan 2 . 5 ( NimbleGen ) . The data were normalized and processed with cubic spline normalization using quantiles to adjust for signal variation among chips [64] . Probe-level summarization using Robust Multi-Chip Analysis ( RMA ) with a median polish algorithm implemented in NimbleScan was used to produce call files . Multiple analyses were performed using the limma package in an R computing environment [65] . Genes with P ≤ 0 . 05 ( significant ) were collected and selected further for genes with expression > 1 or < −1 in at least one MAT genetic background compared to another background such as WT ( MAT1-1;MAT1-2 ) , or MAT-null ( ΔMAT1-1;ΔMAT1-2 ) . The entire dataset from the microarrays was deposited in the NCBI Gene Expression Omnibus ( GEO ) database ( http://www . ncbi . nlm . nih . gov/geo ) under the accession number GSE58543 . The DEGs identified in this study were examined for significant enrichment of functional categorization using GO analysis [66] . The GO term enrichments were calculated with GOMINER [66 , 67] ( http://www . geneontology . org/ , http://discover . nci . nih . gov/gominer/ ) . The 11 , 603 genes were matched to A . nidulans FGSC A4 sequencing assembly ( Aspergillus Genome Database; http://www . aspergillusgenome . org/ ) with score 30 and up by BlastP analysis , and were used as total gene set in gominer for GOMINER analysis . Gominer first categorize each gene according to their GO terms and mode of gene expressions either down- or up- regulation . It calculates p-values with the one-sided Fisher exact test for the number of categorized GO terms in the total . The GO terms with p-value less than 0 . 05 were considered significantly enriched among the DEGs , and used in further analysis . DNA constructs to delete individual DEGs from the genomes of F . graminearum Z3643 or PH-1 were created using a split marker recombination procedure , as previously described [68] . Generally , the 5′ and 3′ flanking regions of the target gene , which were amplified by PCR using the primers listed in S15 Table , were mixed with the gen or hygB cassettes , which were amplified from pII99 and pBCATPH using the primer pairs Gen-for/Gen-rev and Hyg-for/Hyg-rev , respectively . The final split markers were amplified from the mixture using the new nested primer sets ( S15 Table ) . The amplified products were added into WT F . graminearum protoplasts for transformation , as previously described [14 , 49] . Gene deletion was confirmed using either DNA gel blot hybridization or PCR ( S11 Fig ) .
|
The production of sexual propagules via a self-fertile mating strategy in Fusarium graminearum , an important cereal pathogen , is essential for overwintering and dissemination during the recurrent disease cycle caused by this fungus . Genome-wide microarray analyses allow the identification of gene sets that are regulated by the mating-type ( MAT ) loci , which is a master regulator of sexual reproduction in F . graminearum . By employing in-depth and high-throughput functional analyses , the current study provides novel insight into our understanding of the regulation of sexual developmental processes by the MAT loci . MAT genes , which are located at two linked MAT loci , play important roles in even the late stages of sexual development by controlling regulatory pathways involving several sexual-specific transcription factors and putative RNA interference regulators . This study could be significant both practically and fundamentally because of the ecological impact of sexual reproduction by F . graminearum during disease development in the field .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
A Large-Scale Functional Analysis of Putative Target Genes of Mating-Type Loci Provides Insight into the Regulation of Sexual Development of the Cereal Pathogen Fusarium graminearum
|
Activated Cdc42 kinases ( Acks ) are evolutionarily conserved non-receptor tyrosine kinases . Activating somatic mutations and increased ACK1 protein levels have been found in many types of human cancers and correlate with a poor prognosis . ACK1 is activated by epidermal growth factor ( EGF ) receptor signaling and functions to regulate EGF receptor turnover . ACK1 has additionally been found to propagate downstream signals through the phosphorylation of cancer relevant substrates . Using Drosophila as a model organism , we have determined that Drosophila Ack possesses potent anti-apoptotic activity that is dependent on Ack kinase activity and is further activated by EGF receptor/Ras signaling . Ack anti-apoptotic signaling does not function through enhancement of EGF stimulated MAP kinase signaling , suggesting that it must function through phosphorylation of some unknown effector . We isolated several putative Drosophila Ack interacting proteins , many being orthologs of previously identified human ACK1 interacting proteins . Two of these interacting proteins , Drk and yorkie , were found to influence Ack signaling . Drk is the Drosophila homolog of GRB2 , which is required to couple ACK1 binding to receptor tyrosine kinases . Drk knockdown blocks Ack survival activity , suggesting that Ack localization is important for its pro-survival activity . Yorkie is a transcriptional co-activator that is downstream of the Salvador-Hippo-Warts pathway and promotes transcription of proliferative and anti-apoptotic genes . We find that yorkie and Ack synergistically interact to produce tissue overgrowth and that yorkie loss of function interferes with Ack anti-apoptotic signaling . Our results demonstrate how increased Ack signaling could contribute to cancer when coupled to proliferative signals .
Activated Cdc42 kinases ( Acks ) are non-receptor tyrosine kinases that are evolutionarily conserved . The founding member of this family is human ACK1 , which was identified as a protein that binds to CDC42 in its active GTP bound form [1] . Since this discovery Ack homologs have been found in chordates , arthropods and nematodes . Ack family members can be divided into three structural categories based on the presence or absence of four conserved domain motifs ( Figure 1A ) . All Ack family members contain an amino-terminal tyrosine kinase domain that is flanked by a sterile alpha motif ( SAM ) and a Src homology 3 ( SH3 ) domain . The carboxy-terminal half of these kinases contains short proline rich sequences , but lacks any identifiable domains , with the exception of two tandemly repeated ubiquitin-associated ( UBA ) domains located at the extreme carboxy-terminus [2]–[4] . ACK1 UBA domains have been shown to interact with both mono and poly-ubiquitinated proteins [5]–[7] and are thought to play a role in ACK1 protein turnover [6] . The Caenorhabditis elegans Ack homolog , Ark-1 , contains no UBA domains , placing it in a class by itself . The other two Ack structural classes can be distinguished by the presence or absence of a Cdc42/Rac interactive binding ( CRIB ) domain . Human ACK1 and Drosophila PR2 are representative members of the CRIB domain containing structural class , while human TNK1 and Drosophila Ack are members of the structural class lacking a conserved CRIB domain . Variants containing a CRIB domain bind GTP liganded CDC42 , but this interaction does not appear to directly influence Ack activity in vitro [2] . Human ACK1 is the most well characterized member of the Ack family . Early studies uncovered a role for ACK1 in the promotion of internalization and down-regulation of activated epidermal growth factor ( EGF ) receptor . ACK1 tyrosine phosphorylation is enhanced and ACK1 is co-localized with EGF receptor after EGF stimulation [8] , [9] . Knockdown of ACK1 reduces the rate of EGF receptor degradation following EGF stimulation [5] . While on the surface these data suggest that ACK1 merely serves as a negative regulator of growth factor signaling , ACK1 activation may additionally propagate downstream signaling . Recent studies support this latter alternative by uncovering a role for ACK1 as a positive transducer of cell surface receptor signaling that promotes growth and survival by ACK1 mediated phosphorylation and activation of downstream components , including AKT [10] and the androgen receptor [11] . A pro-survival role for Ack function is consistent with reported links between activation of Ack family members and cancer genesis and metastasis . Several somatic missense mutations have been identified in ACK1 from cancer tissue samples that increase ACK1 autophosphorylation and promote cellular proliferation and migration [12] . Amplification of the ACK1 gene in tumors correlates with a poor prognosis , and ACK1 overexpression in cancer cell lines increases invasiveness in a mouse metastasis model [13] , while knockdown of ACK1 reduces the migration of human breast cancer cells [14] , [15] . Activated ACK1 has been detected in advanced human prostate cancers [13] , [16] where it has been shown to phosphorylate three cancer relevant substrates in prostate cancer cell lines: WWOX [16] , AKT [10] , and androgen receptor [11] . WWOX spans the FRA16D chromosomal fragile site that is frequently disrupted in human cancers [17]–[19] . While the molecular function of WWOX is not known , it has been shown that the growth of tumor cells lacking WWOX is strongly inhibited by restoring WWOX expression [20] . ACK1 phosphorylation of WWOX leads to the polyubiquitination and degradation of WWOX , which correlates with a tumorigenic role [16] . AKT is a serine/threonine kinase whose activity promotes cell survival and proliferation , while deregulation of the AKT signaling pathway is commonly associated with cancer [21] . ACK1 activation results in tyrosine phosphorylation and apparent activation of AKT in a PI3K independent mechanism [10] . Finally , the activity of the androgen receptor is required for growth of prostate cells . In advanced stages of prostate cancer , these cells lose their dependence on androgens for activation of this receptor to become androgen independent prostate cancer [22] , [23] . ACK1 has been found to phosphorylate the androgen receptor , promote androgen independent growth of prostate cells , and activate transcription of androgen inducible genes in the absence of androgen [11] . Less is known about the function of Ack family members lacking CRIB domains , and published studies on TNK1 describe conflicting functions . TNK1 overexpression in cell culture lines inhibits cell growth in a kinase dependent manner [24] . Mutant mice having deletions in the kinase domain of TNK1 develop spontaneous tumors at a high frequency , which is thought to originate from hyperactivation of Ras signaling [25] , [26] and suggests that TNK1 functions as a tumor suppressor . In contrast to this function , TNK1 was identified as a potentially oncogenic tyrosine kinase in a mutagenesis screen [27] and activated TNK1 was found in Hodgkin's lymphoma [28] . It is possible that these conflicting findings reflect tissue specific responses or complex dosage sensitivity to TNK1 loss and gain of function . In order to better understand the physiological role of Ack family members and determine how Ack might contribute to cancer , we conducted genetic and biochemical experiments in the model organism Drosophila melanogaster . Our studies focus on Drosophila Ack , which has a domain structure resembling human TNK1 ( Figure 1A ) , but shares significantly higher sequence identity with ACK1 in all conserved domains including the kinase domain activation loop . We find that Drosophila Ack possesses potent anti-apoptotic properties that function downstream of EGF receptor signaling through an unknown mechanism . This activity is dependent on Ack kinase function and can be further stimulated by increased Ras signaling . We have conducted a protein interaction study and find that many of the same proteins that associate with human ACK1 also bind to fly Ack . The influence of these proteins on Ack anti-apoptotic activity was tested , and we determined that the adapter protein Drk ( GRB2 ) is required for this activity , while the transcriptional co-activator protein yki ( YAP ) functions synergistically with Ack to promote cell survival and massive tissue overgrowth . Our findings support both anti-apoptotic and proliferative roles for Ack family members , which may contribute to cancer genesis and progression .
The Drosophila compound eye has been used as an experimental system to assess apoptotic gene function and dissect apoptotic signaling pathways [34] , [35] . Ack null flies have eyes that appear normal as do flies that overexpress Ack or kinase inactive Ack ( AckK156A ) using a promoter that drives expression in the developing eye disc ( GMR-Gal4 , UAS-Ack or AckK156A , Figure 2A–2E ) . A previous study reported that eye expression of either Ack or a kinase dead variant AckK156A using GMR-Gal4 resulted in massive disorganization of the eye [4] . We find that we can replicate these phenotypes by doubling the dose of Ack or AckK156A through the introduction of a second copy of the Ack transgene ( Figure 2F and 2G ) , indicating that our Ack transgenes express proteins at comparatively lower levels . Therefore , any eye size changes seen can be attributed to a genetic interaction with Ack and not due to Ack overexpression itself . Reaper ( rpr ) , head involution defective ( hid ) and grim [34] , [36] , [37] are three cell death genes that function as antagonists of Drosophila inhibitor of apoptosis ( DIAP ) [38] . The BIR domain of DIAP binds to initiator and effector caspases to block their proteolytic activity [39] . IAP antagonists competitively bind to the BIR domain to dissociate bound caspases [40] , leading to caspase activation and apoptotic cell death . Expression of hid using GMR-GAL4 results in a reduced eye size ( Figure 2H ) . Co-expression of the baculovirus caspase inhibitor p35 with hid completely suppresses the small eye phenotype ( Figure 2I ) , demonstrating that the reduction of eye size caused by hid expression is due to the activation of caspases , resulting in apoptosis of cells that make up the eye during development . To determine if Ack contributes to the regulation of cell survival , we assessed the ability of various Ack transgenes and alleles to modify the small eye phenotype induced by hid expression . Co-expression of Ack with hid greatly suppresses the hid induced small eye phenotype , demonstrating that Ack may function in a survival role ( Figure 2J ) . Kinase inactive AckK156A expression fails to modify the hid small eye phenotype , indicating that Ack kinase activity is required for inhibition of hid induced eye size reduction ( Figure 2K ) . When Ack function is removed , either genetically or by RNAi , enhancement of the hid small eye size is observed ( Figure 2L and 2M ) . Additionally , we assessed the ability of Ack to suppress the small eye phenotype induced by rpr expression in the eye . Expression of wildtype Ack or kinase inactive AckK156A suppresses the small eye phenotype induced by rpr expression ( Figure 2N–2P ) , suggesting that non-kinase functions of Ack may be important for suppressing rpr function . Single copy loss or RNAi mediated knockdown of Ack expression in rpr expressing fly eyes results in flies that fail to eclose . Dissection of the pharate adults contained within the pupa cases revealed stunning phenotypes . Gene dosage reduction of Ack in a GMR-rpr background results in eyes that are partially ablated ( Figure 2Q ) . A few female escapers eclosed ( no male escapers were found ) that had eye sizes slightly smaller than GMR-GAL4; GMR- rpr ( Figure 2R ) . Further reduction of Ack dosage by RNAi in the GMR-rpr background results in complete ablation of the eye ( Figure 2S ) with no escapers . In both cases , the ablated eye area is replaced by a hole in the head cuticle where eye tissue would normally reside ( Figure 2Q and 2S ) . Further dissection of the heads of these animals determined that the white tissue visible in place of the eye is not actually eye tissue , but is instead the optic lobe region of the Drosophila brain , which in some animals protrudes outside of the head ( as in Figure 2S ) . Given the shape and location of the holes , we hypothesize that the eye developed to a point and then was partially or completely consumed by massive activation of apoptosis prior to deposition of lens cuticle during pupa development . We conclude from these experiments that Ack has anti-apoptotic or proliferative properties capable of overcoming the effects of hid and rpr induced apoptosis . Finally , overexpression of PR2 or knockdown of PR2 using RNAi fails to significantly modify the hid or rpr induced small eye phenotypes ( Figure S1 ) . We conclude that unlike Ack , PR2 is either not involved in survival regulation or may require additional factors , such as interaction with Cdc42-GTP , for activation of its kinase activity . Ack mediated modification of the hid induced small eye phenotype could be achieved through two mechanisms . Ack could function in an anti-apoptotic role to block pro-apoptotic events initiated by overexpression of hid or rpr . Alternatively , Ack activity could stimulate cellular proliferation such that apoptotic cells are replaced by additional cell division of surviving cells . Again , the Drosophila eye serves as an ideal system to distinguish between these possibilities . Patterning of the Drosophila eye occurs during the third instar larval stage . During pattern formation , a wave of differentiation known as the morphogenetic furrow ( MF ) moves anteriorly across the eye imaginal disc [41] . Cells divide asynchronously ahead of the furrow; however , cell division following the furrow is tightly regulated and only occurs once more in a synchronous band that follows behind the furrow known as the second mitotic wave ( SMW ) [42] . The GMR synthetic promoter drives expression of proteins in cells behind the morphogenetic furrow . Expression of hid via the GMR promoter has been shown to increase apoptotic cells posterior to the furrow , but it also induces an additional region of cell division called the zone of compensatory proliferation ( ZCP ) that is posterior to the SMW ( Figure 3B ) [43] , [44] . Use of the GMR synthetic promoter to express hid and Ack transgenes allows us to assess whether Ack expression leads to inhibition of apoptosis or stimulation of proliferation by TUNEL staining and BrdU incorporation , respectively , in the third instar eye disc . Expression of hid in the eye disc produces an increase in TUNEL positive cells in the region of the SMW and in the most posterior region of the disc containing the ZCP ( compare Figure 3A and 3B ) . Expression of Ack in the eye disc substantially reduces the number of TUNEL positive cells in both regions ( Figure 3C and 3D ) , while expression of kinase inactive AckK156A appears to increase the number of TUNEL positive cells ( Figure 3E and 3F ) . Ack or AckK156A expression does not appear to affect the number of BrdU positive cells . These data demonstrate that Ack kinase activity functions in an anti-apoptotic manner ( see Figure 3G for quantification ) . It is possible that Ack specifically functions within the eye to modulate programmed cell death activation . To address this we adapted a system that had been developed to study regenerative growth in imaginal discs [45] . In this system UAS-rpr is driven by rotund ( rn ) -GAL4 , which is active in several imaginal discs including the wing disc . Reaper expression in this system is further modulated due to the inclusion of a GAL80ts transgene , which allows for the temporal activation of UAS controlled transgenes by raising the temperature from 18°C to 30°C . We determined that a three hour exposure to 30°C followed by a four hour recovery at 18°C was sufficient to produce an intermediate amount of programmed cell death in the wing disc pouch as assessed by TUNEL ( Figure S2A ) . Expression of Ack in this background decreased the number of TUNEL positive cells in the wing disc ( Figure S2B ) while expression of kinase inactive AckK156A or dosage reduction of Ack using the Ack86 allele increased TUNEL positive cells ( Figure S2C and S2D ) . We conclude that Ack anti-apoptotic function is not limited to the Drosophila eye . EGF receptor signaling regulates promotion of apoptosis by hid expression in the Drosophila eye [35] , [46] . Specifically , EGF receptor activation leads to the activation of mitogen activated protein kinase ( MAPK ) , via the Ras/Raf/MEK pathway . In Drosophila , activated MAPK directly phosphorylates hid on multiple serine and threonine residues , which inactivates the pro-apoptotic function of hid [35] . A schematic representation of this pathway is shown ( Figure 4J ) . Expression of hidAla5 , a mutant in which the serine or threonine of 5 putative MAPK sites have been changed to alanine , induces apoptosis which can not be suppressed by MAPK activity [35] . Vertebrate Ack studies have implicated ACK1 in the turnover and regulation of EGF receptor signaling [5] , [7] , [47] . Therefore , it is possible that Ack may influence apoptosis by modulating EGF receptor activity itself or downstream components that ultimately stimulate MAPK . In order to test this possibility , we assessed the ability of Ack expression to block programmed cell death stimulated by hidAla5 expression in the Drosophila eye . Expression of hidAla5 using the GMR promoter produces a reduced eye size due to the induction of apoptosis ( Figure 4A ) . The eye size reduction caused by hidAla5 is greater than that seen by expression of hid ( compare Figure 4A and 2H ) . This variation could simply be due to differences in expression levels of the hid transgenes caused by genomic insertion site positional effects . Alternatively the differences could be explained by the activity of endogenous MAPK , which is able to partially block apoptosis induced by hid , but not by hidAla5 . We selected two proteins that are known to inhibit EGF receptor signaling or Ras/Raf/MEK signaling . Argos inhibits EGF receptor activation by competitively binding extra-cellular EGF , while Gap1 attenuates Ras signaling by facilitating the conversion of Ras-GTP to the inactive Ras-GDP form [48] , [49] . Reducing the protein levels of these inhibitors increases signaling through the EGF receptor pathway , resulting in increased MAPK activation . Reduction of protein levels can be achieved by reducing gene dosage through introduction of a single copy of a strong hypomorphic allele . Single copy loss of either argos or Gap1 has been found to block hid induced apoptosis [35] . However , single copy loss of argos or Gap1 is unable to block apoptosis induced by hidAla5 expression and fails to suppress the small eye size ( Figure 4D and 4G and [35] ) . If Ack kinase activity blocks apoptosis by enhancing signaling through the Ras/Raf/MEK pathway , then we would expect that Ack expression would be unable to suppress apoptosis induced by hidAla5 expression . However , we find that expression of Ack is able to suppress the small eye that is induced by both hid and hidAla5 ( compare Figure 2H to 2J and Figure 4A to 4B ) . In both cases , Ack expression induces a similar fold increase in eye size: roughly a 1 . 8 fold increase in eye area for hid and a 1 . 6 fold increase for hidAla5 . Taken together , our data demonstrate that Ack kinase activity blocks programmed cell death induced by hid through a MAPK independent mechanism . In vertebrates , ACK1 has been shown to be tyrosine phosphorylated and localized to the EGF receptor upon EGF stimulation [5] , [50] , [51] . This is largely thought to lead to the internalization and degradation of the EGF receptor . Our findings raise the possibility that in addition to EGF receptor turnover , Ack activation may signal to uncharacterized downstream components to regulate programmed cell death . By using hidAla5 to induce programmed cell death , we eliminate the pro-survival signaling of the EGF receptor through MAPK . If EGF signaling activates Ack , then we would expect that the introduction of mutations that activate the EGF receptor or downstream components would result in an increase in the ability of Ack to suppress apoptosis . This is in fact what we observe: gene dosage reduction of argos or Gap1 combined with expression of Ack further suppresses the hidAla5 induced small eye phenotype ( compare Figure 4B to 4E and 4H ) . These data suggest that Ack may not be directly activated by the EGF receptor , but may instead be activated at the level of Ras-GTP or further down the Raf/Mek/MAPK pathway . Finally , suppression of apoptosis is absolutely dependent on the kinase activity of Ack , as AckK156A shows no suppression of the hidAla5 small eye phenotype either alone or in conjunction with mutations that enhance EGF receptor or Ras signaling ( Figure 4C , 4F and 4I ) . Next we sought to determine if Ack activity is able to block programmed cell death that occurs during normal developmental tissue patterning . The Drosophila compound eye is composed of 800 simple eye units called ommatidia comprising eight photoreceptors , four cone cells and two primary pigment cells . Each ommatidium is surrounded by a regular hexagonal array of 12 interommatidial cells comprising six secondary pigment cells , three tertiary pigment cells and three bristles ( Figure 5A ) . The lattice of interommatidia cells starts to form at 20% pupal development and is completed by 33% of pupal development . During this time period , excess interommatidial cells are selectively eliminated by programmed cell death . Ack expression results in an increase in the number of bristles without significantly affecting the number of pigment cells as assessed at 42% pupa development ( Figure 5B and 5H ) . Expression of AckK156A also increases the number of bristles within the interommatidia lattice , but results in a 33% decrease in pigment cell number ( Figure 5C and 5H ) . Single copy loss of argos or Gap1 does not affect bristle or pigment cell number ( Figure 5D , 5E and 5H ) on their own , but produces a 25% increase in pigment cell number when combined with Ack expression ( Figure 5F–5H ) . These data further support an anti-apoptotic function for Ack that can be activated by EGF/Ras signaling and suggest that kinase inactive AckK156A expression enhances programmed cell death activation in a background not containing overexpression of DIAP antagonists . Our results indicate that Ack kinase activity is anti-apoptotic and that this activity can be enhanced by activation of the EGF receptor or its downstream signaling components such as Ras . The proteins that link EGF/Ras signaling to Drosophila Ack activation and the downstream targets of Ack are not known . Several vertebrate ACK1 interacting proteins have been identified ( see [52] for a comprehensive list ) and some of these ( including AKT , GRB2 , NCK , NEDD4 and WWOX ) are attractive candidates for regulation or transduction of Ack anti-apoptotic signaling . We utilized a biochemical approach to identify Ack interacting proteins from Drosophila Schneider 2 ( S2 ) cells ( see Materials and Methods for details ) . We identified several Ack interacting proteins that are fly orthologs of known vertebrate ACK1 interacting proteins , including Clathrin , Drk ( GRB2 ) , Hsp83 ( HSP90 ) , Dock ( NCK ) , WASp ( WAS ) and SH3PX1 ( SNX9 ) . The top ten hits based on Mascot ( probability ) score that specifically co-purify with Ack and are absent from control preparations are shown ( Table 1 ) . RNAi lines for each of the proteins in Table 1 as well as Dock , Akt1 , Nedd4 and Wwox were obtained and used to determine if knockdown of these proteins influences the ability of Ack to rescue hidAla5 induced programmed cell death . Weak or no modification of eye size was seen for the candidate Ack interacting proteins CG4169 , Gale , Hsp23 and Dock . The ACK1 interacting proteins not identified in our dataset ( Akt1 , Nedd4 and Wwox ) also failed to robustly modify the eye size in this background ( see Figure 6 ) . Additionally , we obtained loss and gain of function alleles for Akt1 , but could detect only minor enhancement or suppression , respectively , of hid or hidAla5 induced programmed cell death , which appeared to produce an additive effect when combined with Ack expression ( Figure S3 ) . SH3PX1 RNAi suppresses the hidAla5 induced small eye phenotype in the Ack expressing background , while Hsp26 and Magi RNAi enhance the small eye phenotype . These data indicate that SH3PX1 , Hsp26 and Magi may function directly with Ack or in a parallel pathway to regulate cell survival . Hsp83 , Yorkie and Drk had the largest effect and were further analyzed: RNAi knockdown of Hsp83 and Yorkie in a hidAla5 and Ack expressing background resulted in flies that failed to eclose from pupa cases , while Drk knockdown produced adults that had greatly reduced eye size ( Figure 6E ) . Based on our results , we elected to further characterize the requirement of Drk and yki for Ack anti-apoptotic function . Drk is an adaptor protein made up of SH2 and SH3 domains and is the fly ortholog of vertebrate GRB2 and C . elegans Sem5 . In both vertebrates and C . elegans , these proteins have been shown to interact with their Ack equivalents to negatively regulate EGF receptor signaling [50] , [53] . We find in the absence of Ack overexpression , Drk RNAi does not influence hid or hidAla5 induced apoptosis as assessed by modification of eye size ( compare Figure 7A to 7D and 7B to 7E ) . Expression of Ack suppresses hidAla5 induced cell death ( compare Figure 7B to 7C ) , and remarkably this suppression can be eliminated by RNAi knockdown of Drk ( compare Figure 7B and 7C to 7F ) . We conclude from these experiments that Ack requires Drk function for transmission of its anti-apoptotic signal . Drk immunoprecipitation followed by western analysis shows that Drk and Ack exist in a physical complex , and this interaction can be enhanced by increasing Ack expression levels ( Figure 7G ) . We further find that the Ack that co-precipitates with Drk is tyrosine phosphorylated , as assessed by western analysis , but we are unable to detect any tyrosine phosphorylation on Drk itself ( Figure 7H ) . Taken together , these data support a model in which Drk is not an Ack substrate , but serves as an adapter protein that functions to position Ack to receive and propagate anti-apoptotic signals . Yorkie ( yki ) is the Ack interacting protein that has the second highest Mascot score . Yki has not been previously identified as an Ack family member binding protein , but is an attractive candidate due to its role as a transcriptional co-activator that interacts with the TEAD/TEF family protein Scalloped to control expression of proliferative ( Cyc E ) and anti-apoptotic genes ( DIAP ) [54] , [55] . Nuclear yki is shuttled to the cytoplasm by the FERM-domain containing protein Expanded [56] . Cytoplasmic yki is subject to inactivation by the Salvador-Warts-Hippo signaling pathway , which functions to regulate organ size [57] . Yki phosphorylation by the serine/threonine kinase Warts leads to yki interaction with cytoplasmic 14-3-3 proteins and yki sequestration from the nucleus [58] . Since yki is a transcriptional co-activator of both proliferative and anti-apoptotic genes , it is tempting to speculate that Ack functions to promote the nuclear localization of yki to enhance transcriptional activation of yki target genes . We conducted a series of experiments to test this hypothesis and further characterize the interactions between Ack and yki . We have shown that expression of Ack in the Drosophila eye blocks native apoptosis ( Figure 3C ) and produces a slightly larger and mildly rough eye ( Figure 8B ) . Expression of yki in the eye produces a strong overgrowth phenotype in which eye tissue protrudes out of the normal curvature of the eye ( Figure 8C ) . Co-expression of Ack and yki reveals a synergistic effect , producing massive overgrowth ( Figure 8D ) . We draw two conclusions from these results . First , Ack and yki expression in the eye result in different phenotypes , suggesting that Ack may not activate yki transcriptional activity . Second , Ack activity is not limited to prevention of apoptosis , but can also enhance proliferation when combined with proliferative signals . Next we sought to determine if Ack and yki expression resulted in similar suppression of hid induced programmed cell death in the eye . We find that Ack and yki result in similar magnitudes of suppression of hid small eye size ( Figure 8E–8G ) ; however , yki expression in the hid background produces eyes displaying a loss of eye pigment whereas Ack expression results in uniformly pigmented eyes . Co-expression of Ack and yki in the GMR-hid background further suppresses the hid small eye phenotype , but still results in an eye with comparable pigment loss to yki suppression of hid alone ( Figure 8H ) . The fact that we observe different results for Ack ( Figure 8F ) versus yki ( Figure 8G ) suppression of the hid induced small eye phenotype suggests that Ack suppresses apoptosis via a mechanism that is independent of yki mediated transcriptional co-activation . Based on our data , it is unlikely that Ack enhances yki nuclear localization and subsequent activation of yki transcriptional targets . To confirm this we assessed the effect that Ack has on yki subcellular localization . Ack-mCherry expressed in R-cells is excluded from the nucleus and localizes throughout the cytoplasm . The highest concentrations of Ack are at apical membrane surfaces and in puncta near the nuclear envelope and within axons ( Figure S4A ) . Yki-GFP expressed in R-cells has a localization pattern similar to Ack-mCherry , exhibiting nuclear exclusion with cytoplasmic , apical membrane and axonal puncta localization; however , the highest levels are seen ringing the nucleus with puncta apparent on or near the nuclear envelope ( Figure S4B ) . Co-expression of Ack-mCherry and yki-GFP reveals overlapping expression patterns in the apical membrane regions and both axon and nuclear localizing puncta ( Figure S4C–S4E ) . Co-expression of Ack and yki does not result in increased nuclear localization of either protein but does increase the yki localization to puncta surrounding the nucleus . Consistent with these data , we detected no increases in beta-galactosidase expression in the yki transcriptional activity reporter line ex697 when Ack is overexpressed ( Figure S5A–S5F ) . These data are consistent with Ack and yki participating in physical interactions outside of the nucleus that do not lead to nuclear import of either protein or transcriptional up-regulation of yki targets . We tested the effect of protein dosage on the ability of Ack or yki to suppress hid induced apoptosis . Single copy loss of yki dominantly suppresses Ack anti-apoptotic function ( compare Figure 8F to 8J ) , while single copy loss of Ack has no effect on the ability of yki to suppress the hid induced small eye phenotype ( compare Figure 8G to 8K ) . Single copy loss of yki did not noticeably modify the hid induced small eye in the absence of Ack suggesting that yki loss is specifically affecting Ack function ( Compare Figure 8E to 8I ) . Finally , we employed yki RNAi to further reduce the levels of yki in the eye . Yki RNAi in the GMR-hid background resulted in flies that die as pharate adults . Expression of exogenous Ack in this background rescues lethality and produces eye sizes smaller than yki single copy loss ( Figure 8l ) . Taken together these data support a model in which Ack anti-apoptotic function is activated by yki protein . Yes associated protein ( YAP ) is the vertebrate homolog of yki . Mutant versions of YAP have been generated and studied in the Drosophila eye . The YAP S127A mutation prevents phosphorylation by LATS family kinases , thereby increasing YAP nuclear localization . The YAP S94A mutation disrupts the binding of YAP to TEAD family DNA binding proteins . Expression of YAPS127A in the fly eye leads to an overproliferation phenotype , while expression of the double mutant YAPS94A/S127A produces eyes that are normal in appearance due to an inability to bind Sd and activate YAP/yki target gene transcription [59] . We obtained YAPS94A/S127A and found that co-expression of YAPS94A/S127A with Ack leads to an increase in eye size ( compare Figure 8M and 8N ) , indicating that YAP can also genetically interact with Ack to enhance proliferation even when it cannot serve as a transcriptional co-activator with Scalloped .
Drosophila melanogaster has two Ack family members: Ack and PR2 . We have found that Ack possesses anti-apoptotic properties , while PR2 either does not possess anti-apoptotic properties or requires activators not present in our assay system . While Ack may appear to be more closely related to vertebrate TNK1 because both proteins lack a CRIB domain , Ack is most similar to vertebrate ACK1 based on sequence identity of all shared protein domains . Additionally we find that many of the proteins that interact with ACK1 also interact with fly Ack . Therefore , we feel that the conclusions drawn in this study will likely be applicable to vertebrate ACK1 function . We have shown that Ack function can suppress programmed cell death induced by hid or rpr expression in the developing eye and wing discs . Our overexpression studies reveal that Ack kinase activity is required for suppression of apoptosis induced by hid but is unnecessary for suppression of rpr-induced apoptosis . We further show that Ack loss of function enhances cell death induced by expression of both of these genes and determine that Ack is critically required for the survival of rpr expressing eye tissue . The molecular mechanisms underlying these differential requirements are not known . Hid and rpr are known to function in a multimeric protein complex on the mitochondria outer membrane to promote apoptosis [60] . Both hid and rpr are able to stimulate apoptosis by competing with initiator and effector caspases for DIAP binding , but rpr additionally induces DIAP auto-ubiquitination leading to DIAP degradation . While Ack kinase activity may be important for aspects of hid complex regulation , it is tempting to speculate that the UBA domains of Ack may play a critical role in the modulation of DIAP or rpr ubiquitination and stability . EGF receptor signaling has been shown to activate ACK1 in vertebrates , and we find that EGF signaling enhances the anti-apoptotic function of Ack in Drosophila . ACK1 negatively regulates EGF receptor signaling by stimulating endocytosis of activated receptor complexes . We have evidence supporting that Drosophila Ack , in conjunction with SH3PX1 , functions in a similar manner , which will be described elsewhere . In Drosophila , EGF signaling is anti-apoptotic through the activation of MAPK , which phosphorylates and inactivates hid . If Ack affected apoptosis exclusively through attenuation of EGF signaling , then we would expect that Ack loss of function would be anti-apoptotic while gain of function would be pro-apoptotic , which is opposite to what we observe . By using the hidAla5 mutant , we demonstrate that Ack does not modulate programmed cell death through activation of MAPK . The anti-apoptotic function of Ack is surprisingly robust compared to other proteins that we have tested . Our studies show that activity of the kinase domain contributes to Ack anti-apoptotic function . Based on this , we conclude that Ack propagates anti-apoptotic signals by phosphorylating downstream targets . Several ACK1 substrates have been identified that are attractive candidates for the regulation of programmed cell death: the putative tumor suppressor WWOX , the apoptosis inhibiting protein kinase AKT and the caspase-cleaved ubiquitin E3 ligase NEDD4 . Akt1 loss and gain of function alleles ( Figure S3 ) and Nedd4 RNAi ( Figure 6K ) fail to significantly modify hid induced small eye phenotypes . Wwox RNAi is able to suppress hid induced apoptosis but not nearly as robustly as Ack expression ( Figure S6 ) . Since ACK1 mediated phosphorylation of WWOX leads to WWOX destruction , we would predict that Wwox RNAi would phenocopy Ack expression in our assay system , but it is unable to reproduce the magnitude of Ack anti-apoptotic function . This does not rule out Wwox as an anti-apoptotic substrate target of Ack , but it demonstrates that Wwox is not the only cell death relevant substrate of Ack . It is worth noting that hid and rpr act fairly late within the programmed cell death pathway , being just a step upstream of initiator caspase activation . Therefore , Ack must target substrates that have activities influencing hid , rpr or events at the level of caspase activation . We identified several Ack physically interacting proteins using a tandem affinity purification strategy . Many of these have vertebrate homologs that have been previously determined to interact with ACK1 . We found that Drk and yki have the most pronounced effect on Ack's anti-apoptotic properties and chose to further characterize their contribution to Ack signaling . Drk is the fly ortholog of vertebrate GRB2 , which has previously been described as an ACK1 and TNK1 interacting protein . In the case of TNK1 , GRB2 is tyrosine phosphorylated by TNK1 , which disrupts the ability of GRB2/SOS complexes to activate Ras [26] . This does not appear to be the case for Drosophila Ack , because even though Drk forms a complex with tyrosine phosphorylated Ack , we find no evidence of tyrosine phosphorylation on Drk . Rather , our data support that Drk association with Ack is required for Ack anti-apoptotic properties . We propose that Drk SH3 domains likely interact with PXXP motifs in the C-terminal half of Ack similar to the interaction described in vertebrates [50] . This interaction could then lead to the recruitment of Ack into protein complexes required for Ack activation . Yki is a transcriptional co-activator that regulates expression of genes with proliferative and anti-apoptotic functions . The vertebrate homolog of yki is Yes Associated Protein ( YAP ) , which has not previously been identified as an ACK1 or TNK1 interacting protein . Yki and YAP studies have focused primarily on the pathways that regulate their function as transcription factors . Given the role of yki and YAP in transcriptional control of proliferative and anti-apoptotic genes , it would seem likely that Ack activity leads to enhancement of yki function . However , this does not appear to be the case because Ack overexpression does not lead to increased yki nuclear localization or increased expression of yki target genes . Rather , our data indicate that yki directly interacts with Ack in the cytoplasm and functions to regulate Ack activity . In support of this , we have found that Ack colocalizes with yki , and yki dosage reduction suppresses Ack anti-apoptotic function . Yki contains two WW domains , which may interact with conserved PPXY motifs that are present in the region flanked by the SH3 and UBA domains of Ack family members . In vertebrates , these PPXY motifs have been shown to interact with WWOX , which also contains two WW domains [16] . Further studies are required to define how yki and Ack interact . Yki expression in the fly eye produces an overgrowth phenotype that is indicative of its role in regulating proliferation . Ack overexpression produces a slightly larger eye due to inhibition of apoptotic events that occur during normal eye development . Simultaneous expression of yki and Ack results in a synergistic effect that produces enormous eyes . These results reveal that in addition to anti-apoptotic function , Ack can also enhance proliferation . This illustrates how increased Ack signaling could contribute to cancer when coupled to proliferative signals . Indeed , our results are consistent with recent reports of ACK1 activating somatic mutations [12] and gene amplification [13] being associated with human cancers . At present the key anti-apoptotic substrates of Ack and their mechanisms of action remain to be determined . With their discovery will come a better understanding of Ack signaling and potentially new targets for cancer interventions .
Standard Drosophila genetic technique was used for all crosses and transgenesis . All flies were raised at 25°C with 70% relative humidity . Fly lines used are as follows: Oregon R , GMR-Gal4/CyO , GMR-hidG1/CyO , GMR-p35 , P{lacW}Gap1/TM3 , argosΔ7/TM3 , UAS-Dcr-2/CyO , UAS-Dcr-2 , P{PZ}Akt104226/TM3 , GMR-rpr/TM6B , UAS-yki-GFP ( Bloomington Stock Center ) ; UAS-PR2 and Ack86/TM3 ( a gift from Nicholas Harden , Simon Fraser University ) ; GMR-hidAla5/CyO ( a gift from Andreas Bergmann , M . D . Anderson Cancer Center ) ; UAS-yki and FRT42 , ykiB5/CyO ( a gift from Nic Tapon , London Research Institute ) ; UAS-YAPS127A and UAS-YAPS94A/S127A ( a gift from Zhi-Chun Lai , Pennsylvania State Univeristy ) ; rn-GAL4 , UAS-rpr , tub-GAL80ts/TM6B , GAL80 ( a gift from Ken Irvine , Rutgers University ) ; UAS-PR2 RNAi , UAS-Ack RNAi , UAS-SH3PX1 RNAi , UAS-CG4169 RNAi , UAS-Gale RNAi , UAS-Drk RNAi , UAS-HSP 26 RNAi , UAS-HSP 23 RNAi , UAS-HSP 83 RNAi , UAS-yki RNAi , UAS-14-3-3 epsilon RNAi , UAS-Magi RNAi , UAS-Dock RNAi , UAS-Nedd4 RNAi , UAS-Wwox RNAi ( Vienna Drosophila RNAi Center , VDRC ) . UAS-Ack constructs were produced by sub-cloning Ack from pMT/V5-Ack plasmid ( a gift from Jack Dixon , UCSD ) into EcoRI/XhoI sites in the pUASt vector . Transgenic flies ( UAS-Ack/TM6B , UAS-AckK156A/TM6B and UAS-Ack-mCherry ) were made by microinjecting pUASt-Ack DNA constructs into yw flies . GMR-hid and GMR-hidAla5 were recombined with GMR-Gal4 to aid in the Drosophila eye size assays . Newly eclosed males were dissected in PBS and transferred to PBX ( PBS+0 . 1% Triton X-100 ) containing 4% formaldehyde for 20 minutes . Testis were washed with PBX , incubated with anti-cleaved caspase-3 ( cell signaling , 1∶75 ) overnight at 4°C followed by anti-rabbit Alexa 488 ( 1∶500 , Invitrogen ) secondary antibody for 3 hours . F-actin was stained using 1 unit of Alexa 568 phalloidin ( Invitrogen ) and nuclei were stained with Hoechts used at 1∶10 , 000 dilution ( Thermo Scientific ) . Samples were mounted in VECTASHIELD and images were acquired using a Disk Scanning Unit-enabled Olympus BX61 microscope equipped with a Hamamatsu DCMA-API camera . GMR-Gal4 , GMR-hid/CyO or GMR-Gal4 , GMR-hidAla5/CyO virgin females were crossed with male flies of various genotypes . Color images were taken by removing the heads and mounting them with rubber cement onto 1 µl micropipettes and imaged submerged in water using a Zeiss Discovery . V12 dissecting microscope equipped with a color camera and a plan S 1 . 0X FWD 81 mm objective . All images were acquired and processed identically . Images were taken at 85× magnification , cropped & rotated using Adobe Photoshop and then scaled uniformly using Microsoft PowerPoint . Images were acquired using a Zeiss LSM 710 confocal microscope . For immuno and direct fluorescence , 3rd instar larva were dissected in PBS and fixed in 4% paraformaldehyde for 30 minutes . Eye discs expressing GFP and mCherry fusion proteins were washed prior to mounting in VECTASHIELD ( Vector Laboratories ) and fluorescent imaging . For analysis of non-tagged proteins , the eye discs were blocked in PBTN ( 10% normal goat serum in PBT ( PBS+0 . 5% Triton X-100 ) ) and incubated overnight in PBTN containing antibodies directed against Ack ( JCD2 , 1∶1000 ) and beta-galactosidase ( 40-1a , 1∶50 , Developmental Studies Hybridoma Bank , University of Iowa ) followed by one hour incubations with goat anti-mouse Alexa 568 ( 1∶500 ) and goat anti-rabbit Alexa 488 ( 1∶500 ) secondary antibodies . For TUNEL and BrdU analysis , 3rd instar larva imaginal discs were dissected in PBS and incubated in BrdU incorporation solution ( 5 µl/ml BrdU ( Sigma ) in PBS ) for 1 hour at room temperature ( RT ) . Eye discs were then washed with PBS and fixed with 4% paraformaldehyde for 30 minutes . The samples were blocked overnight at 4°C with PBTN . TUNEL labeling: Eye discs were incubated in 99 mM Na-Citrate+0 . 1% Trition X-100 for 30 minutes at 65°C , washed and incubated in TUNEL assay solution ( In Situ Cell Death Detection Kit , Roche ) for 1 hour at 37°C rotating in the dark . Eye discs were then washed in PBT and once in PBS then treated with 2N HCl for 30 minutes and neutralized with 100 mM Borax for 2 minutes . They were then washed with PBT and re-blocked with PBTN for 1 hour at RT . BrdU Staining: tissue was incubated overnight at 4°C in anti-BrdU primary antibody ( 1∶20 , G3G4 , Developmental Studies Hybridoma Bank , University of Iowa ) , washed and incubated for 3 hours at RT with goat anti-mouse Alexa 568 ( 1∶500 ) secondary antibody . Eye discs were washed twice with PBT , twice with PBS and then mounted in VECTASHIELD . rn-GAL4 , UAS-rpr , tub-GAL80ts/TM6B , GAL80 flies were crossed to yw , UAS-Ack , UAS-AckK156A or Ack86 and the progeny were raised at 18°C until third instar larva appeared . Male third instar larva were isolated and transferred to new vials that contained approximately 1 ml of mechanically churned food . Vials were then placed in a 30°C water bath for 3 hours and then returned to 18°C for four hours . Wing discs were dissected , fixed and subjected to TUNEL analysis as described above . Crosses were established at 22°C and larvae allowed to pupate to 42% pupal development . Retinas were dissected in standard saline buffer ( 2 mM KCl , 128 mM NaCl , 4 mM MgCl2 , 1 . 8 mM CaCl2 , 36 mM sucrose , 5 mM HEPES pH 7 . 1 ) . Immediately after dissection , retinas were immersed in fixative ( 10 mM periodate , 75 mM lysine , 2% paraformaldehyde in 1× PBS ) with 0 . 05% saponin for 20 minutes . Following fixation , retinas were washed in PBX and incubated with mouse anti-Armadillo ( 1∶30; DSHB ) and Alexa 568 phalloidin ( Molecular Probes ) with 5% goat serum overnight at 4°C . Retinas were then washed in PBST and incubated in Alexa 646 ( Molecular Probes ) and Alexa 488 conjugated secondary antibodies for 3 hours at room temperature . Retinas were then washed with PBST and mounted in mounting medium ( 0 . 25% n-propyl gallate , 50% glycerol in PBS ) . Confocal images were taken and the images were inverted and printed . For easier analysis the pigment cells were false colored . Schneider 2 ( S2 ) Drosophila cells were transfected with pMT/V5-6xHis-Flag-Ack and pCoHygro in Schneider's S2 medium ( Invitrogen ) . Stably transfected cells were obtained by hygromycin selection ( 200 µg/ml , Invitrogen ) . Both control S2 cells ( non-transfected ) and 6xHis-Flag-Ack expressing S2 cells were subjected to the following purification procedure and mass spectrometry analysis . Cells were grown in a liter spinner flask and induced with 0 . 7 mM CuSO4 for 1 day . Cells were harvested and suspended in RIPA buffer ( 150 mM NaCl , 1% Igepal CA-630 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , 50 mM Tris pH 8 . 0 , . 2 mM sodium orthovanadate , 10 mM NaF , 0 . 4 mM EDTA pH 8 , 10% glycerol ) then lysed by sonication . 6xHis-Flag Ack was captured using Flag M2 resin ( Sigma ) and Ack was eluted with 100 µg/ml 3× flag peptide ( Sigma ) . A second level of purification was achieved by adding a Phospho-tyrosine resin ( protein A with cross-linked 4G10 antibody ( Millipore ) ) to the flag elution . The final elution was with 100 mM triethanolamine ( TEA ) pH 11 . 5 and dried using a speedvac . To prepare samples for mass spectrometry , purified proteins were resuspended in 50 mM ammonium bicarbonate with 0 . 2% RapiGest ( Waters ) . 5 mM tris ( 2-carboxyethyl ) phosphine was added and the sample was heated at 56°C for 30 minutes . After heating , 15 mM Iodoacetamide was added and vortexed for 1 hour in the dark . 1 µg trypsin was added to digest the protein overnight at 37°C . 0 . 5 ul of concentrated HCl was added to degrade the RapiGest . The sample was then centrifuged and dried using a speedvac and subsequently resuspended in 5% acetonitrile , 0 . 1% trifluoroacetic acid prior to liquid chromatography and electrospray ionization-MS/MS analysis . An Agilent 1100 nano-HPLC system was used to separate peptides using Zorbax C18 trap and 75 um×150 mm capillary columns . Peptides were eluted with a gradient of increasing acetonitrile in 0 . 1% formic acid at a flow of 300 nL/min and injected into a Thermo LTQ-Orbitrap using a nanoelectrospray source . The Orbitrap was operated in data-dependent tandem MS mode with one MS scan followed by 3 tandem MS scans and a dynamic exclusion window of two minutes . Data were searched against a database of Drosophila melanogaster protein sequences using both Sorcerer ( Sage-N Research ) and Mascot ( MatrixScience ) search software . Similar results were obtained with both . Database searches were set with a mass tolerance of 25 ppm , full tryptic cleavage , one allowed mis-cleavage , and carbamidomethyl cysteine modification . Identified proteins were considered specific Ack binding partners only if the following stringent criteria were met: no peptides from the protein were observed in the control sample; at least 3 unique peptides were identified with Mascot ion scores above the designated threshold significance value; and Mascot protein scores were above 100 . S2 cells and S2 cells expressing exogenous Ack were grown to confluency , counted and plated in a 6-well dish containing S2 medium at a concentration of 1×106 cells/1 ml/well . Ack dsRNA ( 15 µg ) was added to a well with S2 cells and incubated for 45 minutes before the addition of 2 ml S2 media supplemented with 10% FBS to all wells . After 24 hours , 0 . 7 mM CuSO4 was added to the Ack overexpressing line to induce Ack expression . The cells were harvested after an additional 48 hours and lysed in RIPA buffer by sonication . Cleared lysate was achieved from a tandem centrifugation procedure of 10 minutes at 4°C on a tabletop centrifuge and then transferred to an ultracentrifuge for 30 minutes at 100 , 000×g at 4°C . The lysates were incubated with anti-Drk antibody ( a gift from Efthimios Skoulakis , Alexander Fleming Biomedical Sciences Research Center , Greece ) on ice for 1 hour before adding 50 µl slurry of immobilized protein A resin ( G-Biosciences ) for 1½ hours at 4°C with continuous rotation . The samples were washed with RIPA buffer . Drk IP samples were eluted in SDS-PAGE sample buffer and subjected to SDS-PAGE . The protein was transferred to an immobilon-P transfer membrane ( Millipore ) and blocked for 1 hour with 5% milk . Drk and Ack were detected using anti-Drk 12201 antibody ( 1∶3 , 000 , overnight ) and anti-JCD2 ( 1∶1 , 500 , 1 hour; [61] ) primary antibodies and goat anti-rabbit secondary antibody ( 1∶10 , 000 ) for 1 hour and detected with ECL detection reagent ( Thermo Scientific ) .
|
A number of recent studies have uncovered an involvement of Ack family members in human cancer . The majority of these studies focus on human ACK1 and suggest that ACK1 regulates diverse cancer-relevant biological functions , including stimulation of proliferation , blocking programmed cell death , and enhancing metastasis . It is unclear from these studies whether these biological outcomes are directly controlled by ACK1 activity or if they are indirect consequences of ACK1 signaling . Using Drosophila as a model organism , our study demonstrates that Ack serves to promote cell survival by blocking programmed cell death: a mechanism of eliminating excess , damaged , or cancerous cells . We further find that Ack activity functions synergistically with cell growth signals to produce massive cellular overgrowth . Our findings define the physiological role of Ack proteins and add further support to the value of Ack family members as therapeutic drug targets for the treatment of cancer .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"medicine",
"cell",
"death",
"basic",
"cancer",
"research",
"signaling",
"in",
"selected",
"disciplines",
"gene",
"function",
"oncology",
"tyrosine",
"kinase",
"signaling",
"cascade",
"signaling",
"in",
"cellular",
"processes",
"oncogenic",
"signaling",
"ras",
"signaling",
"biology",
"antiapoptotic",
"signaling",
"molecular",
"biology",
"signal",
"transduction",
"apoptotic",
"signaling",
"cascade",
"cell",
"biology",
"genetics",
"molecular",
"cell",
"biology",
"genetics",
"and",
"genomics",
"signaling",
"cascades"
] |
2012
|
Drosophila Activated Cdc42 Kinase Has an Anti-Apoptotic Function
|
The ever-increasing availability of transcriptomic and metabolomic data can be used to deeply analyze and make ever-expanding predictions about biological processes , as changes in the reaction fluxes through genome-wide pathways can now be tracked . Currently , constraint-based metabolic modeling approaches , such as flux balance analysis ( FBA ) , can quantify metabolic fluxes and make steady-state flux predictions on a genome-wide scale using optimization principles . However , relating the differential gene expression or differential metabolite abundances in different physiological states to the differential flux profiles remains a challenge . Here we present a novel method , named REMI ( Relative Expression and Metabolomic Integrations ) , that employs genome-scale metabolic models ( GEMs ) to translate differential gene expression and metabolite abundance data obtained through genetic or environmental perturbations into differential fluxes to analyze the altered physiology for any given pair of conditions . REMI allows for gene-expression , metabolite abundance , and thermodynamic data to be integrated into a single framework , then uses optimization principles to maximize the consistency between the differential gene-expression levels and metabolite abundance data and the estimated differential fluxes and thermodynamic constraints . We applied REMI to integrate into the Escherichia coli GEM publicly available sets of expression and metabolomic data obtained from two independent studies and under wide-ranging conditions . The differential flux distributions obtained from REMI corresponding to the various perturbations better agreed with the measured fluxomic data , and thus better reflected the different physiological states , than a traditional model . Compared to the similar alternative method that provides one solution from the solution space , REMI was able to enumerate several alternative flux profiles using a mixed-integer linear programming approach . Using this important advantage , we performed a high-frequency analysis of common genes and their associated reactions in the obtained alternative solutions and identified the most commonly regulated genes across any two given conditions . We illustrate that this new implementation provides more robust and biologically relevant results for a better understanding of the system physiology .
The turnover rates of metabolites through a pathway are called fluxes , and genome-wide intracellular metabolic fluxes are the ultimate regulator of cellular physiology . Perturbations on the normal physiology , such as those that occur in a disease state , directly influence the metabolic fluxes . The well-established experimental approach for determining these metabolic fluxes is 13C metabolic flux analysis , though this experimental technique that directly measures metabolite levels is costly and time-consuming , such that computational tools for flux prediction have become a very popular alternative . Genome-scale metabolic models ( GEMs ) , which essentially associate an organism’s genotype with its phenotype , integrate genomic information with known information about metabolite levels to comprehensively describe an organism's metabolism [1] . These models can predict metabolic fluxes , growth rates , or the fitness of gene knockouts using constraint-based approaches , which mainly require the knowledge of network stoichiometry that is available from the annotated genome sequences and metabolic pathway databases . One of the most routinely used constraint-based approaches is flux balance analysis ( FBA ) , which relies on the stoichiometry and optimization principles to predict the steady-state metabolic flux distribution according to an objective function in a given metabolic network [2] . Due to network complexity , FBA commonly results in a span of alternative optimal solutions indicating different flux distributions with the same objective value rather than a unique steady-state flux distribution profile , and then selects one of these solutions at random to present back to the user , which is a major limitation of this method . To remedy this , it has been shown that integrating additional layers of constraints , such as thermodynamics , can effectively reduce the overall solution space of feasible flux distributions in an organism to limit the number of alternative solutions [3 , 4] . With the growing availability of high-throughput data for different organisms under a wide range of genetic or environmental perturbations , GEMs became popular because of their ability to incorporate omics data as additional regulatory constraints for FBA problems . Because GEMs associate a genotype with a phenotype , it is essential to understand that a single genome can result in thousands of different physiologies through different regulatory mechanisms . Therefore , the integration of static snapshots of the metabolism , obtained from transcriptomic and metabolomic data , provides more biologically relevant constraints for the system and helps to increase the precision of the flux prediction , therefore better deducing the observed physiology . However , despite the high number of methods that have been introduced in recent years for the integration of omics data into constraint-based metabolic models , the enhanced prediction of flux profiles using omics data , particularly in cases using multi-omics data , is still far from being resolved . Recently , these methods , their scopes , and limitations were extensively reviewed [5] , and the authors concluded that using gene-expression data does enhance flux predictions , though they inferred that the accurate predictions of the physiology is not achievable with the available reviewed methods . The existing methods for integrating gene-expression data into GEMs can be classified into two categories with the first relying on the integration of absolute gene-expression data into GEMs . This includes techniques such as gene inactivity moderated by metabolism and expression ( GIMME ) [6] and the use of continuous and discrete formulations to find a flux distribution that is consistent with given context-specific gene-expression data , including integrative metabolic analysis tools ( iMAT ) [7 , 8] [5 , 9–11] . However , the assumption that absolute gene-expression data can be directly correlated with flux values is questionable and might not hold true for all genes . Moreover , these methods require user-defined thresholds to identify and categorize the expression levels of metabolic genes ( high , moderate , or low expression ) , and the results are sensitive to the set thresholds . These drawbacks motivated the development of ( ii ) the second class of methods , which integrate the relative gene-expression data while aiming to maximize the correlation between differential changes in gene-expression and reaction fluxes . The underlying assumption for this class of methods is that the relative changes in gene expression between two conditions correlate with the resulting differential flux profiles [12 , 13] . The increasing availability and quality of metabolomic data have promoted the development of methods that can be integrated into GEMs to refine model reconstruction , to reduce the solution space of feasible fluxes , and to better predict the physiological state of a system . These methods , their scope , and their limitations have been reviewed by Töpfer et al . [14] . One of these methods , thermodynamic-based flux balance analysis ( TFA ) , integrates the absolute metabolite concentration data into GEMs , as the metabolite concentrations are intrinsically associated with the Gibbs free energy of metabolic reactions [3 , 4] . Another available method is gene inactivation moderated by metabolism , metabolomics , and expression ( GIM3E ) , an extension of the GIMME algorithm with added metabolomic data in addition to gene-expression data [15] . However , this method only considers the presence/absence of metabolites to refine the model , therefore preventing a full utilization of the quantitative metabolomic data . A time-resolved expression and metabolite-based prediction of flux values , named TERM-FLUX , integrates time-series expression and metabolomic data , and predicts flux distribution for a given time point t . [16] . However , the application of TERM-FLUX is limited to studies with time-series data , which are not widely available . More recently , a method for the integration of relative metabolite levels for flux prediction , iReMet-flux , has been introduced to predict differential fluxes at the genome-scale [17] , and it requires an assessment of the differential changes of all existing metabolites in a GEM . This limits its application , as metabolomic data are mostly measured not at a genome-wide level but rather for only a few metabolites in a system . For multi-omic data , methods have recently been introduced for integrating different layers of data , such as genomic , transcriptomic , proteomic , and fluxomic , into metabolic models [18] or multi-scale models [19] . However , a method that couples the thermodynamic constraints into GEMs with relative transcriptomic and metabolomic data is not yet available . To address this deficiency , we herein propose a novel method , termed Relative Expression and Metabolite Integration ( REMI ) , to integrate relative expression and relative metabolite abundance data into thermodynamically curated GEMs . REMI is the first method that integrates thermodynamics together with relative gene-expression and metabolomic data as constraints for FBA . We demonstrate that REMI’s ability to integrate different layers of constrictive data significantly reduces the solution space of feasible fluxes . REMI also extensively enumerates alternative optimal and sub-optimal solutions , bringing a robustness and flexibility to the flux distribution analysis . We applied REMI to a GEM of E . coli to estimate the central carbon metabolism flux measurements that were determined by 13C metabolic flux analysis ( 13C-MFA ) and were provided by two independent experimental studies [20 , 21] . Although there is limited number of such fluxomic datasets for validation and the measurements are not available at a genome-wide level , our results suggest that the integration of gene-expression , metabolite abundance , and thermodynamic data within REMI’s optimization framework allows for improved flux predictions . Comparing REMI’s predictions with a similar method ( GX-FBA [12] ) , we also show that REMI has on average a 32% higher Pearson correlation coefficient ( r = 0 . 79 ) indicating a more precise exploration of organismal metabolism under wide-ranging conditions .
The underlying assumption of the REMI method is that the perturbation of gene-expression and metabolite levels influences the flux levels in the metabolic network . To this end , REMI maximizes the consistency between relative experimentally observed changes in gene expression and metabolite abundance with the flux levels ( the objective function of the REMI constraint-based method ) . The maximum consistency is then calculated as an integer number , called the maximum consistency score ( MCS ) . This represents the maximum number of constraints that can be incorporated into a FBA model from a given set of constraints ( gene-expression or metabolite abundance levels ) while ensuring that the model still achieves the required metabolic functionalities and remains feasible . MCS is a unique number , however , in that the complex nature and interconnectivity of metabolic networks can result in several alternative solutions for a given MCS , meaning that numerous combination of different constraints from the input data could result in the same MCS . The theoretical maximum consistency score ( TMCS ) indicates the number of genes ( or metabolites or both ) with available experimental data that can potentially be integrated into the model , and MCS indicates the number of these available constraints that could be consistently integrated into the model . We first applied REMI to the integration of eight datasets from Ishii et al . [20] , which included genome-wide transcriptomics together with some metabolomic data obtained for one reference condition and seven different conditions or mutations , into an E . coli model . After integrating the gene-expression data of each condition into the model and comparing it with the reference model , we computed TMCSs , MCSs , and the number of alternative solutions for the REMI-Gex method ( without thermodynamic constraints ) and the REMI-TGex method ( with thermodynamic constraints ) . In contrast to other methods , REMI finds all possible alternative solutions of a given maximum consistency score , which involves all possible combinations of the given set of constraints that always result in a feasible model . These alternative solutions provide flexibility in the biological interpretation of the results as they are equally consistent with the provided experimental data ( applied as constraints to the model ) . Note that in the GEM analysis , the alternate flux solutions are conventionally considered as equivalent phenotypic states [22] . In this study , however , alternative solutions represent the equivalent states of the maximum consistency between gene-expression ( or metabolite abundance or both ) data and the flux levels . Therefore , each feasible alternative solution provides an opportunity to analyze and interpret the given phenotypic state based on the condition-specific omics data , from a different standpoint . We further integrated the available metabolomics measurements into the E . coli model using REMI-TGexM and obtained the MCS for the integrated metabolites as well as the global maximum consistency score ( GMCS ) , which encompasses both genes and metabolites ( Table 1 ) . Although different pairs of conditions showed a very close TMCS variability across the seven case studies based on gene-expression data integration ( mean = 103 . 6 , standard deviation [sd] = 0 . 5 ) and on metabolic data ( mean = 4 . 7 , sd = 0 . 5 ) , the MCS significantly varied across the four REMI methods: REMI-Gex ( mean = 58 . 7 , sd = 4 . 1 ) , REMI-TGex ( mean = 49 . 7 , sd = 3 . 5 ) , REMI-GexM ( mean = 63 . 3 , sd = 4 ) , and REMI-TGexM ( mean = 54 . 3 , sd = 3 . 3 ) ( Table 1 ) . As Table 1 shows , the gene-expression and metabolite abundance constraints for the deregulated metabolites were consistent across the conditions . Therefore , the REMI-TM and REMI-GexM consistency scores sum up to the REMI-TGexM consistency score . This means that there is no conflict between the gene-expression and metabolite abundance data and that they can be co-integrated without confronting each other . We found in our results that MCS was noticeably lower than TMCS , suggesting that the assumption that the relative changes in gene expression correlate with those in fluxes does not always hold . This is probably because the relative changes in gene expression depend on the mechanism of post-transcriptional and translational processes which are currently not captured in metabolic models . However , the number of enumerated alternative solutions highly differs across the conditions in all four methods: REMI-Gex ( mean = 80 . 6 , sd = 80 . 3 ) , REMI-TGex ( mean = 104 . 6 , sd = 168 . 5 ) , REMI-GexM ( mean = 156 . 1 , sd = 241 . 8 ) , and REMI-TGexM ( mean = 25 . 1 , sd = 26 . 4 ) ( Table 1 ) , which suggests that the numbers of alternative solutions are condition-specific , as expected . As shown in the Table 1 , wherever the sd is very high , for example sd = 241 . 8 in REMI-GexM for the rpe vs Ref case , we observe a high number of alternative solutions ( n = 735 in this case ) . Different conditions ( mutations ) alter the cell metabolism differently , leading to different levels of metabolic adaptations and metabolic flux rerouting . Hence , we speculate that the differences in flux rerouting across conditions results in differences in the numbers of alternative solutions across the seven relative conditions . Note that for REMI-TM and consequently for REMI-M , the constraints for all the deregulated metabolites were consistently integrated into the model , so we found only one solution for the REMI-TM models without any alternative solution . For the metabolomic integration , the GMCS was higher in the REMI-TGexM models compared to REMI-TGex because in REMI-TGexM , the GMCS was computed based on both relative metabolite ( Table 1; Metabolites ) and relative gene-expression levels ( Table 1; Genes ) , whereas the MCS for the REMI-TGex model was computed based on only relative expression levels . We further investigated the consistency between gene-expression and metabolomic data and whether the data contradicted each other in certain scenarios . All the available experimental metabolomic data ( Table 1; TMCS and MCS ) were integrated using the REMI-TGexM method for the pgm vs Ref , gapC vs Ref , zwf vs Ref , wt5 vs Ref , and wt7 vs Ref comparisons . We observed that the number of alternative solutions for these five cases was identical between REMI-TGexM and REMI-TGex . This implies that the relative expression constraints and the relative metabolite constraints were not contradictory for these five cases . However , in rpe vs Ref and pgi vs Ref , all the metabolic data were integrated in the model , but the number of alternative solutions differed ( and in the case of rpe vs Ref was noticeably reduced ) between REMI-TGexM and REMI-TGex . To see if this indicated a contradiction , further investigation into the alternative solutions revealed that in the rpe vs Ref and wt5 vs Ref comparisons , REMI-TGexM and REMI-TGex have the same set of constraints , which means that the constraints from metabolomics and expression data were not contradictory . However , we found that the metabolomics integration resulted in a reduction in the number of alternative solutions ( Table 1 ) . We hypothesized that further integration of metabolomics ( on the top of the gene-expression constraints ) imposed a flux rerouting in the metabolic network . As REMI allows enumerating all the possible alternative solutions for a given consistency score , we further interrogated the alternative solutions by High-frequency constraint ( HFC ) analysis . The results of this analysis indicate the core constraints that consistently operate in all the alternative solutions ( the constitutive part of all solutions ) . Meaning that such core constraints certainly perturb fluxes within each pair of conditions . Therefore , these constraints could potentially be the indicators of the regulators of the condition-specific metabolism , which assist biologist in determining which metabolic subsystems to deregulate or to mutate . We believe that the capability to analyze and identify these regulators is a key advantage of REMI . As shown in the Table 1 , the computed HFCs differ across conditions for all four cases: REMI-Gex ( mean = 52 , sd = 5 . 4 ) , REMI-TGex ( mean = 44 . 4 , sd = 4 . 5 ) , REMI-GexM ( mean = 56 . 9 , sd = 5 . 4 ) , and REMI-TGexM ( mean = 49 . 6 , sd = 4 . 2 ) . Constraints that were common amongst all the alternative solutions , indicating key regulators , were the potential candidates for further investigations . After analyzing HFCs across conditions and between the four cases , we found that a reaction catalyzed by glycolate oxidase ( GLYCTO4 ) from the alternate carbon metabolism and another reaction from the murine recycling pathway ( MDDEP4pp ) were always deregulated in the pgm , gapC , zwf , rpe , pgi , and wt7 conditions . These reactions are likely key regulators of mutation in E . coli because they were found to be deregulated in all mutant conditions . To study the effect of thermodynamics on the model , we compared the reduction in solution space for the predicted flux profiles from the REMI-TGex and REMI-Gex methods when coupled with the gene-expression data ( Table 1 ) . The MCS was consistently reduced in the REMI-TGex model compared to REMI-Gex for all pairs of conditions , as REMI-TGex eliminates flux solutions that are not thermodynamically feasible . To better illustrate the positive influence of thermodynamic constraints in reducing the solution space , we show the example of pgm vs Ref as a case study , where we obtained MCS = 56 in REMI-Gex and MCS = 49 in REMI-TGex ( Table 1 ) . First , we enforced the models to satisfy any given consistency score ( 56 and 49 in this example ) by adding a new constraint , which would further allow us to perform conditional FVA . Then , we performed the FVA that satisfies the consistency score ( MCS = 56 ) in REMI-Gex and the consistency score ( MCS = 49 ) in REMI-TGex . Comparing the FVA results of REMI-Gex and REMI-TGex revealed that there exist 45 reactions in REMI-Gex that operate in a thermodynamically infeasible direction and which also contribute to the MCS = 56 . The flux ranges of these reactions are shown in S1 Table and indicate that the TGex method is indeed eliminating the infeasible solutions to enrich for more relevant results . For more clarification , two reactions out of the 45 are shown as examples in S1 Fig . As expected , the flux ranges for these reactions are less flexible for the REMI-Gex ( MCS = 56 ) compared to the REMI-TGex ( MCS = 49 ) , which confirms some extent of the thermodynamic infeasibility in the REMI-Gex predictions as infeasible flux ranges directly indicate the model infeasibility . On the other words , if we integrate thermodynamic constraints to the model and allow the consistency score ( MCS = 56 ) then the model certainly generates infeasible solutions . To investigate whether the higher consistency score caused thermodynamic infeasibility in the REMI-Gex , we performed a FVA of REMI-Gex while forcing lower consistency scores ( MCS = 49 and 10 ) . We found that the flux ranges of reactions became more flexible at lower consistency scores in the REMI-Gex model compared to the REMI-TGex model ( S1 Fig ) , indicating that if both REMI-TGex and REMI-Gex have the same consistency scores , the REMI-Gex cannot allow thermodynamic infeasibility . In contrast , if the consistency score is higher in the REMI-Gex compared to the REMI-TGex , then it leads to thermodynamic infeasibility . The same results were obtained for all other reactions ( S1 Table ) . To further benchmark REMI with the available experimental data , we used a second data set ( 2 overexpression compared to the ref condition ) from an independent study where the role of metabolic cofactors , such as NADH and ATP in different aspect of metabolism is studied by overexpressing NADH oxidase ( NOX ) and the soluble F1-ATPase in E . coli [21] . REMI integrated the gene-expression data from Holm et al . [21] into the E . coli model , and a summary of the results is shown in Table 2 . Like the previous analysis , we observed a reduction in the MCS value within REMI-TGex as compared to REMI-Gex , as REMI-Gex satisfies fluxes that were not thermodynamically feasible . The number of alternative solutions highly differs between NOX overexpression and ATPase overexpression for both REMI-TGex and REMI-Gex , which is likely due to the condition-specific regulations ( NOX vs ATPase overexpression ) that do not necessarily involve the same set of deregulated genes . To investigate the influence of thermodynamic constraints on flux ranges , we identified the overlapping constraints ( HFCs ) across all the alternative solutions and then enforced them to be active to build the most consistent model . An active HFC satisfies differential gene expression ( or metabolite levels ) between two conditions form a given experimental data . Thus , for each condition , we built the most consistent model despite having many alternatives . We next performed FVA on the REMI-Gex and REMI-TGex models . As REMI is based on pair-wise relative constraints ( for two conditions ) and builds two models that are then compared , as opposed to modifying one solution based on a given condition , we obtained two FVA solutions , i . e . one for each condition . We identified less bidirectional reactions ( BDRs ) in the REMI-TGex case compared to the REMI-Gex case ( Table 3 ) , which means that thermodynamic constraints reduce the solution space and consequently the number of BDRs . This is consistent with the fact that thermodynamic constraints eliminate infeasible reaction directionalities . The number of BDR reductions differs across conditions , and we identified the highest BDR reduction for the rpe vs . Ref case and the lowest BDR reduction for the NOX vs . Ref case , which therefore indicates more reduction in the feasible flux solution space in the rpe vs . Ref case compared to the NOX vs . Ref case . For the all comparisons , we found a further reduction in BDRs upon the integration of relative metabolomic data into the REMI-TGex model . In most of the cases , we found a similar decrease in BDRs , which means that the metabolomic data further constrained the solution space . Except for the wt7 vs Ref case , we observed a decrease in BDRs for all cases that were constrained by metabolites and expression data together ( GexM ) as compared to only expression ( Gex ) data . Unexpectedly and unlike all the other cases , by incorporating metabolomics data for the wt7 vs Ref case , we found an increase of one reaction in the BDRs . This suggests that for the wt7 vs Ref case the integration of gene expression and metabolites reroutes fluxes through the metabolic networks differently compared to other cases . As expected , we consistently find a reduction in BDRs for the REMI-TM model ( thermodynamics and relative metabolomics ) compared to without thermodynamics ( the REMI-M model ) . This is in agreement with the fact that integrating thermodynamic constraints into a model eliminates infeasible reaction directionalities and consequently the flux feasible optimal space . To further illustrate the positive influence of thermodynamic constraints in reducing the optimal solution space , we performed a relative flexibility ( Materials and Methods ) analysis using the REMI-TGex and REMI-Gex methods . To perform a relative flexibility analysis , a reference model is compared to a target model to investigate the relative flux reduction . For a reference , we used the iJ01366 model without integrating any data , meaning that the reference model implies only mass balance constraints . We took the pgi vs . Ref case as an example to demonstrate the average relative flexibility ( ARF ) reduction at a global ( e . g . all reactions ) level as well as at the subsystem level . For the pgi vs . Ref case , we found a 10% , 20% , 50% , 77% , and 80% reduction in the global ARF in REMI-M , REMI-Gex , REMI-TGex , REMI-GexM , and REMI-TGexM models compared to the reference model , respectively ( Fig 2A ) . We found 40% and 80% more reduction in the global ARF for the REMI-TGex and REMI-TGexM models compared to REMI-Gex ( Fig 2A ) , which was expected as the REMI-TGex and REMI-TGexM models are more constrained by thermodynamic and metabolomic data compared to REMI-Gex . We further analyzed the ARF at the subsystem/pathway level to investigate the reduction in ARF for each specific subsystem using the REMI-TGex and REMI-Gex methods . Consistently , each subsystem for the REMI-TGexM and REMI-TGex models was more reduced than REMI-Gex ( Fig 2B ) . For REMI-TGex and REMI-TGexM , we observed a remarkable ARF reduction in the glycerophospholipid metabolism , lipopolysaccharide biosynthesis , murein recycling and biosynthesis , and the biomass and maintenance function subsystems . We further performed the same analysis for the pgi vs . Ref , rpe vs . Ref , pgm vs . Ref , wt5 vs . Ref , wt7 vs . Ref , NOX vs . Ref , and ATPase vs . Ref data ( S2 Fig ) . We found a similar reduction in ARF for REMI-TGex and REMI-TGexM compared to REMI-Gex for the cases of gapC vs . Ref and zwf vs . Ref and found a small reduction in pgi vs . Ref and rpe vs . Ref ( S2 Fig ) . We identified a remarkable reduction in ARF ( more than 90% ) across all the comparisons using the REMI-TGexM method for the glycerophospholipid metabolism , murein recycling , and lipopolysaccharide biosynthesis/recycling subsystems ( S2 Table ) . This suggests that these subsystems are more perturbed based on our available gene-expression and metabolite level data , which indicates that they might be key regulator pathways for the studied mutations . To demonstrate the efficacy of the REMI methods in reducing the solution space and therefore predicting flux profiles close to the experimental measurements , we compared the flux predictions of the REMI-Gex , REMI-TGex , and REMI-TGexM methods with those of the alternative , previously used GX-FBA method and compared both methods to the available experimental measured fluxes from 13C experiments . In both GX-FBA and REMI methods , it is assumed that the differential changes in gene expression correlate with flux changes . GX-FBA uses a linear optimization to address the problem , while the REMI method employs mixed-integer linear optimization that enables enumerating alternative states of the maximum consistency . Additionally , unlike GX-FBA method , REMI allows the integration of thermodynamics and relative metabolite abundances . To implement the GX-FBA method , we integrated the relative gene-expression datasets into the iJO1366 model using GX-FBA and computed the flux distributions . For the comparisons , we computed two metrics: 1 ) the Pearson correlation between the predicted and measured intracellular fluxes , and 2 ) the average percentage error ( see Materials and Methods ) between the measured and predicted fluxes . A good prediction requires a noticeable correlation and a small average percentage error . The results of the first set of experimental data [20] ( pgm vs . Ref , rpe vs . Ref , zwf vs . Ref , wt5 vs . Ref , and wt7 vs . Ref ) showed a considerably improved flux prediction for the REMI-Gex , REMI-TGex , REMI-TGexM , and REMI-GexM models as compared to the GX-FBA method , indicated by Pearson correlation and average percentage error ( Fig 3A ) . The GX-FBA and REMI-Gex methods predicted a similar flux correlation for the experimental fluxes for the pgi vs . Ref and gapC vs . Ref cases ( Fig 3A ) . For the second set of experimental data [21] ( Nox vs . Ref and ATPase vs . Ref ) , REMI-TGex predicted better correlation than REMI-Gex and GX-FBA , and the average percentage error of GX-FBA was higher than that of REMI-TGex and REMI-Gex ( Fig 3B ) . On average , across all nine comparisons ( excluding references ) we found that the REMI-Gex method has 32% higher Pearson correlation coefficient compared to the GX-FBA method , which indicates a remarkable improvement in the flux prediction . Since the REMI methods use an additional objective that is the minimization of the sum of fluxes ( see Materials and Methods ) , we modified GX-FBA to imply the minimization of the sum of fluxes as an objective in order to perform an unbiased comparison . This modified GX-FBA prediction agreed less with the experimental results than the REMI predictions ( S3 Fig ) , meaning that REMI outperforms GX-FBA in terms of predictions . REMI also has two advantages over GX-FBA and other relative expression methods in that , first , we do not need to estimate a reference flux distribution a priori , because two flux distributions for two different conditions are obtained in the same optimization framework in REMI ( see Materials and Methods ) , second , REMI enumerates alternative solutions at the MCS , providing a higher confidence when investigating and analyzing results . Generating two separate flux distributions for the two compared conditions allows REMI to be more suitable to study the differential flux analysis between two conditions , and the extensive enumeration of alternative solutions provides robustness and flexibility in the biological interpretations of the provided data . Although all REMI methods were in relative agreement with the experimental fluxomic measurements , we did not observe a significant difference in the predicted results of REMI-Gex , REMI-TGex , and REMI-TGexM . However , as the fluxomic measurements were very limited around the central carbon metabolism , we cannot draw any overarching conclusions about the accuracy of REMI from these results , as this could only indicate that the major fluxomic differences occur in pathways outside of this one specific metabolic pathway . We believe that to investigate the differences in flux predictions across REMI methods , fluxomic and metabolomic measurements will be required on a grander scale , such as the genome level .
Eleven total sets of experimental data that had been previously integrated into the genome-scale model ( GEM ) of E . coli by Kim et al . [23] and were originally obtained from two independent studies done by Ishii et al . ( 8 datasets ) [20] and Holm et al . ( 3 datasets ) [21] were used for the evaluation of the REMI methodology . The three datasets from Holm et al . [20] comprise genome-wide transcriptomic data together with fluxomic data ( 21 measured fluxes ) collected from three experimental conditions: wildtype E . coli , cells overexpressing NADH oxidase ( NOX ) , and cells overexpressing the soluble F1-ATPase ( ATPase ) . The eight datasets from Ishii et al . [20] include genome-wide transcriptomic , fluxomic ( 31 measured fluxes ) , and metabolomic ( 42 metabolites ) data obtained under eight different experimental conditions: wildtype E . coli cells cultured at different growth rates of 0 . 2 , 0 . 6 , and 0 . 7 per hour along with single-gene knockout mutants of the glycolysis and pentose phosphate pathway ( pgm , pgi , gapC , zwf , and rpe ) . All analyses were performed using IJO1366 , the latest GEM of E . coli [24] . The model comprises 2 , 583 reactions , 1 , 805 metabolites , and 1 , 367 genes . The REMI code is implemented in Matlab R2016a , and it is available on GitHub at https://github . com/EP-LCSB/remi . Mixed-integer linear programming ( MILP ) problems were solved using the CPLEX solver on an Intel 12-core desktop computer running Mac . It has been previously shown that thermodynamic constraints not only effectively reduce the solution space of FBA by eliminating the thermodynamically infeasible fluxes from the solution space , but also allow the integration of metabolite concentrations . This provides important links between mass and energy balance and the phenotypic characteristics of the organism . The thermodynamic constraints , as depicted in Eq ( 1 ) , were integrated into the IJO1366 model [3] . The standard Gibbs free energy ΔrGi° without corrections for the pH and ionic strength was estimated using the group contribution method [25] . For each reaction of a GEM , the Gibbs free energy of the reaction ( ΔrGi′ ) was computed , which considers the charge and the activity ( xj ) of each metabolite j given the pH , the metabolite concentration range , and the ionic strength at the cellular compartment where the reaction occurs . We used the gene-protein-reaction ( GPR ) association rules acquired from the E . coli GEM to translate the relative gene-expression levels ( being relatively up- or downregulated ) to the differential and relative flux values of corresponding reactions . GPRs are not mapped as one gene to one reaction , meaning there are many cases in which one gene is mapped to several reactions and multiple genes are mapped to a single reaction , which are depicted with “and” and “or” affiliations , respectively . To capture this , we followed the same procedure for the mapping between GPRs and reactions fluxes introduced by Fang et al [13] . In that study , the authors showed that using a geometric mean of expression ratios where several genes are jointly required for a complex reaction to occur , is the most efficient way to capture the condition that all reaction ratios are required . For the isoenzyme case when any of the several potential genes are sufficient to carry out the reaction , the arithmetic mean of reaction ratios of the genes is suggested , as it captures the minimum condition where any of the reaction ratios is required . However , other types of assumptions were used by other methods for mapping GPRs to fluxes [7 , 12 , 15 , 23 , 26] . For example , minimum expression value is used for complex reactions and maximum or sum of expression values are used for isoenzyme case [23 , 26] . In REMI , if the reaction R is associated with two genes ( g1 “and” g2 ) , the expression level ratios for genes g1 and g2 in the two corresponding conditions are calculated to obtain the geometric mean of the g1 and g2 ratios . Whereas , if the reaction R is associated with two genes ( g1 “or” g2 ) , the arithmetic mean of the obtained expression data ratios is calculated . Thus , from GPR associations , REMI computes the so-called tentative “reaction flux ratios” to further constrain the model . For the metabolomic data , the ratio of metabolite concentration for each metabolite ( if available ) is calculated for any two given conditions . To evaluate whether a reaction or metabolite was up- or downregulated , we sorted the ratios ( calculated as explained in the previous section ) , and selected the top 5% as upregulated and the bottom 5% as downregulated . The fold changes greater than two are considered as significant in many studies . For comparison purposes , we used the two-fold change as the cut-off threshold to identify the significant gene expression and metabolite changes . We found that across all mutants , the set of significant changes identified with our threshold of the top and bottom 5% encloses the corresponding set identified with the two-fold change , meaning that the proposed cutoff criterion is conservative . However , this threshold is a user-defined parameter and one could use different threshold cutoff . For a given metabolic network that includes R reactions and M metabolites , bidirectional reactions are decomposed into forward and backward reactions to allow all fluxes to have positive values . Assuming that S is a stoichiometry matrix , Smr is the stoichiometric coefficient associated with the metabolite m ( m = 1 , … , M ) in reaction r ( r = 1 , … , R ) . Positive and negative stoichiometric coefficients of metabolites signify the substrate or products of a reaction . A binary variable zr was assigned to each reaction r to ensure a positive flux vr ( Eq ( 2 ) ) through the reaction r , and when zr = 0 , there was no flux . An additional constraint was formulated using Eq ( 3 ) to ensure that only one reaction directionality could be active and carry flux . α and β indicate the forward and reverse directions of a reaction . In REMI , two models are described for each given condition . For both models , we constrained the cellular growth rate to be at least 0 . 1 mmol g-1 DW-1 h-1 , to ensure that the model is able to synthesize all the biomass building blocks required for the cellular growth . Throughout the manuscript , the terms “wildtype” and “mutant” are used to better differentiate between the two conditions ( or models ) when describing the REMI framework . REMI can , however , be used for any two given conditions and is not restricted to the wildtype and mutant labels . Eq ( 4 ) specifies the mass balance constraints for the wildtype and mutant conditions at the steady state . The relative information about the gene-expression levels or metabolite levels between the two given experimental conditions was formulated as additional constraints and integrated into the two representative models of the conditions . To do this , binary variables for the up- and downregulated reactions were assigned as u and d , respectively , where n is the total number of up- and downregulated reactions . For the upregulated reactions , a higher flux was enforced in the mutants as compared to the wildtype , while for downregulated reactions , a higher flux was enforced for the wildtype as compared to the mutant . For u upregulated and d downregulated reactions , a total of n binary variables were generated ( B1 , …Bi , …Bn ) , where Bi = 1 indicates the up- or downregulation of a reaction . Next , n constraints ( Eqs ( 6 and 7 ) ) were added to enforce a basal flux in both the wildtype and mutant conditions . For u upregulated reactions , constraints ( Eq ( 8 ) ) were added to ensure a mutant flux could be higher ( p*vrwild ) than a wildtype flux , where p is a reaction ratio between the wildtype and mutant ( computed from gene-expression ratio ) . Constraints were added ( Eq ( 9 ) ) for d downregulated reactions that ensured a mutant flux was lower compared to a wildtype flux . In Eq ( 10 ) , n constraints were added to form the boundary for the slack variables that are used in Eqs ( 8 ) and ( 9 ) , where ε = 10−5 , M′ = 1000 . In GEMs , gene-level perturbations can mediate both reactions and their subsequent metabolites . Available studies show a correlation between gene changes and metabolite changes and infer that perturbations at the metabolite level are formed from perturbations in genes or reaction levels [27 , 28] . Thus , if experimental evidence shows remarkable changes in a given metabolite abundance level across two conditions , the assumption is that there is an imbalance in the incoming or outgoing fluxes around that metabolite . If the experimental data indicates that a metabolite is upregulated , it is assumed in REMI that either the sum of production ϕp in condition 2 is greater than the ϕp in condition 1 or the sum of consumption ϕc in condition 2 is less than the ϕc in condition 1 ( Fig 4B ) . Due to mass balance , ϕp and ϕc will be equal . In Eq ( 11 ) , the sum of production and of consumption of a metabolite i is shown , where the metabolite is produced by reactions 1 and 2 and is consumed by reactions 3 and 4 ( Fig 4A ) . Based on available experimental measurements of metabolite abundance , REMI finds the total number ( n’ ) of up- and downregulated metabolites , where u’ and d’ are up- and downregulated metabolites , respectively . For an up-regulated metabolite i ( i . e . in the mutant vs . wildtype ) , either more production or less consumption is enforced in the mutant compared to the wildtype using Eqs ( 12 ) and ( 13 ) . In Eq ( 12 ) , a binary variable ( Bi ) is introduced , which switches to production if Bi = 1 and to consumption if Bi = 0 . Similarly , for downregulated metabolites i , less production or more consumption is enforced in the mutant compared to the wildtype ( Eqs ( 13 ) and ( 14 ) , see supplementary description for more detail ) . Based on the assumption that alterations in gene-expression or metabolite levels within two different physiological conditions results in differential flux profiles , REMI defines such alterations as constraints and integrates them accordingly into the two metabolic models corresponding to the two conditions . However , as additional constraints reduce the solution space of FBA , particularly in the case of multi-omic data integration , the resulting models might not be feasible . Therefore , the objective function ( Eq ( 15 ) ) was formulated in such a way as to obtain feasible models with a maximum agreement between the relative expression and metabolite levels and their corresponding constraints . Eq ( 15 ) maximizes the agreement with experimental data using mathematical optimization principles subject to Eqs ( 5 ) – ( 10 ) , where n is the total number of up- and downregulated reactions . The maximum consistency score ( MCS ) is the sum of the binary variables ( Eq ( 15 ) ) in the outcome of the optimization that is formulated in REMI . An aforementioned mathematical optimization model ( objective function ( 15 ) subject to Eqs 1–9 ) allows us to maximize the total number of consistent reactions between the differential gene-expression or metabolite levels with the differential flux profiles between two models and to obtain a maximum consistency score ( MCS ) . Depending on the flexibility of the model , many alternative flux distribution profiles for a given MCS , and subsequently MCS-n , are possible . MCS and MCS-n represent optimal and suboptimal consistency , respectively . To enumerate alternative solutions , integer cut constraints ( Eq ( 16 ) [29] were used as follows: ∑i=1nB′iBi≤ ( ∑i=1nB′i ) −1 ( 16 ) The left-hand side of Eq ( 16 ) determines the number of up- and downregulated reactions in the current solution that carries fluxes in the first MCS solution . The right-hand side represents the number of reactions that carry fluxes in MCS-1 . The inequality ensures that the new solution differs at least by one new reaction that carries flux compared to the previous solution . Repeating this procedure allows the enumeration of alternative solutions for each MCS . To concurrently integrate both the relative gene-expression data and the relative metabolite levels , an integrated mathematical optimization model was built with a global objective function ( Eq ( 17 ) subject to a combined set of constraints , i . e . Eqs ( 5 ) – ( 14 ) . This optimization model was then solved to maximize the objective , which is the combined consistency score of the two sets of constraints . GlobalConsistencyscore ( GCS ) =Maximize∑i=1n+n′Bi , ( 17 ) where n and n’ represent a total number of up- and downregulated metabolites and up- and downregulated genes , respectively . To compare the REMI-predicted fluxes with the experimentally measured ones , predicted flux distribution profiles were required . To obtain such predicted flux profiles , all the alternative solutions at MCS were first enumerated . REMI method optimizes consistency and identifies alternative sets of consistency . Then , for each consistency set we build a model by fixing binary variables which enforces constraints are applied in the model . Then , an additional optimization was performed by minimizing the sum of the fluxes for each alternative solution to obtain a representative flux profile for benchmarking REMI against the experimental flux measurements . To effectively compare the predicted in silico fluxes from REMI with the corresponding 13C-determined in vivo intracellular fluxes , the following two metrics were used: the uncentered Pearson correlation coefficient ( Eq 18 ) , and the average percentage error in predicted fluxes ( Eqs ( 18 ) – ( 20 ) ) . The uncentered Pearson correlation is a good metric for the flux comparison , as fluxes are usually not centered , and it has been used for comparing two flux vectors [23] . In Eq ( 18 ) , vi and vm are the in silico and measured vectors of the fluxes , respectively . The correlation coefficients +1 and -1 indicate a strong positive and negative linear relationship between vi and vm , and the 0 correlation coefficient indicates no linear relationship between vi and vm . The average percentage error has been used in the GX-FBA method [12] to compare two fluxes . In Eq ( 19 ) , the dr is used to measure the relative deviation between the two fluxes in two conditions , where x and y correspond to the flux of a given reaction in condition 1 and condition 2 , respectively . Since |dr| lies between 0 and 1 , one can consider dr as a percentage flux change from condition 1 to condition 2 . The average ( per reaction ) percentage error , e , in the predicted in silico fluxes was calculated using Eq ( 20 ) , where diinsilico and diexp indicate relative deviation in predicted in silico flux using methods such as REMI and GX-FBA , and experimentally measured flux and N represent the number of reactions with available experimental flux data . For a given system , the FBA results in a solution space of optimal flux profiles , and the magnitude of this solution space indicates the metabolic flexibility of the system . The integration of the thermodynamic knowledge of reactions as well as condition-specific experimental data , e . g . gene-expression or metabolomic data , constrains the metabolic system to a less flexible one . Thus , the solution space and the subsequent range of the metabolic responses are reduced . Comparing and quantifying the relative flexibility of a metabolic system before and after constraint is a decent indication of the effectiveness of the data integration [30] . Performing a flux variability analysis ( FVA ) outlines the flux variability range of each reaction in the system for the two conditions as follow: FRi1=[vmin , i1 , vmax , i1] ( 21 ) FRi2=[vmin , i2 , vmax , i2] ( 22 ) The relative flexibility ( RF ) for reaction i is calculated using the following equation: RFi=[ ( vmin , i2−vmax , i2 ) / ( vmin , i1−vmax , i1 ) ] ( 23 ) where FRi1 and FRi2 represent the flux variability range of reaction i at each of the two conditions , one condition is usually designated as a reference condition or reference state , such as when comparing the relative flexibility of a metabolic system with ( condition 1 ) and without ( condition2 ) thermodynamic constraints . The value of RF that is computed for each reaction i reflects the relative changes in the flux variability range of one condition compared to the other condition . The global relative flexibility change between two given condition is then computed by averaging the Fi values for each reaction i that carry flux in the reference state .
We developed the computational tool , REMI , which combines gene-expression , metabolomics , and thermodynamics constraints with the mass balance constraints imposed in metabolic models to predict phenotypic changes in an organism upon environmental or genetic perturbations . As the integration of these three additional physiological constraint results in a highly reduced flexibility of the predicted optimal flux profiles , REMI enhances the quality of the computationally predicted fluxes . REMI’s novel formulation permits the extensive enumeration of alternative solutions because there exist several alternative sets of pathway that result in the same phenotype due to the complexity and interconnectivity of metabolic networks , meaning that the results provided by REMI more accurately reflect natural biological states than previously existing methods . Within several examples , we showed the effectiveness of incorporating thermodynamic data with gene-expression and metabolomics in reducing the flexibility of predicted optimal flux profiles . This means that we can obtain manageable set of physiological consistent hypothesis and physiological interpretations which have a higher confidence as they are consistent with a larger set of data . Applying REMI to experimental data has shown that there is not always a full consistency between gene-expression and metabolomic data , which shows that there is still much to learn about how gene expression and metabolism are linked . The application of REMI goes beyond the study of physiology of a mutant versus a wild-type cell presented in this work . With a slight modification in the formulation , REMI can be employed for investigating the physiology of several mutants simultaneously against the wild type physiology within a single optimization . Although in this study we showcased REMI for constraining internal fluxes , REMI can also be applied to study the perturbation of external fluxes and metabolites whenever omics data are available . This has a potential application in studying the overflow metabolism , e . g . , the acetate overflow in fast-growing E . coli or the Warburg effect in the cancer cells . Furthermore , REMI can also be used to investigate metabolism of diseased states compared to the healthy one , where numerous sets of omics data are available . Various REMI methods introduced in this work permit a wide range of applications depending on the type of available data ( thermodynamics , single or multi omics data ) . However , whenever gene-expression , metabolite abundance , and thermodynamic data are available , our results suggest that the most extensive data integration method , REMI-TGexM , provides the best results and the most reduced optimal solution space . As systematic multi-omics integration remains a challenge , REMI opens the possibility of not only multi-omics integration , but also the identification of the crosstalk between the various omics present in a system .
|
The recent advances in omics technologies have provided us with an unprecedented abundance of data spanning genomes , global gene expression , and metabolomes . Though these advancements in high-throughput data collection offer an excellent opportunity for a more thorough understanding of metabolic capacities of a wide range of species , they have caused a considerable gap between “data generation” and “data integration . ” . In this study , we present a new method named REMI ( Relative Expression and Metabolomic Integrations ) that enables the co-integration of gene expression , metabolomics and thermodynamics data as constraints into genome-scale models . This not only allows the better understanding of how different phenotypes originate from a given genotype but also aid to understanding the interactions between different types of omics data .
|
[
"Abstract",
"Introduction",
"Results",
"Materials",
"and",
"methods",
"Discussion"
] |
[
"cell",
"physiology",
"enzymes",
"metabolic",
"networks",
"enzymology",
"cell",
"metabolism",
"phosphatases",
"metabolomics",
"metabolites",
"network",
"analysis",
"thermodynamics",
"computer",
"and",
"information",
"sciences",
"proteins",
"metabolic",
"pathways",
"gene",
"expression",
"adenosine",
"triphosphatase",
"physics",
"biochemistry",
"cell",
"biology",
"genetics",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"metabolism"
] |
2019
|
Enhanced flux prediction by integrating relative expression and relative metabolite abundance into thermodynamically consistent metabolic models
|
To uncover the processes and mechanisms of cellular physiology , it first necessary to gain an understanding of the underlying metabolic dynamics . Recent studies using a constraint-based approach succeeded in predicting the steady states of cellular metabolic systems by utilizing conserved quantities in the metabolic networks such as carriers such as ATP/ADP as an energy carrier or NADH/NAD+ as a hydrogen carrier . Although such conservation quantities restrict not only the steady state but also the dynamics themselves , the latter aspect has not yet been completely understood . Here , to study the dynamics of metabolic systems , we propose adopting a carrier cycling cascade ( CCC ) , which includes the dynamics of both substrates and carriers , a commonly observed motif in metabolic systems such as the glycolytic and fermentation pathways . We demonstrate that the conservation laws lead to the jamming of the flux and feedback . The CCC can show slow relaxation , with a longer timescale than that of elementary reactions , and is accompanied by both robustness against small environmental fluctuations and responsiveness against large environmental changes . Moreover , the CCC demonstrates robustness against internal fluctuations due to the feedback based on the moiety conservation . We identified the key parameters underlying the robustness of this model against external and internal fluctuations and estimated it in several metabolic systems .
In recent decades , mathematical modeling of metabolic systems has been intensively explored in the field of systems biology [1–6] . Several studies have demonstrated that metabolic systems can be quantitatively predicted using mathematical models [7–10] . A particularly important feature for the mathematical modeling of a metabolic system is that some conserved quantities characterize the system . The total number of atoms remains unchanged before and after the reactions , allowing for mass conservation . Furthermore , several coenzymes act as the carriers of molecules and energy through various reactions , e . g . , ATP/ADP as an energy carrier and NADH/NAD+ as a hydrogen carrier [11 , 12] . The total concentrations of the carriers represent the conserved quantities in the steady state , which is referred to as moiety conservation [11] , by which various coenzyme-related reactions have to be balanced . These constraints , i . e . , the law of mass conservation and moiety conservation , restrict the solution space and facilitate the analysis of complicated metabolic networks . One of the most successful approaches used for the modeling of metabolic systems to date is the constraint-based analysis of metabolic fluxes based on the stoichiometry that describes conservation laws in metabolic fluxes [2 , 5] . In this approach , the metabolic fluxes are assumed to be in a steady state , meaning that , for any metabolite pool , the fluxes governing its synthesis and degradation are balanced . Owing to its predictive power and advantage of not requiring detailed information on the kinetic parameters , a steady state-based metabolic modeling approach has been applied in various studies analyzing the characteristics of metabolic systems . However , in contrast to the steady-state solution , the characteristics of the dynamic behavior of metabolic systems have not been elucidated . In a fluctuating environment , a metabolic system is not in a steady state , and thus its dynamic behavior should be analyzed over time . In a biological system , the producing and decomposition reactions of such carriers are also faster than cellular growth . Since the conservation laws are considered to be maintained for a longer time than several cell cycles , the law of mass conservation and moiety conservation restrict both the steady state and the transient dynamics . Thus , the constraints of metabolic networks should be taken into account when studying metabolic dynamics . In particular , the moiety conserved coenzymes should have particularly strong effects on the various pathways in analyzed metabolic systems due to their recycling [11] . For example , in the glycolytic pathway , ATP is transformed into ADP by phosphofructokinase , while ADP is recycled to ATP in downstream reactions . During the anaerobic growth of microorganisms , NAD+ is transformed into NADH by glyceraldehyde-3-phosphate dehydrogenase in upper glycolysis , while NADH is recycled to NAD+ through the fermentation pathways [1] . These reactions involving carriers must be balanced not only in the steady state but also in the dynamic states , and yet the effect of such balance remains unknown . Several previous studies have analyzed the characteristics of metabolic pathways with moiety conservation . Reich and Sel’kov [11] derived simple dynamical system models of metabolic system , and pointed out that recycling of the moiety conservations represents the skeleton of energy metabolism . Such a moiety-conserved cycle was formulated using mass-action kinetics and was analyzed with dynamical systems theory . They found that the cycle of moiety conservation works as a positive feedback to produce the high-energy carrier autocatalytically if its concentration is low , whereas a higher concentration of the high-energy carrier limits its own production . Following these pioneering works , the behavior of the moiety-conserved metabolic cycle has been extensively studied [13 , 14] . Although the majority of these studies focused only on the analysis of steady states , some researchers addressed the dynamic behavior of the moiety-conserved cycle after environmental changes [15 , 16] . These studies demonstrated transient switching behavior based on a simple model with mass-action kinetics without considering complex formation among enzymes , substrates , and cofactors . Recently , some theoretical studies demonstrated that complex formation in a chemical reaction system can cause transient abnormal behavior , which could not be obtained using simple mass-action kinetics [17] . This finding suggested that the complex dynamic behavior of metabolic systems , especially those including moiety conservation , can also be captured by considering complex formation . Following this idea , in the present study , we focused on metabolic reaction pathways with carrier recycling ( also known as a “turbo design” [18] ) . To investigate the dynamic behavior of nonlinear metabolic systems with the law of mass conservation and moiety conservation , we here explore a simple motif designated as a carrier cycling cascade ( CCC ) , which simultaneously considers the dynamics of substrates and coenzymes . We formulate the CCC with complex formation to analyze the dynamics of both substrates and carriers simultaneously . The CCC shows slow relaxation into the steady state against external perturbations around a critical point where the relaxation speed follows the power law . We analyze the origin of the slow relaxation using dynamical systems theory . Additionally , the CCC maintains robustness of a metabolic state with intrinsic noise . We demonstrate that the negative feedback through the conserved carrier provides such robustness , i . e . , the moiety-conserved cycle can reduce fluctuations in the concentrations of metabolites . We further discuss the relationship between microscopic feedback and macroscopic dynamics , and estimate the effect of this feedback using experimentally determined parameters .
The steady-state characteristics in a moiety-conserved cycle were intensively studied by Reich and Sel’kov [11] . However , most of these proposed models are too complicated to effectively extract the essence of the dynamic features . The metabolite flow is usually branched , and the same coenzyme can be utilized at multiple steps . Hence , to investigate the effects of a cycling carrier on the dynamics of metabolic systems , we here focus on the simplest cascade , the CCC , which models ATP-ADP cycling in the glycolysis pathway and NADH-NAD+ cycling in the fermentation pathway ( Fig 1 ) . Here , we refer to the high-energy and low-energy carriers as the active and inactive carriers , respectively . The CCC consists only of active carrier-consuming and -producing steps . Our model is similar to the simplest skeleton proposed by Reich and Sel’kov [11] . In contrast to the previous study , which considered the mass-action dynamics without complex formation , we here include the complex-formation process . We assume that the enzyme , substrate , and coenzyme form the complex , while the enzyme is saturated under the ordinal metabolic condition [19 , 20] and does not appear in the equation explicitly . Original ordinary differential equations ( ODEs ) are given as seven mass-action kinetic equations ( see section 1 in S1 Text ) , which can be reduced to five ODEs by eliminating the association and dissociation reactions between coenzymes and substrates adiabatically . d [ m 0 ] d t = k in - k c [ m 0 ] [ c ] K 0 + [ c ] - k leak [ m 0 ] , ( 1a ) d [ m 1 ] d t = k c [ m 0 ] [ c ] K 0 + [ c ] - k p [ m 1 ] [ c * ] K 1 + [ c * ] , ( 1b ) d [ m 2 ] d t = k p [ m 1 ] [ c * ] K 1 + [ c * ] - k out [ m 2 ] , ( 1c ) d [ c ] t d t = - k c [ m 0 ] [ c ] K 0 + [ c ] + k p [ m 1 ] [ c * ] K 1 + [ c * ] , ( 1d ) d [ c * ] t d t = k c [ m 0 ] [ c ] K 0 + [ c ] - k p [ m 1 ] [ c * ] K 1 + [ c * ] , ( 1e ) [ c ] t = [ c ] + [ m 0 ] [ c ] K 0 + [ c ] , ( 1f ) [ c * ] t = [ c * ] + [ m 1 ] [ c * ] K 1 + [ c * ] , ( 1g ) where mi is the i-th metabolite , c and c* are active and inactive carriers , respectively , and [x] denotes the concentration of x . m0 is supplied and diluted with rates kin and kleak , respectively , and m2 is diluted with rate kout . c and m0 make complex cm0 , and c* and m1 make complex c* m1 . The active carrier is consumed with rate kc when m1 is transformed from m0 , and is produced with rate kp when m2 is transformed from m1 . [c]t and [c*]t represent the total concentration of active and inactive carriers as [c]t = [c] + [cm0] and [c*]t = [c*] + [c*m1] , respectively . K0 and K1 represent the dissociation constants between c and m0 and between c* and m1 , respectively . Here , we considered the condition in which the total concentrations of the carriers are conserved . There are multiple reactions between the active carrier-producing and -consuming reactions in the actual metabolic networks , while these reactions are reversible and not rate-limiting under ordinal metabolic conditions [16 , 18 , 20] . Then , multi-step reactions can be reduced to the present form . The conserved quantities in the CCC model are as follows: Note that the number of independent conserved quantities is two because of c pool = c sum + c diff * . In metabolic networks , some types of inputs will be in a CCC , e . g . , changes in the influx and efflux rates of the first metabolite , changes in the ratio between active and inactive carriers due to the environmental changes , and the new synthesis and degradation of a carrier . Here , we consider the condition in which the total carrier concentration is conserved so that the cpool does not change . Therefore , we consider changes in kin and the ratio between active and inactive carriers as inputs . Initially , we changed the influx rate . We allowed the system to approach a steady state , and then changed kin at time = 0 ( Fig 2A ) . When kin is altered , the concentrations of all molecules do not change in the timescales of the active carrier consuming and producing reactions ( kc and kp are set as 1 . 0 ) . Accordingly , the changing of [m0] is slower than the timescale of kc and kp , while the concentrations of other components do not change as in the quasi-steady state . When [m0] decreases to become sufficiently small , the concentrations of the others drastically change with a fast timescale , i . e . , the timescale of relaxation in [m0] is approximately one thousand times that of kc and kp . Consequently , the system reaches the true steady state , i . e . , the other components relax into the steady-state values in the timescale of kc and kp . When we change the ratio between [c] and [c*] , which is considered the change in csum , a similar type of behavior is observed ( S1 Fig ) . Although the concentrations of molecules without m0 change when the ratio changes , the system can reach the quasi-steady state quickly . Then , [m0] changes slowly and the other components subsequently change rapidly . The timescale of the relaxation depends on the input strength . We defined the relaxation time τ as the time when the sum of differences of the concentrations of all molecules from the previous time point falls below the threshold ( 10−7 ) . When we change kin , τ diverges at the point where k in = k in th ( Fig 2B , inset ) . The relaxation time is proportional to the inverse of the difference between kin and k in th ( Fig 2B ) so that the relaxation time critically slows down around the critical point . Note that although the relaxation time is also critically prolonged in the vicinity of saddle-node bifurcation , the exponent is 1/2 [21] and differs from the obtained result . When the relaxation time slows down , the relaxation manner of [m0] becomes linear with time rather than exponential ( S2 Fig ) . This can be observed in both cases where kin and csum change in the system ( Fig 2A and S1 Fig ) . If both the active carrier-producing and -consuming reactions are reversible , similar slow dynamics appears , while the switch from the quasi-steady state to the true steady state is smooth for small kout values ( see S3 Fig and section 2 in S1 Text ) . This may be due to inhibition of the active carrier-producing reaction by binding of m2 to c* . Therefore , the slow relaxation is considered to be a general behavior of the CCC in response to different environmental changes . To elucidate the process of the slow relaxation , we reduced the number of arguments in our model by using conserved quantities . Although our model had five ODEs even when the association and dissociation reactions are eliminated adiabatically , [m2] is involved only in the equation for [m2] and not in other equations . Moreover , two equations can be eliminated because of two independent conserved quantities . Hence , the model can be reduced to the following two ODEs: d [ m 0 ] d t = k in - k c [ m 0 ] [ c ] K 0 + [ c ] - k leak [ m 0 ] , ( 2a ) d [ m 1 ] d t = k c [ m 0 ] [ c ] K 0 + [ c ] - k p [ m 1 ] [ c * ] K 1 + [ c * ] , ( 2b ) [ c ] = - β + { β 2 + 4 K 0 ( c sum - [ m 1 ] ) } 1 / 2 2 , ( 2c ) [ c * ] = - γ + { γ 2 + 4 K 1 ( c pool - c sum + [ m 1 ] ) } 1 / 2 2 , ( 2d ) where β = [m0] + K0 − csum + [m1] and γ = K1 + csum − cpool . The nullclines for [m0] and [m1] are presented in Fig 3 , in the case of kleak = 0 . By altering kin , the vertical position of the nullcline for [m0] is changed as well . The CCC behavior changes drastically in the vicinity of the critical point . When kin is smaller than k in th , a stable fixed point appears ( Fig 3A ) , and when kin is close to k in th , relaxation to the fixed point is critically slowed down due to the approach of two nullclines . In this case , a slow manifold is located on the nullcline for [m1] , which is nearly parallel to the nullcline for [m0] . In the region where two nullclines are parallel , the relaxation speed across the slow manifold is proportional to the distance from the nullcline for [m0]; i . e . , the speed of change in [m0] is given as a constant that is proportional to k in th - k in . Hence , the degradation of m0 is the rate-limiting process when the distance between the nullclines is sufficiently small , and the relaxation time is proportional to the inverse of k in th - k in ( Fig 2B ) , while the relaxation manner of m0 depends linearly on time ( S2 Fig ) . However , when kin is larger than k in th , nullclines do not cross anywhere and [m0] diverges ( Fig 3C ) . In particular , if kin is equal to k in th , the points on the overlapped nullclines become neutrally stable fixed points ( Fig 3B ) . When kin is fixed and csum is changed , both of the nullclines move . Moreover , in this case , the nullclines approach each other for the critical csum value ( see S4 Fig ) , and the slow relaxation occurs . Here , k in th can be obtained analytically for kleak = 0 ( see section 3 in S1 Text ) . k in th = k c k p 2 ( k c + k p ) { c pool + c sum + α + ( c pool - c sum + α ) 2 + 4 c sum α } . ( 3 ) where α = kc K1/ ( kc + kp ) . For the limit of K1 → 0 , i . e . , m1 can perfectly bind to c* and never dissociate , Eq ( 3 ) becomes: k in th = { k c k p c sum ( k c + k p ) ( c pool > c sum ) , k c k p c pool ( k c + k p ) ( c pool < c sum ) . ( 4 ) Eq ( 4 ) represents the maximal capacity of the flux of CCC . Therefore , when the influx exceeds the capacity , the CCC can become jammed , which leads to the appearance of slow dynamics . If kleak is not zero , the nullcline for [m0] is tilted , and a fixed point is obtained for large kin ( S5 Fig ) . For a system with a finite kleak value , k in th cannot be defined . However , in the parameter range where kin is close to the k in th for the system with no kleak , the fixed point value changes drastically and slow relaxation appears , which cannot be represented by bifurcation . For organisms , robustness against small fluctuations in nutrient uptake is important for maintenance of the intracellular environment . However , in the case of a sudden decrease in the concentration of nutrients , which may lead to starvation , stress-resistant systems should be activated [22 , 23] . Here , we analyzed the frequency responses of the CCC using a cyclic nutrient uptake rate kin ( t ) with different amplitudes ( Fig 4 ) . The CCC demonstrates the low-pass filter characteristics with a sharp cut-off frequency for both weak and strong inputs ( Fig 4A ) . For the case when kin ( t ) is higher than k in th , m0 accumulates and the concentrations of the others does not change over time . For the case when kin ( t ) is lower than k in th , m0 decreases slowly , while the concentrations of the other components remains the same . Here , m0 concentration represents a buffer , and the sharp cut-off frequency can be achieved . The timescale of the decrease depends on the distance between kin and k in th , and thus the filter characteristics depend on the amplitude of inputs ( Fig 4B ) . For inputs with a smaller amplitude , the cut-off frequency becomes lower , but when a larger amplitude of inputs is used , the CCC demonstrates the response against the higher frequency of inputs . For larger concentrations of m0 , the rates of enzymatic reactions , i . e . , the active coenzyme-consuming and -producing reactions , are considered to be constant due to saturation . Therefore , the nullcline for [m1] can be considered to be nearly constant , while that for [m0] can be regarded as a linear equation of [m0] due to a leak term ( S5 Fig ) . When two nullclines are close , the flow rate along the nullclines can be approximated as the distance between the two nullclines , which is slower than the flow rate approaching the nullclines . Accordingly , the two-dimensional dynamics can be reduced into one-dimensional dynamics along the nullcline for [m1] . We analytically obtained the frequency response of the dynamics . When kleak is not as large , the two-dimensional dynamics ( Eqs ( 2a ) and ( 2b ) ) can be reduced into one-dimensional dynamics of [m0] ( see section 4 in S1 Text ) . d [ m 0 ] d t = - k leak k c [ m 0 ] - k c c sum - k in ( t ) k c + k c c sum ( k c + k p ) . ( 5 ) For kin ( t ) as a sinusoidal k in ( t ) = A in cos ( 2 π f t ) + k in 0 , when [m0] is saturated , the cut-off frequency is given as: 2 π f = k leak k c A in k c k in 0 - ( k c + k leak ) k in th - k leak k c . ( 6 ) The estimated cut-off frequency fits well with the simulation result ( red dashed line in Fig 4B ) , suggesting that the complex dynamics can be reduced to one-dimensional dynamics due to the conserved quantities , and that the necessity for robustness against external fluctuation is determined by the conditions underlying this saturation . Note that Eq ( 5 ) is independent of a form of function of the influx rate . Hence , if the influx rate is regulated by downstream products , the dynamics of the CCC are reduced into one-dimensional dynamics . Although the filter characteristics ( Eq ( 6 ) ) depend on the regulation , robustness by the slow dynamics will always be achieved . To investigate the robustness of the CCC against the intrinsic noise caused by the stochasticity of biochemical reactions , we calculated the stochastic dynamics of the original full model with consideration of complex formation using the Gillespie algorithm [24] , which can generate a statistically possible trajectory of the solution of stochastic equations and is often used for modeling biochemical reactions . We calculated the long trajectories of the solution and several statistics from the given trajectories . When the cpool is larger than the critical value , the average number of m1 is almost constant for different cpool values . In this condition , the Fano factor , which is the ratio of the variance to the average , is approximately 1 , as in similar non-catalytic reactions . However , below the critical point where the flux is limited by the concentration of the carrier , i . e . , the first metabolite is saturated , the Fano factor of the number of m1 decreases below 1 ( Fig 5A ) . This suggests that the carrier cycling can reduce the variance of the concentration of intermediate metabolites . To investigate the mechanisms underlying the decrease in the intrinsic noise , we analyzed the probabilistic dynamics of the number of m1 molecules , n ( see section 5 in S1 Text ) . Under the not-saturated condition , i . e . , when the m0 concentration is lower than the maximal coenzyme concentration cmax , which is the same as csum when csum < cpool , only the consumption rate of m0 is proportional to n but the production rate is not . Therefore , the steady-state distribution of the number of m1 is given as a Poisson distribution [25] in the limit of K0 → 0 and K1 → 0; i . e . , the metabolites can bind to coenzymes perfectly . Thus , the Fano factor becomes 1 , which is similar to the previously reported condition [26] . In the saturated condition , i . e . , the m0 concentration is higher than cmax , the production rate becomes kc ( cmax − n ) due to conservation of the coenzyme , while the consumption rate remains the same as above . Here , both the production and consumption rates are proportional to n , which is considered the feedback-regulated production through the conserved quantity . The steady-state distribution represents the binomial distribution [25] and the Fano factor is given as: σ 2 < n > = 1 - k c k c + k p . ( 7 ) Therefore , the fluctuation is reduced depending on the kc and kp values ( Fig 5B ) , and the recycling of carriers improves the signal-to-noise ratio by feedback regulation via the moiety conservation . This effect does not depend on the concentration of the coenzyme as long as the metabolite is saturated before the coenzyme consumption step . The feedback can also reduce the fluctuation in the active and inactive carrier concentrations . Note that when both the production and consumption rates of m1 obey the Michaelis-Menten reaction with different coenzymes , the number of m0 shows a random walk in the range of 0 to ∞ , or becomes zero , or diverges , depending on the parameter settings . In any case , the Fano factor never falls below 1 ( see section 5 in S1 Text ) .
Here , we proposed a minimal model of metabolic systems that includes recycling of the moiety-conserved carriers with complex formation between carriers and other metabolites . In transient dynamics , the conserved quantities constrain a dimension of orbits moving on to the steady state , similar to the steady-state solution space , and dynamics in the restricted dimension demonstrate various phenomena . These effects can be summarized by two properties: 1 ) jamming and 2 ) feedback , which are followed by the slow relaxation and robustness against internal and external fluctuations . We demonstrated that the relaxation dynamics in the CCC are decelerated by the jamming when the nutrient uptake rate is close to the capacity of the cascade . Such slow relaxation to the steady state has also been discussed in other enzymatic networks [17] . From the viewpoint of dynamical systems theory , this jamming is due to a restriction of the phase space and the closure of nullclines by the conservation of carrier concentration . In this situation , the concentrations of internal metabolites , the final product , and active and inactive carriers are almost constant and can be drastically different from the concentrations in the steady state . Hence , if a metabolic system seems to be in a steady state on a short timescale after an environmental change , this system will not always be in the true steady state but rather in a quasi-stable state . Consequently , the previous theoretical studies considering steady metabolic states were not able to analyze the cellular metabolism in these quasi-stable states . Such jamming in the metabolic flow is likely to be observed in several metabolic systems , including the glycolytic cascades . In fact , some mutants of the budding yeast exhibit growth defects , which were explained due to the abnormal accumulation of intermediate metabolites based on computer simulations [18] . Such time evolution is consistent with our results caused by the jamming , while we further uncovered the origin of this phenomenon using dynamical systems theory . This study suggests the possibility that the native metabolic cascade resides in the vicinity of the critical point and can be easily jammed by genetic perturbations , which will be validated in future studies . Furthermore , we suggest that the jamming mechanism may represent the mediator between molecular- and organism-level timescales [27]; i . e . , the jamming slows down the fast enzymatic turnover and may determine the timescale of physiological behaviors . The slow relaxation can help organisms maintain their cellular condition against changes in the nutrient condition and will work as a memory of past environmental change . For example , the slow timescale might be related to a slow response of the relaxed strain against carbon and amino acid starvation [22 , 23] . We have demonstrated the robustness of metabolic systems against external fluctuation , which opens the door for further theoretical studies of the quasi-steady state of metabolic dynamics . The concentrations of some metabolites and carriers are maintained throughout small fluctuations in the influx rates . At the same time , the CCC is responsible for the large alterations in the influx rate . Therefore , the CCC shows both robustness against small fluctuations and responsiveness to large fluctuations; i . e . , the CCC does not respond to small and short-term changes in the nutrient uptake rate , but it can rapidly respond to larger alterations in nutrient uptake rates . This robustness and responsiveness are most likely due to the wide range of the quasi-steady states in the phase space and the input-dependent timescales of relaxation . These two properties depend on the non-linear dynamics owing to the complex formation that is not in the vicinity of a fixed point , which has not been investigated using a constraint-based approach or with steady-state analysis of the previous model . A relationship between robustness and responsiveness has also been studied in other systems [28] and will be further investigated in various biological systems . To measure quasi-steady state characteristics , the dynamics of metabolites should be measured non-invasively for a long time; however , the real-time measurement techniques of metabolites are insufficient at present . Thus , we expect that the continuous development of new measurement techniques will help to validate our predictions experimentally . We demonstrated that an internal metabolite regulates its production via feedback mechanisms by the moiety conservation . The noise in the concentrations of metabolites and carrier can be reduced drastically . This feedback mechanism is important for stabilization of the metabolite and carrier concentrations at the finite constant values during the slow relaxation . When this feedback is missing , this noise cannot be reduced microscopically . Therefore , the concentration of the internal molecule should decrease to zero or diverge in the saturated condition ( S6 Fig and section 5 in S1 Text ) . This suggests that the feedback mechanism involving the carrier cycling may be responsible for the existence of the quasi-steady states . There are several important parameters contributing to the robustness against both external and internal fluctuations . In particular , kc , the turnover rate of an enzyme in the active carrier-consuming reaction , and kp , that in the active carrier-producing reaction , are important . We estimated these turnover rates using data from previously published studies: the glycolytic pathway of Escherichia coli during continuous aerobic cultivation [3 , 29] and the lactic acid fermentation pathway of Lactococcus lactis during anaerobic cultivation with an 80% lactic acid yield [30–32] . In both pathways , the enzyme levels necessary for each reaction are considered to be sufficient and not a rate-limiting factor , which we assumed in our model , and allowed us to consider the kinetics of substrate and carrier alterations subsequently . Based on our estimation , kc and kp in the glycolytic pathway are given as 50 s−1 and 12 s−1 , respectively , while the Fano factor of the internal metabolite is estimated at nearly 0 . 2 . For the lactic acid fermentation pathway , kc and kp are given as 1 . 2 s−1 and 4 . 9 s−1 , respectively , and the Fano factor is estimated at 0 . 85 ( see Table 1 ) . Our estimations are too simplified to allow for a quantitative discussion about the cellular metabolic process , because actual metabolic processes are more complicated than represented by our model . However , the properties observed with our model are preserved even if the details of the model change . Indeed , if two CCCs are coupled through a common carrier pool , the slow relaxation to the steady state is observed ( see S7 and S8 Figs , and section 6 in S1 Text ) . This suggests that the moiety conservation pool can underlie the jamming and the feedback processes that determine the dynamics of more complicated metabolic networks than the CCC . If there are some branches in the CCC , our results would still be qualitatively reproduced , although some quantities would no longer be conserved but rather given as quasi-steady-state values . It has been reported that such branches will change the steady-state characteristics [11] , e . g . , the bistability and hysteresis , and the slow dynamics might appear in transient dynamics approaching each stable fixed point . We expect that further non-trivial dynamic phenomena will be observed in more complicated metabolic networks with moiety conservation via the jamming and feedback . To investigate the effects of conserved carriers in more complicated metabolic networks quantitatively , both theoretical and experimental investigations are required in the future .
Our analyses suggest that enzyme turnover rates in metabolic pathways are the essential contributors to the robustness of metabolic dynamics . We estimated the parameters described in actual metabolic pathways , i . e . , the glycolytic pathway of Escherichia coli during continuous aerobic cultivation [29] and the lactic acid fermentation pathway of Lactococcus lactis during anaerobic cultivation with an 80% lactic acid yield [30] , and calculated the effects of carrier cycling . We approximated the kinetics of each reaction using the Michaelis-Menten equation . Hence , the concentrations of a substrate and carrier , the dissociation constant , and the speed of flux are necessary for the estimation of kc and kp . For kc of phosphofructokinase ( PKF ) in the glycolytic pathway , the concentrations of fructose 6-phosphate ( F6P ) as a substrate and ATP as a carrier used are 3 . 21 × 10−2 mM and 8 . 28 × 10−1 mM , respectively [29] . The dissociation constant is given as 0 . 16 mM [3] . The flux was estimated from the glucose uptake rate , which is 1 . 8 mM s−1 under the same conditions described previously [29] . However , the entire amount of glucose is not catalyzed by PKF , and 20–40% of glucose is considered to be involved in the pentose phosphorylation pathway [29] . Here , we considered this leak to be 20% , and the flux is estimated as 1 . 8 × 0 . 8 = 1 . 44 mM s−1 . Subsequently , kc is estimated to be approximately 50 s−1 , using the Michaelis-Menten form . For kp of pyruvate kinase ( PK ) , the concentrations of phosphoenolpylvate ( PEP ) and ADP are 1 . 49 × 10−1 mM and 9 . 65 × 10−1 mM , respectively [29] . The dissociation constant is 0 . 26 mM [3] . We assumed that the flux of PK is similar to that of PKF , and kp is estimated to be approximately 12 s−1 . From the estimated kc and kp , σ2/<n> of the internal metabolite is determined to be 0 . 2 . In the same manner , kc and kp in the lactic acid fermentation pathway can be estimated . For kc of glyceraldehyde phosphate dehydrogenase ( GAPDH ) , the concentrations of glyceraldehyde 3-phosphate ( GAP ) and NAD+ are 6 . 0 mM and 8 . 4 mM , respectively [30] . The dissociation constant is 0 . 2 mM [31] and the flux is 5 mM s−1 [30] . Therefore , kc is estimated to be 0 . 85 s−1 . For kp of lactate dehydrogenase ( LDH ) , the concentration of pyruvate and NADH are estimated as 1 mM and 0 . 7 mM , respectively [30] . The dissociation constant is 0 . 08 mM [32] , while the flux is given as 4 . 4 mM s−1 [30] . Therefore , kp is estimated to be 4 . 9 s−1 , and the σ2/<n> of the internal metabolite is determined to be 0 . 85 .
|
Although a metabolic shift is essential for the adaptation of cells or organisms to environmental changes , the transient behaviors of metabolic systems are poorly understood . When describing the time development of metabolic systems , there are several conserved quantities to consider due to balances of metabolic reactions , e . g . , the cycling of coenzymes . Such conserved quantities limit the possible changes in the metabolic state and can generate non-trivial dynamical behaviors . We here propose a minimal motif of metabolic reactions that includes coenzyme recycling to investigate the effect of conserved quantities on metabolic dynamics . We demonstrate that the dynamics with this motif intrinsically show slow relaxation to the steady state after environmental changes . Moreover , this motif can maintain robustness against external and internal fluctuations owing to the conservation of coenzymes . Overall , these results suggest that the complex metabolic dynamics generated by coenzyme recycling are beneficial to organisms .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Models",
"and",
"methods"
] |
[
"relaxation",
"time",
"coenzymes",
"enzymes",
"metabolic",
"networks",
"enzymology",
"systems",
"science",
"mathematics",
"metabolites",
"network",
"analysis",
"enzyme",
"metabolism",
"enzyme",
"chemistry",
"relaxation",
"(physics)",
"computer",
"and",
"information",
"sciences",
"proteins",
"metabolic",
"pathways",
"dynamical",
"systems",
"physics",
"biochemistry",
"biology",
"and",
"life",
"sciences",
"cofactors",
"(biochemistry)",
"physical",
"sciences",
"metabolism"
] |
2017
|
Metabolic dynamics restricted by conserved carriers: Jamming and feedback
|
Many choice situations require imagining potential outcomes , a capacity that was shown to involve memory brain regions such as the hippocampus . We reasoned that the quality of hippocampus-mediated simulation might therefore condition the subjective value assigned to imagined outcomes . We developed a novel paradigm to assess the impact of hippocampus structure and function on the propensity to favor imagined outcomes in the context of intertemporal choices . The ecological condition opposed immediate options presented as pictures ( hence directly observable ) to delayed options presented as texts ( hence requiring mental stimulation ) . To avoid confounding simulation process with delay discounting , we compared this ecological condition to control conditions using the same temporal labels while keeping constant the presentation mode . Behavioral data showed that participants who imagined future options with greater details rated them as more likeable . Functional MRI data confirmed that hippocampus activity could account for subjects assigning higher values to simulated options . Structural MRI data suggested that grey matter density was a significant predictor of hippocampus activation , and therefore of the propensity to favor simulated options . Conversely , patients with hippocampus atrophy due to Alzheimer's disease , but not patients with Fronto-Temporal Dementia , were less inclined to favor options that required mental simulation . We conclude that hippocampus-mediated simulation plays a critical role in providing the motivation to pursue goals that are not present to our senses .
Would you prefer a can of beer today or a bottle of champagne in one week ? Intertemporal choices , involving trade-offs between short-term and long-term outcomes , are pervasive in everyday life . The propensity to favor short-term pleasures defines a form of impulsivity that may have dramatic consequences on professional careers or family relationships . How can some people resist the attraction of short-term pleasures and pursue long-term goals , while others easily succumb and compromise their ultimate expectations ? This issue has been tackled in the recent years using functional neuroimaging techniques to explore neural activity during intertemporal choices [1] , [2] . Most studies implemented binary choices derived from behavioral economics paradigms , in which subjects have to choose between smaller-sooner and bigger-later monetary payoffs . Choice data could be fitted with a hyperbolic decay function , which characterizes how monetary payoffs are discounted over time and hence captures individual impulsivity [3]–[5] . Neural data suggested that recruitment of the dorsal prefrontal cortex is crucial to resist the attraction of immediate rewards , which is mediated by ventral prefronto-striatal circuits [6]–[8] . However , paradigms employing monetary rewards may miss some essential processes that crucially determine intertemporal choices in everyday life . A long time ago , Aristotle pointed out that “when some desirable object is not actually present to our senses , exerting its pull on us directly , our motivation to strive to obtain it is driven by our awareness of its ( memory or fantasy ) image” [9] . Along the same lines , some more recent authors suggested that imagining future situations might help in providing a motivation that counters the attraction of immediate pleasures [10]–[12] . Imagining future situations involves recomposing elements stored in episodic memory and hence recruiting the medial temporal lobe ( MTL ) regions . Indeed , these regions , with the hippocampus as a key component , are thought to be implicated in both recalling past episodes and imagining future episodes [13] , [14] . This idea was principally suggested by the observation of patients with MTL damage , who exhibit parallel impairment in episodic memory and future simulation [15]–[18] . The MTL general function has consequently been conceptualized as episodic thinking or mental time travelling [11] , [19]–[22] . Therefore , favoring long-term goals should involve not only the dorsal prefrontal cortex but also the medial temporal regions , as subjects engage in imagining future episodes . The aim of the present study was to uncover the role of the hippocampus in the conflict defined by Aristotle between temptations that strike our senses and fictions that we have to generate . It has been argued that such conflict between tangible and simulated options represents the most typical case of intertemporal choice we have to make in ecological situations [12] . We therefore extended previous intertemporal choice paradigms by showing concrete options ( food , culture , and sport items ) with two modes of presentation: some options were accompanied with pictures and thus immediately observable through vision , whereas other options were only described textually and thus required mental simulation . We first verified in a pilot behavioral study that participants assign higher values ( likeability ratings ) to the options imagined with more details . Then we used functional MRI to analyze neural activity elicited by option presentation and choice response , which were separated in time . Our prediction was that in ecological situations , which opposed simulated to observable rewards , hippocampus activity would be associated with higher value assigned to the delayed option . This was not expected in control conditions using the same difference in time ( immediate versus delayed options ) , but no difference in the presentation mode . Thus , hippocampus activation would explain a significant part of intersubject variability in the propensity to favor imagined outcomes , irrespective of delay . To complete our demonstration , we intended to establish a critical link with hippocampus anatomical structure , and not only a correlation with functional activation . First , we regressed the degree of impulsivity exhibited by our healthy participants in the ecological choices against grey matter density measured by structural MRI , using voxel-based morphometry ( VBM ) analysis . The prediction was that subjects preferring imagined outcomes would show increased grey matter density in the hippocampus . Second , we tested on the same intertemporal choice task patients with Alzheimer's disease ( AD ) , who represent the prototypical case of episodic memory impairment due to hippocampus degeneration [23] , [24] . As controls we included elderly healthy subjects and patients with moderate behavioral variant of fronto-temporal dementia ( bvFTD ) , another degenerative disease that preferentially affects the prefrontal cortex ( PFC ) [25] , [26] . The prediction was that AD patients should make more impulsive choices in the ecological situation , when delayed options have to be simulated and hence need the hippocampus to attain higher values , relative to control groups and conditions .
In a first behavioral pilot experiment , we verified that the quality of simulation indeed enhanced the values assigned to textually described options . Participants ( n = 15 ) of Experiment 1 first performed the monetary and episodic intertemporal choice tasks and then were asked how many details they imagined when reading each Sim option . Number of details was used as a proxy for simulation richness , as was implemented in studies that established the link between episodic memory and future simulation deficits [16] , [32] , [33] . Simulation richness was significantly correlated across trials to subjective likeability ratings ( Figure 2A , one-sample t test on individual robust regression coefficients , t14 = 8 . 69 , p<0 . 001 ) . Consistently , when considering ecological trials ( contrasting Sim to Obs options ) , subjects were significantly more prone to favor the delayed option when it was simulated with higher richness ( one-sample t test on individual robust regression coefficients , t14 = 6 . 47 , p<0 . 001 ) . The same relation between valuation and richness was observed across subjects: participants who reported having imagined more details gave higher ratings to Sim options ( Figure 2B , robust regression , t13 = 1 . 99 , p<0 . 05 ) . Experiment 1 therefore confirmed that simulation richness is a crucial factor in the ability to favor delayed options when opposed to directly observable options . Participants of the main MRI study ( n = 20 , Experiment 2 ) performed the same monetary and episodic intertemporal choice tasks . In both experiments , the observed choices were well predicted by the difference in discounted value between the two options ( Figure 2C ) . We then examined whether behavioral performance was consistent across the episodic and monetary tasks , in order to establish that subjective ratings could be discounted similarly to classical payoffs . We found that impulsive choices were obtained with similar frequency in the monetary and episodic tasks ( 46 . 32%±8 . 26% and 50 . 90%±17 . 57% of impulsive choices , respectively ) . Moreover , nonimpulsive choice rate was significantly correlated across participants between the two tasks ( Figure 2D , robust regression , t32 = 2 . 58 , p<0 . 01 ) , arguing in favor of a common underlying impulsivity trait . Consistently , the discount factor k was significantly correlated across individuals between monetary and episodic tasks ( robust regression , t32 = 3 . 57 , p<0 . 001 ) . Thus , the form of impulsivity that is characterized by steeper discounting with delay was observed in the episodic as well as in the monetary task . Focusing on the episodic task , we verified that the model fit was equally good in all conditions , to ensure that the imaging contrasts reported hereafter were valid . Prediction scores were calculated as the percentage of trials in which the option with the higher value estimate was chosen . Importantly , there was no significant difference in prediction scores between control and ecological trials ( 80 . 22%±1 . 31% and 78 . 47%±1 . 37% , paired t test: t33 = −1 . 26 , p>0 . 2 ) . The average difference in estimated values of chosen and nonchosen options was also very similar ( control , 3 . 40±0 . 21; ecological , 3 . 46±0 . 20; paired t test: t33 = −0 . 28 , p>0 . 3 ) . Overall , behavioral results validate our original episodic task , suggesting that participants weigh delays as they would do in classical economic paradigms ( following hyperbolic discounting ) but valuate delayed options in proportion to their simulation richness . Only the episodic task , not the monetary task , was used in the following analyses . All activations reported below survived family-wise error ( FWE ) correction for multiple comparisons , either for the whole brain ( noted WBC ) or for a small volume ( noted SVC ) corresponding to anatomical delineation of the hippocampus ( noted HC ) . We started with the identification , using GLM1 ( see Methods ) , of brain regions encoding values and choices across participants . We first looked for brain regions that parametrically encode option values in the episodic task , collapsing all trial types ( Figure S1 ) . This brain valuation system encompassed numerous regions ( all pFWE_WBC<0 . 05 ) , such as the ventromedial prefrontal cortex ( VMPFC ) , lateral parietal cortex ( LPC ) , posterior cingulate cortex ( PCC ) , and dorsolateral prefrontal cortex ( DLPFC ) . We then analyzed the activity recorded during choice period , looking for regions that reflect choosing delayed options ( nonimpulsive choices ) , regardless of trial type . A large prefrontal network , extending from the bilateral DLPFC to the dorsomedial prefrontal cortex ( DMPFC ) , was significantly more activated ( pFWE_WBC<0 . 05 ) for nonimpulsive than for impulsive choices ( Figure 3A , left ) . Next we searched for regions that would be specifically recruited for nonimpulsive choices in the ecological trials ( Figure 3A , right ) . The interaction between choice and condition elicited specific activation in the left hippocampus ( L_HC , pFWE_SVC<0 . 01; bi_HC pFWE_SVC = 0 . 05 ) . An ROI analysis ( Figure 3B ) confirmed the dissociation between prefrontal regions ( DLPFC and DMPFC ) , which were more activated for nonimpulsive choices in both control and ecological trials , and left hippocampus , which was specifically engaged when subjects made nonimpulsive choices in the ecological trials ( one-tailed paired t tests , p<0 . 05 ) . Thus , the analysis of brain activity suggests that the hippocampus is specifically involved in choosing delayed options when they need to be simulated , against immediate options that are directly observable ( i . e . , in Obs/Sim trials ) . This is in line with the hypothesis that hippocampus activity is proportional to simulation richness and therefore to the value of simulated options . This hypothesis predicts the observed absence of hippocampus activation in the two control conditions , for different reasons . In Obs/Obs trials , there is no need for simulation , and hence no need for hippocampus activation . In Sim/Sim trials , there are two options to simulate , but their value is on average the same for impulsive and nonimpulsive choices . This is why the contrast between impulsive and nonimpulsive choices yields no activation in the hippocampus . We explored interindividual differences , in order to provide additional evidence for the role of the hippocampus in resisting impulsive choices during ecological trials . We first took advantage of the intersubject variability in the rate of impulsive choices . Using GLM2 ( see Methods ) , we specifically looked for regions in which the correlation between the nonimpulsive minus impulsive choice contrast and the nonimpulsive choice rate was higher in the ecological compared to the control condition ( Figure 4A , left ) . This analysis , controlling for factors such as age , gender , and global correlation , revealed a significant cluster in the left hippocampus ( L_HC , pFWE_SVC<0 . 05; bi_HC , pFWE_SVC = 0 . 07 ) . Post hoc analysis confirmed that the signal extracted from this L_HC ROI was positively correlated across subjects with nonimpulsive choice rate in the ecological condition ( robust regression , t17 = 3 . 9 , p<0 . 001 , Figure 4A , right ) . Thus , subjects who exhibited less impulsivity in ecological choices had stronger hippocampus activation when choosing delayed options . To provide further insight into the relationship between neural activity and behavioral impulsivity , we investigated correlations across individuals between the values inferred from the behavior and the activation measured at the time of option display ( Figure 4B ) . More precisely , the two variables tested for these correlations were the difference in value estimates between delayed and immediate options ( which we termed “behavioral valuation” ) and the difference in activation between delayed and immediate options ( which we termed “neural valuation” ) . We used GLM3 ( see Methods ) to search for regions showing higher correlation in ecological than in control conditions ( Figure 4B , left ) . We found a significant cluster in the left hippocampus ( L_HC , pFWE_SVC<0 . 01; bi_HC , pFWE_SVC<0 . 05 ) . Thus , the individual propensity to value delayed options more than immediate options was linked with more activation in the hippocampus for delayed than for immediate option presentation . Post hoc analysis of the signal extracted from the L_HC ROI ( Figure 4B , right ) confirmed that intersubject correlation between behavioral and neural valuation was significantly positive in the ecological trials ( robust regression , t17 = 2 . 72 , p<0 . 05 ) . We next examined whether interindividual differences in brain structure could account for the propensity to favor nonimpulsive options in our ecological condition . For this we performed VBM analysis ( see Methods ) on T1-weighted anatomical scans ( n = 18 ) . As a first step , we tested whether grey matter ( GM ) density in the L_HC ( ROI from Figure 4A , top ) could account for nonimpulsive choice rate . We found a significant correlation in the ecological trials ( robust regression , t16 = 1 . 84 , p<0 . 05 ) but not in the control trials ( robust regression , t16 = 011 , p>0 . 4 ) . Similar results were obtained for the right or bilateral hippocampus . We also performed a whole-brain analysis that directly regressed nonimpulsive choice rate in the ecological condition against individual segmented GM maps , controlling for age , gender , and total intracranial volume ( Figure S2 ) . We found a small set of significant clusters , among which the hippocampus ( R_HC , pFWE_SVC<0 . 05; bi_HC , pFWE_SVC = 0 . 07 ) . We then examined whether the link from brain structure to behavioral choice could be mediated by brain activity . For this we extracted the nonimpulsive versus impulsive contrast from the L_HC ROI ( Figure 4A , top ) , for both the ecological and control conditions . These functional contrasts were regressed against segmented GM maps , controlling for age , gender , and total intracranial volume . We then searched for regions were GM density was more correlated with functional contrast in the ecological than in the control condition . We again found a significant cluster ( Figure 5A , left ) in the hippocampus ( L_HC , pFWE_SVC = 0 . 059 ) . Post hoc ROI analysis ( Figure 5A , right ) confirmed that intersubject correlation between GM density and functional activation was significantly positive in the ecological condition ( robust regression , t16 = 4 . 64 , p<0 . 001 ) . Thus , GM density in the left hippocampus accounted for both the individual propensity to favor delayed options in our ecological condition and the related hippocampus activation ( nonimpulsive minus impulsive contrast in the ecological condition ) . A simple explanation of these statistical dependencies is that hippocampus activation mediates the relationship between GM density and behavioral choice ( Figure 5B ) . To test this hypothesis , we performed a mediation analysis ( see Methods ) . Results revealed that , when including hippocampus activation as a mediator , the direct path from anatomy to behavior was no longer significant . On the contrary , all the links of the indirect path ( anatomy to activation to behavior ) were significant ( all p<0 . 05 ) . One could argue that the VBM results may somehow confound the fMRI results , because a region with more neurons will show more activation , irrespective of functional implications . This is only true of course if those neurons are concerned with the contrast used to elicit functional activation , which we precisely intended to demonstrate . Furthermore , fMRI data established hippocampus functional implication in nonimpulsive choice not only in between-subject correlation but also in within-subject contrast , which cannot be driven by anatomical variations across participants . To summarize , the functional and structural MRI data suggest that subjects with higher GM density in the hippocampus show more pronounced hippocampus activation in ecological trials and therefore better resistance to impulsivity . We note , however , that the statistical links demonstrated so far have no directionality . It could be argued that less impulsive subjects have higher hippocampal activation because they tend to simulate future options with more details . The same reasoning can apply to hippocampus anatomy , if we assume that activating a brain structure can increase its density . To eliminate the possibility that the anatomo-functional properties of the hippocampus are just a by-product of subjects liking future options , we investigated the consequence of hippocampal damage . To assess whether intact hippocampus is necessary for preventing choice impulsivity , we compared the performance of AD patients to that of bvFTD patients and elderly controls in the episodic intertemporal choice task . We first verified that AD and bvFTD patients recruited in the Pitié-Salpêtrière neurology wards ( see Table 2 for demographic and clinical details ) presented with a differential atrophy in the hippocampus . To this aim , we compared T1-weighted anatomical scans from 55 AD and 48 bvFTD patients using a two-sample t test on segmented GM maps , controlling for age , gender , mini-mental state ( MMS ) , and total intracranial volume ( see Methods ) . AD patients had reduced GM density in a large cluster ( pFWE_WBC<0 . 05 ) that extended bilaterally from medial temporal regions to parietal lobules ( Figure 6A , top ) , with a local maximum in the hippocampus ( bi_HC , pFWE_SVC<0 . 05 ) . Reciprocally , bvFTD patients had reduced GM density in a large prefrontal cluster ( pFWE_WBC<0 . 05 ) , mostly in the ventral and medial PFC areas ( Figure 6A , bottom ) . This VBM analysis therefore confirmed that patients diagnosed with AD or bvFTD in our neurology wards were indeed characterized by specific neuro-degeneration patterns in temporo-parietal versus prefrontal regions , respectively . We then administered the episodic intertemporal choice task ( experiment A , see Methods ) to 20 AD patients , 14 bvFTD patients , and 20 elderly controls ( CTL ) ( see Table 3 for characteristics ) . AD and bvFTD patients were matched for global cognitive ability measured with Mini-Mental State examination ( two-sample t test , t33 = 0 . 34 , p>0 . 3 ) . However , as expected , AD patients had more difficulty with episodic memory in the Free and Cued Selective Reminding Test ( total free recall , two-sample t test , t24 = 2 . 47 , p<0 . 01 ) and bvFTD patients with executive functions in the Frontal Assessment Battery ( FAB score , two-sample t test , t33 = 2 . 70 , p<0 . 01 ) . Increasing delay significantly decreased the proportion of nonimpulsive choices in all groups ( one-sample t tests; CTL , t19 = 6 . 43 , p<0 . 001; AD , t19 = 5 . 90 , p<0 . 001; bvFTD , t14 = 2 . 85 , p<0 . 01 ) . A first notable difference ( Figure 6B , left ) was that both groups of patients were on average more impulsive compared to healthy controls ( two-sample t tests , AD versus CTL , t38 = 2 . 21 , p<0 . 05; bvFTD versus CTL , t33 = 4 . 30 , p<0 . 001; bvFTD versus AD , t33 = 2 . 29 , p<0 . 05 ) . The key test was the comparison of nonimpulsive choice rate between the ecological condition ( Obs/Sim ) , in which the delayed option had to be mentally simulated , and the control condition ( Obs/Obs ) , in which the delayed option could be visually observed . Unfortunately , the Group×Condition interaction tested with a global ANOVA did not reach the significance threshold ( F2 , 52 = 12 . 52 , p = 0 . 14 ) . However , an exploratory analysis using t tests in each group separately fulfilled our predictions: AD patients were significantly more impulsive in the ecological condition ( one-sample t test , AD , t19 = 2 . 61 , p<0 . 001 ) , not control subjects or bvFTD patients ( one-sample t tests , CTL , t19 = 0 . 95 , p>0 . 3; bvFTD , t14 = 0 . 97 , p>0 . 1 ) . Also , the difference between healthy controls and AD patients was significant for ecological but not for control trials ( two-sample t tests , ecological , t38 = 2 . 84 , p<0 . 001; control , t38 = 1 . 20 , p>0 . 1 ) , whereas the difference between healthy controls and bvFTD patients was significant in both conditions ( two-sample t tests , ecological , t33 = 4 . 16 , p<0 . 001; control , t33 = 4 . 02 , p<0 . 001 ) . Thus , pathological impulsivity was specifically revealed by the ecological condition in AD patients , but was exhibited irrespective of condition in bvFTD patients . In order to confirm the specific deficit observed in AD patients , we modified the task by removing sport and culture options , which proved not suitable for aged and diseased subjects , and by shortening the delays , such that they were more adapted to elderly patients ( Experiment B , see Methods ) . We recruited another 15 AD patients and 15 control subjects to try and replicate the results in an independent sample ( Figure 6B , right ) . Crucially , the Group ( AD versus CTL ) ×Condition ( ecological versus control ) interaction was this time significant ( F1 , 28 = 12 . 52 , p<0 . 01 ) , Thus , AD patients made more impulsive choices than healthy controls , and this difference was driven by the ecological condition ( two-sample t test , t28 = 5 . 10 , p<0 . 001 ) . Because the AD and CTL groups were not well matched in age , we verified that the group factor still explained the difference in impulsivity between ecological and control conditions when inserted into a GLM that also included a regressor for age . The GLM fit confirmed that group was a significant factor ( t28 = 3 . 12 , p<0 . 01 ) but not age ( t28 = 0 . 57 , p>0 . 5 ) . Thus , the behavior of AD patients suggests that damage to the hippocampus had an impact on intertemporal choices , since future options that needed mental simulation were no longer favored .
In this study , we extended standard delay discounting paradigms investigating intertemporal choices between smaller-sooner and bigger-later monetary payoffs . First , we developed an episodic choice task using more concrete options such as food , sport , and culture items . Second , we manipulated the mode of presentation to investigate ecological choices where the immediate option is directly observable , while the delayed option requires simulating a future episode . Behavioral data showed that richness of mental simulation is a crucial factor in valuating and hence choosing delayed options during ecological intertemporal conflicts . Imaging data revealed that interindividual variability in the propensity to favor simulated options can be explained by hippocampus functional activation during both valuation and choice , which in turn can be explained by the hippocampus grey matter density . Patient data demonstrated that AD , which is characterized by hippocampus atrophy , exacerbates impulsivity specifically when delayed options require mental simulation . Taken together , these results provided a strong support to our central hypothesis that the hippocampus helps valuating imagined outcomes , which reduces impulsivity in the context of ecological intertemporal choices [10]–[12] . In the following paragraphs we discuss behavioral data , functional MRI data , structural MRI data , and patient data , successively . Behavioral data were modeled using a hyperbolic decay function to discount values with delays and a softmax rule to estimate choice likelihood . This model arguably provides a good account of intertemporal choices [4] , [5] , [34] and has become standard in the recent neuroeconomic literature [1] , [2] , [6] , [28] , [29] . We found that hyperbolic discounting provided an equally good fit of the monetary and episodic tasks—that is , whether we take objective financial payoffs or subjective likeability ratings as proxies for values . Note that because values were objective payoffs in one task and subjective ratings in the other , we could not assess whether episodic options attenuate discounting , as was shown previously [30] , [31] . However , comparing with other discounting models would go beyond the scope of this study . Indeed , our focus was on how the brain assigns values to simulated options , not on how the brain discounts these values with delays . Importantly , the percentage of nonimpulsive choices , as well as the adjusted discount factor k , was correlated across subjects between the two tasks . This indicates that the same impulsivity trait , characterized as steepness of delay discounting , explains a significant part of variance in both financial and more concrete choices . Steepness of delay discounting might therefore be dissociated from the ability to simulate future episodes , which affected values irrespective of delays . Consistent with this idea , it was recently reported that a patient with impaired future simulation , due to hippocampal damage , exhibited normal discounting in a standard , monetary intertemporal choice task [35] . Imaging data corroborate previous findings [6] , [7] , [36] , that nonimpulsive choices involved dorsal prefrontal regions ( DLPFC and DMPFC ) , in both ecological and control trials . The novelty is the dissociation of hippocampus activation , which was specifically observed during ecological nonimpulsive decisions . Most hippocampus activations reported here were predominant on the left side , but survived small volume correction on both sides , when using bilateral masks of the hippocampus , independently defined from anatomical criteria . Thus , whereas the dorsal prefrontal cortex seems involved in preventing impulsivity during various types of choice , the hippocampus is specifically recruited for selecting simulated future options against directly observable options . To our knowledge , this study is the first to implement the conflict between delayed options represented in episodic systems and immediate options represented in perceptual systems . Let us discuss the reasons why the hippocampus was activated by the contrast of nonimpulsive versus impulsive choice in this ecological situation , specifically . This contrast isolates choices where the simulated option was preferred and therefore imagined with greater detail , which according to our working hypothesis is underpinned by higher hippocampus activity . The same contrast did not activate the hippocampus in control conditions when the two options were observed , or when they were both simulated , for different reasons . When the two options can be represented in perceptual systems , there is no purpose for hippocampus activation since mental simulation is not required . When the two options are simulated they are presumably represented in episodic systems , but with similar richness irrespective of the choice , hence the absence of hippocampus activation when contrasting nonimpulsive and impulsive choices . Although the idea that books elicit more imagination than movies seems well shared , it may be argued that pictures could also have yielded some simulation ( for instance , imagining oneself consuming the food ) . Thus , observed and simulated options might not differ radically but rather in the degree of simulation needed for a proper valuation . We did not implement the contrast between immediate-simulated and delayed-observable options , because it would make little sense with respect to choice problems encountered in the real life . Indeed , immediately available options cannot be far from our senses , and our senses cannot directly perceive future events . We would nonetheless argue that it is the simulation process , and not the temporal frame ( future against present ) , that determines the recruitment of hippocampus; otherwise , we would have observed hippocampus activation during nonimpulsive choices in the control conditions . Yet we do not take position on which subprocess of mental simulation ( such as retrieving pieces of information , reassembling these pieces into a new structure , representing the affective content of the simulation etc . ) was implemented by the hippocampus . We also acknowledge that , due to practical constraints , our immediate-observable options were not truly obtainable at the moment of choice ( but only after the fMRI session was over ) . The contrast with future-simulated options might have been even more powerful had subjects been confronted with the real object physically present at a reachable distance , either because it would have had more concreteness [12] or because it would have triggered Pavlovian consummatory processes [37] . Finally , let us emphasize that behavioral and fMRI data provide no indication about the direction of causality between valuation and simulation . The correlation observed in the behavior between simulation richness and likeability rating could reflect the fact that subjects imagined in more details what they liked in the first place . In this framework , variations of hippocampus activity could represent a by-product ( and not a cause ) of the valuation process . It was therefore crucial to assess the existence of a directional link from the simulation to the valuation process—that is , to demonstrate the necessity of hippocampus recruitment for valuating simulated options , which we did with patient studies . To demonstrate that hippocampus-mediated simulation could explain some part of choice impulsivity , we explored interindividual variability . We found that the individual difference in value between delayed and immediate options ( behavioral valuation ) was correlated with the individual difference in activation between delayed and immediate options ( neural valuation ) in the hippocampus specifically during ecological trials . Thus , higher values assigned to delayed options correlated with both richer future simulation in behavioral data and stronger hippocampus activation in neuroimaging data . This is in line with the hypothesis that hippocampus-mediated future simulation helps valuating textually described options . This functional feature was corroborated by the anatomical observation that participants with higher grey matter density in the hippocampus have a higher propensity to make nonimpulsive choices in the ecological situation . Moreover , the relation between hippocampus anatomy and choice impulsivity was mediated by differential hippocampus activation in nonimpulsive versus impulsive choices . We therefore suggest that , on top of an impulsivity trait that relates to how delays are weighted and hence would affect any form of intertemporal choices , some individual variability in resisting observable rewards and favoring simulated options relies on the hippocampus structure and function . Again , the role of the hippocampus would not be to adjust the impact of delay but to provide a simulation that would make the delayed option more attractive . One obvious clinical implication is that patients suffering from hippocampal damage , such as in AD , might encounter difficulties in pursuing long-term goals , due to deficient future simulation . To examine this possibility , we compared AD patients to patients with bvFTD , which is also a degenerative disease that progressively induces dementia in the aged person . Unfortunately , the overlap with the patients who performed our episodic intertemporal choice task was only partial , precluding direct correlations between neural degeneration and behavioral performance . Yet a direct comparison of brain anatomy between groups showed that GM density reduction preferentially affected prefrontal regions in bvFTD patients and MTL regions ( including the hippocampus ) as well as parietal areas in AD patients . This pattern could be expected from previous studies that compared AD and bVFTD groups to patients with mild cognitive impairment or to healthy controls [38]–[41] . Nevertheless , the clear-cut dissociation obtained here was not trivial , because of the common pathological features shared by the two pathological conditions [42] . This direct comparison between patient groups is certainly more stringent than the traditional comparison with healthy controls . Thus , we can reasonably assume that our AD patients had hippocampus degeneration , which validates our prediction that they should be particularly impulsive in the ecological condition . We note , however , that these patients also had other atrophic brain regions , particularly in the parietal cortex . Therefore , the patient study alone cannot be conclusive on hippocampus contribution to choice impulsivity . Nevertheless , the hippocampus appears as the most parsimonious candidate , given that it was also implicated in fMRI and VBM studies . Furthermore , results of the mediation analysis suggest that the impact of anatomical damage on choice impulsivity is mediated by the inability to activate the hippocampus during ecological conflicts . The behavioral performance showed the following dissociation: AD patients exhibited pathological impulsivity in the ecological situation specifically , whereas bvFTD patients were found impulsive in all situations . The impulsivity observed in bvFTD patients accords well with the disinhibition syndrome that is classically reported in this variant of FTD [43]–[45] . It remains unclear whether their impulsivity emerges from a deficit in valuating options or in controlling choices . On the contrary , AD patients showed normal behavior in control conditions , ruling out any general impairment in valuation or choice . Instead , their impulsivity was revealed when no visual support was provided for the delayed option , which hence required simulation to be properly valuated . This enhanced impulsivity was driven by the food domain , which arguably proposes more tangible options than culture and sport domains , and hence a better contrast between immediately available rewards and simulated future events . The idea that AD patients are impulsive might seem counterintuitive to clinicians , as AD has rather been associated with apathy [46]–[48] . Apathy and impulsivity are not incompatible , however; indeed they frequently coexist in the same patients . We suggest that the inability to simulate future situations might also explain a lack of motivation in AD patients , precisely for long-term goals that cannot be visualized in their immediate environment . We replicated the demonstration of specific impulsivity during ecological intertemporal conflict in AD patients , using a short version of the task that can be administered in a few minutes . The test was therefore robust enough to overcome the fact that food preferences may vary across patients . This short test might prove useful in detecting motivational disorders in AD patients , and possibly in distinguishing AD from other degenerative diseases .
|
Economic theory assumes that we assign some sort of value to options that are presented to us in order to choose between them . In neuroscience , evidence suggests that memory brain regions , such as the hippocampus , are involved in imagining novel situations . We therefore hypothesized that the hippocampus might be critical for evaluating outcomes that we need to imagine . This is typically the case in intertemporal choices , where immediate rewards are considered against future gratifications ( e . g . , a beer now or a bottle of champagne a week from now ) . Previous investigations have implicated the dorsal prefrontal cortex brain region in resisting immediate rewards . Here we manipulated the mode of presentation ( text or picture ) , such that options were represented either in simulation or in perception systems . Functional neuroimaging data confirmed that hippocampal activity lends a preference to choosing simulated options ( irrespective of time ) , whereas dorsal prefrontal cortex brain activity supports the preference for delayed options ( irrespective of presentation mode ) . Structural neuroimaging in healthy subjects and in patients with brain atrophy , due to Alzheimer's disease ( with hippocampal damage ) or Fronto-Temporal Dementia ( with damage to the prefrontal cortex ) , further demonstrated the critical implication of the hippocampus . Individuals with higher neuronal density in the hippocampus , but not in the dorsal prefrontal cortex , were more likely to choose future rewards that have to be mentally simulated .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion"
] |
[] |
2013
|
A Critical Role for the Hippocampus in the Valuation of Imagined Outcomes
|
Filamentous fungi produce diverse secondary metabolites ( SMs ) essential to their ecology and adaptation . Although each SM is typically produced by only a handful of species , global SM production is governed by widely conserved transcriptional regulators in conjunction with other cellular processes , such as development . We examined the interplay between the taxonomic narrowness of SM distribution and the broad conservation of global regulation of SM and development in Aspergillus , a diverse fungal genus whose members produce well-known SMs such as penicillin and gliotoxin . Evolutionary analysis of the 2 , 124 genes comprising the 262 SM pathways in four Aspergillus species showed that most SM pathways were species-specific , that the number of SM gene orthologs was significantly lower than that of orthologs in primary metabolism , and that the few conserved SM orthologs typically belonged to non-homologous SM pathways . RNA sequencing of two master transcriptional regulators of SM and development , veA and mtfA , showed that the effects of deletion of each gene , especially veA , on SM pathway regulation were similar in A . fumigatus and A . nidulans , even though the underlying genes and pathways regulated in each species differed . In contrast , examination of the role of these two regulators in development , where 94% of the underlying genes are conserved in both species showed that whereas the role of veA is conserved , mtfA regulates development in the homothallic A . nidulans but not in the heterothallic A . fumigatus . Thus , the regulation of these highly conserved developmental genes is divergent , whereas–despite minimal conservation of target genes and pathways–the global regulation of SM production is largely conserved . We suggest that the evolution of the transcriptional regulation of secondary metabolism in Aspergillus represents a novel type of regulatory circuit rewiring and hypothesize that it has been largely driven by the dramatic turnover of the target genes involved in the process .
Filamentous fungi produce diverse repertoires of small molecules known as secondary metabolites ( SMs ) [1] . SMs include widely used pharmaceuticals such the antibiotic penicillin [2] , the cholesterol-reducing drug lovastatin [3] , and the immunosuppressant cyclosporin [4] , as well as potent mycotoxins , such as aflatoxin [5] and fumonisin [6 , 7] . SMs play key ecological roles in territory establishment and defense , communication , and virulence [8–12] . The genes involved in fungal SM pathways are often physically linked in the genome , forming contiguous SM gene clusters [13] . These gene clusters are typically characterized by a backbone gene , such as those encoding nonribosomal peptide synthetases ( NRPSs ) , polyketide synthases ( PKSs ) , hybrid NRPS-PKS enzymes , and prenyltransferases , whose protein products are responsible for synthesizing the proto-SM . Additional genetic components of SM gene clusters include genes for one or more tailoring enzymes that chemically modify SM precursors , transporter genes responsible for exporting the final product , and transcription factors that drive expression of the remaining genes in the gene cluster . For example , the gene cluster responsible for the synthesis of the mycotoxin gliotoxin in the opportunistic human pathogen Aspergillus fumigatus contains 13 genes including a non-ribosomal peptide synthase ( gliP ) , multiple tailoring enzymes ( gliI , gliJ , gliC , gliM , gliG , gliN , gliF ) , a transporter gene ( gliA ) , a transcription factor ( gliZ ) , and a gliotoxin oxidase gene that protects the fungus from the harmful effects of gliotoxin ( gliT ) [14 , 15] . Filamentous fungi exhibit a huge amount of SM biochemical diversity . Individual SMs are often known to be produced by only one or a handful of species , and the SM chemotypic profiles of closely related fungi are typically non-overlapping [1 , 16] . For example , the meroterpenoid fumagillin , originally isolated from A . fumigatus , has only been detected in A . fumigatus and some isolates of Penicillium raistrickii [17 , 18] . The gene cluster required for its production appears to be conserved in the A . fumigatus close relative , Aspergillus fischerianus , though only intermediate compounds have been detected from cultures of this and other closely related species [19–21] . In some genera , including Aspergillus [22] , the extent of fungal SM distribution is so taxonomically narrow that SM chemotypic profiles have been used as unequivocal species-level identifiers . As might be expected given their key roles in fungal ecology , SM production–and as a consequence SM gene cluster transcriptional activity–is tightly controlled by a complex network of master SM regulators triggered by a wide variety of environmental cues such as temperature , light , pH , and nutrient availability [23] . Among the master SM regulators identified to date are members of the fungal-specific Velvet protein family , which regulate SM production in a light-dependent manner in the model filamentous fungus Aspergillus nidulans [24–27] . The founding member of the Velvet family , VeA , stimulates production of diverse types of SMs in various fungal genomes under dark conditions , and has been shown to regulate gliotoxin , fumagillin , fumitremorgin G , and fumigaclavine C gene cluster expression and metabolite production in A . fumigatus [28] . Recently , a VeA-dependent regulator of secondary metabolism , MtfA , was identified in A . nidulans , which–unlike VeA–is localized in the nucleus regardless of light conditions [29] . MtfA regulates terrequinone , sterigmatocystin , and penicillin in A . nidulans; in A . fumigatus , MtfA is necessary for normal protease activity , and virulence assays using the moth Galleria mellonella suggest it plays a role in pathogenicity [30] . In addition to regulating SM , both of these regulators have been linked to the regulation of asexual and sexual development . Timing of SM production with developmental changes is well established in filamentous fungi , and the presence/absence of certain SMs has been linked with developmental changes [31–33] . It has been suggested that regulators that coordinate SM and development allow filamentous fungi to support more “complex” lifestyles through the production of a much greater diversity of natural products than their unicellular yeast relatives , which lack veA as well as backbone synthesis genes necessary for SM production [33–35] . Remarkably , both veA and mtfA appear to be broadly conserved in filamentous fungi with non-overlapping SM profiles [29 , 36] . We used four well-studied organisms from the fungal genus Aspergillus , a highly diverse genus and producer of some of the most iconic SMs , including gliotoxin and penicillin , to investigate the evolutionary variability in the distribution of SM gene clusters and its interaction with these two broadly conserved global transcriptional regulators that differ in their response to light , veA and mtfA . Our evolutionary analyses show that although both the SM gene clusters as well as their gene content are poorly conserved between Aspergillus species , explaining the narrow taxonomic distribution and distinctiveness of their SM profiles , the effects of the global transcriptional regulators on SM production in response to environmental cues are largely conserved across these same species . In contrast , examination of the role of veA and mtfA in development , a process that involves genes that are highly conserved between the two species and whose regulation is intimately linked to SM regulation , yields a different pattern; whereas the role of veA is conserved , mtfA regulates development in the homothallic A . nidulans but not in the heterothallic A . fumigatus . Although rewiring has been well documented in yeast and animal regulatory networks [37–44] these studies largely concern rewirings of regulatory signals between otherwise conserved regulators and target genes . Our finding that VeA , and to a certain extent MtfA , regulate non-homologous SM pathways in both A . fumigatus and A . nidulans suggests a new type of rewiring in which conserved regulators control a conserved biological process , even though the underlying genes and pathways that make up the biological process are not themselves conserved .
The genomes of A . fumigatus , A . nidulans , Aspergillus oryzae , and Aspergillus niger contain 317 , 498 , 725 , and 584 secondary metabolic genes , respectively , which are organized in 37 , 70 , 75 , and 78 corresponding secondary metabolic gene clusters ( S1 Table ) [45] . We considered SM gene clusters to be conserved between species if greater than half of the genes in the larger gene cluster were orthologous to greater than half of the genes in the smaller gene cluster . Even with this very liberal definition of gene cluster conservation , we found that no SM gene clusters were conserved across all four species . Moreover , 91 . 9–96 . 1% of SM gene clusters were specific to each species , with only 7 SM gene clusters conserved between any species ( Fig . 1A , S2 Table ) . Only one two-gene cluster is completely conserved between two species ( A . fumigatus 27 and A . oryzae 9 ) . Although none of these conserved SM gene clusters have chemically characterized products , the fact that only one of the 7 gene clusters appear to be 100% conserved in only two of the four Aspergillus species , suggests that the SM chemotypic profiles of the four species are non-overlapping . While only one ( at the 100% level ) or very few ( at the 50% level ) conserved SM gene clusters can be identified in comparisons between any of these four species , SM gene clusters do contain genes whose orthologs are parts of other , non-homologous , SM gene clusters . For example , the 25 genes in the sterigmatocystin gene cluster in A . nidulans , one of the largest SM gene clusters present in the genomes analyzed , have orthologs in 25 SM gene clusters in the other three species as well as inparalogs in 8 other A . nidulans SM gene clusters ( Fig . 1B , S3 Table ) . However , in all but one case , less than 20 . 0% ( 5 genes ) of the sterigmatocystin gene cluster is present in the other gene cluster . The only exception is the truncated aflatoxin gene cluster of A . oryzae , which shares 11 orthologs with the ST gene cluster . Although the A . oryzae aflatoxin gene cluster is non-functional [46–48] , the evolutionary conservation between the aflatoxin and sterigmatocystin gene clusters is reflected in the fact that sterigmatocystin is the penultimate precursor product of the aflatoxin biosynthetic pathway [49] . Given the remarkable lack of conservation of SM gene clusters between Aspergillus species , we next tested whether the genes belonging to these clusters were also less conserved by comparing their degree of evolutionary conservation to that of genes involved in primary metabolism . To determine the percentage of species-specific orthogroups ( see Methods ) involved in SM and primary metabolism , we first determined the number of orthogroups with genes annotated to these functional categories that contained no genes in any other species ( i . e . , all orthogroups that contain only a single gene or only putative inparalogs ) . To determine the conservation of gene function across species , we determined the number of orthogroups in each genome that are either entirely species-specific or whose members in other genomes are annotated to different functional categories . We found that SM orthogroups were significantly far less conserved than primary metabolic orthogroups in all four genomes examined ( adjusted P < 1e-10 for all combinations; S4 Table ) . Specifically , the percentage of species-specific primary metabolic orthogroups ranged between 7 . 5 and 15 . 4% ( Fig . 2 ) . When we considered conservation of function ( i . e . whether orthogroups contained genes annotated to primary metabolism ) , we saw a slight and non-significant increase in the percentage of species-specific orthogroups ( 8 . 9–18%; Fig . 2 ) . In contrast , 25 . 0% to 40 . 5% of orthogroups containing SM backbone genes were species-specific; considering functional conservation had a negligible impact on these percentages ( Fig . 2 ) . Orthogroups containing genes in SM clusters exhibited similar percentage ranges for species-specificity ( 26 . 0–38 . 1%; Fig . 2 ) . Strikingly , the orthologs of many SM genes in a given species were not in SM gene clusters in any of the other species; specifically , 53 . 7% ( in A . fumigatus , with 37 SM clusters ) to 74 . 7% ( in A . niger , with 77 SM gene clusters ) of orthogroups containing an SM gene from a particular genome had no orthologs in SM clusters in any of the other genomes . We next examined the function of the conserved secondary metabolic regulator VeA by performing RNA sequencing [19 , 48 , 50] of ΔveA and wild-type ( WT ) A . fumigatus strains TSD1 . 15 [51] and CEA10 and A . nidulans strains TXFp2 . 1 and TRV50 . 2 [52] to identify genes and biological processes that are differentially regulated in ΔveA vs WT in the two species . Of the 9 , 783 transcribed genes in the A . fumigatus genome , 1 , 546 ( 15 . 8% ) were over-expressed and 1 , 555 ( 15 . 9% ) were under-expressed in the ΔveA vs WT analysis in A . fumigatus ( Tables 1 , S5 ) . We observed very similar numbers of genes differentially regulated in the A . nidulans ΔveA vs WT analysis; out of 10 , 709 genes in the A . nidulans genome , 1 , 165 ( 10 . 9% ) were over-expressed and 1 , 671 genes ( 15 . 6% ) were under-expressed . In total , approximately 32% and 26% of protein coding genes were differentially regulated in ΔveA compared to WT in A . fumigatus and A . nidulans , respectively . To characterize the broad biological processes that these differentially regulated genes are involved with , we performed GO term enrichment analysis using the Aspergillus GOSlim term hierarchy [53 , 54] . Remarkably , the same four GO terms , namely secondary metabolic process , carbohydrate metabolic process , oxidoreductase activity , and extracellular region , were significantly enriched in under-expressed genes in both A . nidulans and A . fumigatus , showing that VeA is a positive regulator of the same processes in both species ( Fig . 3 ) . Over-expressed genes in A . fumigatus were significantly enriched for twelve GO terms potentially related to cell growth , namely ribosome biogenesis , cellular amino acid metabolic process , translation , rna metabolic process , structural molecule activity , helicase activity , rna binding , transferase activity , nucleus , nucleolus , ribosome , and cytosol . Five of these twelve terms were also significantly enriched in A . nidulans ( ribosome biogenesis , cellular amino acid metabolic process , rna metabolic process , nucleus , and nucleolus ) . Over-expressed genes were present in the remaining seven terms in A . nidulans but did not show statistically significant enrichment ( S6 Table ) . In addition to the enrichment of the GO term secondary metabolic process in the differentially expressed genes from the ΔveA vs WT comparison in both A . nidulans and A . fumigatus , a large portion of the 317 A . fumigatus SM cluster genes and 498 A . nidulans SM cluster genes was also differentially expressed in the ΔveA strains . In A . fumigatus , 98 genes ( 30 . 9% ) were under-expressed and 38 ( 12 . 0% ) were over-expressed; in A . nidulans , 184 genes ( 36 . 9% ) were under-expressed and 67 ( 13 . 5% ) were over-expressed ( S5 Table ) . Interestingly , all constituent genes of several SM gene clusters were differentially expressed . For example , all genes in the A . fumigatus pseurotin A gene cluster and all genes in the A . nidulans asperthicin cluster were under-expressed ( S10 Table ) . We next examined the role of the recently identified SM regulator MtfA [29 , 30] in A . fumigatus and A . nidulans by performing RNA sequencing and differential gene expression analysis of ΔmtfA vs WT strains of both species ( A . fumigatus tTDS4 . 1 ΔmtfA [30] and CEA10 , A . nidulans TRVp ΔmtfA and TRV50 . 2 [29] ) . In contrast to our findings with veA , we found an approximately 10-fold difference in the percentage of genes regulated in both species ( Tables 1 , S5 ) . Thirty-six genes were over-expressed ( 0 . 4% ) and 63 ( 0 . 6% ) were under-expressed in the A . fumigatus ΔmtfA vs WT analysis , whereas in the A . nidulans ΔmtfA vs WT analysis 400 genes were over-expressed ( 3 . 7% ) and 568 were under-expressed ( 5 . 3% ) . To determine the functional categories impacted by mtfA deletion both species , we performed GO term enrichment analysis on the genes differentially expressed between ΔmtfA and WT strains . Under-expressed as well as over-expressed genes in A . nidulans were significantly enriched for secondary metabolic process , toxin metabolic process and oxidoreductase activity , suggesting that MtfA is involved in positive and negative regulation of different secondary metabolites ( Fig . 3 ) . Over-expressed genes in A . nidulans were also significantly enriched for asexual developmental processes , namely developmental process and asexual sporulation . Under-expressed genes in A . fumigatus were significantly enriched for two of the three processes as in A . nidulans , namely secondary metabolic process and oxidoreductase activity . However , over-expressed genes in A . fumigatus were not significantly enriched for any GO terms; some over-expressed genes were present in the secondary metabolic process term , though this was not statistically significant ( S6 Table ) . Examination of the 317 A . fumigatus SM cluster genes and 498 A . nidulans SM cluster genes showed that they too were also differentially expressed in both A . fumigatus and A . nidulans ΔmtfA mutants compared to the WT strains . In A . nidulans , 107 of the 498 SM cluster genes ( 21 . 5% ) were under-expressed and 32 ( 6 . 4% ) were over-expressed ( S5 Table ) . In some SM gene clusters , such as in the dba cluster , all genes were under-expressed in ΔmtfA A . nidulans relative to WT ( S10 Table ) . In contrast , many fewer SM cluster genes were differentially expressed in A . fumigatus ΔmtfA; 32 out of the 317 genes ( 10 . 1% ) were under-expressed and 4 ( 1 . 3% ) were over-expressed ( S5 Table ) . Despite this much smaller number of differentially expressed SM genes , at least one entire SM gene cluster ( the pseurotin A cluster ) was under-expressed in A . fumigatus ΔmtfA relative to ST ( S10 Table ) . To examine whether gene conservation and involvement in a biological process ( secondary metabolism and development ) correlated with conservation of regulation by VeA and MtfA , we tested whether orthologous and species-specific genes in A . nidulans and A . fumigatus showed the same or different responses in ΔveA vs WT and ΔmtfA vs WT analyses ( Fig . 4 ) . Even though both species contain large numbers of species-specific genes in their SM gene clusters , the proportions of differentially expressed SM cluster genes in both ΔveA A . nidulans and ΔveA A . fumigatus were remarkably similar ( Figs . 1A , 4A; S5 Table ) . Examination of SM cluster genes that are differentially expressed suggests that even when genes in SM clusters have orthologs , these orthologs were often differentially regulated by ΔveA . Entire SM pathways that are over- or under-expressed in one species may have orthologs with completely different expression patterns in the other species , as in the case of the dba SM gene cluster in A . nidulans ( S2 Fig ) . Specifically , of the 184 under-expressed SM cluster genes in ΔveA A . nidulans , 64 genes ( 34 . 8% ) had an ortholog in A . fumigatus ( S7 Table ) . Of these 64 genes , 45 ( 70 . 3% ) had at least one differentially expressed ortholog in A . fumigatus , and 37 ( 57 . 8% ) had at least one similarly under-expressed ortholog in A . fumigatus ( S7 Table ) . Fewer SM genes were over-expressed in either ΔveA A . nidulans or A . fumigatus; of the 67 over-expressed genes in A . nidulans , only 14 ( 20 . 9% ) had orthologs in A . fumigatus . Of these 14 genes , 5 had at least one differentially expressed ortholog in A . fumigatus , and 3 had at least one similarly over-expressed ortholog in A . fumigatus ( S7 Table ) . When mtfA was deleted , fewer SM cluster genes were differentially expressed in A . fumigatus than in A . nidulans . Of the 107 under-expressed genes in ΔmtfA A . nidulans , 36 ( 33 . 6% ) had an ortholog in A . fumigatus ( Fig . 4B , S7 Table ) . Unlike veA , however , only 6 of these conserved genes had differentially expressed orthologs in A . fumigatus . Finally , of the 32 over-expressed genes in ΔmtfA A . nidulans , 2 of the 12 genes with orthologs in A . fumigatus had orthologs that were differentially expressed . Apart from their involvement in the global regulation of SM , both VeA and MtfA are also involved in the regulation of asexual and sexual development . In contrast to genes involved in SM , genes involved in asexual and sexual development in Aspergillus have been shown to be highly conserved across the genus [55] . Of the 490 genes annotated to the GO term developmental process in A . nidulans , 462 have at least one ortholog among the 478 genes annotated to this term in A . fumigatus . In ΔveA A . nidulans , 72 developmental genes are under-expressed and 32 are over-expressed . Of the 72 under-expressed genes , 66 ( 91 . 7% ) have an ortholog in A . fumigatus; of these 66 orthologs , 30 were differentially expressed in both species ( Fig . 4C , S7 Table ) . There were fewer over-expressed developmental genes in ΔveA A . nidulans , but they showed similar trends; 31 of the 32 over-expressed genes have an ortholog , 15 of which had differentially expressed orthologs in A . fumigatus and 11 of which had over-expressed orthologs . In contrast with veA , many more developmental genes were differentially expressed in A . nidulans ΔmtfA ( 35 ) than in A . fumigatus ΔmtfA ( 1 ) . While 4 of the 6 under-expressed genes and 28 of the 29 over-expressed genes in A . nidulans had orthologs in A . fumigatus , none of these orthologs were differentially expressed ( Fig . 4D , S7 Table ) .
All but one of the SM gene clusters present in the four Aspergillus species we examined were species-specific . Even when we used a very low threshold of 50% evolutionary conservation , we found that there were no clusters conserved in all four species , one cluster conserved in three species , and a very small number conserved between pairs of species ( Fig . 1A ) . Consistent with our results , an examination of the close relatives A . fumigatus , A . fischerianus , and Aspergillus clavatus , using an 80% threshold of evolutionary conservation , also found relatively small numbers of conserved SM gene clusters [56] . Remarkably , variation in SM gene cluster content can sometimes be even strain-specific; for example , a single SM gene cluster has been shown to vary in its presence in A . fumigatus isolates [57] , and several SM gene clusters also appear to vary between isolates of A . niger [58] . Surveys of fungal genomes show that for any given fungus there are many more gene clusters than known SMs , suggesting that the currently characterized SMs might be only a small fraction of the SMs that a fungus can produce [56 , 59] . For example , 33 of the 37 putative SM gene clusters in A . fumigatus have no characterized products , despite evidence from metabolomics surveys suggesting that the fungus produces many SMs [19 , 45 , 60] . Interestingly , most known SMs are produced by only a handful of species [16] and SM profiles between closely related species are quite distinct [17 , 61] , both indications that the typical taxonomic distribution of SMs is narrow . Thus , the observed extremely high degree of taxonomic narrowness of SM gene clusters is not only consistent with the distribution of known SMs , but also suggests that this distribution is likely to be typical of all the SMs produced by a fungus . SM gene clusters are largely species-specific , so one hypothesis is that their constituent genes are also species-specific . Another alternative is that their constituent genes are conserved but have been reshuffled and become members of different SM gene clusters in different species . Our analysis shows support for both hypotheses . For example , we found that a much larger fraction of the genes comprising SM gene clusters is species-specific than of genes participating in primary metabolism ( Fig . 2 ) , which is likely explained by extensive gene duplication and loss [35 , 62 , 63] de novo gene emergence [64] , horizontal gene transfer [65–69] , as well as by very high sequence divergence . However , a considerable fraction of SM genes was not species-specific . In several cases , the orthologs of SM genes in one Aspergillus species were found residing in non-homologous SM pathways in another species; for example , the 25 genes in the sterigmatocystin gene cluster in A . nidulans have orthologs in 25 distinct SM gene clusters in the other three species ( Fig . 1B ) . Surprisingly , the orthologs of many genes that resided within an SM gene cluster in one species were not in an SM gene cluster in another species ( Fig . 2 ) . The majority of these orthologs either lack annotation or are annotated as being involved in primary metabolism , consistent with a model in which the gene content of SM gene clusters is formed or altered through the recruitment and incorporation of native genes involved in primary metabolism and other essential cellular processes . By comparing genome-wide gene expression of deletion mutants of veA and mtfA with wild-type strains in both A . fumigatus and A . nidulans we assessed the degree to which their regulatory roles in controlling secondary metabolism and development are conserved ( Fig . 4A and 4C ) . The involvement of VeA in regulating both processes is highly conserved in both A . fumigatus and A . nidulans . Remarkably , however , the downstream SM genes regulated by VeA are different between the two species; this difference is not fully accounted by the fact that VeA regulates many species-specific SM genes . Many genes that are regulated by VeA in one species are present but are not regulated by VeA in the other ( Fig . 4A ) . This conservation in regulatory logic ( i . e . , conservation in the regulation of secondary metabolism ) despite the dramatic change in the genes involved may be explained by considering the number and complexity of VeA’s interacting partners as well as VeA’s putative transcription factor function , which in combination offer many degrees of freedom for changes in the regulation of specific SM genes and pathways in either species . Specifically , VeA is known to have many interacting partners [33] including the Velvet family protein VelB , which it transports from the cytoplasm to the nucleus , where both proteins form a trimeric complex with LaeA that regulates secondary metabolism production and development [24] . Furthermore , VeA also functions outside of this complex [24 , 28]; it interacts with red light-sensing proteins in the nucleus [27] and with other methyltransferases [70] , and it has been hypothesized that it may also act as a scaffold protein for the recruitment of additional transcriptional regulators [33 , 36] . Finally , recent analysis has shown that the Velvet domain is a DNA-binding domain , and that Velvet family proteins may act as direct transcriptional regulators [71] . The involvement of MtfA in regulating SM in both A . fumigatus and A . nidulans is also conserved , although the numbers of SM genes that are under its control differ considerably between the two species . As with veA , comparing expression patterns between ΔmtfA and WT identified differentially expressed species-specific and conserved SM genes . Again like veA , many conserved genes that are regulated by MtfA in one species are not regulated in the other ( Fig . 4B ) . This rewiring could either be due to a divergence in the signal that targets these genes for regulation by MtfA , a C2H2 zinc finger transcription factor [29] , or its interacting partners . Although not much is known about MtfA’s interacting partners , indirect support for their presence comes from our finding that genes involved in SM are significantly over-represented in both the under-expressed and the over-expressed gene sets in A . nidulans ( Fig . 3 ) , suggesting that the regulatory effect of MtfA in certain regulatory partnerships is positive and in others negative . Interestingly , our results also suggest that MtfA genetically interacts with ( and acts downstream of ) VeA in A . nidulans , but not in A . fumigatus [28–30] , as its expression is decreased in ΔveA vs WT in A . nidulans but not in ΔveA vs WT in A . fumigatus ( S5 Table ) . Finally , MtfA appears to regulate development in the homothallic A . nidulans , but not in the heterothallic A . fumigatus ( Figs . 3 , 4 ) . Unlike veA , deleting mtfA influenced the expression of developmental genes only in A . nidulans but not in A . fumigatus . This is consistent with the loss in A . fumigatus or gain in A . nidulans of the signal that directs MtfA or its downstream targets to regulate developmental processes . VeA’s central role in coordinating SM and development under dark conditions as well as its large number of interacting partners likely explain in part why many more genes are differentially expressed in its absence in both A . nidulans and A . fumigatus than in the absence of MtfA . Gaining a more complete understanding of the molecular mechanisms through which VeA and MtfA regulate downstream targets is an interesting line of future inquiry that harbors significant promise in elucidating how VeA and MtfA have been rewired in different fungal species . However , it is likely that VeA and MtfA globally regulate SM and developmental processes not only directly but also indirectly , through interactions with other regulatory proteins; thus , obtaining a full mechanistic understanding of the rewiring of this regulatory circuit will likely also require characterization of VeA and MtfA’s interacting partners and their regulatory functions . It is abundantly clear that changes in transcriptional regulation , also known as rewiring , are a major driver of phenotypic divergence [40 , 43] . It is also becoming increasingly clear that rewiring of regulatory circuits can also take place in the absence of phenotypic change [42 , 72 , 73] . Irrespective of whether it leads to phenotypic change or not , rewiring is typically thought to occur either through changes to the transcriptional regulators or through changes to the regulatory signals between the regulators and the target genes . An example of rewiring due to changes to the transcriptional regulator is offered by the galactose pathway , which is controlled by Gal4p in the baker’s yeast Saccharomyces cerevisiae , but by the non-related Cph1p in the human commensal Candida albicans [37 , 74] . The acquisition or loss of specific motifs in otherwise conserved transcriptional regulators can also lead to rewiring [75 , 76] , as can acquisition or loss of interacting regulatory partners [41] . An example of changes in regulatory signals is offered by the small percentage of Mcm1 regulated genes shared between S . cerevisiae and C . albicans [77] , which can be largely attributed to high rates of cis-regulatory sequence gain and loss that Mcm1 binds to . This is likely to be the major explanation in cases that involve conserved regulators , conserved target genes , and conserved phenotypes [78 , 79] . Although the evidence for the regulatory impact of VeA and MtfA on secondary metabolism and development is genetic in our case ( i . e . , it is not known whether VeA or MtfA directly bind to target gene cis-regulatory elements or whether they control their expression indirectly ) , the evolution of VeA and MtfA regulation on development is consistent with rewiring that involves changes either to the transcriptional regulator or the regulatory signals between the regulator and the target genes ( Fig . 5A , B ) . In the case of VeA , the rewiring is not associated with phenotypic change–genes involved in development are regulated by VeA in both A . fumigatus and A . nidulans , whereas in the case of MtfA the rewiring might be associated with the differences in sexual and asexual development between the two species [55 , 80] . In contrast to the rewiring examples mentioned so far , the major driver of change in the evolution of VeA and MtfA regulation on SM appears to be the dramatic turnover in the target genes involved in the process , which has resulted in their high degree of species specificity and their even higher species-specific presence in SM gene clusters ( Figs . 2 and 5C , D ) . Although it seems reasonable to exclude the transcriptional regulators as potential co-drivers of the rewiring because the key domains of both VeA and MtfA are conserved across Aspergillus [29 , 36] , it is possible that the regulatory signals or the differential presence or absence of interacting regulatory partners may have also contributed to rewiring . However , given the vital ecological importance of secondary metabolites in fungal ecology , it seems reasonable to hypothesize that the evolution of the regulatory circuit governing secondary metabolism in filamentous fungi is primarily driven by the likely extreme evolutionary pressure imposed on each fungus to produce its own unique blend of secondary metabolites .
All genome sequences and annotations for A . nidulans FGSC A4 s10-m02-r03 , A . fumigatus AF293 s03-m04-r11 , A . oryzae RIB40 s01-m08-r21 and A . niger CBS 513 . 88 s01-m06-r10 were taken from the Aspergillus Genomes Database ( AspGD ) [54] . Groups of orthologous genes ( orthogroups ) for these four genomes were taken from AspGD’s orthology assignments for 16 Aspergillus species , which were generated using a Jaccard clustering approach [81] . AspGD orthogroups contain groups of genes that are thought to have descended from the Aspergillus common ancestor; genes from the same species that are part of a given orthogroup are defined as in-paralogs that have duplicated at some later point after the species diverged from the Aspergillus common ancestor . Species-specific genes , which were absent from AspGD orthogroups , were organized into species-specific orthogroups using the MCL algorithm in combination with all-versus-all protein BLAST search [82] . Proteins with BLAST hits with 60% query and subject coverage , an e-value of less than 1e-5 , and a percent identity of greater than 60% were subsequently clustered in MCL with an inflation parameter of 2 and were considered species-specific orthogroups . Proteins that did not pass the BLAST cutoffs were considered single-gene , species-specific orthogroups . Genes involved in secondary metabolism were taken from a previous study that expertly annotated secondary metabolic gene clusters in the four species under study [45] . Manually curated gene cluster boundaries were used when available . Primary metabolism genes were annotated using a previously described enzyme classification pipeline which utilizes KEGG Enzyme Commission annotations [62] . Genes involved in development were determined from all genes in A . fumigatus and A . nidulans annotated to the GO term developmental process ( GO:0032502 ) in AmiGO [83] . This data was accessed on 2014–07–19 . The strains used in this study include A . fumigatus CEA10 , TSD1 . 15 ( ΔveA ) and TTDS4 . 1 ( ΔmtfA ) [30 , 51] and A . nidulans TRV50 . 2 [52] , TXFp2 . 1 ( ΔveA ) generated in this study , and TRVpΔmtfA [29] . Deletion and wild-type strains presented isogenic genetic backgrounds , and all strains used in this study are prototrophs . Many A . nidulans studies have used a veA partial deletion [84] . For the present study we generated a strain with a complete deletion of the veA coding region , TXFp2 . 1 ( ΔveA ) . This strain was constructed as follows . First , The veA deletion cassette was obtained by fusion PCR as previously described [85] . A 1 . 4 kb 5’ UTR and a 1 kb 3’ UTR veA flanking regions were PCR amplified from wild type FGSC4 genomic DNA with primers veA_comF and AnidveA_p2 , and ANVeASTagP3 and ANVeASTagP4 primers sets , respectively ( S8 Table ) . The A . fumigatus pyrG ( pyrGA . fum ) selectable marker was amplified with AnidveA_p5 and ANVeASTagP6 primers from plasmid p1439 . The 5’ and 3’ UTR fragments were then PCR fused to pyrGA . fum to generate the veA replacement construct using primers AnidveA_P7 and AnidveA_P8 . The deletion cassette was transformed into A . nidulans RJMP1 . 49 strain [86] . The resulting colonies were then transformed with the pSM3 plasmid containing the A . nidulans pyroA to generate a prototroph with a ΔveA background . This strain was confirmed by DNA analysis and designated as TXFp2 . 1 . All strains were grown in liquid stationary cultures in Czapek-Dox medium ( Difco ) in the dark . The experiments were carried out with two replicates . After 72 hours of incubation at 37°C mycelial samples were harvested , immediately frozen in liquid nitrogen and lyophilized . Total RNA was isolated from lyophilized mycelia using the directzol RNA MiniPrep Kit ( Zymo ) according to the manufacturer’s instructions . RNA then was quantified using a nanodrop instrument . Expression patterns of veA and mtfA were verified in the A . fumigatus and A . nidulans wild types as well as in the deletion mutants by qRT-PCR prior to RNA sequencing ( S1 Fig ) , confirming the absence of transcripts in the deletion mutants . RNA-Seq libraries were constructed and sequenced at the Vanderbilt Technologies for Advanced Genomics Core Facility at Vanderbilt University using the Illumina Tru-seq RNA sample prep kit as previously described [28 , 48 , 50] . In brief , total RNA quality was assessed via Bioanalyzer ( Agilent ) . Upon passing quality control , poly-A RNA was purified from total RNA and the second strand cDNA was synthesized from mRNA . cDNA ends were then blunt repaired and given an adenylated 3’ end . Next , barcoded adapters were ligated to the adenylated ends and the libraries were PCR enriched , quantified , pooled and sequenced an on Illumina HiSeq 2500 sequencer . Two biological replicates were generated for each strain sequenced . Raw RNA-seq reads were trimmed of low-quality reads and adapter sequences using Trimmomatic using the suggested parameters for single-end read trimming [87] . After read trimming , all samples contained between 20–30 million reads . The smallest sample contained 19 . 9 million and the largest contained 30 . 6 million reads; the average sample contained 24 . 0 million reads ( S9 Table ) . Trimmed reads were aligned to A . nidulans and A . fumigatus genomes using Tophat2 using the reference gene annotation to guide alignment and without attempting to detect novel transcripts ( parameter—no-novel-juncs ) [88] . Reads aligning to each gene were counted using HTSeq-count with the intersection-strict mode [89] . Differential expression between ΔveA vs WT and ΔmtfA vs WT strains of A . fumigatus and A . nidulans were determined using the DESeq2 software , which normalizes read counts by library depth [90] . Genes were considered differentially expressed if their adjusted P-value was less than 0 . 1 and their log2 fold change was greater than 1 or less than-1 . GO term enrichment was determined for over- and under-expressed genes in all four conditions tested ( A . nidulans and A . fumigatus ΔveA vs WT and ΔmtfA vs WT ) using the Cytoscape plugin Bingo [91 , 92] . To allow for a high-level view of the types of differentially expressed gene sets , the Aspergillus GOSlim term subset developed by AspGD was used . The Benjamani-Hochberg multiple testing correction was applied , and terms were considered significantly enriched if the adjusted P-value was less than 0 . 05 . Fisher’s exact tests were performed using the R function fisher . test with a two-sided alternative hypothesis [93] . P-values were adjusted for multiple comparisons using the R function p . adjust with the Benjamini-Hochberg multiple testing correction [94] . Figures were created using the R plotting system ggplot2 [95] and circos [96] .
|
Filamentous fungi produce a highly diverse cadre of secondary metabolites , small molecules whose potent toxic activities are integral to the fungal lifestyle . Most secondary metabolites are narrowly taxonomically distributed , whereas the transcriptional regulators that control their production , alongside with controlling other key processes such as development , are broadly conserved . To gain insight into the evolution of the regulatory circuit governing secondary metabolism and development , we examined the evolution of the genes and pathways underlying these processes as well as the evolution of their transcriptional regulation in the filamentous fungal genus Aspergillus , a prolific and important producer of secondary metabolites . We discovered that although secondary metabolic genes and pathways are poorly conserved across Aspergillus , their regulation is largely conserved . In contrast , the regulation of highly conserved developmental genes is divergent . These results point to a new type of rewiring that has occurred during the evolution of the regulatory circuit governing secondary metabolism and development in Aspergillus in which conserved regulators control a conserved biological process ( secondary metabolism ) , even though the underlying genes and pathways that make up the biological process are not themselves conserved .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
Examining the Evolution of the Regulatory Circuit Controlling Secondary Metabolism and Development in the Fungal Genus Aspergillus
|
Piwi-interacting RNAs are a diverse class of small non-coding RNAs implicated in the silencing of transposable elements and the safeguarding of genome integrity . In mammals , male germ cells express two genetically and developmentally distinct populations of piRNAs at the pre-pachytene and pachytene stages of meiosis , respectively . Pre-pachytene piRNAs are mostly derived from retrotransposons and required for their silencing . In contrast , pachytene piRNAs originate from ∼3 , 000 genomic clusters , and their biogenesis and function remain enigmatic . Here , we report that conditional inactivation of the putative RNA helicase MOV10L1 in mouse spermatocytes produces a specific loss of pachytene piRNAs , significant accumulation of pachytene piRNA precursor transcripts , and unusual polar conglomeration of Piwi proteins with mitochondria . Pachytene piRNA–deficient spermatocytes progress through meiosis without derepression of LINE1 retrotransposons , but become arrested at the post-meiotic round spermatid stage with massive DNA damage . Our results demonstrate that MOV10L1 acts upstream of Piwi proteins in the primary processing of pachytene piRNAs and suggest that , distinct from pre-pachytene piRNAs , pachytene piRNAs fulfill a unique function in maintaining post-meiotic genome integrity .
Piwi-interacting RNAs ( piRNAs ) are a diverse class of gonad-specific small interfering RNAs that bind to members of the Piwi subfamily of Argonaute proteins . One common function of piRNAs in all species studied so far is the silencing of transposable elements , which is essential for the protection of genome integrity during germ cell development [1]–[3] . Distinct from miRNAs and siRNAs in origin , length , structure , and biogenesis , piRNAs are generated by dicer-independent processing of long precursor transcripts , however , the precise mechanisms of their biogenesis remain largely unclear [4] , [5] . In mice , the Piwi family has three members: Miwi ( Piwil1 ) , Mili ( Piwil2 ) , and Miwi2 ( Piwil4 ) . These Piwi genes exhibit different developmental expression patterns in testis . While Miwi2 is expressed in fetal and perinatal germ cells [6] , the expression of Miwi is restricted to pachytene spermatocytes and round spermatids in adult testes [7] . Mili is expressed from the fetal germ cell stage onwards through the round spermatid stage [8] . Two developmentally distinct populations of piRNAs are expressed in mouse male germ cells at the pre-pachytene and pachytene stages . Pre-pachytene piRNAs are mostly derived from transposable elements and are associated with MILI and MIWI2 in fetal and perinatal male germ cells [6] , [9] , [10] . Pachytene piRNAs originate from ∼3000 genomic clusters [11] and bind to both MILI and MIWI [12]–[17] . Interestingly , more than 90% of MILI- and MIWI-bound pachytene piRNAs shared identical 5′end sequences [18] . As a result , most MILI- and MIWI-bound pachytene piRNAs map to the same genomic clusters [18] . The biogenesis of piRNAs involves primary and secondary processing mechanisms [1] , [2] . Pre-pachytene piRNAs derive from precursor transcripts that are cleaved into putative primary piRNA intermediate molecules by a yet unknown primary processing mechanism , followed by loading onto MILI for further processing . In embryonic germ cells , the endonuclease ( slicer ) activity of MILI is required for the secondary piRNA processing mechanism , which amplifies MILI-bound piRNAs through an intra-MILI ping-pong loop and generates all MIWI2-bound secondary piRNAs [19] . In this feed-forward ping-pong model , Piwi proteins with piRNAs complimentary to retroelement-derived transcripts drive transcript cleavage and piRNA amplification [6] , [9] , [10] , [19] . In contrast , the biogenesis of pachytene piRNAs only engages the primary processing mechanism , i . e . the presumptive cleavage by an unknown nuclease and eventual processing of the precursor transcript into mature piRNAs [5] , [17] , [20] , [21] . Therefore , pachytene piRNAs provide a simple and ideal system for dissecting the mysterious primary processing mechanism in mammals [11] , [13]–[16] . We and others previously demonstrated that MOV10L1 , a putative RNA helicase , interacts with all mouse Piwi proteins and is required for biogenesis of pre-pachytene piRNAs [22] , [23] . MOV10L1 homologues are evolutionarily conserved among insects ( Armi in Drosophila melanogaster ) , plants ( SDE3 in Arabidopsis thaliana ) , and vertebrates ( MOV10 and MOV10L1 ) . Arabidopsis SDE3 is required for post-transcriptional gene silencing [24] . Drosophila Armi is essential for the maturation of RISC ( RNA-induced silencing complex ) and miRNA-mediated silencing [25] , [26] . Armi is also relevant to the piRNA pathway , evident from the loss of specific piRNAs and the activation of retrotransposons in armi mutants [27] , [28] . Specifically , Armi plays an essential role in the primary piRNA processing pathway [29] . In contrast to Drosophila and Arabidopsis with a single Mov10l1 homologue , the vertebrate genome encodes two genes ( Mov10 and Mov10l1 ) , which apparently arose by gene duplication . MOV10 is ubiquitously expressed and associates with Ago proteins , forming part of the purified human RISC [30] , [31] . Depletion of MOV10 in cultured cells leads to reduced miRNA-mediated silencing [30] . We initially identified MOV10L1 as a putative RNA helicase that is specifically expressed in mouse germ cells [32] , [33] . Disruption of Mov10l1 leads to meiotic arrest , de-repression of transposable elements , and depletion of both MILI- and MIWI2-associated perinatal piRNAs [22] , [23] . Apparently , MOV10 and MOV10L1 function in the miRNA and the piRNA pathway , respectively , due to specialization after gene duplication during vertebrate evolution . The existing piRNA pathway mouse mutants either fail to deplete all pachytene piRNAs or exhibit meiotic arrest prior to the pachytene stage , leaving the biogenesis and role of pachytene piRNAs largely unexplored . Inactivation of either Mili or Miwi2 causes postnatal meiotic arrest at the leptotene/zygotene stage in the male germline [8] , [34] . Similarly , other piRNA pathway mutants , such as Ddx4 ( Vasa ) , Mael , Gasz , Tdrd9 , Mov10l1 , and Mitopld , also exhibit early meiotic arrest in males [22] , [35]–[40] . Inactivation of Miwi leads to spermiogenic arrest at the round spermatid stage [7] . However , MILI-associated pachytene piRNAs are abundant in Miwi-deficient testes [17] , [18] . Therefore , a mouse mutant containing pachytene spermatocytes , but lacking all pachytene piRNAs ( both MILI- and MIWI-bound piRNAs ) has not been available to specifically study the function of pachytene piRNAs . In this study , we have specifically and completely depleted the pachytene piRNA population in the male germline of Mov10l1 mutant mice , uncovering a novel function for pachytene piRNAs in maintaining post-meiotic genome integrity .
MOV10L1 , a putative RNA helicase , interacts with all three mouse Piwi proteins , and is an essential component of the piRNA pathway [22] . To explore the biogenesis and function of pachytene piRNAs , we disrupted MOV10L1 function specifically during and after male meiosis using Cre-mediated inactivation of a conditional Mov10l1 allele ( Mov10l1fl ) ( Figure 1A and Figure S1 ) at the following stages: after postnatal day 7 ( Neurog3-Cre ) [41] , at the pachytene stage ( Hspa2-Cre ) [42] , and in post-meiotic spermatids ( Prm-Cre ) [43] . Cre-mediated recombination of the conditional Mov10l1 allele deletes the RNA helicase domain , producing a truncated protein termed MOV10L1Δ . In male Mov10l1fl/- Neurog3-Cre mice resulting from intercrosses of Mov10l1fl/fl mice with Neurog3-Cre mice [41] , Cre-mediated disruption of Mov10l1 was first detected in testes at postnatal day 9 ( leptotene/zygotene spermatocytes ) , with a decrease in the abundance of the full-length MOV10L1 protein in the mutant testes compared with the wild type ( Figure S2A ) . Mov10l1fl/- Neurog3-Cre males were sterile , with substantially smaller testes ( 140±10 . 5 mg/pair at 2–4 months of age ) compared to age-matched wild-type mice ( 189±18 . 4 mg/pair ) ( Student's t test , p<0 . 0008 ) . In contrast to seminiferous tubules from wild-type mice ( Figure 1B ) , tubules from Mov10l1fl/- Neurog3-Cre mutant mice lacked elongated spermatids , while earlier germ cell stages including pachytene spermatocytes and round spermatids were present ( Figure 1C ) . Acrosome staining with the anti-ACRV1 antibody identified spermiogenic arrest at the step 4 spermatid stage . Therefore , very different to the meiotic arrest observed in male germ cells with ubiquitous deletion of Mov10l1 [22] , [23] , postnatal disruption of Mov10l1 mediated by Neurog3-Cre causes post-meiotic spermiogenic arrest ( Figure 1C ) , revealing that MOV10L1 plays an essential role during the post-meiotic stages of spermatogenesis . To distinguish consequences of inactivation of MOV10L1 during the pachytene stage from those resulting from disruption at earlier stages such as in differentiating spermatogonia , we generated Mov10l1fl/- Hspa2-Cre mice in which Cre is expressed specifically in spermatocytes , particularly pachytene cells ( Figure 1A ) [42] . Deletion of MOV10L1 in Mov10l1fl/- Hspa2-Cre mice occurred by postnatal day 14 , apparent from a decrease in the abundance of the full-length MOV10L1 protein in the mutant testes ( Figure S2B ) . Notably , Mov10l1fl/- Hspa2-Cre males were also sterile . Although testes ( 159±24 mg/pair ) from 2–3 month old Mov10l1fl/- Hspa2-Cre mice were slightly smaller than those from Mov10l1+/− males ( 182±26 mg/pair ) ( Student's t test , p<0 . 2 ) , histological analysis revealed spermiogenic arrest at the round spermatid stage ( Figure 1D ) . The most advanced spermatids in Mov10l1fl/- Hspa2-Cre males were late round spermatids at step 8 . The arrest of spermiogenesis at early and late round spermatid stages in Mov10l1fl/- Neurog3-Cre and Mov10l1fl/- Hspa2-Cre mutant mice , respectively , demonstrates that MOV10L1 is required for the differentiation of post-meiotic germ cells . The temporal delay in the spermiogenic arrest in Mov10l1fl/- Hspa2-Cre testes is likely due to the late onset of Hspa2-Cre expression , which may allow residual MOV10L1 to persist longer . The round spermatid arrest in Mov10l1fl/- Neurog3-Cre and Mov10l1fl/- Hspa2-Cre testes could be due to disruption of MOV10L1 function during the pachytene stage of meiosis , or at early spermatid stages . To define the requirement for MOV10L1 more precisely , we disrupted Mov10l1 with Cre recombinase under the control of the protamine 1 ( Prm ) promoter , which is only expressed in post-meiotic spermatids [43] . Mov10l1fl/- Prm-Cre males exhibited normal fertility but a slight reduction in testis weight ( Table S1 ) . Histological analysis of testes from Mov10l1fl/- Prm-Cre males revealed normal spermiogenesis ( Figure S3 ) . These genetic studies demonstrate that disruption of MOV10L1 function at the pachytene stage causes spermiogenic arrest . Isolation and radiolabeling of total testicular small RNAs from adult Mov10l1fl/- Neurog3-Cre testes showed that mutant testes were devoid of pachytene piRNAs ( Figure 2A ) . Immunoprecipitation experiments further revealed that both MILI- and MIWI-associated pachytene piRNAs were absent in the mutant ( Figure 2B , 2C ) . As Mov10l1 mutant testes contained less MIWI protein than wild-type testes , we performed serial dilutions of immunoprecipitated complexes to rule out the possibility that the observed loss of MIWI-bound piRNAs was due to the detection limit of the assay . However , MIWI-associated piRNAs were detectable in wild-type testes even when MIWI protein was not detectable ( Figure 2C , lane 4 ) , indicating a specific depletion of pachytene piRNAs in the testes from Mov10l1fl/- Neurog3-Cre mice . Moreover , the abundance of pachytene piRNAs was sharply reduced in Mov10l1fl/- Hspa2-Cre testes ( Figure 2D ) . In addition , Northern blotting showed that individual pachytene piRNAs ( piR1 , piR2 , and piR3 ) were absent in testes from Mov10l1fl/- Neurog3-Cre mice and dramatically reduced in abundance in testes from Mov10l1fl/- Hspa2-Cre mice ( Figure 3 ) . As expected , the abundance of individual pachytene piRNAs was not affected in the testes from Mov10l1fl/- Prm-Cre mice ( Figure 3 ) . Therefore , MOV10L1 function is essential for the biogenesis of all pachytene piRNAs . Pre-pachytene piRNAs are present in mitotic germ cells such as spermatogonia ( Figure 1A ) . Because Neurog3-Cre initiated the disruption of Mov10l1 at post-natal day 9 , we anticipated that the production of pre-pachytene piRNAs would not be affected in Mov10l1fl/- Neurog3-Cre testes . To test this hypothesis , we performed immunoprecipitation of postnatal day 10 testis lysates with anti-MILI antibodies . Postnatal day 10 testes do not contain pachytene spermatocytes , and express only MILI but not other Piwi proteins ( Figure 1A ) . Consequently , all MILI-bound piRNAs in postnatal day 10 testes are pre-pachytene piRNAs [10] . We found that pre-pachytene piRNAs were present in postnatal day 10 Mov10l1fl/- Neurog3-Cre testes ( Figure 2E ) . Furthermore , Northern blot analysis showed that abundance of a specific pre-pachytene piRNA was not reduced in adult testes from Mov10l1 mutant mice , regardless of whether deletion had been mediated by Neurog3-Cre , Hspa2-Cre , or Prm-Cre ( Figure 3 ) . These data demonstrate that pre-pachytene piRNA production is not affected in the Mov10l1 conditional mutant testes . We next examined the consequences of the loss of pachytene piRNAs on the localization of piRNA pathway components such as MILI , MIWI , TDRD1 , and GASZ . In wild-type pachytene spermatocytes , these proteins localize to cytoplasmic nuage granules ( also called inter-mitochondrial cement ) ( Figure 4A , 4C , 4E , 4G ) [7] , [8] , [37] , [44] . Strikingly , in Mov10l1-deficient pachytene spermatocytes , these four proteins congregated to one extremely large novel perinuclear polar “granule” ( Figure 4B , 4D , 4F , and 4H ) . Further analyses revealed immunoreactivity of the polar granule to a cocktail of antibodies against mitochondrial proteins ( OXPHOS ) , demonstrating co-localization of mitochondria with MILI in polar granules of Mov10l1-deficient pachytene spermatocytes ( Figure S4 ) . Electron microscopy ( EM ) analysis confirmed that in Mov10l1-deficient pachytene spermatocytes , mitochondria form a single cluster ( Figure 4J ) , in contrast to their random distribution in wild-type pachytene cells ( Figure 4I ) . Consistent with a recently described role for MitoPLD , a mitochondrial surface protein , in the piRNA pathway [39] , [40] , these data strongly suggest a novel but yet unknown role for mitochondria in the biogenesis of pachytene piRNAs and/or a function for pachytene piRNAs in the cytoplasmic organization and distribution of mitochondria and piRNA pathway protein components . We next examined the status of chromatoid bodies , which are large and dynamic ribonucleoprotein aggregates prominent in haploid spermatids . Chromatoid bodies contain various RNA regulatory proteins as well as piRNA pathway components , but their precise function remains unclear [45] . Wild-type round spermatids contained one prominent chromatoid body , visualized by EM as a multi-lobular electron-dense nuage ( Figure S5 ) . In Mov10l1-deficient spermatids , however , the chromatoid body was fragmented ( Figure S5 ) . A similar fragmentation of chromatoid bodies has been observed in other mouse mutants of RNA processing pathway proteins with male infertility phenotype , implying importance of their structural integrity ( Miwi , Tdrd5 , and Tdrd6 ) [7] , [18] , [46] , [47] . The introduction of DNA double strand breaks ( DSBs ) into the germ cell genome takes place as part of the chromatin remodeling process occurring at the elongating spermatid stage ( Figure 5A ) . This chromatin remodeling process is initiated by the replacement of canonical histones first with transition proteins and eventually by protamines . Concurrently , nucleosomal DNA supercoils must be resolved , presumably by topoisomerase IIB ( TOP2B ) . TOP2B generates DNA double-strand breaks ( DSBs ) , relaxes supercoils , and subsequently religates DNA ends [48] . DSBs trigger a DNA damage response , resulting in the phosphorylation of histone H2AX ( γH2AX ) . In wild-type testis , histone H2AX phosphorylation is therefore detectable in several germ cell stages that undergo changes in their chromatin configuration , including elongating spermatids ( Figure 5B ) , but it is absent from round spermatids . Intriguingly , round spermatids from Mov10l1fl/- Neurog3-Cre testes exhibit a high degree of DNA damage visualized by γH2AX ( Figure 5C ) . This could be due to a developmental progression of piRNA-deficient round spermatids to the “elongating” spermatid stage without apparent morphological change . However , the absence of both TOP2B and PRM2 ( protamine 2 ) in γH2AX-positive round spermatids from Mov10l1 mutant testes indicated that these round spermatids were not undergoing chromatin remodeling , excluding that γH2AX-positivity was due to TOP2B activity ( Figure 5E , 5G ) . Secondly , DNA damage might be induced by de-repressed transposable elements active in piRNA-deficient round spermatids . Genetic studies have shown that the piRNA pathway is required for silencing of retrotransposons such as LINE1 and IAP in pre-pachytene germ cells [19] . However , quantitative RT-PCR analysis revealed no de-repression of LINE1 ( Figure 6B , 6C ) or IAP in Mov10l1fl/- Neurog3-Cre testes , confirmed by immunofluorescent analyses of testis sections with anti-LINE1 and anti-IAP antibodies ( Figure S6 ) . Therefore , pachytene piRNAs are not required for silencing of LINE1 and IAP retrotransposons , although we cannot rule out the possibility that other transposable elements might be de-repressed in Mov10l1-deficient round spermatids . Notably , we did not observe γH2AX foci in round spermatids from Rnf17-deficient mice , in which spermatogenesis is also arrested at the round spermatid stage [49] but piRNA biogenesis does not appear to be severely affected ( data not shown ) . These results suggest that the DSBs observed in round spermatids from Mov10l1fl/- Neurog3-Cre testes and Mov10l1fl/- Hspa2-Cre testes are not a direct consequence of their developmental arrest . Rather , these observations suggest that the piRNA pathway , specifically MOV10L1 and pachytene piRNAs , play a yet undefined role in maintaining genome integrity in post-meiotic round spermatids . We previously found that MOV10L1 interacts with all three Piwi proteins ( MILI , MIWI , and MIWI2 ) [22] . The low abundance of MOV10L1Δ in both Mov10l1−/− ( ubiquitous null mutant ) testes [22] and adult Mov10l1 conditional mutant testes ( Figure S2A ) precluded co-immunoprecipitation ( IP ) experiments to ascertain if deletion of the helicase domain affected interaction of the truncated protein with Piwi proteins in vivo . However , a peak expression of MOV10L1Δ protein in post-natal day 20 Mov10l1fl/- Neurog3-Cre testes , with a level highly exceeding that of the remaining wild-type MOV10L1 ( Figure S2A ) , allowed us to perform co-immunoprecipitation of day 20 testicular extracts with anti-MILI and anti- MOV10L1 antibodies ( Figure 6A ) . While wild-type MOV10L1 ( due to the lack of Neurog3-Cre expression in spermatogonia ) could be detected as a very faint band in the MILI immunoprecipitate as expected , the much more abundant MOV10L1Δ was absent ( Figure 6A , Lane 6 ) . Notably , MILI was not detectable in the MOV10L1/MOV10L1Δ immunoprecipitate , and the level of MIWI was extremely low ( Figure 6A , Lane 4 ) . These results suggest that , apart from its putative enzymatic activity , the RNA helicase domain of MOV10L1 is also essential for its association with MILI and MIWI , and that piRNA production could be affected by disruption of the MOV10L1-Piwi interactions . Pachytene piRNAs are derived from only one strand of genomic clusters [11] , [13]–[16] , prompting the hypothesis that a single long primary piRNA transcript is made from each cluster and is cleaved into intermediate RNAs by an unknown Dicer-independent mechanism [5] , [20] , [21] . Due to their large size and low abundance , detection of these precursor transcripts requires RT-PCR analysis , with the exception of a ∼10 kb piLR ( piRNA like small RNA ) transcript that can be visualized on Northern blots of testicular extracts [50] . As the depletion of pachytene piRNAs in Mov10l1fl/- Neurog3-Cre testes may be due to a blockade of pachytene piRNA precursor processing , we examined the abundance of precursors of four pachytene piRNAs ( piR1 , piR2 , piR3 , and piLR ) by RT-PCR assays ( Figure S7 ) . All four precursors accumulated substantially in Mov10l1fl/- Neurog3-Cre testes , at 8 to 20 fold increased levels ( Figure 6B , 6C ) . As expected , abundance of the pre-pachytene piRNA precursor ( cluster 10 ) [10] and the miRNA precursor Pri-let7g remained constant ( Figure 6B , 6C ) . These data suggest that MOV10L1 is required for the primary processing of precursor transcripts and thus plays an essential role in the early steps of the piRNA biogenesis pathway , i . e . primary processing and loading onto Piwi proteins ( Figure 6D ) .
We have identified MOV10L1 as the only factor known to date that is required for the production of all pachytene piRNAs in mouse . As the biogenesis of pachytene piRNAs only involves the primary processing pathway , our conditional Mov10l1 mutants provide a unique opportunity to delineate this enigmatic component of piRNA biogenesis in mammalian species . Presumably , long piRNA precursor transcripts are first cleaved into intermediate molecules , and then processed into mature piRNAs ( Figure 6D ) . Observations that the Drosophila Armi-Piwi-Yb complex is associated with a population of 25–70 nt piRNA intermediate-like ( piR-IL ) molecules support this hypothesis [20] . Furthermore , recent biochemical studies using silkworm ovarian cell lysate have shown that intermediate piRNA molecules with 5′ U are specifically loaded onto Piwi proteins and then trimmed from the 3′end to generate mature piRNAs [21] . Here , we show that , in the mouse male germline , postnatal disruption of Mov10l1 does not affect the expression of Piwi proteins ( MILI and MIWI ) but causes a complete loss of pachytene piRNAs , demonstrating that MOV10L1 functions upstream of Piwi proteins in the piRNA biogenesis pathway . Consistent with its homology to Drosophila Armi [25] , [26] , MOV10L1 is therefore a master regulator of piRNA biogenesis in mouse . This notion is further supported by the dramatic accumulation of pachytene piRNA precursors in the Mov10l1 mutant testes . As MOV10L1 interacts with Piwi proteins , we postulate that MOV10L1 may facilitate the loading of intermediate piRNA molecules onto the Piwi proteins in mouse ( Figure 6D ) . In spermatocytes , proteins of the piRNA pathway such as MILI , MIWI , TDRD1 , MAEL , and GASZ , localize to the nuage - inter-mitochondrial cement [7] , [8] , [36] , [37] , [44] , [51] . However , the functional significance of the physical association of nuage with mitochondria in germ cells is poorly understood . MitoPLD , a mitochondrial signaling protein , is essential for nuage formation and piRNA production , suggesting an important role for mitochondria in these mechanisms [39] , [40] . In this study , we find an unusual polar congregation of piRNA pathway proteins ( such as MILI , MIWI , TDRD1 , and GASZ ) . Similar to wild-type MOV10L1 , truncated MOV10L1Δ is distributed diffusely through the cytoplasm of pachytene spermatocytes; therefore the polar coalescence of the other piRNA pathway components in MOV10L1-deficient pachytene cells is likely caused by the absence of pachytene piRNAs . However , as the association of Piwi-MOV10L1 is disrupted in the Mov10l1 mutant , it is also possible that the localization of Piwi proteins and their interacting partners has become perturbed as a consequence of this disruption . The unusual polar congregation of piRNA pathway proteins with mitochondria in Mov10l1 mutant spermatocytes suggests that MOV10L1 and/or pachytene piRNAs are essential for nuage formation and proper mitochondria distribution . Consistent with such a role , we find that the chromatoid body , a prominent nuage in spermatids , is fragmented in pachytene piRNA-deficient mutant cells . This previously unknown role in organelle distribution shows that pachytene piRNAs are intricately integrated in the inter-dependent relationships among piRNA production , nuage formation , and mitochondria organization that are essential for male germ cell maturation . A recent study has shown that MIWI is an RNA-guided RNase with slicer activity that directly cleaves transcripts of the LINE1 retrotransposon [18] . Miwi-deficient and MiwiADH ( slicer inactive ) mutant testes , in which MIWI is either absent or lacks slicer activity , exhibit substantial accumulation of LINE1 transcripts and protein . In Mov10l1fl/- Neurog3-Cre testes , however , LINE1 RNA levels are not affected . One possible explanation for these differential effects on LINE1 abundance could be that , in Mov10l1fl/- Neurog3-Cre testes , MIWI is catalytically intact and may function as a slicer through pachytene piRNA-independent mechanisms . Moreover , MIWI directly binds to spermiogenic mRNAs , independent of piRNAs [17] . Although previous genetic studies of piRNA pathway mutants show that perturbation of pre-pachytene piRNAs causes meiotic arrest and de-repression of LINE1 and IAP retrotransposons , the functions of pachytene piRNAs have remained elusive . Our study on the role of Mov10l1 and the piRNA pathway during later stages of meiosis and spermiogenesis demonstrates that pachytene piRNAs fulfill distinct and essential functions during post-meiotic stages of male germ cell development . Most importantly , the massive DNA damage observed in piRNA-deficient round spermatids in the absence of de-repression of LINE1 and IAP transposable elements suggests that the integrity of the post-meiotic germ cell genome remains highly prone to damage , and that pachytene piRNAs fulfill a protective role at this stage by yet undefined mechanisms .
Mice were maintained and used for experimentation according to the guidelines of the Institutional Animal Care and Use Committee of the University of Pennsylvania . Neurog3-Cre , Hspa2-Cre , and Prm-Cre mice were purchased from the Jackson Laboratory ( Stock numbers: Neurog3-Cre , 006333; Hspa2-Cre , 008870; Prm-Cre , 003328 ) . Mov10l1fl/fl mice were generated previously [22] . Genotyping for Mov10l1 and Cre alleles was performed separately on genomic DNA isolated from tails . The anti-MOV10L1 antibody was generated previously [22] . Other antibodies used were MILI ( Abcam ) , MIWI ( Abcam , or gifts from R . Pillai ) , GASZ ( M . M . Matzuk ) , LINE1 ORF1p ( S . L . Martin ) , IAP ( B . R . Cullen ) , TDRD1 ( S . Chuma ) , TOP2B ( Santa Cruz Biotechnology ) , PRM2 ( SHAL ) , and ACTB ( Sigma-Aldrich ) . Mouse testicular extract preparation , immunoprecipitation , and 5′ end-labeling of piRNAs were performed as described previously [22] . Antibodies were described previously [22] . Northern blot analyses were performed as previously described with modifications [14] . Total RNAs were isolated from mouse testes using Trizol reagent , separated by 15% denaturing polyacrylamide gel , and electro-blotted onto GeneScreen Plus hybridization membrane . Membranes were UV crosslinked and hybridized with 32P end-labeled oligonucleotide probes in Ultrahyb Oligo Buffer ( Ambion Cat#8663 ) at 42°C . Probes for detecting pachytene piRNAs , a pre-pachytene piRNA , or microRNA were perfectly complementary to their sequences: probe-piR1: AAAGCTATCTGAGCACCTGTGTTCATGTCA; probe-piR2: ACCAGCAGACACCGTCGTATGCATCACACA; probe-piR3: ACCACTAAACATTTAGATGCCACTCTCA; probe-let7g: TACTGTACAAACTACTACCTCA; pre-pachytene piRNA probe ( derived from sense SINE B1 ) : 5′-TGGCTGTCCTGGAACTCACTYTGT [10] . After hybridization , membranes were washed three times at 42°C in 2×SSC buffer containing 0 . 5% SDS , or stripped by boiling in 0 . 1×SSC containing 0 . 1% SDS . Membranes were exposed to a phosphor imager screen for autoradiography . For histology , testes were fixed in Bouin's solution overnight , processed , sectioned , and stained with hematoxylin and eosin . Immunofluorescence was performed on frozen sections of testes fixed in 4% paraformaldehyde as previously described [52] . EM followed a standard protocol used at the Electron Microscopy Resource Laboratory of the University of Pennsylvania . PCR primers for piRNA precursor transcripts were chosen from genomic clusters to which each piRNA was mapped [10] , [14] , [50] . PCR primers and PCR product sizes are listed in Table S2 .
|
Small non-coding RNAs play critical roles during development and in disease . The integrity of the germline genome is of paramount importance to the wellbeing of offspring and the survival of species . Piwi-interacting RNAs ( piRNAs ) are a class of small non-coding RNAs abundantly expressed in the gonad . Compared to microRNAs and small-interfering RNAs ( siRNAs ) , the biogenesis and function of piRNAs remain poorly understood . Here we have identified MOV10L1 , a putative RNA helicase , as a master regulator of piRNA biogenesis in mouse . We find that production of pachytene piRNAs requires MOV10L1 . Blockade of pachytene piRNAs disrupts germ cell development and results in defects in post-meiotic genome integrity . Therefore , mutations in MOV10L1 and other piRNA pathway components may contribute to male infertility in humans .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"developmental",
"biology",
"biology"
] |
2012
|
Blockade of Pachytene piRNA Biogenesis Reveals a Novel Requirement for Maintaining Post-Meiotic Germline Genome Integrity
|
Recent advances in top-down mass spectrometry enabled identification of intact proteins , but this technology still faces challenges . For example , top-down mass spectrometry suffers from a lack of sensitivity since the ion counts for a single fragmentation event are often low . In contrast , nanopore technology is exquisitely sensitive to single intact molecules , but it has only been successfully applied to DNA sequencing , so far . Here , we explore the potential of sub-nanopores for single-molecule protein identification ( SMPI ) and describe an algorithm for identification of the electrical current blockade signal ( nanospectrum ) resulting from the translocation of a denaturated , linearly charged protein through a sub-nanopore . The analysis of identification p-values suggests that the current technology is already sufficient for matching nanospectra against small protein databases , e . g . , protein identification in bacterial proteomes .
When Church et al . [1] proposed to use nanopores for sequencing biopolymers , they had envisioned both DNA and proteins sequencing . However , the progress in protein sequencing turned out to be much slower since it is more difficult to force proteins through a pore systematically and measure the resulting signal [2] . These difficulties underlay the experimental and computational challenges of Single Molecule Protein Identification ( SMPI ) . Nanopores promise single molecule sensitivity in the analysis of proteins , but an approach for the identification of a single protein from its nanospectrum has remained elusive . The most common approach to nanopore sequencing relies on the detection of the ionic –current blockade signal ( nanospectrum ) that develops when a molecule is driven through the pore by an electric field . Preliminary work [3 , 4] was limited to analyzing protein conformations in pure solutions rather than identifying proteins in a mixture . Subsequent steps demonstrated that nanopores can detect protein phosphorylations [5] as well as conformations and protein-ligand interactions [6] . Recent studies on combining nanopores with aptamers have shown limited success for protein analysis [7] . Proposals for electrolytic cell with tandem nanopores and for single molecule protein sequencing have been made , but not yet implemented [8–11] . Recently , the sequence of amino acids in a denatured protein were read with limited resolution using a sub-nanometer-diameter pore , sputtered through a thin silicon nitride membrane [12] . Protein translocations through the pore modulated the measured ionic current , which was correlated with the volumes of amino adids in the proteins . However , the correlation was imperfect , making it difficult to solve the problem of reconstructing a protein from its nanospectrum with high fidelity . Developing computational and experimental methods for analyzing nanospectra derived from a electrical signals that produced when a protein translocates through a sub-nanopore could enable a real-time sensitive approach to SMPI that may have advantages over top-down mass spectrometry for protein identification . Despite difficulty and expense ( requiring especially powerful magnets ) to implement it , top-down mass spectrometry has been used in a few labs around the world to identify intact proteins and their proteoforms . However , it is about 100-fold less sensitive than bottom-up mass spectrometry , which can be used to detect attomoles of material [13] . In stark contrast , a sub-nanopore has been used to discriminate residue substitutions in a single molecule with low fidelity [12] . Similar to mass-spectrometry , where de novo protein sequencing ( based on top-down spectra ) remains error-prone [14 , 15] , the challenge of de novo deconvoluting nanospectra into amino acids sequences of proteins is currently unsolved . However , protein identification based on top-down spectra ( i . e . , matching a spectrum against all proteins in a protein database ) is a well-studied topic . For example , top-down protein identification tools ProsightPC [16] and MS-Align+ [17] reliably identify proteins , report p-values of resulting Protein-Spectrum Matches ( PrSMs ) , and even contribute to improving gene annotations by discovering previously unknown proteins [18] . In this paper , we describe the first algorithm for protein identification based on nanospectra derived from current blockades associated with denaturated , charge linearized translocation of protein through pores with sub-nanometer diameters . Our Nano-Align algorithm matches nanospectra against a protein database , identifies Protein-Nanospectrum Matches ( PrNMs ) , and reports their p-values . Our analysis revealed that the typical p-values of identified PrNMs vary from 10−4 to 10−6 , which is already sufficient for a limited analysis of nanospectra against small bacterial proteomes . The software is publicly available at http://github . com/fenderglass/Nano-Align .
The details regarding the experiments and methods used to acquire electrical current blockade signals from the translocation of single protein molecules through sub-nanopores have been described elsewhere [12] . To summarize , first , a pore with sub-nanometer cross-section was sputtered through thin silicon nitride membrane supported on a silicon chip using a tightly focused , high-energy electron beam in a scanning transmission electron microscope ( Fig 1 ) . The thickness of the membranes ranged from 8 to 12nm . Then the silicon chip supporting the membrane was embedded in a multiport microfluidic device that allowed for independent electrical access to the cis and trans-sides of the sub-nanopore by two Ag/AgCl electrodes . To perform electrical measurements , the sub-nanopore was immersed in 0 . 2 − 0 . 3 M NaCl and a transmembrane voltage in the range between 300 − 700 mV was applied . The resulting pore current was measured using an Axopatch 200B amplifier controlled with Clampex 10 . 2 software . Finally , recombinant denatured protein , along with 2 ⋅ 10 − 3% sodium dodecyl sulfate that imparted a nearly uniform negative charge to the protein , were added to the microfluidic reservoir ( c . a . 20 fmoles of protein ) and subsequently blockades in the open pore current associated with single molecules translocating through the pore were observed . It was determined that a lower transmembrane bias voltage improved the signal-to-noise ratio ( SNR ) and lengthened the median duration of the blockades , but it also increased the propensity for the pore to clog . Multi-level events associated with residual native protein structure or multiple molecules competing for the pore were occasionally observed , but were manually culled from the data pre- analysis [12] . Five proteins were analyzed by measuring the blockade currents through sub-nanopores: a recombinant chemokine CCL5 of length 68 AAs; two variants of the H3 histone designated as H3 . 2 and H3 . 3 , which consist of the chain of 136 AAs , differing only by residue substitutions at positions 32 , 88 , 90 and 91; a tail peptide of the H3 histone ( residues 1-20 ) and a fourth histone , H4 of length 103 AAs . More details about the datasets are given at the ‘Datasets’ section below . When a single molecule of protein translocates through the sub-nanopore , its amino acids block the flow of ions , causing a change in the open pore current Iopen . The fraction of occupied pore volume Vmol/Vpore ( where Vpore and Vmol are volumes of the pore and molecule inside this pore , respectively ) was assumed to be proportional to the fractional blockade current , which is calculated as |I − Iopen|/Iopen , where I is the raw current during the translocation . The raw signal measurements from the pore were pre-processed as follows: first , the discretized pore signal , sampled at 250 kHz , was split into the separate blockades , each one representing a translocation of a single protein ( Fig 2 ) Only events with sufficient duration to detect single-AA duration features were selected . Typical blockade duration analyzed here ranged from 1 to 20 milliseconds , as shorter times did not permit accurate discrimination of intra-event features due to the measurement bandwidth . The mean fractional blockade current varied from 0 . 05 to 0 . 5 for different nanospectra . Recorded signals exhibited fluctuations that were associated with different structural features of a protein translocating through the pore . Since the electrolytic current through the pore is associated with the occupied pore volume , one of the major factors that influences the signal is the volume of amino acids that occupy the sub-nanopore near the waist [19] . The estimates of amino acid volumes were obtained from crystallography data [20] . Since the pore can simultaneously accommodate multiple amino acids , it was assumed that the fluctuations in a blockade were proportional to a linear combination of amino acids volumes in the pore waist . In particular , we found that the mean volume of amino acids yielded a good approximation of the empirical signal values . Thus , given a protein P of length |P| , we split it into overlapping windows of size k ( or k-mers ) and generate a theoretical nanospectrum MV ( P ) as a vector of dimension |P| + k − 1 by taking the average volume of |P| − k + 1 k-mers and extra 2 * ( k − 1 ) shorter prefix and suffix substrings from the beginning and end of a protein . These extra prefix and suffix substrings correspond to the start and the end of a translocation , when the pore is occupied by less than k amino acids . For example , for k = 3 , the “protein” KLMNP results in a vector of length seven corresponding to the following substrings: K , KL , KLM , LMN , MNP , NP , and P . Experimental analysis of peptides with post-translational modifications and mutations [12] revealed changes in the specific regions of the recorded signal traces , that corresponded to approximately four amino acids in length . In addition , simulations of the electric field in a 0 . 5x0 . 5 nm2 diameter , 8 nm thick pore in an SiN membrane indicated that the vast majority of the field was confined within 1 . 5 nm of the pore near the waist at the center of the membrane , which gives roughly the same estimate of the number of amino acids . Thus , the Mean Volume ( MV ) model assumes that each fluctuation in the blockade current corresponds to a read of a quadromer ( short prefixes and suffixes of a protein correspond to shorter mers ) , which results in the best fit ( among all reasonable values of k ) with experimental nanospectra . Generally , the MV model results in theoretical nanospectra correlated with the empirical data . The mean Pearson product-moment correlation coefficient between a consensus of experimental nanospectra ( an average of multiple protein translocations , as described below ) and the corresponding MV model was ranging from 0 . 25 to 0 . 45 for various datasets . However some regions show large deviations between theoretical and experimental nanospectra , which may be associated with additional attributes such as hydrophilicity or charge . In particular , our analysis revealed that such discordant regions were enriched with small amino acids , which have volumes below the median value ( see Fig 3 ) for illustration and ‘Characterizing errors of the models’ section below for the detailed discussion ) . Since we acquired multiple nanospectra originating from multiple known proteins , an alternative approach for generating theoretical nanospectra was to use a supervised learning paradigm . We used a Support Vector Regression ( an SVM-based regressor ) to establish the correspondence between a k-mer inside the pore and a signal it generates [21] . Given an empirical nanospectrum E recorded from a protein P , we tiled P into overlapping quadromers qi and discretized E into |P| + 3 points . Thus , each qi had an associated experimental signal value ei . Next , the feature space of the model has to be defined . Following the ideas of the MV model , it is natural to assume that blockade current is affected by the composition of amino acids in a quadromer , rather than their order ( however , the dependence might be non-linear ) . As many of the 20 proteinogenic amino acids have similar volumes , we partitioned them into four volume groups ( Fig 4 ) and defined a feature vector fi of a quadromer qi as the composition of amino acids from each group ( as a tuple of length four ) . For example , an amino acid quardromer GQLD has zero amino acids from Large group ( > 0 . 2nm3 ) , two from Intermediate group ( between 0 . 15 and 0 . 2 nm3 ) , one from Small group ( between 0 . 11 and 0 . 15nm3 ) and one from Minuscule group ( < 0 . 11nm3 ) , and is converted to a feature vector ( 0 , 2 , 1 , 1 ) . This choice of the feature space reduced the overfitting effect and increased coverage of the training dataset ( there are only 35 distinct quadromer compositions in the defined feature space versus 204 = 160 000 amino acid quadromers ) . Using a set of pairs ( fi , ei ) we trained an SVR regressor with the Radial Basis Function kernel ( implemented in an open-source library libsvm [22] ) . The Support Vector Regression ( SVR ) model takes a peptide P as input and outputs an SVR-based theoretical nanospectrum SVR ( P ) ( Fig 3 ) . The mean Pearson correlation coefficient between the theoretical and empirical nanospectra ( consensus ) for the SVRmodel was varying from 0 . 38 to 0 . 68 for different datasets , confirming the improvement over the MV model . The parameters of the SVR model were chosen through cross validation experiments and are equal to C = 1000 , γ = 0 . 001 , and ϵ = 0 . 01 . The analysis of error patterns of the SVR model revealed a bias in the signal estimation that was correlated with the hydrophilicity of the amino acids ( see ‘Characterizing errors of the models’ section ) . Also , Bhattacharya et al . [23] recently reported that water molecules affect the signal of DNA translocating through the nanopore since hydrophilic amino acids are more likely to acquire a water molecule and change the effective volume [24] . Thus , it is desirable to include amino acid hydrophilicity into the model . Motivated by these finding , we explored an alternative approach for supervised learning by using the Random Forest ( RF ) regression [25 , 26] for theoretical nanospectra generation . In comparison to the SVR model , the resulting Random Forest ( RF ) model is more robust to outliers and exhibit less overfitting [27] , which allowed us to use the volumes of all 20 amino acids as features . According to this RF model , each quadromer qi from the training set is converted to a feature vector fi , where each element of the vector is a pair of volume and hydrophilicity of the corresponding amino acid . We used an open source implementation of the Random Forest regressor from Scikit-learn package [28] to build the described model . The model performed well on the training sets , but the accuracy was poor on the test proteins with different amino acid sequence and composition . This was mainly caused by the fact that only a few among all possible amino acid quadromers were observed in the training sets . However , under assumption that nanopore current does not depend on the order of amino acids , it is possible to significantly expand the training sets by randomly permuting amino acids within quadromers . Specifically , prior to model we randomly permuted each fi vector , leaving the same corresponding qi value . This dataset expansion significantly improved the performance of the RF model on to training testing datasets . See Fig 3 for examples of theoretical nanospectra in the MV , SVR , and RF models . Given an experimental nanospectrum S and a protein P , we transformed S into a vector S → by splitting S into |P| + 3 regions and taking the average value inside each of them . The vector S → was then normalized by subtracting the mean and dividing by the standard deviation . Under the hypothesis that P has generated S → , we estimated the proportion of explained variance by computing R2 coefficient of determination between S → and the model output . Given a database of proteins DB , a protein P ( S , DB ) is defined as a protein with the maximum R2 against S among all proteins from DB . A pair formed by the protein P ( S , DB ) and the nanospectrum S defines a putative Protein-Nanospectrum Match ( PrNM ) . Single protein correlation analysis indicated that proteins were correlated more with themselves on average ( Fig 5 ) . In contrast , we did not observe such correlation in the open pore current , indicating that there is an inherent signal in blockades . However , electrolytic current through the pore is affected by many factors , such as uncorrelated time-dependent fluctuations in the ionic current and electrical instrument noise , which results in noisy nanospectra . Averaging multiple nanospectra from the same protein resulted in significant noise reduction and increased accuracy of PrNM identification . This effect is similar to improvements in peptide identifications that are achieved by clustering of mass spectra in traditional proteomics [29 , 30] . Typically , clustering of 5 − 10 nanospectra results in a consensus nanospectrum that significantly improves the signal-to-noise ratio over a single nanospectrum ( the mean Pearson correlation coefficients between theoretical and empirical nanospectra increased 1 . 5—2-fold for various datasets ) . Since each of the existing datasets of nanospectra originated from a single pure protein , we randomly partitioned the dataset of nanospectra into clusters and performed identification of consensus nanospectra instead of a single nanospectrum . In traditional proteomics , the precursor mass assists top-down protein identification since it greatly reduces the computational space that has to be searched in the protein database . Likewise , information about the protein length would be very useful for SMPI , but estimating the protein length based on a nanospectrum originating from a sub-nanopore is a non-trivial problem since the existing experimental protocol does not control the translocation speed that may vary widely as evident from the blockade duration . Our analysis revealed that protein translocations modulate the blockade current , which was captured by the measurements . Each blockade , associated with the translocation of a protein showed a characteristic number of fluctuations during the duration of the blockade . It turned out that the fluctuation frequency ( described below ) was correlated with the protein length and the other features , such as amino acid composition . We explored a possibility of the separation of a sample of nanospectra into clusters corresponding to different proteins . From a sample of different proteins , we estimated the fluctuation frequency of each nanospectrum as the number of peaks ( local maximums ) divided by the duration of the blockade . The distribution of fluctuation frequencies ( Fig 5b ) revealed that each protein in our datasets has a characteristic peak in the distribution . To separate the nanospectra into clusters based on the fluctuation frequency one can apply the Gaussian Mixture model to estimate the protein lengths from nanospectra and to improve the efficiency of SMPI . Analyzing a mixture of multiple proteins is conceptually harder than analyzing the existing experimental datasets of nanospectra that all originated from pure protein solutions . Since it is unknown what protein gives rise to what nanospectrum in a mixture , it is difficult to cluster nanospectra for a reliable identification . Further , orientation of each molecule must be deduced prior to clustering since each protein can translocate through the pore in two different directions . However , it is possible to cluster nanospectra based on their estimated fluctuation frequency to differentiate proteins with different lengths . As multiple proteins may have a similar length , it is important to further split some length-based clusters into finer protein-based clusters . We believe , that this could be done by applying clustering algorithms which automatically estimate the number of clusters ( e . g . Affinity Propagation [31] ) . Evaluating the results of clustering in the case of complex mixtures was problematic since all available experimental datasets of nanospectra were generated from the pure protein solutions .
We benchmarked Nano-Align using nanospectra from five short human proteins: H3 . 2 , H3 . 3 , H4 , CCL5 and H3 tail peptide ( Table 1 ) . The nanospectra from H3 . 2 , H3 . 3 and H4 were acquired using the two similar pores whereas the nanospectra for CCL5 and H3 tail were acquired using two different pores with different sizes . The proteins were split into three pairs: ( CCL5 , H3 tail ) , ( H4 , H3 . 2 ) and ( H3 . 3 , H3 . 2 ) . For each pair of proteins , the SVR and RF models were trained using the protein with higher number of nanospectra and the accuracy of identifications was estimated using the other protein from the pair . The first two pairs represented proteins that were very different in both length and amino acid composition , thus minimizing the overfitting effect . The third pair represented highly similar proteins , that only differ in four amino acids . To evaluate the accuracy of SMPI , we constructed decoy protein database for each dataset from the correct protein and randomly generated proteins of the same length and amino acid composition as the correct protein . The size of decoy database varied from 105 to 5 ⋅ 106 for different datasets , depending on the identification accuracy and the number of nanospectra in the dataset . The p-value of a PrNM was approximated as the percentage of proteins from the database scoring higher than the correct protein against the given nanospectrum . Below we show results for the SVR and RF models only since they turned out to be significantly more accurate than the MV model for all datasets . Fig 6 shows median p-values for SVR and RF models as a function of the number of nanospectra in a cluster . As expected , both models showed the improvement in the accuracy with the increase in the cluster size . The p-values for the pair ( CCL5 , H3 tail ) were high for both models ( 0 . 03–0 . 05 for a consensus of size 10 ) . However , the dataset ( H4 , H3 . 2 ) showed a significant improvement for the RF model ( p-values of the order of 10−4 for a consensus of 10 nanospectra ) , while the accuracy of the SVR model was comparable to the previous dataset . Finally , the RF model showed high accuracy on ( H3 . 3 , H3 . 2 ) dataset , with p-values below 10−5 for the nanospectra clusters of size five . The RF model consistently outperformed the SVR model on the datasets that were generated using pores of similar sizes , which suggests that the decision trees are better suited for SMPI due to their robustness against outliers . Also , amino acid hydrophilicity proved to be a valuable predictor of the pore signal . The RF model performed slightly worse than the SVR model on the dataset generated using two different pores , suggesting that it is more sensitive to the experimental conditions . The fact that the RF model performed better on the proteins that were more similar to the training proteins is not surprising , but rather highlights the importance of choice of the training set , which should have substantial coverage of the data . Additionally , we benchmarked the RF model performance using a database containing real human proteins . We extracted all proteins of length between 100 and 160 from the human proteome ( about 20% of the human proteome ) and performed the identification of H3 . 3 spectra against this reduced database . On average , the true protein was ranked five against all other proteins ( for a cluster of size five ) . An example of database hits is given in the Table 2 . Interestingly , all high-scoring proteins belong to H3 histone family and differ by only few amino acids . While the search space was artificially reduced , this experiment already provides a justification for analysis of unknown nanospectra against small bacterial proteomes , after further improvements in the protein length estimation discussed above . For each of the three models ( MV , SVR and RF ) we measured the bias with respect to different features of amino acids . Using H3 . 2 dataset ( that provides the best amino acid coverage ) we calculated the signed error defined as the mean difference between the empirical and theoretical nanospectra . For each amino acid , the signed error was measured among the associated quadromers . Fig 7 shows the volume-related bias of the MV model . This bias could be explained by the fact that larger amino acids have more influence on the pore signal than smaller amino acids . The SVR model and RF model show no bias with respect to amino acid volumes . A similar analysis revealed a bias with respect to amino acid hydrophilicity in the SVR model . The MV model did not show a clear dependence , possibly due to the dominant effect of the volume bias . The RF model showed no statistically significant bias related to hydrophilicity .
We presented the first algorithm for Single Molecule Protein Identification using a signal generated by a protein translocation through a sub-nanopore . We also proposed three models for generating theoretical nanospectra and concluded that the Random Forest model results in the most accurate identifications . The typical estimated p-values of identification accuracy were ranging from 10−4 to 10−6 , which is already sufficient for a limited analysis of nanospectra against small bacterial proteomes containing a few thousands proteins . The comparison of algorithm performance on different datasets suggests that the model sensitivity will further improve when more nanospectra originated from different proteins become available . Cysteine ( Cys ) was the highest source of error in all three models for H3 . 2 . Likewise , Cys was an above average source of error in CCL5 [12] but , it was a below average source of error in the similar sequence of CXCL1 . Thus , it seemed unlikely that only the size affects the error . On the other hand , both Cys and Met , which exhibit higher number of prediction errors are at the high end of the hydropathy index and have only few waters ( 4 and 10 , respectively ) binding them [32] , which may indicate that water affects the blockade current . In addition , it has been speculated that charge could also affect the duration and magnitude of a blockade [12 , 33] . Whereas it seems likely that both charge and water play a role in the blockade current , measurements and the MV model testing these ideas have been inconclusive so far . While SMPI is currently not in a position to compete with top-down proteomics , this technology is still in its infancy . Furthermore , due to the inherent single molecule sensitivity , there are several avenues of research that can be addressed uniquely by SMPI that offer protein-discrimination from very small samples ( attomoles ) . Thus , SMPI has a potential to emerge as a new technology for accurate protein identification .
|
Protein identification is the key step in many proteomics studies . Currently , the most popular technique for intact protein analysis is top-down mass spectrometry which recently enabled high-throughput identification of many proteins and their proteoforms . However , this approach requires large amounts of materials and is currently limited to short proteins , typically less than 30 kDa . On the other hand , nanopore sensors promise single molecule sensitivity in protein analysis , but an approach for the identification of a single protein from its blockade current ( nanospectrum ) has remained elusive , since the signal from the sensors relates to the amino acid sequence of the protein in a poorly understood way . In this work we describe the first algorithm for protein identification based on nanospectra associated with translocation of proteins through pores with sub-nanometer diameters . While identification accuracy currently does not allow reliable processing of complex protein samples yet , we believe , that the rapidly improving experimental protocols along with the new computational algorithms will transform into a viable protein identification approach in the near future .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"protein",
"transport",
"sequencing",
"techniques",
"cell",
"processes",
"dna-binding",
"proteins",
"materials",
"science",
"molecular",
"biology",
"techniques",
"materials",
"by",
"structure",
"research",
"and",
"analysis",
"methods",
"amino",
"acid",
"analysis",
"proteins",
"biological",
"databases",
"proteomics",
"recombinant",
"proteins",
"histones",
"molecular",
"biology",
"molecular",
"biology",
"assays",
"and",
"analysis",
"techniques",
"biochemistry",
"proteomic",
"databases",
"cell",
"biology",
"proteomes",
"database",
"and",
"informatics",
"methods",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"mixtures",
"protein",
"sequencing"
] |
2017
|
Single-molecule protein identification by sub-nanopore sensors
|
Quantitative trait locus ( QTL ) analysis is a powerful tool for mapping genes for complex traits in mice , but its utility is limited by poor resolution . A promising mapping approach is association analysis in outbred stocks or different inbred strains . As a proof of concept for the association approach , we applied whole-genome association analysis to hepatic gene expression traits in an outbred mouse population , the MF1 stock , and replicated expression QTL ( eQTL ) identified in previous studies of F2 intercross mice . We found that the mapping resolution of these eQTL was significantly greater in the outbred population . Through an example , we also showed how this precise mapping can be used to resolve previously identified loci ( in intercross studies ) , which affect many different transcript levels ( known as eQTL “hotspots” ) , into distinct regions . Our results also highlight the importance of correcting for population structure in whole-genome association studies in the outbred stock .
Quantitative trait locus ( QTL ) analysis has been the primary tool for geneticists to study complex genetic traits in experimental organisms . However , while such QTL mapping has great power to identify loci controlling the traits , resolution of mapping is usually quite low and as a result few candidate genes have been successfully identified using this approach . The use of molecular phenotypes , in particular gene expression levels , as quantitative traits for mapping , coupled with the ability to measure thousands of such traits simultaneously , has added a tremendous spark to the field of complex trait genetics . The integration of expression QTL ( eQTL ) with complex clinical traits using statistical modeling has allowed the identification of genes and pathways involved in a variety of complex traits . Some of the recent successes of this integrative approach have been identification of causal genes underlying the QTL for clinically relevant trait [1]–[3] , the identification of genomic loci regulating the expression of biological pathway genes[4] , the identification of genomic hotspots harboring master regulators [5]–[7] , and prioritization of candidate genes underlying physiological trait QTLs [8] . Moreover , mathematical models have been developed to construct gene expression networks [9] , [10] , deduce the causal relationship between different components of the network [11] , and understand the transcriptional regulation of the genes [12] . Despite these successes , such integrative genomic approaches using F2 populations suffer from the same limitation that has hindered the success of the traditional physiological trait QTL mapping , namely lack of resolution in mapping [13] . To overcome the lack of resolution problem , Flint and colleagues recently investigated the use of outbred stocks of mice to simultaneously detect and fine map physiological trait QTLs [14]–[16] . In the first of the two recent studies , they used 790 outbred mice ( MF1 ) to study the genetics of behavioral traits and successfully mapped three QTLs within a 1cM region [14] . In the second study , the authors extended this approach to multiple traits and mapped 97 metabolic and human disease related phenotypes to intervals of 2 . 8 Mb ( average 95% confidence interval ) by using over 2000 heterogeneous stock mice [15] . The success of these studies prompted us to investigate the potential use of outbred mice for eQTL studies , where many validated quantitative trait genes for expression traits have been identified . In this report , we present the results of a whole genome association study for the liver gene expression profiling of 110 MF1 mice and compare the results obtained in this population with previously published linkage studies in F2 mice [17] .
A total of 110 outbred MF1 mice were studied for whole genome transcript levels in liver and subjected to genotyping using the Affymetrix 5K Mouse Chip . From the 5024 SNPs on this array , 1813 SNPs ( about one third of total SNPs ) had a minor allele frequency of 5% or greater , and were used for the analyses described below . The average and median distance between neighboring SNPs for these 1813 markers were 1 . 38 Mb and 0 . 57 Mb respectively . To determine the percent coverage of the genome by these 1813 SNPs , we used the average distance of the adjacent markers which had an r squared value of 0 . 9 or higher . This analysis showed that on average adjacent markers will have such a high LD when within 1 . 03Mb . Based on this , approximately 3000 informative evenly spaced markers are needed to allow coverage of the entire genome . This calculation is based on the assumption that the haplotype blocks in this population are all about the same size . However , as we show below , this is not true for some regions of the genome , and thus the estimate is very approximate . Given the non-uniform distribution of the 1813 markers across the genome the genetic coverage for these set of markers is 72% of the genome . This means that in the whole genome association analysis described in this report we would expect to miss 28% of the signals . Gene expression measurements were performed on RNA isolated from liver using Illumina's mouse whole genome expression BeadChip ( MouseRef-8-v1 Expression BeadChip ) ( see Materials and Methods ) . We applied two filtering criteria to the 24048 probes on the microarray . The first filtering criteria was based on the detection p-value calculated for each probe ( see Materials and Methods ) . This filtering step eliminated any probe with low signal which could be due to nonspecific hybridization . The second filtering criterion was based on the recent report by Walter et al [18] which they showed that for the Affymetrix platform the presence of SNP within the 25mer probe sequence may affect the hybridization of transcripts and lead to artifactual detection of local ( cis ) eQTL . To investigate if this applies to the 50mer probe sequences of the Illumina microarrays used in the current study , we examined the degree of enrichment of SNPs in probes with local eQTL vs probes with no local eQTL . We found that from the 10765 probes with reliable signal and unique genomic location 602 probes had a local eQTL and from these 105 contained at least one SNP ( as determined from the Perlegen SNP database ) within their probe sequences ( 17% ) . In contrast , from the 10163 probes with no evidence for local eQTL 647 ( 6% ) contained one or more annotated SNPs . This means that the proportion of probes with SNPs in the sequence is significantly higher for probes determined to have local eQTL vs probes with no local eQTL ( chi squared statistics p-value for such enrichment was <10−16 ) . These results suggest that , as with Affymetrix arrays [18] , the presence of SNP within the probe sequence on the Illumina microarray might result in artifactual local eQTL detection . Therefore , to overcome such a bias we excluded any probe with one or more SNPs in its sequence . The two filtering criteria combined yielded 10013 probes ( from the original 24048 ) which we used as a starting set for the whole genome association analyses described below . We first computed the degree of linkage disequilibrium in the population of 110 MF1 mice . Pair-wise r2 were calculated among all the SNPs , and the average r2 measures for different ranges were used to look at the LD structure in the population . Visual inspection of the LD between markers within the same chromosome revealed a complex LD structure in the population ( Figure 1A and Figure S1 ) . In particular , it was evident that the extent of LD varied in different regions of the genome . Moreover , although for most regions highest LD was between adjacent markers , in some cases non-adjacent markers showed a higher LD than adjacent ones . To quantify the extent of LD between the markers in this population , we created 100 kb bins of various distances between marker pairs and calculated the average r2 for each bin . As shown in Figure 1B , the average r2 dropped with increasing distance between markers . The average r2 for markers within 2 Mb of each other was 0 . 58 , for markers between 2 to 5 Mb was 0 . 30 , and for markers 5Mb or more apart was 0 . 04 , suggesting that the extensive LD exists over several Mb . For markers located on different chromosomes , the LD was low ( the mean r2 was 0 . 015 and the median was 0 . 008 ) . Despite this low r2 , inspection of the distribution of the chi-squared statistics p-values for expected r2 in absence of LD indicated significant LD between certain pairs . These observations were consistent with our visual inspection of LD maps suggesting the existence of a complex relationship pattern among different loci , presumably due to population substructure within the MF1 stock . To further investigate this we performed hierarchical clustering of mice based on the kinship matrix which we derived from the overall correlation of genotypes between pairs of mice . The results revealed clear evidence of familial clustering ( Figure 1C ) indicating differences in relatedness . In addition , several multi-leveled larger clusters were observed with weaker levels of similarity , suggesting a complex genetic relatedness between the samples . The potential confounding effect from population structure was further supported by the fact that a very large number of expression levels are significantly explained by genetic relatedness between individuals . Using variance component test and at the 5% FDR level , 30 . 2% ( 3027 ) of transcripts were significantly associated with genetic background while only 1 . 5% ( 151 ) are expected by chance at the same threshold . In addition , for 19 . 8% ( 1985 ) of transcripts , more than 50% of variance was explained solely by the genetic background effect . This indicated that correcting for population structure is essential to avoid larger numbers of false positives . In order to correct for population structure we used Efficient Mixed Model Association ( EMMA ) . The underlying statistical algorithm for performing such correction has recently been published [18] and is briefly explained in the Materials and Methods section . In summary , EMMA controls for population structure and familial relatedness by modeling the gene expression on two terms ( plus the random error term ) : one is the SNP genotype and the other is a term which takes into account the population structure . This term , which is estimated based on the genetic similarity of mice in the population , essentially captures the variance attributable to population structure and provides a better estimate of SNP effect and its significance on gene expression . Without partitioning this term , the variance due to the genetic structure will be falsely attributed to the SNP and might result in a false positive association signal . To examine the effects of correcting for familial structure on the genome wide association , we compared the results of the association before and after correcting for population structure using a linear additive model and a linear mixed model . Table 1 shows the results for various FDR thresholds and for both local ( primarily cis ) and distant ( trans ) eQTL . Before correction , regression analysis of gene expression on markers revealed a total of 812 significant associations ( at FDR of 1% , corresponding to the p-val of 2 . 29e-06 ) . From these , 444 ( 55% ) were local and 368 ( 45% ) were distant eQTL . After correcting for population structure , there were about two thirds as many significant associations as found originally ( 478 significant associations at the FDR of 1% corresponding to the p-value of 1 . 05e-06 ) . This result suggested that about one third of the associations found in the absence of correction were false positives due to the relatedness of the mice . The reduction in significant associations among local and distant eQTL , however , was not the same . For local eQTL , there were 18% fewer associations after correction ( 366 vs 444 ) , but for the distant eQTL there was a 70% reduction in the number of associations found ( 112 vs 368 ) . We also examined the results at 5% and 10% FDR thresholds ( Table 1 ) and as with the 1% FDR , we observed a similar pattern where the total number of associations was less after correction and distant eQTL were more affected by this correction than local eQTL . This inflation of p-values resulting from the population structure was also evident from the pattern and number of significant p-values for each transcript ( Figure S2 ) . One of the limitations of using outbred stock for mapping complex traits has been the statistical power issue and the need to include large number of mice in the study [13] . In addition , the presence of population structure between the animals can also have a negative impact on the statistical power . To assess the power of the current study , we performed power calculations under various genetic background ( population structure ) effects ( Figure 2A and Figure S3 ) . As shown in Figure 2A , for minor allele frequency of 0 . 3 , the average minor allele frequency in our population , and the genetic background effect of 0 . 3 at the 10% FDR level ( p-value = e-05 ) , this study has over 60% power to detect QTL typical of what is expected from local eQTL ( 30% variance explained ) as estimated from intercross data ( unpublished data ) . For distant eQTL , however , where the effects are smaller ( typically less than 10% of variance explained ) at the same FDR level the use of 110 related mice will have relatively small power ( <20% ) . These results imply that for eQTL described below the local eQTL detected reflect the majority of true local eQTL present in the population and for distant eQTL there may be a significant number associated with type I and/or type II errors . We next examined the eQTL structure in the MF1 mice . As shown in Table 1 ( and Table S1 ) , 1196 eQTL had significant association at the 10% FDR ( p-value of 2 . 43e-05 ) after correcting for population structure , which greatly exceeded the 119 expected by chance . Among the 1196 eQTL , 24 were due to different probes of the same gene mapping to the same location . This reduced the number of unique eQTL for each gene to 1172 . From these , 471 were classified as local or cis eQTL ( defined as the association peak marker for transcript of a gene mapping to within 10Mb of the physical location of the gene itself ) and 701 were classified as distant or trans eQTL ( The significant results for the whole genome association analysis at various FDRs can be found in Table S1 and the top 10 most significant distant eQTL at 10% FDR are shown in Table 2 ) . From the 1172 eQTL , which belonged to 1093 genes , there were a total of 1019 single gene associations , 69 genes had two associations , and 5 genes had more than two associations . From the 69 genes with two associations , 27 genes had one local and one distant eQTL and 42 genes had two distant eQTL . In general , the p-values for local eQTL ( mean–log of p-value = 11 . 5 ) were more significant than the p-values for distant eQTL ( mean–log of p-value = 5 . 4 ) . Local eQTL are very likely to be due to a variation either within the gene or in the regulatory region in close proximity to the gene . To investigate the resolution achieved in the MF1 data we calculated the distance of the association peak marker to the physical location of the gene for each local eQTL . This analysis revealed that the median distance of the peak markers from the physical location of the gene was 0 . 67 Mb and for 25% of the genes the peak marker was located less than 300 Kb away from the gene itself ( Figure 2B ) . We also searched for the presence of co-localizing distant eQTL ( eQTL “hotspots” ) . To do this , the entire genome was divided into 2 Mb bins ( 1287 total bins ) and the number of significant distant eQTL were counted in each bin . Plotting of the eQTL frequencies at various genomic regions indicated a non-random distribution of the mapping locations ( Figure 2C ) . Several ‘hotspots’ were identified , with the most highly enriched loci on Chromosome 1 ( 17 eQTL ) , Chromosome 4 ( 55 eQTL ) , Chromosome 7 ( 25 eQTL ) , and Chromosome 16 ( 15 eQTL ) . To assess the validity of these hotspots , we randomly grouped mice into two subsets and reanalyzed each subset for the presence of co-localizing distant eQTL . We repeated this procedure four times to test for the preservation of these four highly enriched hotspots in each of the subsets ( Figure S4 ) . From these four hotspots , the Chromosome 4 hotspot was present in 7 of the 8 subsets created . Chromosomes 1 was replicated 4 times , Chromosomes 7 replicated three times , and Chromosome 16 replicated twice in the subsets analyzed . For Chromosomes 1 , 7 , and 16 , the inconsistency in replication could either be due to an artifact of population structure not accounted for by our correction method [20] or to lack of power resulting from doing the analysis on half as many animals as in the original analysis . The presence of hotspots is consistent with the notion that the causal genetic variant is located within a master regulator of gene expression for group of genes . To identify candidate master regulator genes for each of these hotspots , we searched for local eQTL within each region . On Chromosome 1 we found no local eQTL; on Chromosome 16 we found one local eQTL homogentisate 1 , 2-dioxygenase ( Hgd ) ; on Chromosome 7 we found 6 local eQTL including the enhancer binding protein CCAAT/enhancer binding protein alpha ( Cebpa ) , lipolysis stimulated lipoprotein receptor ( Lisch7 ) , peptidase D ( Pep4 ) , coiled-coil domain containing 123 ( Ccdc123 or 2610507L03Rik ) , androgen regulated gene RP2 ( Nudt19 or D7Rp2e ) , a Rho GTPase binding protein rhophilin 2 ( Rhpn2 ) ; and in the most enriched hotspot on Chromosome 4 we only found one local eQTL methylthioadenosine phosphorylase ( Mtap ) . Since MF1s offer a higher mapping resolution , this resource may be used to fine map eQTL previously identified in other crosses . The limitation is that not all the eQTL found previously will be present in the MF1 population . Here we sought to compare the mapping results in the MF1 population by empirically estimating the fraction of eQTL detected in this study compared to what was found in a previously published cross from our laboratory . For this comparison we used the previously reported eQTL study of the liver tissue for the F2 intercross population generated between C57BL/6J . ApoE−/− and C3H/HeJ . ApoE−/− parental strains ( herein referred to as the BxH cross ) [17] . In order to make a direct comparison we used the Entrez-Gene accession IDs to map the probes across the Illumina and the Agilent microarrays . From the 10013 probes used in the MF1 genome wide association analysis 8437 had unique Entrez-Gene IDs , 8036 of which were also represented by one or more probes on the Agilent microarrays used in the BxH cross . Using the genome-wide suggestive LOD score of 3 . 5 , a total of 8111 eQTL were present in the BxH study . From these , 1905 eQTL mapped to within 10Mb of the physical location of the gene and were classified as local eQTL and the remainder ( 6206 ) as distant . Intersection of the local eQTL for the common set of genes in the two studies ( 1905 eQTL in BxH vs 471 eQTL in the MF1 ) identified 163 genes . This amounts to ∼35% of the total local eQTL found in the MF1 study ( 163/471 ) . Intersection of distant eQTL , however , gave a much smaller overlap . From the 760 distant eQTL in the MF1 there were only 9 present in the BxH data ( 7 expected by chance , P = 0 . 22 ) which is about ∼1% of the distant eQTL found in the MF1 study ( Table 3 ) . As discussed previously , the MF1 data has a low statistical power to detect distant eQTL especially at higher p-value cutoffs such as the one we used to compare the two datasets ( P = e-05 ) . Therefore , the lack of overlap between the distant eQTL in the MF1 data and the BxH data can be attributed to both the lack of power associated with detecting distant eQTL in the MF1 study and the conservative p-value cutoff chosen to detect these eQTL . Lowering the cutoff value for significance to 1 . 14e-04 ( 25% FDR ) identified 26 overlaps with the BxH data ( 24 expected by chance , P = 0 . 3 ) and lowering this cutoff further to a nominal p-value of 0 . 001 resulted in 163 distant eQTL overlap between the two studies ( 140 expected by chance , P = 0 . 008 ) . Previous studies suggest that outbred stocks offer a high resolution mapping resource , but these studies did not have prior knowledge of the location of the causal variant for the trait [16] . The presence of common local eQTL , where one can assume , with high confidence , that the causal genetic variant lies within or near by the physical location of the gene , between the BxH and the MF1 data sets provides a unique setting to directly demonstrate the higher mapping resolution attainable in the association study using MF1 outbred stock compared to the linkage study in the F2 population . Figure 3 illustrates the results of such a comparison for 4 shared local eQTL ( Ttf2 , Insig2 , Frzb , and Pparg ) between BxH and MF1 populations . As shown , the MF1 data ( grey curve ) provided much narrower candidate region than the BxH data ( black curve ) . As expected for any local eQTL , the candidate regions in both studies encompassed the genomic region for the gene itself and peaked directly over the physical location of the gene . However , the QTLs in the BxH cross encompassed a much broader region than the association results in MF1 data . These results indicate that the MF1 population yields a much better resolution for eQTL mapping than a traditional cross . We next turned to the distant ( trans-acting ) eQTL . Several eQTL studies with intercross or RI strain mice have observed that many distant eQTL map to the same location on a chromosome , giving rise to what is known as distant eQTL hotspots . Here we illustrate how the MF1 whole genome association analysis can be used to resolve such co-localization of distant eQTL . We had previously detected a distant eQTL hotspot in the middle region of chromosome 6 . A total of 96 unique genes ( 99 probes ) had a distant eQTL at this locus . Based on the assumption that the variation in the expression of the causal gene underlying this hotspot mediates its effects , we identified 8 local eQTL ( Bcl2l13 , Ogg1 , Cidec , Atp2b2 , Pparg , Clstn3 , LOC380687 , C1r ) as primary candidate genes for this hotspot . Thus , judging by the F2 data alone , any one of these 8 local eQTL could be regulating any of the 96 distant eQTL . We turned to the MF1 data and asked , first , if we could resolve the co-localization of the local eQTL on the chromosome 6 locus and , second , if we could resolve the co-localization of the distant eQTL in this region . From the 8 local eQTL in the BxH F2 cross , 3 of them ( Bcl2l13 , Pparg , and Cidec ) were also observed in the MF1 population . Figure 4 shows the mapping of these three local eQTL in the BxH and the MF1 populations . While the mapping results appeared indistinguishable in the BxH study ( Figure 4A ) , association analysis in the MF1 successfully resolved these local eQTL and mapped them near the physical location of each gene ( Figure 4B ) . Out of the three local eQTL , the local eQTL for Cidec mapped to two markers , rs13478971 and rs13478971 , at 112 . 1 and 111 . 2 Mb , respectively , which are ∼1 and ∼2 Mb away from where the physical location of the Cidec gene ( 113 . 3 Mb ) . These were the closest markers to the physical location of Cidec . These results suggested the presence of either a distal regulatory element for this gene or the presence of a closely linked regulatory gene for Cidec on Chromosome 6 . Genotyping with denser markers closer to Cidec location might correctly position the highest peak above the gene itself . Interestingly , the Bcl2l13 transcript levels , in addition to mapping to the nearest marker to the physical location of the gene at 121 . 6 Mb , also showed significant association with markers located at 114 . 2 , 114 . 3 , and 114 . 7 Mb , suggesting the possible presence of an additional distal regulatory locus near the local eQTL for this gene . After resolving the three local eQTL , we examined the distant eQTL which mapped to this locus in the BxH cross and asked which of these distant eQTL mapped to any of these three local eQTL on Chromosome 6 . From the 96 distant eQTL colocalizing to the chromosome 6 locus in the BxH study , 14 were replicated in the MF1 population ( using the nominal p-value cutoff of 0 . 01 ) . Judging by the location of the association peak markers , despite the co-localization in the BxH , these 14 eQTL mapped to varying loci within the chromosome 6 region . In particular , 3 of these genes ( S3-12 , Calr3 , Hmgcl ) mapped over the Pparg locus , another 2 genes ( Gpi1 , Ctps ) mapped over the Cidec locus , and 9 genes had the most significant associations with markers at other than the three local eQTL loci ( Figure S4 ) . For the 5 genes mapping to either the Pparg or the Cidec loci we computed the 50 and 90 percent confidence ( c . i . ) intervals using bootstrapping ( 1000 sample sets ) . The S3-12 and Calr3 50% c . i . span a 1 . 3Mb region over the Pparg locus from 114 . 6Mb to 115 . 9Mb and the 50% c . i . for Hmgcl span a 1 . 9Mb region over the Pparg locus from 112 . 4Mb to 115 . 3Mb ( the 90% c . i . for S3-12 was from 112 . 4Mb to 117Mb , for Calr3 was from 112 . 4Mb to 116 . 2Mb , and for Hmgcl was from 112 . 1Mb to 115 . 9Mb ) . For the two genes with the peak marker at the Cidec locus ( Gpi1 and Ctps ) the 50% c . i . were 3Mb ( from 111 . 2Mb to 114 . 2Mb ) and 1 . 2Mb ( from 111 . 2Mb to 112 . 4Mb ) , respectively , overlapping with the Cidec locus . The 90% c . i . for these two genes were from 110Mb to 114 . 3Mb and from 111 . 2Mb to 114 . 6 Mb , respectively . It is noteworthy that S3-12 is a known Pparg target gene [19] . These results show that the co-localization of eQTL in the BxH study can be successfully resolved with high resolution in the MF1 data . The amount of resolution achieved in our mapping study was limited to the density of the markers used ( average marker density 1 . 37 Mb ) . Next , we asked whether typing more SNPs in a region would enhance the mapping resolution . For this , we focused on the distal locus of chromosome 5 where the results of the whole genome association had identified 3 distant eQTL ( Pbx2 , 2610020N02Rik , and D4Ertd432e ) at the genome wide significance level of 2 . 43e-05 p-value ( 10% FDR ) . The candidate region for this association spanned a 3 . 6 Mb region ( from rs13478570 at 142 . 8 Mb to rs13478583 at 146 . 4 Mb ) with the peak marker at 144 Mb ( rs13478573 ) . To fine map this region each of the 110 animals were genotyped for an additional 5 markers by the PCR-ARMS technique [20] . The primers were designed such that they would be less than 500 kb away from the peak marker or each other ( Materials and Methods ) . The fine mapping results are shown in Figure 5 . For D4Ertd432e ( bottom panel ) and 2610020N02Rik ( top panel ) the fine mapping effort reduced the candidate locus to 1 . 1 Mb ( located between rs29635622 at 143 . 4 Mb and rs32348286 at 144 . 5 Mb ) containing 22 candidate genes . For Pbx2 , the candidate region was reduced to 0 . 7 Mb interval between rs33492148 ( at 143 . 8 Mb ) and rs32348286 ( at 144 . 5 Mb ) containing 14 candidate genes . The 90% c . i . for these three genes ( as determined by 1000 bootstrapping data sets ) spanned a 500 kb region in the interval between 143 . 81Mb and 144 . 52Mb . These results suggest that MF1 population , with a relatively small sample size of 110 , gives a sub-megabase resolution for mapping eQTL .
This report provides a “proof of concept” demonstration of the utility of genome wide association for the identification of genes contributing to complex traits in mice . A number of previous association studies with outbred stocks or different inbred strains have been reported , but these have in most cases not been validated since the underlying genes were not known [14] , [15] , [21]–[23] . We have taken advantage of many local eQTL that have been identified in a recent linkage study in mice to validate the association approach . We also have provided an overall view of LD structure in the MF1 population and have shown the importance of correcting for population structure in association analyses . In this study we used both the local eQTL and the distant eQTL to investigate the attainable high resolution mapping of expression traits in the MF1 population . We used the local eQTL as a proof of concept because with high confidence we can predict where the genetic variant is located ( i , e . near or within the physical location of the gene ) [24] . Therefore , this allowed us to study the level of resolution one can achieve in the MF1 population . The comparative analysis of the four common local eQTL between the BxH F2 and the MF1 mice suggested that one can achieve a mapping resolution below 1 Mb . This is also evident from the fact that in all the local eQTL identified in the MF1 data about half the peak markers for the association mapped within 600 kb from the physical location of the gene . The sub-megabase resolution achieved for the eQTL is also supported by the fine mapping results for the distant eQTL for the chromosome 5 locus as well . These results are also comparable to the previous mapping studies for the behavioral traits in MF1 mice where the reported confidence intervals for 3 closely linked QTLs were between 250 to 750 kb [14] . Clearly , with larger numbers of mice and denser genotyping , we expect the mapping resolution in this population to increase and the confidence intervals to decrease . The number of significant associations observed at various false discovery thresholds in this study ( Table 1 ) were lower than the numbers reported in other genetical genomics studies [5] , [6] , [25] , [26] . This was especially true for distant eQTL which , for most part , have relatively small effects . One of the reasons for this shortcoming is the lack of power associated with the small number of animals used in this study ( 110 mice ) . The presence of population structure , in turn , also negatively impacts the effective size of the animals used in the study . In fact , one of the limitations of using outbred stock for mapping complex traits has been the statistical power issue and the need to include large number of mice in the study [13] . Another important point related to the statistical power issue is the very stringent genome wide significant cut off chosen in the whole genome association analysis due to the multiple testing issue . Without doubt , the use of genome wide cut off value is the appropriate measure for screening significant associations across all the markers and all the gene expression traits in the genome ( over 22 million total tests performed ) . However , in settings where the replication of previously found QTL is under investigation , the hypothesis to be tested is reduced to one trait and several markers along the previously mapped region . Therefore , there should be no need for selecting such a high cutoff value for significance . This was evident in our data when we attempted to resolve the S3-12 and 13 other previously identified chromosome 6 locus distant eQTL . At the genome wide cutoff value , only 2 of these genes were significantly associated with markers on chromosome 6 locus , but with the nominal p-value of 0 . 01 , 12 additional genes also showed evidence of significant association to this locus . The use of less stringent criteria for local QTL studies has also been implemented in other reports and shown to correctly rediscover and validate previously identified QTLs [27] , [28] . Previous genetical genomics studies reported the presence of genomic hotspots where large groups of eQTL collectively map to single loci in the genome [4] , [5] , [29] . In the current study , we were also able to identify such hotspots in the MF1 population , but the number of co-localizing eQTL within each hotspot identified in our study is considerably less than what has been reported before in other crosses [5] , [29] . This is partly due to the lack of power to detect distant eQTL ( as mentioned above ) and partly due to the high resolution mapping achieved in the MF1 population . In general , the presence of eQTL hotspots indicates either the presence of a master regulator gene which regulates the expression of group of genes together or the presence of several tightly linked genes within the hotspot , each of which regulates the expression of subset of the genes which map to this locus [30] . In the case of an F2 population , since the mice carry relatively few recombinations , these two alternatives appear alike and indistinguishable from the mapping data . In the MF1 data , however , since the genome is finely grained , one can resolve a hotspot due to multiple linked genetic variants into its individual components . We used the chromosome 6 locus as an example and were able to resolve the 14 distant eQTL into several groups based on which marker they associated with most strongly . Among these , the group that showed significant association with the Pparg locus comprised three genes one of which ( S3-12 ) has been associated with Pparg gene previously [19] , [31] and another gene ( Hmgcl ) has been shown to be coregulated with Pparg at the transcript level by thyroid hormone [31] . It has been widely acknowledged that standard statistical tests that do not account for population structure or familial relatedness are prone to identify spurious associations [32]–[34] . Different levels of molecular variance between different pairs of individuals are likely to induce different levels of polygenic background effects , invalidating the independence assumption of standard statistical tests such as t-test or ANOVA . Recent studies illustrate that the linear mixed model effectively captures confounding effects due to heterogeneous genetic relatedness [35] , [36] more effectively than previous approaches such as Structured Association [37] , Principal Component Analysis [38] or Genomic Control [39] . In this report we provide additional evidence for how failure to correct for such population structure can result in many more false positive associations . We also show that distant eQTL are more prone to such artifactual associations due to their relatively small effect sizes and higher likelihood of being false positives . Genetical genomics is becoming increasingly popular due to its promise to bridge the gap between the physiological traits and the genetic variations in the population . To fully take advantage of such an approach it is imperative to understand the nature of the association between the transcript level of the gene and the causal genetic variation . Molecular networks underlying the physiological traits cannot be properly constructed without the proper knowledge of the interaction among genes . We believe that the genetical genomics approach coupled with the high precision mapping offered by the MF1 outbred stock will significantly advance the potential for identifying regulatory genes for distant eQTL and provide the necessary components to build such biological networks .
Female MF1 mice , approximately 4–6 weeks of age , were purchased from Harlan ( Indianapolis , Indiana , USA ) . These animals were fed Purina Chow ( Ralston-Purina Co . , St . Louise , MO ) containing 4% fat until 19 weeks of age , and then fed a Western diet ( Teklad 88137 , Harlan Teklad , Madison WI ) containing 42% fat and 0 . 15% cholesterol for ∼14 weeks until they were sacrificed at 33 weeks of age . All mice were maintained on a 12h light/dark cycle . Mice were fasted for 5 hours before being euthanized . For the initial genotyping , the Affymetrix GeneChip Mouse Mapping 5K SNP platform was utilized . The DNA used for the genotyping was isolated from the tail clips of each mouse using the Qiagen's DNeasy tissue kit ( cat# 69506 ) . Overall , a total of 5024 SNPs were genotyped . The genomic location of all the analyzed SNPs were based on the snpdb126 ( http://www . ncbi . nlm . nih . gov/SNP/index . html ) . Fine mapping on the distal region of Chromosome 5 was performed using the PCR-ARMS technique [20] . A total of 6 SNP markers were selected for fine mapping: rs29635622 at 143400193 bp position , rs33318740 at 143817829 bp position , rs33492148 at 143854676 bp position , rs32348286 at 144521447 bp position , rs29524465 at 144918013 bp position , and rs33719947 at 145196793 bp position . These primers were chosen so that the distance between adjacent markers did not exceed 500kb . To carry the PCR-ARMS a set of tetra-primers were designed using the http://cedar . genetics . soton . ac . uk/public_html/primer1 . html website [20] for each marker . The tetra-primer sequences of each marker and the expected band size for each are as follows: 1 ) rs29635622 forward inner primer ( C allele specific ) GCTTATTTGCATACTTTGCGATGTAGAC , reverse inner primer ( T allele specific ) AACTATCCAAATGCACACTGAAGCCA , forward outer primer GTGCTATCTCTTCAGCCCAGAGTGATAT , reverse out primer GAGGAGCGAACCATTCTCTAAAAGTTGT , C-allele size = 133bp , T-allele size = 163 bp , outer-products = 242 bp; 2 ) rs33318740 forward inner primer ( A allele specific ) AAGATGCCGGCCCCAGATTGCCCTGTGA , reverse inner primer ( G allele specific ) GGGAGAAAGCTCCCTGCTTTGTCCAAACTC , forward outer primer GGAAGGTGAGGAGACAGGCTTCCGGCAG , reverse out primer CTTATGGCAAACCACCCTGCCCAGCAGA , A-allele size = 193 bp , G-allele size = 162 bp , outer-products = 297 bp; 3 ) rs33492148 forward inner primer ( T allele specific ) TTCTGTCTTAATTGAGCCCATATGAAAAT , reverse inner primer ( G allele specific ) GCACATTCTTCCAGACTCTGCATATC , forward outer primer ACTCTTGACAAAGAAGAATGCTTGCTTT , reverse out primer ATGTTTTGGCTAAGCACAATCCTACTCT , T-allele size = 194 bp , G-allele size = 172 bp , outer-products = 311 bp; 4 ) rs32348286 forward inner primer ( A allele specific ) AGGACTGCCACAGGCCAGCATCTCACA , reverse inner primer ( G allele specific ) AGTGTATCTATCAGGTGAATTCCAGTAGTC , forward outer primer AAGCCAAGCTGTCTCCAAGTCCTAGAAA , reverse out primer ACCACTTGAAGCCTGATTAAAATGTGCC , A-allele size = 213 bp , G-allele size = 175 bp , outer-products = 331 bp; 5 ) rs29524465 forward inner primer ( A allele specific ) CCTCTAATCTCCTGAGGATTGGAACA , reverse inner primer ( G allele specific ) TCATTTGGACACTAGAGCTTCTTCATTATC , forward outer primer TCGGAGAGACAGTTGTCTGTTAGGTTTA , reverse out primer GACAATGACGAAAAGACAAGTCACTTCT , T-allele size = 116 bp , G-allele size = 127 bp , outer-products = 187 bp; 6 ) rs33719947 forward inner primer ( C allele specific ) GTACATGTTCTTTTAAAATTATTAATCGAC , reverse inner primer ( T allele specific ) AGAAAAAGACACTCCTTGGAGCATGA , forward outer primer CTGTGATTTAAAGGCGTGCTAGTACTAC , reverse out primer GTGAGAGAGAGAGTCTGGGAATATTTCT , C-allele size = 177 bp , T-allele size = 157 bp , outer-products = 278 bp . Each of the 110 MF1 mice were genotyped for these markers by PCR and the products were separated by 4% agarose gel and visualized by ethidium bromide staining . RNA extraction was performed on the liver tissue obtained from each animal at the time of sacrifice , using Qiagen's RNeasy kit ( cat# 74104 ) . For the gene expression measurements , Illumina's Mouse whole genome expression BeadChips ( MouseRef-8-v1 Expression BeadChip ) were used . All amplifications and hybridizations were performed according to Illumina's protocol by the Southern California Genome Consortium microarray core laboratory at UCLA . In brief , 100 ng of total RNA was first reverse transcribed to cDNA using Ambion cDNA synthesis kit AMIL1791 and then converted to cRNA and labeled with biotin . 800ng of biotinylated cRNA product is hybridized to prepared whole genome arrays and allowed to incubate overnight ( 16–20 hrs ) at 55 degrees C . Arrays were washed then stained with Cy3 labeled streptavadin . Excess stain wass removed by washing and the arrays were dried and scanned on an Illumina BeadScan con-focal laser scanner . Data normalization was performed using the rank invariant method by the Bead Studio software . After normalization , all gene expression data were log2 transformed . To filter genes , we selected probes which met two criteria; 1 ) the probes exhibited a reliable signal and 2 ) the probes contained no annotated SNP within their sequence . The former was determined according to the Illumina Bead Studio output . The detection value is equal to 1-probability that a signal level is due to nonspecific hybridization . This value can be interpreted as the probability of seeing a certain signal level without specific probe-target hybridization . For filtering we excluded any probe which had a detection value of lower than 0 . 95 in greater than eleven ( 10% ) or more animals . To select against the bias in hybridization due to probe design , as described by Walter et al [40] , we excluded any probe which in blast search had 100% sequence identity to more than one location of the genome or had at least one SNP within it . To determine this we aligned the genomic location of all the probes against the genomic location of ∼8 million SNPs available on the Perlegen database ( http://mouse . perlegen . com/mouse/index . html ) . Any probe which contained a SNP which was polymorphic between the proposed ancestors of the MF1 mice ( I , e . C57BL/6J , DBA/2J , C3H/HeJ , AKR/J , I/LnJ , BALB/cJ , RIII/J , and A/J ) was excluded . This filtering step resulted in the exclusion of 2160 probes . The remaining 10013 probes were used as the starting set for the whole genome association analysis . The gene expression data are deposited to the Gene Expression Omnibus ( http://www . ncbi . nlm . nih . gov/geo/ ) at NCBI . These data are deposited under the accession number GSE10280 . LD and FDR were calculated using R software algorithms . To compute the pair-wise LD between markers , we used the LD function in the Genetics package , which includes output of the chi square p-values for marker independence which we used to test for LD between markers on different chromosomes . For the FDR calculation we used the q-value package in R [41] . Due to the computational complexity associated with evaluating q-values for >20 million p-values , we computed the FDRs by taking the average FDR for 100 samples each containing 5 million randomly selected p-values from the original 22 , 069 , 649 calculated p-values . We first computed the genetic similarity matrix between the individual mice as the fraction of shared alleles ( identity-by-state ) for each pairs , and visualized it with heatmap R package . A complex multi-leveled population structure and genetic relatedness is observed in the genetic similarity matrix . We applied the following variance component test to estimate the variance explained by genetic background and assess the statistical significance . where y is the vector of expression values of a gene , and μ is mean , and e is random errors following an identical and independent normal distribution with Var ( e ) = σe2I . u is a vector of random variables accounting for the effect from genetic background . u follows a multivariate normal distribution with Var ( u ) = σg2K , where K is the genetic similarity matrix described above . The fraction of variance explained by genetic background is computed as previously suggested with tr ( SKS ) / ( n-1+tr ( SKS ) ) , where S = I-J/n , and J is a square matrix consisting of ones from the REML estimate of H1 . The likelihood ratio test is performed by comparing the maximum likelihood of two hypotheses . The likelihood difference 2* ( l1-l0 ) asymptotically follows a 1:1 mixture of the chi-squared distribution with zero and one degree-of-freedom [42] . The false discovery rate is estimated conservatively by setting π0 = 1 [41] . To account for population structure and genetic relatedness in association mapping , we applied the following standard linear mixed model as previously suggested [18] , [35] , [36] . where y , μ , u , and e are same as described above , and x is the genotype vector of a marker represented in additive model , and β is a marker effect . A standard F test was performed to test H1:β≠0 against H0:β = 0 after estimating restricted maximum likelihood ( REML ) variance components as described [18] , [35] , [36] . We applied EMMA ( Efficient Mixed Model Association ) as a R implementation of linear mixed model . Since EMMA is orders of magnitude faster than other implementations commonly used , we were able to perform statistical analyses for all pairs of transcripts and genome wide markers in a few hours using a cluster of 50 processors .
|
In rodents , as in humans , traits such as obesity or diabetes are under the influence of many genes spread throughout the genome . Using linkage analysis , the locations of the major contributing genes can be mapped only to very large regions of chromosomes , usually encompassing hundreds of genes . This has made it difficult to identify the underlying genes and mutations . Another approach , analogous to genome-wide association in human populations , is to use association analyses among outbred stocks of mice . In this proof-of-principle article , we make use of common variations that locally perturb gene expression to demonstrate the greatly improved mapping resolution of association in mice . Our results indicate that association analyses in mice are a powerful approach to the dissection of complex traits and their underlying molecular networks .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genetics",
"and",
"genomics/genomics",
"genetics",
"and",
"genomics/animal",
"genetics",
"genetics",
"and",
"genomics/gene",
"expression",
"genetics",
"and",
"genomics/genome",
"projects",
"genetics",
"and",
"genomics"
] |
2008
|
High-Resolution Mapping of Gene Expression Using Association in an Outbred Mouse Stock
|
Interactions between an organism and its environment can significantly influence phenotypic evolution . A first step toward understanding this process is to characterize phenotypic diversity within and between populations . We explored the phenotypic variation in stress sensitivity and genomic expression in a large panel of Saccharomyces strains collected from diverse environments . We measured the sensitivity of 52 strains to 14 environmental conditions , compared genomic expression in 18 strains , and identified gene copy-number variations in six of these isolates . Our results demonstrate a large degree of phenotypic variation in stress sensitivity and gene expression . Analysis of these datasets reveals relationships between strains from similar niches , suggests common and unique features of yeast habitats , and implicates genes whose variable expression is linked to stress resistance . Using a simple metric to suggest cases of selection , we found that strains collected from oak exudates are phenotypically more similar than expected based on their genetic diversity , while sake and vineyard isolates display more diverse phenotypes than expected under a neutral model . We also show that the laboratory strain S288c is phenotypically distinct from all of the other strains studied here , in terms of stress sensitivity , gene expression , Ty copy number , mitochondrial content , and gene-dosage control . These results highlight the value of understanding the genetic basis of phenotypic variation and raise caution about using laboratory strains for comparative genomics .
A major focus of genetic study is to elucidate the effects of genetic variation on phenotypic diversity . The evolution of phenotypes is often driven by environmental factors and the interactions between each organism and its environment . Recently , there has been a renewed interest in characterizing the diversity and ecology of organisms long used in the laboratory as models for biological study . Yeast , worms , flies , and mice have been studied on a molecular level for decades and have provided many insights into basic biology . However , most of our knowledge base exists for only a handful of domesticated lines . Little is known about the natural ecology of these organisms or the degree to which individuals of each species vary within and between natural populations . The budding yeast Saccharomyces cerevisiae exists in diverse niches across the world and can be found in natural habitats associated with fruits , tree soil , and insects , in connection with human societies ( namely through brewing and baking ) , and in facultative infections of immuno-compromised individuals [1] . These yeasts are transported by insect vectors and likely through association with human societies . Recent population-genetic studies have begun to explore the genetic diversity of S . cerevisiae strains [2]–[5] . These studies have demonstrated little geographic structure in natural yeast populations and relatively low sequence diversity , particularly within vineyard strains . It has been proposed that low sequence diversity in this species may be due to a more recent common ancestor compared to other yeasts [6] . Genomic comparisons also suggest low rates of outcrossing between strains [7] , which may limit the fixation of genetic differences under selection by reducing effective population sizes [8] . Although the genetic diversity of S . cerevisiae populations is emerging from large-scale sequencing projects , the phenotypic diversity within and between yeast populations has been less systematically studied . Myriad studies have characterized strain-specific differences in specific phenotypes to identify the genetic basis for phenotypes of interest ( for example , those related to wine making [9] , thermotolerance [10]–[12] , sporulation efficiency [13]–[16] , drug sensitivity [17]–[19] , and others [20]–[25] ) . The degree to which these phenotypes vary across diverse strains has not been systematically explored . Other genomic studies have investigated variation in genomic expression across strains , with the goal of investigating the mode and consequence of gene-expression evolution [26]–[30] . These studies demonstrated significant variation in gene expression between strains , and in some cases pointed to the genetic basis for those differences [27] , [31]–[35] . However , each study investigated only a few strains , typically vineyard strains . The broader phenotypic variation across diverse yeast strains and populations , particularly natural isolates , is largely uncharacterized . Here we investigated the variation in stress sensitivity and genomic expression in a large panel of Saccharomyces strains . We quantified the sensitivity of 52 strains collected from diverse niches to 14 environmental conditions and measured genomic expression in 18 of these strains growing in standard medium . We observe a large amount of phenotypic variation , both in terms of stress sensitivity and gene expression . Associations among phenotypes revealed relationships between environmental conditions and among yeast strains . One case in particular suggests that genetically diverse strains collected from oak soil have undergone selection for growth in a common niche . This study provides a representative description of expression variation and stress sensitivity within and across yeast populations , particularly non-laboratory strains , setting the stage for elucidating the genetic basis of this variation .
Fay and Benavides conducted a population-genetic study of 81 Saccharomyces strains by analyzing ∼7 kb of coding and non-coding sequence from each isolate [2] . We characterized the phenotypic diversity of 52 of these strains , shown in Figure 1 . This set included natural isolates from European vineyards , yeasts collected from African palm-wine fermentations , commercial wine- and sake-producing strains , clinical yeasts , natural isolates collected from African and Asian fruit substrates , strains from oak-tree soil and exudates from the Northeastern United States , three common lab strains , and other isolates ( see Table S1 and [2] for references ) . We also characterized two haploid S . cerevisiae strains ( RM11-1a and YJM789 ) and three other Saccharomyces species ( S . paradoxus , S . mikatae , and S . bayanus ) for which whole-genome sequence is available [36] , [37] . Each strain was grown under 31 different conditions representing 14 unique environments , chosen to provoke diverse physiological responses . These environments varied in nutrient composition , growth temperature , and presence of toxic drugs , heavy metals , oxidizing agents , and osmotic/ionic stress . Cells were grown on solid medium in the presence of each environmental variable , and viability was scored relative to a no-stress control for each strain ( see Materials and Methods for details ) . The results reveal a tremendous amount of phenotypic diversity in environmental sensitivity ( Figure 2 ) . Although there were similarities between strains , no two strains were exactly alike in phenotypic profile . Each displayed a propensity for growth under at least one environment and sensitivity to one or more conditions . Some strains were generally tolerant to stressful environments across the board . For example , strain Y2 , originally collected from a Trinidadian rum distillery , and clinical isolates YJM454 and YJM440 were tolerant of most of these conditions , while the S . bayanus strain used in our study was sensitive to nearly all stresses tested . Several strains , including commercial sake-producing strains , showed a wide standard deviation of growth scores across the stresses , reflecting that they were either highly sensitive or highly resistant to different stresses . In contrast , most vineyard isolates grew moderately well in most of the environments examined ( see Discussion ) . Exploration of the range of strain sensitivities measured for each environment also suggested common and unique features of Saccharomyces' habitats . Collectively , this set of strains showed the greatest variation in copper sulfate tolerance , sodium chloride resistance , and freeze-thaw survival , implicating these as niche-specific features not generally experienced by yeast . In contrast , strains showed the least variation ( but some variability nonetheless ) for growth on non-fermentable acetate , in minimal medium lacking supplemental amino acids , and at 37°C . Presumably , defects in respiration , prototrophy , and growth at physiological temperature represent a significant selective disadvantage , regardless of the particular niche . Hierarchical clustering of the phenotype data revealed interesting relationships between groups of strains . In particular , several groups of strains displayed similar profiles of stress sensitivity across the environments tested ( Figure 2 ) . As a group , the sake-producing strains were extremely resistant to lithium chloride but sensitive to copper sulfate , calcium chloride , cadmium chloride , and SDS detergent ( p<0 . 005 based on 10 , 000 permutations , see Materials and Methods ) ; indeed , this group was slightly more sensitive to stress in general . Many of the vineyard strains shared specific phenotypes , including resistance to copper sulfate , as previously noted for other vineyard strains [26] , [38] , [29] . The group of laboratory strains was also highly resistant to copper sulfate as well as sodium and lithium chloride . In contrast , strains collected from oak soil were particularly sensitive to copper sulfate and sodium chloride but highly resistant to freeze-thaw stress ( p<0 . 005 , 10 , 000 permutations ) . The similarities in phenotypic profiles could arise through selection ( either directional or purifying ) due to shared selective pressures across strains living in the same environment . Alternatively , phenotypic similarity could result simply if the strains are genetically related due to a recent common ancestor . For example , many of the lab strains are closely related , since a large fraction of their genomes is derived from a common progenitor [39] , [40] . We wished to distinguish between these possibilities for other strain groups . Natural selection can be inferred by comparing the population genetic structure ( FST ) to an analogous measure of phenotypic structure ( QST ) [41] , [42] . A deviation from unity suggests that either divergent ( QST/FST>1 ) or purifying ( QST/FST<1 ) selection has occurred across populations . We wished to analyze each subpopulation separately , and therefore we devised a simple alternative approach to identify deviations from neutral phenotypic variation . We calculated the average pairwise phenotypic distance over the average pairwise genetic distance for pairs of strains collected from the same environment ( ‘sake’ , ‘vineyard’ , ‘oak’ , ‘clinical’ , ‘natural’ or ‘other fermentation’ ) . This ratio was compared to the ratio of distances calculated for pairs of strains between niche groups , generating the parameter P/G . A P/G ratio = 1 is expected under neutrality , where the phenotypic to genetic distance is equal for within-group versus between-group comparisons . In contrast , a value of P/G<1 suggests that the strains within the group are more similar in phenotype than would be expected under the neutral model , whereas a ratio >1 indicates that the strains are phenotypically more variable than expected based on their genetic relatedness . The results provide evidence of both selection and shared ancestry for different groups of strains . First , the P/G ratio did not deviate significantly from unity for strains in the ‘clinical’ , ‘natural’ , or ‘other fermentation’ groups ( average P/G = 1 . 02+/−0 . 22 ) , nor did it deviate significantly for randomized simulations ( data not shown ) . In contrast , P/G was 4 . 2 and 3 . 0 for sake strains and vineyard strains , respectively . Thus , the similarity in their phenotypes likely arises due to their recent divergence from a common ancestor . Interestingly , these P/G values were significantly higher than expected by chance ( p<0 . 0001 from 10 , 000 permutations ) , suggesting that the strains show more phenotypic variation than expected . This could arise if strains have experienced diversifying selection for disparate phenotypes , although it could also result if genetic distances are underrepresented or skewed due to limited sequence data . In contrast , strains collected from oak-tree exudates and soil are phenotypically more similar than would be expected under a neutral model . We observed a P/G ratio of 0 . 31 ( p = 0 . 0013 from 10 , 000 permutations ) , indicating that phenotypic variation within this group is lower than expected based on the strains' genetic relatedness . This suggests that the strains have undergone selection for growth in a common environment ( see Discussion ) . Consistent with this model , the S . paradoxus strain YPS125 , also collected from Northeastern oak flux [6] , is phenotypically more similar to S . cerevisiae strains collected from that environment ( pairwise R of 0 . 61 , 0 . 66 , and 0 . 77 to YPS1000 , YPS1009 , and YPS163 , respectively ) than to the other S . paradoxus strain in our collection ( R = 0 . 51 ) . At least some of the phenotypes shared by these strains are likely important for their ability to thrive in their niche ( see Discussion ) . Numerous studies have characterized differences in genomic expression between individual strains of yeast , typically vineyard and lab strains [13] , [26]–[31] , [34] , [43] . To more broadly survey the variation in genomic expression across populations , we measured whole-genome expression in 17 non-laboratory strains compared to that in the diploid S288c-derived strain DBY8268 , using 70mer oligonucleotide arrays designed against the S288c genome . The long oligos used to probe each gene minimize hybridization defects due to sequence differences from S288c . We verified this by hybridizing genomic DNA from 6 strains of varying genetic distance from S288c: indeed , fewer than 5% of the observed gene expression differences described below could be explained by defective hybridization to the arrays ( see Materials and Methods ) . Therefore the vast majority of measured expression differences are due to differences in transcript abundance . A striking number of yeast genes showed differential expression from the laboratory strain in at least one other strain ( Figure 3A ) . Of the ∼5 , 700 predicted S . cerevisiae open reading frames , 2680 ( ∼47% ) were statistically significantly altered in expression ( false discovery rate , FDR = 0 . 01 ) in at least one non-laboratory strain compared to S288c , with an average of 480 genes per strain . At an FDR of 0 . 05 , over 70% of genes were significantly altered in expression in at least one non-lab strain ( Table 1 ) . The number of expression differences is comparable to that observed by Brem et al . , who reported over half of yeast genes differentially expressed between the vineyard strain RM11-1a and S288c [27] . However , closer inspection revealed that many of these expression differences were common to all of the non-laboratory strains ( Figure 3A ) , revealing that these expression patterns were unique to S288c . This group was enriched for functionally related genes , including those involved in ergosterol synthesis , mitochondrial function , respiration , cell wall synthesis , transposition , and other functions ( Table 2 ) . Many of these functional groups were also reported by Brem et al . , who noted that multiple categories ( including ergosterol synthesis and mitochondrial function ) can be linked to a known polymorphism in the Hap1p transcription factor [44] . Indeed , the expression differences specific to S288c were enriched for targets of Hap1p ( p<10−11 , hypergeometric distribution ) as well as targets of Hap4p ( p<10−6 ) [45] , which regulates genes involved in respiration . Hence , many of the observed expression differences may result because of S288c-specific physiology ( see Discussion ) . For a more representative description of expression variation in non-laboratory strains , we sought to represent the expression differences in a way that was not obscured by S288c . First , we identified genes whose expression varied significantly from the oak strain YPS163 . Second , we identified transcripts whose abundance varied from the mean of all non-laboratory strains ( see Materials and Methods ) . Although the mean expression value of each gene is merely an arbitrary reference point , this data transformation serves to remove the effect of S288c from each array while maintaining the statistical power to identify expression differences . Roughly 1330 ( 23% ) of yeast genes varied in expression in at least one non-laboratory strain relative to the mean of all strains , while 953 ( 17% ) of genes varied significantly from YPS163 ( FDR = 0 . 01 ) . In both cases , two thirds of significant expression differences were specific to only one strain ( Figure 3B and 3C ) . The number of genes with statistically significant expression differences from the mean ranged from 30 ( in vineyard strain I14 ) to nearly 600 ( in clinical isolate YJM789 ) , with a median of 88 expression differences per strain . The number of expression differences did not correlate strongly with the genetic distances of the strains ( R2 = 0 . 16 ) . However , this is not surprising since many of the observed expression differences are likely linked in trans to the same genetic loci [27] , [31] , [34] , [35] , [43] . Consistent with this interpretation , we found that the genes affected in each strain were enriched for specific functional categories ( Table S4 ) , revealing that altered expression of pathways of genes was a common occurrence in our study . We noticed that some functional categories were repeatedly affected in different strains . To further explore this , we identified individual genes whose expression differed from the mean in at least 3 of the 17 non-laboratory strains . This group of 219 genes was strongly enriched for genes involved in amino acid metabolism ( p<10−14 ) , sulfur metabolism ( p<10−14 ) , and transposition ( p<10−47 ) , revealing that genes involved in these functions had a higher frequency of expression variation . Differential expression of some of these categories was also observed for a different set of vineyard strains [26] , [28] , and the genetic basis for differential expression of amino acid biosynthetic genes in one vineyard strain has recently been linked to a polymorphism in an amino acid sensory protein [35] . We also noted that the 1330 genes with statistically variable expression in at least one non-laboratory strain were enriched for genes that contained upstream TATA elements [46] ( p = 10−16 ) and genes with paralogs ( p = 10−6 ) but under-enriched for essential genes [47] ( p = 10−25 ) . The trends and statistical significance were similar using 953 genes that varied significantly from YPS163 . Thus , genes with specific functional and regulatory features are more likely to vary in expression under the conditions examined here , consistent with reports of other recent studies [30] , [43] , [48] , [49] ( see Discussion ) . Expression from transposable Ty elements was highly variable across strains . However , Ty copy number is known to vary widely in different genetic backgrounds [50] , [51] , suggesting that these and other observed expression differences could be due to copy number variations in particular strains . Indeed , numerous expression differences could be linked to known gene amplifications in S288c , such as ASP3 , ENA1 , CUP1 , and hexose transporters [52] , [51] . We quantified the contribution of increased copy number to the observed increases in gene expression relative to S288c in 6 of our strains . In general , ∼2–5% of expression differences could be wholly or partially explained by differences in gene copy number ( see Materials and Methods ) . YPS1009 was an exception to the trend , since nearly 20% of genes with higher expression could be attributed to increased copy number - most of these genes reside on Chromosome XII . In fact , more than 80% of genes on Chromosome XII met our criteria for increased copy number ( Figure S1A ) , indicating that the entire chromosome is duplicated in this strain . Another example of chromosomal aneuploidy is evident in strain K9 , for which Chromosome IX appears amplified ( Figure S1B ) . Whole-chromosome aneuploidy has been frequently observed in strains growing under severe selective pressure ( for example [53]–[56] . Interestingly , the majority of genes on these duplicated chromosomes do not show elevated transcript abundance in the respective strains . In fact , only ∼25% of genes with increased copy number in each strain showed elevated expression ( defined at FDR = 0 . 01 or as genes whose expression is >1 . 5× over S288c ) . This is in stark contrast to previous studies demonstrating little dosage compensation in S288c in response to gene amplification and chromosomal aneuploidy , leading to the conclusion that yeast does not have a mechanism for dosage compensation . [53] , [54] , [57] . Instead , our results suggest that some form of feedback control acts to normalize the dosage of most genes in non-laboratory yeast strains . The remaining quarter of amplified genes may be inherently exempt from this feedback mechanism . Alternatively , relaxed feedback may occur for specific amplifications if the resulting transcript increase provides a selective advantage to the strain in question . Indeed , 15–40% ( depending on the strain ) of genes lacking feedback control show at least 1 . 5× higher expression beyond what can be accounted for by gene amplification alone , indicating that the expression differences are affected by both gene dosage and regulatory variation . These genes are excellent candidates for future studies of adaptive changes . As observed for gene expression , we found that some genomic amplifications were common across all 6 strains compared to S288c . All strains showed decreased Ty1 copy number , ranging from 2–15× lower than S288c . This is consistent with previous studies that showed higher Ty1 copy number ( including active and partial Ty elements ) in S288c compared to wine strains and natural isolates [50] , [51] , [58] . Most strains also showed even lower Ty1 transcript abundance , beyond what could be explained by copy number variations . Thus , in addition to a higher Ty content , S288c also shows higher expression from Ty genes , perhaps reflecting elevated rates of retrotransposition under the conditions studied here . In contrast , all strains showed higher copy number of the mitochondrial genome compared to S288c , typically elevated 2–3× but nearly 7× higher in clinical strain YJM789 . The most likely explanation is that these strains harbor more mitochondria than S288c , a fact confirmed in vineyard strain RM11-1a by mitochondrial staining [25] . In addition to revealing phenotypic diversity within and between yeast populations , natural variation can also uncover new insights into the effects of each environment on cellular physiology . For example , we noted correlations between environments based on the distribution of strain-sensitivity scores . The most likely explanation is that these stresses have similar effects on cellular function , and thus strains display similar sensitivities to them . Resistance to sodium chloride and lithium chloride or tolerance of ethanol and elevated temperature were highly correlated ( R = 0 . 66 at p<0 . 0001 and R = 0 . 51 at p<0 . 0006 , respectively , based on 10 , 000 permutations ) , consistent with the known effects of these stress pairs on ion concentrations or membrane fluidity/protein structure , respectively . Other relationships were not previously known , including the correlation between sensitivity to SDS detergent and the heavy metal cadmium ( R = 0 . 64 , p<0 . 0001 ) and between ethanol and caffeine tolerance ( R = 0 . 59 , p<0 . 0001 ) . In contrast , resistance to freeze-thaw stress was anticorrelated to sodium chloride resistance ( R = −0 . 35 , p = 0 . 006 ) , suggesting antagonistic outcomes of the same underlying physiology . These relationships point to commonalities in the cellular consequences inflicted by these environments that will be the subject of future investigations of stress-defense mechanisms . We also conducted an associative study to identify gene expression patterns correlated with environmental sensitivity across the 17 non-laboratory strains ( see Materials and Methods for details ) . As basal expression differences could significantly contribute to the inherent ability of cells to survive a sudden dose of stress , the results point to genes whose expression is related to , and perhaps causes , the phenotypes in question . Among the top genes associated with copper sulfate resistance was the metallotheionein CUP1 , important for copper resistance and known to have undergone tandem duplications in copper-resistant strains [59] , [60] . Of the genes whose expression was correlated to sodium chloride tolerance , nearly 20% are known to function in Na+ homeostasis and/or osmolarity maintenance ( including RHR2 , COS3 , SIS2 identified through genetic studies [61]–[63] and JHD2 , SRO7 , YML079W , YOL159C , TPO4 , UTH1 implicated in high-throughput fitness experiments in S288c [64] ) . Thus , these and likely other genes whose expression is highly correlated with each stress-sensitivity profile play a functional role in surviving that condition . Other correlations were not expected . Ethanol and caffeine tolerance were both correlated to the expression of genes encoding transmembrane proteins ( p<0 . 003 , hypergeometric distribution ) , perhaps related to the effect of these drugs on membrane fluidity . Sensitivity to the cell-wall damaging drug Congo Red was significantly correlated to the expression of genes involved in mitochondrial function and translation , respiration , and ATP synthesis ( p<10−13 ) , revealing a link between mitochondria/respiration and the cell wall . Although these connections will require further characterization , they demonstrate the power of using natural diversity to uncover previously unknown relationships between stresses and cellular processes .
This study demonstrates the vast amount of phenotypic variation in Saccharomyces strains collected from diverse natural habitats , used in industrial processes , and associated with human illness . Considering the phenotypic responses to the conditions studied here provides insights into the relationships between specific strains and their niches . For example , the wide variance in growth scores of sake-producing strains indicates that they are either highly resistant or sensitive to the different environments studied here , suggesting that they may be specialized for growth in the defined conditions of sake fermentation . In contrast , many of the vineyard isolates survived relatively well in most of the conditions tested . This may reflect their ability to thrive in more variable , natural environments and may also have facilitated their dispersal into new environments in a manner associated with human interactions [5] . Geographic dispersal might also explain the higher-than-expected phenotypic diversity of vineyard strains , which might be driven by diversifying selection ( suggested by our analysis ) due to unique pressures imposed after expansion into new environments . Although many of the phenotypic differences we observed are probably neutral , providing no benefit or disadvantage to the strains in question , some are likely to provide a selective advantage . Copper-sulfate resistance in European vineyard strains may have arisen through positive selection , since copper has long been used as an antimicrobial agent in vineyards and orchards [1] , [65] . Another example may apply to the oak strains studied here . Our simple metric comparing phenotypic to genetic diversity in strains collected from similar environments suggests that oak strains are phenotypically more similar than expected based on their genetic relationship . Formally , this could arise if multiple traits are evolving neutrally ( but slower than the genetic drift represented by the sequences used here ) since the strains diverged from a distant , common progenitor . However , the fact that S . paradoxus oak isolate YPS125 is phenotypically more similar to S . cerevisiae oak strains than the other S . paradoxus isolate in our analysis instead supports that these strains have undergone selection for growth in a common environment . One intriguing phenotype is freeze-thaw resistance , which may be important to survive the wintry niche from where these strains were collected . Consistent with this hypothesis , we have recently isolated numerous Sacharomycete strains ( including S . cerevisiae ) from Wisconsin oak exudates , of which 86% ( 19/22 ) are freeze-thaw tolerant ( DJK and APG , unpublished data ) . Ongoing studies in our lab are dissecting the genetic basis for this phenotypic difference . In addition to stress sensitivity , gene expression also varies significantly across yeast populations . More than a quarter of yeast genes varied in expression in at least one non-laboratory strain under the conditions studied here . Consistent with other recent reports [30] , [48] , [49] , [66] , we find that genes with specific structural or functional characteristics ( including nonessential genes and those with upstream TATA elements and paralogs ) show higher levels of expression variation across strains . This has previously been interpreted as a higher rate of regulatory divergence for genes with these features , either in response to selection [48] or mutation accumulation [49] . However , these features are also common to genes whose expression is highly variable within the S288c lab strain grown under different conditions ( [67] and data not shown ) , particularly those induced by stressful conditions [46] , [68] . It is also notable that genes with TATA elements show higher ‘noise’ in gene expression within cultures of the same strain [69] , [70] . Thus , an alternative , but not necessarily mutually exclusive , hypothesis is that the expression of these genes is more responsive to environmental or genetic perturbations , again consistent with previous studies [66] , [30] , [48] , [49] . We have conducted our experiments under ‘common garden’ lab conditions in attempt to minimize environmental contributions to expression phenotypes . However , because each strain may have evolved for growth in a unique environment , each may in fact respond differently to the same growth conditions used here . Indeed , this may explain the prevalence of metabolic genes in our set of genes showing variable expression in multiple strains , since many of these strains have not evolved for growth in highly artificial laboratory media . Emerging from our analysis is the fact that S288c is phenotypically distinct from the other non-laboratory strains studied here . This strain displays extreme resistance to specific stresses , harbors fewer mitochondria , contains more transposable elements , and shows unique expression of many genes compared to all other strains investigated ( a direct comparison of the number of differentially expressed genes in S288c is difficult due to the different statistical power in calling these genes ) . We have also found that this strain has an aberrant response to ethanol , since it is unable to acquire alcohol tolerance after a mild ethanol pretreatment , unlike natural strains [71] . It is likely that additional responses found in natural strains have been lost or altered in this domesticated line . The progenitor of S288c was originally isolated from a fallen fig in Merced , California , and sequence analysis indicates that S288c is genetically similar to other natural isolates [1]–[3] . A recent study by Ronald et al . counters the proposal that S288c has undergone accelerated divergence during its time in the laboratory [72] . Instead , our results suggest that the strain has evolved unique characteristics through inadvertent selection for specific traits ( such as growth on artificial media ) and population bottlenecks . Thus , the laboratory strain of yeast may not present an accurate depiction of natural yeast physiology . Indeed , no single strain can be used to accurately represent the species , a note especially important for comparing phenotypes across species . Complete exploration of an organism's biology necessitates the study of multiple genetic backgrounds to survey physiology across populations . Despite its limitations , the lab strain offers nearly a century of detailed characterization , along with powerful genetic and genomic tools . A useful approach is to complement studies on laboratory strains with investigations of natural variation . By characterizing stress sensitivity in a large set of strains , we have leveraged the power of natural diversity to uncover new relationships between stresses and to reveal previously unknown connections between genes , stresses , and cellular processes . These connections lead to hypotheses about stress defense mechanisms that can often be dissected using the valuable tools provided by the lab strain . Application of genomic techniques to characterize natural yeast strains will foster such studies while revealing additional insights into genetic and phenotypic variation in Saccharomyces .
Strains used in this study and references are found in Table S1 . In addition to sequence data from [2] , an additional 5 , 305 bp of noncoding DNA was sequenced for 41 S . cerevisiae strains over 8 intergenic sequences ( GENBANK accession numbers EU845779 - EU846095 ) for a total of 13 , 016 bp over 13 loci . Phylogenetic analysis shown in Figure 1 was performed on the combined sequence set using the program MrBayes [73] . Evolutionary distances were estimated using the Jukes-Cantor ( JC ) model based on 2 , 056 bp noncoding sequence data present in all strains; results and significance were very similar when the distance was based on 9 , 334 bp of noncoding sequence excluding only pairwise-deletion data [74] . Strains with evolutionary distances equal to zero over this subsequence ( but clearly non-zero when all sequence was assayed ) were set to 0 . 00001 to facilitate permutation calculations . Paralogs were defined as genes with a BLAST E-value score <10−100 . Yeast strains were grown in YPD medium at 30°C to an optical density of ∼0 . 3 in 96-well plates . Three 10-fold serial dilutions were spotted onto YPD agar plates containing the appropriate stress , as well as a YPD plate for a no-stress control . Cells were also plated onto minimal medium [75] or YP-acetate . In the case of freeze-thaw stress , 200 µl cells was frozen in a dry ice/ethanol bath for two hours or left on ice as a control before spotting onto YPD plates . Cells were grown for 2–3 days at 30°C unless otherwise noted , and viability of each dilution was scored relative to the no-stress control for each strain . All experiments were done in at least duplicate over 2–3 doses of most stresses ( see Table S2 for raw data and stress doses ) . Final resistance scores were summed over the 3 serial dilutions then averaged over replicates and stress doses , providing a single score ranging from 0 ( no growth ) to 6 ( complete growth ) for each strain and each stress condition . For Figure 2 , strains were clustered based on phenotypes using the Pearson correlation and UPGMA clustering [76] . Correlations between stresses were calculated based on the Pearson correlation between strains , excluding 14 strains of highly similar genetic distance ( JC<0 . 0008 ) . Phenotypes specific to groups of strains collected from similar environments ( see Table S1 for groupings ) were calculated based on the median growth score of strains in that group . Significance was estimated by 10 , 000 permutations of strain-group labels , scoring the frequency of observing a median growth score equal to or greater than that observed . A parameter , P/G , was calculated to compare the similarity in phenotype to the similarity in genotype for strains within and between niche groups . The average pairwise phenotypic distance , taken as the Pearson distance ( 1 – Pearson correlation ) between phenotype vectors , was divided by the average pairwise JC distance for strains within a niche group . This value was divided by the same ratio calculated for all pairs of strains between niche groups ( see Table S1 for niche groupings ) . Significance was estimated based on 10 , 000 random permutations of strain-group labels . The distribution of P/G ratios from randomized trials was centered on 0 . 99; furthermore P/G was ∼1 . 0 for strains in the ‘clinical’ , ‘natural’ , and ‘other fermentation’ groups , reflecting either neutral drift for these groups or that these strains were inappropriately grouped together into somewhat amorphous categories . Seventeen strains ( including B1 , I14 , M22 , M8 , PR , RM11-1a , K1 , K9 , YJM308 , YJM789 , YJM269 , Y12 , SB , Y1 , Y10 , YPS1009 , and YPS163 ) were chosen for whole-genome expression analysis . Cells were grown 2–3 doublings in YPD medium to early log-phase in at least biological triplicate . Cell collection , RNA isolation , and microarray labeling and scanning were done as previously described [77] , using cyanine dyes ( Flownamics , Madison , WI ) and spotted DNA microarrays consisting of 70mer oligos representing each yeast ORF ( Qiagen ) . For all arrays , RNA collected from the denoted strain was compared directly to that collected from the diploid S288c lab strain DBY8268 , with inverse dye labeling used in replicates to control for dye-specific effects . At least three biological replicates were performed for all comparisons . Data were filtered ( retaining unflagged spots with R2>0 . 1 ) and normalized by regional mean-centering [78] . Genes with significant expression differences ( compared to the S288c control , strain YPS163 , or the mean expression across all strains ) were identified separately for each strain with a paired t-test ( or unpaired t-test in reference to YPS163 ) using the BioConductor package Limma v . 2 . 9 . 8 [79] and FDR correction [80] , taking p<0 . 01 as significant unless otherwise noted ( see Table S3 for limma output and Figure S2 for a comparison of the statistical power for each strain ) . All microarray data are available through the NIH Gene Expression Omnibus ( GEO ) database under accession number GSE10269 . Array-based comparative genomic hybridization ( aCGH ) was performed in duplicate on six strains ( K9 , M22 , RM11-1a , Y10 , YJM789 , and YPS1009 ) relative to the DBY8268 control as previously described [81] , using amino-allyl dUTP ( Ambion ) , Klenow exo-polymerase ( New England Biolabs ) , and random hexamers . Post-synthesis coupling to cyanine dyes ( Flownamics ) was performed using inverse dye labeling in replicate experiments . Technical variation in hybridization was defined as the mean+2 standard deviations ( a log2 value of 0 . 3 ) of all spot ratios , based on triplicate comparisons of DBY8268 to DBY8268 genomic DNA . For non-lab strains compared to DBY8268 , genes with negative aCGH ratios outside the range of technical variation on both duplicates were defined as those affected by copy number and/or hybridization defects . Transcript levels within 0 . 45 ( 3 standard deviations of technical variation ) of the aCGH ratio were identified as those largely explained by copy number and/or hybridization defects – on average , fewer than 5% of genes with statistically significant ( FDR = 0 . 01 ) differential expression compared to DBY8268 fell into this class . Genes with a positive aCGH ratio >0 . 7 in log2 space were defined as genes with increased copy number in each non-lab strain . All microarray data are available through the NIH Gene Expression Omnibus ( GEO ) database under accession number GSE10269 . A vector of relative phenotype scores was generated by dividing scores from Figure 2 by the score measured for DBY8268 . The Pearson correlation between this vector and the measured expression vector for each strain relative to DBY8268 was calculated for all genes in the dataset . Genes whose expression was correlated above or below what was expected by chance ( p<0 . 01 ) were defined based on 100 permutations of each of the ∼6 , 000 expression vectors .
|
Much attention has been given to the ways in which organisms evolve new phenotypes and the influence of the environment on this process . A major focus of study is defining the genetic basis for phenotypes important for organismal fitness . As a first step toward this goal , we surveyed phenotypic variation in diverse yeast strains collected from different environments by characterizing variations in stress resistance and genomic expression . We uncovered many phenotypic differences across yeast strains , both in stress tolerance and gene expression . The similarities and differences of the strains analyzed uncovered phenotypes shared by strains that live in similar environments , suggesting common features of yeast niches as well as mechanisms that different strains use to thrive in those conditions . We provide evidence that some characteristics of strains isolated from oak tree soil have been selected for , perhaps because of the shared selective pressures imposed by their environment . One theme emerging from our studies is that the laboratory strain of yeast , long used as a model for yeast physiology and basic biology , is aberrant compared to all other strains . This result raises caution about making general conclusions about yeast biology based on a single strain with a specific genetic makeup .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"genetics",
"and",
"genomics/genomics",
"genetics",
"and",
"genomics/microbial",
"evolution",
"and",
"genomics",
"genetics",
"and",
"genomics/comparative",
"genomics",
"genetics",
"and",
"genomics/gene",
"expression",
"genetics",
"and",
"genomics/functional",
"genomics"
] |
2008
|
Variations in Stress Sensitivity and Genomic Expression in Diverse
S. cerevisiae Isolates
|
Severe acute respiratory syndrome coronavirus ( SARS-CoV ) encodes a papain-like protease ( PLpro ) with both deubiquitinating ( DUB ) and deISGylating activities that are proposed to counteract the post-translational modification of signaling molecules that activate the innate immune response . Here we examine the structural basis for PLpro's ubiquitin chain and interferon stimulated gene 15 ( ISG15 ) specificity . We present the X-ray crystal structure of PLpro in complex with ubiquitin-aldehyde and model the interaction of PLpro with other ubiquitin-chain and ISG15 substrates . We show that PLpro greatly prefers K48- to K63-linked ubiquitin chains , and ISG15-based substrates to those that are mono-ubiquitinated . We propose that PLpro's higher affinity for K48-linked ubiquitin chains and ISG15 stems from a bivalent mechanism of binding , where two ubiquitin-like domains prefer to bind in the palm domain of PLpro with the most distal ubiquitin domain interacting with a “ridge” region of the thumb domain . Mutagenesis of residues within this ridge region revealed that these mutants retain viral protease activity and the ability to catalyze hydrolysis of mono-ubiquitin . However , a select number of these mutants have a significantly reduced ability to hydrolyze the substrate ISG15-AMC , or be inhibited by K48-linked diubuiquitin . For these latter residues , we found that PLpro antagonism of the nuclear factor kappa-light-chain-enhancer of activated B-cells ( NFκB ) signaling pathway is abrogated . This identification of key and unique sites in PLpro required for recognition and processing of diubiquitin and ISG15 versus mono-ubiquitin and protease activity provides new insight into ubiquitin-chain and ISG15 recognition and highlights a role for PLpro DUB and deISGylase activity in antagonism of the innate immune response .
Ubiquitin ( Ub ) , a 76-amino-acid protein , is the building block for a set of versatile , post-translational modifications that regulate a number of cellular pathways , including many processes associated with combating viral infection [1] . Through the action of activating and conjugating enzymes , the C-terminus of ubiquitin is covalently attached to the ε-amino group of lysine side chains on target proteins , forming an isopeptide bond . The most common ubiquitin modifications are extended to form chains of ubiquitin molecules , linked through a lysine side chain on the proximal ubiquitin and the C-terminus of the neighboring distal ubiquitin . The complexity associated with ubiquitin modifications arises from the ubiquitin lysine residue participating in the polyubiquitin chain . There are 7 lysine residues on ubiquitin , and most are believed capable of forming homotypic chains that mediate different linkage-dependent cellular pathways [2] . The most prevalent ubiquitin linkage and most widely studied , the K48-based chain , directs the modified protein to the proteasome for degradation [3] . Another well characterized Ub modification , the K63-based chain , is commonly associated with regulating endocytic processes , the DNA-damage response , and innate immune response pathways [2] . More recently , the recognition of linear ubiquitin chains has been implicated in the activation of NFκB signaling [4] . In addition to the many possible modifications associated with ubiquitin , a structurally homologous family of proteins comprised of the ubiquitin superfold , the ubiquitin-like proteins ( UBLs ) , can also be conjugated to target proteins to mediate different responses [5] . Examples of ubiquitin-like tags include the SUMO family of proteins , NEDD8 , FAT10 , and ISG15 . Of interest to this study is the interferon-stimulated gene 15 ( ISG15 ) , which is comprised of two tandem ubiquitin-like folds . Conjugation of ISG15 to target proteins ( ISGylation ) is dramatically up-regulated following cellular stimulation by interferons or viral infection [6] . As effector molecules , ubiquitin chains and UBLs such as ISG15 must be recognized by downstream components , through protein-protein interactions , to elicit the appropriate cellular response . The differentiation of various linkages and UBLs by these interacting proteins is particularly important since the nature of the ubiquitin and UBL linkages dictates the downstream response . Recognition of a specific chain type can be accomplished by various ubiquitin binding domains [7] , with one or several of these domains coexisting on a ubiquitin/UBL-interacting protein . As with all other post-translational modifications , ubiquitin chains and UBLs can also be removed from modified proteins by deconjugating enzymes [8] , thereby reversing the effect of the modification on the protein . The human genome encodes approximately 100 deubiquitinating ( or deubiquitylating ) enzymes ( DUBs ) and other deconjugating enzymes ( e . g . deISGylating ) with specificities ranging from linkage-specific to target-specific to promiscuous [8] . The structural basis for how a DUB or downstream Ub/UBL-binding partner deciphers such subtly different modifications is of particular interest . In recent years , a number of cellular and viral proteins have been identified that exploit or interfere with cellular ubiquitin/UBL modifications [9] , [10] , [11] , [12] , [13] . Structural analysis of cellular and viral DUBs in association with ubiquitin chains has revealed the molecular basis for interaction with specific Ub-linkages [14] , [15] , [16] . Distinct sites on both host ( USP21 ) and viral DUBs ( vOTU ) were shown to mediate interaction between the DUB and various Ub-linkage chains . Recently , exploiting knowledge of the interacting residues to guide mutagenesis studies allowed researchers to engineer the equine arterivirus ovarian tumor domain ( vOTU ) DUB enzyme with tailor-made specificity to either polyubiquitin or ISG15 [17] . Here , we explore this strategy to uncouple the viral protease polyprotein processing activity of the papain-like protease ( PLpro ) of Severe Acute Respiratory Syndrome Coronavirus ( SARS-CoV ) from its DUB and deISGylating activities . SARS-CoV emerged into the human population in 2002–2003 from a bat reservoir to cause a pandemic infecting more than 8000 people with a case/fatality ratio of 10% [18] . The Middle East Respiratory Syndrome cornonavirus ( MERS-CoV ) also recently entered the human population and has infected over 130 individuals and so far has killed over 58 individuals . Identifying targets for antiviral therapies and viral factors that contribute to delayed or reduced innate immunes responses are important for developing strategies to control CoV infections . PLpro , required for proteolytic processing of the SARS-CoV replicase polyprotein [19] , [20] , structurally resembles the ubiquitin specific protease ( USP ) class of human deubiquitinating enzymes ( DUBs ) [21] , [22] and has been shown to have both deubiquitinating and deISGylating activities [23] , [24] , [25] . Though the exact cellular targets for these accessory activities are unknown , PLpro has been directly linked to suppressing interferon-β ( IFNβ ) production by the innate immune response during virus infection [26] . More specifically , pathways leading to the activation of both IRF3 and NFκB , essential transcription factors of the ifnb gene , are blocked in the presence of PLpro [27] , [28] . Both K48- and K63-linked ubiquitin chains , as well as ISG15 modifications , have been shown to play important roles in regulating IRF3 and NFκB activation [29] . Here we investigate the molecular recognition of monoubiquitin , K48- and K63-linked ubiquitin chains , and ISG15 by SARS-CoV PLpro , and test whether disrupting these interactions alters antagonism of the innate immune response .
Due to the structural and functional similarities of PLpro with human deubiquitinating enzymes [21] , we sought to examine the structural requirements for ubiquitin recognition by PLpro via X-ray crystallography . In order to co-crystallize PLpro with ubiquitin , a semisynthetic version of ubiquitin containing a C-terminal aldehyde functional group was employed for crystallization . Ubiquitin aldehydes ( Ubals ) can modify the catalytic cysteine of deubiquitinating enzymes in a covalent but reversible manner , and can account for a 300 , 000-fold increase in binding affinity relative to unmodified ubiquitin [30] . The PLpro-Ubal structure was determined to a resolution of 2 . 75 Å ( Table S1 ) , with one PLpro and one Ubal monomer comprising the asymmetric unit . The substrate-free ( unbound ) structure of PLpro , which we previously described , presents a large , unobstructed binding surface for ubiquitin , but access to the enzyme's active site is blocked by a flexible , glycine-hinged β-turn [21] . As expected for ubiquitin-specific proteases , the body of ubiquitin makes contacts with the palm and fingers regions of PLpro ( Figure 1A ) , with little perturbation to the overall enzyme structure . Due to the considerably smaller size of PLpro's palm and fingers regions relative to those of its human USP counterparts , HAUSP ( USP7 ) and USP14 , the ubiquitin molecule buries much less surface area upon binding to PLpro , forming a 900 Å2 protein-protein interface , compared to a 3600 Å2 HAUSP-ubiquitin interface [31] . The interactions at the PLpro-Ubal protein-protein interface are comprised of van der Waals contacts and both direct and water-mediated hydrogen bonds ( Figure 1B ) . As observed for HAUSP [31] and USP2 [32] interactions with ubiquitin , a number of water molecules line the protein-protein interface including a loop of four ubiquitin residues ( A46-Q49 ) that is involved in a number of polar interactions with the palm of PLpro ( Figure 1C ) . Directly adjacent to this loop , the hydrophobic patch of ubiquitin ( I44 , V70 , and L8 ) , often the focal point of many ubiquitin-protein interactions [33] , makes extensive contacts with the side chains of M209 and P248 on PLpro ( Figure 1D ) . Despite the complete burial of PLpro's palm domain and a significant portion of the body of ubiquitin , the vast majority of PLpro-ubiquitin interactions involve the 5 C-terminal residues of ubiquitin ( R72-G76 ) . In this area , 12 intermolecular hydrogen bonds align and direct the ubiquitin C-terminus to the catalytic cysteine ( C112 ) in the active site of PLpro ( Figure 2A ) . The flexible beta-turn loop that blocks the tunnel to the PLpro active site in the unbound enzyme undergoes a significant conformational change and becomes significantly more ordered as it interacts with the C-terminus of ubiquitin . The electron density associated with this loop is well resolved ( Figure 2B ) and the motion of the loop opens the active site of PLpro to accommodate and to hydrogen bond to the side chain of R74 on ubiquitin ( Figure 2C ) . The large number of hydrogen bonds and other contacts involved in the interaction between PLpro and the 5 C-terminal residues of ubiquitin ( RLRGG ) suggest that a significant amount of binding energy is contributed by these 5 amino acids of ubiquitin . Indeed , the 5-amino acid peptide RLRGG-AMC is hydrolyzed by PLpro with a kcat/Km = 5 . 5×103 M−1 s−1 . However , the Km value for this peptide substrate is too large to measure experimentally , and the catalytic rate of hydrolysis of this substrate is approximately 13 . 6-fold lower than that of ubiquitin-AMC ( kcat/Km = 7 . 5×104 M−1 s−1 ) [25] . Therefore , other regions of PLpro distant from the active site are important for the binding and catalysis of ubiquitin and UBL substrates . Though PLpro recognizes and cleaves ubiquitin well , presumably at a faster rate than it cleaves its own viral polyprotein substrate [25] , further analysis of PLpro catalyzed reaction kinetics reveals that it has a 30- to 50-fold greater specificity for another ubiquitin-like modifier , ISG15 [24] ( Figure 3 ) , a molecule consisting of two tandem ubiquitin-like domains involved in modulating the innate immune response . Although PLpro has a measurable rate of hydrolysis activity towards Ub-AMC over a concentration range of 1000 nM , it cannot be saturated indicating that mono-ubiquitin interacts weakly with the enzyme ( Figure 3B ) . In contrast , the ISG15-AMC substrate is hydrolyzed at a much more significant rate than Ub-AMC and saturation of SARS-CoV PLpro is observable suggesting that ISG15-AMC is the preferred substrate in terms of catalytic efficiency . To further probe the recognition and catalysis of different ubiquitin and UBL proteins by PLpro , we performed a series of kinetic studies to determine the ability of PLpro to hydrolyze/deconjugate a series of different polyubiquitin chains ( Figure 3A ) and to hydrolyze the AMC group from the substrates Ub-AMC and ISG15-AMC ( Figure 3B ) . In addition , we determined whether some of the products of these reactions ( Ub , ISG15 , K48-Ub2 and K63-Ub2 ) would serve as inhibitors of PLpro ( Figure 3C ) . The results of a time-course comparison between PLpro cleavage of K48- and K63-linked pentaubiquitin chains ( Ub5 ) demonstrate that PLpro rapidly degrades K48-linked chains , with only moderate activity towards K63-linked chains ( Figure 3A ) . The processing of linear tetra-ubiquitin chains by PLpro was also investigated but no activity was observed under any of the experimental conditions tested ( Figure S1 ) . PLpro is therefore able to distinguish between K48-linked , K63-linked and linear polyubiquitin chains , with a significant preference for K48 linkages . Despite PLpro's strong preference for processing K48-Ub chains , the rapid breakdown of K48-Ub5 into its smaller Ubx components does not follow an unbiased trend . Initially , PLpro rapidly cleaves K48-Ub5 into Ub4 , Ub3 , Ub2 and Ub species . Rather surprisingly though , the Ub4 and Ub3 products continue to be rapidly converted to mono- and di-ubiqutin but the levels of Ub2 do not diminish significantly over 3 hours ( Figure 3A , top gel ) . Di-ubiqutin remains a primary product of the K48-Ub5 cleavage reaction even following extended incubation suggesting that the generation of Ub2 during chain cleavage leads to product inhibition . These initial observations led to the hypothesis that PLpro preferentially recognizes a diubiquitin species on the distal side of the isopeptide bond of K48-linked polyubiquitin chains , a pattern that does not extend to K63-based chains ( Figure 3A , bottom gel ) . To further explore this hypothesis , we next determined the extent of inhibition of PLpro hydrolysis of the RLRGG-AMC substrate by mono-Ub , K48-Ub2 , K63-Ub2 and ISG15 ( Figure 3C ) . Over a concentration range of 20 µM , PLpro does not bind to K63-Ub2 or to mono-ubiquitin with sufficient affinity to inhibit PLpro activity . However , ISG15 displays moderate inhibition , and K48-Ub2 dramatically inhibits PLpro activity ( IC50 = 2 . 7±0 . 2 µM ) , implying that ISG15 and K48-Ub2 bind to PLpro significantly more tightly than mono-Ub and K63-Ub2 . With the structure of PLpro bound to ubiquitin aldehyde in hand , we sought to elucidate the potential binding modes of K48-Ub2 and ISG15 to PLpro . Because ISG15 closely resembles a di-ubiquitin molecule ( Figure S2 ) , we hypothesized that PLpro may have a second or extended binding site that recognizes a second ubiquitin or ubiquitin-like domain of ISG15 or K48-Ub2 . These potential binding or recognition sites for polyubiquitin chains on SARS-CoV PLpro are illustrated schematically in Figure 4 . We have chosen here to use nomenclature similar to that used for describing how proteases recognize their peptide substrates . Each amino acid of the substrate distal to the peptide cleavage site is designated P1 , P2 , P3 etc , and each amino acid proximal to the cleavage site is designated P1′ , P2′ , P3′ etc . The recognition subsites for each of these amino acids on the protease are designated S1 , S2 , S3 etc and S1′ , S2′ S3′ etc . Here , we define each ubiquitin or ubiquitin-like domain distal to the isopeptide cleavage site as Ub1 , Ub2 , Ub3 etc , and each ubiquitin or ubiquitin-like domain proximal to the isopeptide cleavage site as Ub1′ , Ub2′ etc . We also define each recognition site or surface on PLpro as either SUb1 , SUb2 etc . We established in the studies presented in Figure 3 that PLpro greatly prefers the binding of K48-Ub2 over mono-ubiquitin and that this recognition must stem from a portion of the second ubiquitin molecule ( Ub2 ) . Previous work by Lindner et al . [24] showed that removal of the N-terminal ubiquitin-like domain ( equivalent to Ub2 ) of ISG15-AMC leads to a 6-fold loss in SARS-CoV PLpro recognition of ISG15 , verifying that this N-terminal domain contributes major determinants to the specificity of PLpro for ISG15 . This information , in conjunction with the knowledge that PLpro does not bind with appreciable affinity to K63-Ub2 ( Figure 3C ) , provided the initial parameters for constructing molecular models of PLpro in complex with ISG15 and K48-Ub2 ( Figure 5 ) . Because the crystal structures of K48-linked [34] , [35] and K63-linked [36] , [37] ubiquitin chains and of ISG15 [38] have been reported , a simple molecular alignment of these structures with the Ubal molecule bound to PLpro reveals important similarities between K48-Ub2 and ISG15 relative to K63-Ub2 ( Figures 5A , 5B and 5C ) . It is widely established that K63-linked ubiquitin chains form an extended , almost linear , conformation [36] , [37] , whereas K48-linked chains have a proclivity for more compact arrangements [34] , [35] . The positioning of K63 on the PLpro-bound Ubal directs the second ubiquitin molecule ( Ub2 ) of a diubiquitin chain away from the body of PLpro ( Figure 5C ) , particularly if the K63 chain remains in the favored extended conformation . This K63-linked model prevents further interactions between the second ubiquitin ( Ub2 ) and its SUb2 subsite ( region indicated by the blue curved line in Figure 4 ) on PLpro . Conversely , the position of K48 on the PLpro-Ubal structure , buried in the palm domain of PLpro , dictates that any diubiquitin chain will likely make extensive contacts with PLpro at both the SUb1 and the SUb2 subsites , though the isopeptide bond joining the two ubiquitin molecules ( Ub1 and Ub2 ) is too flexible to restrict the positioning of the second Ub2 molecule to a predefined SUb2 of PLpro ( Figure 5B ) . Finally , the structure of ISG15 is more rigid than K48-Ub2 , and direct alignment of the C-terminal domain of ISG15 ( equivalent to Ub1 of a diubiquitin chain ) with PLpro-Ubal projects the N-terminal domain of ISG15 ( equivalent to the Ub2 of a diubiquitin chain ) onto a “ridge” defining the perimeter of the fingers domain of PLpro ( Figure 5B ) , which is the potential SUb2 subsite . This ridge region or SUb2 subsite , the white shaded residues in Figures 5A , 5B and 5C , can be accessed by the Ub2 domains of both K48-Ub2 and ISG15 . To test the validity of these initial models proposed in Figure 5 , and to further refine the positioning of ISG15 and K48-Ub2 on the structure of PLpro , we surveyed the uppermost region of the SUb2 ridge of PLpro for amino acids that potentially interact with ISG15 or K48-Ub2 . Seven amino acids were chosen “to cast a wide net” over this potential SUb2 binding surface with the purpose of perturbing ISG15 or K48-Ub2 interactions with PLpro through regional scanning , site-directed mutagenesis ( Figure 5D ) . To ensure that the mutations did not alter the catalytic activity of PLpro or its ability to bind mono-ubiquitin , all purified mutants were first tested for their catalytic activity towards RLRGG-AMC and Ub-AMC substrates ( Figure 6 ) . Mutants that retained high activity towards these substrates , but lost catalytic activity with ISG15-AMC and/or lost the ability to be inhibited by K48-Ub2 , were further investigated ( Figure 6 ) . Most notably , two mutants , E71A and H74A , were fully active with Ub-AMC but their ability to either hydrolyze ISG15-AMC or be inhibited by K48-Ub2 was perturbed ( Figure 6 ) . The E71A mutation caused a 50% drop in activity with ISG15-AMC but did not affect the inhibitory potency of K48-Ub2 , whereas the H74A mutant had 58% activity with ISG15-AMC , relative to wild-type PLpro , and was significantly less inhibited by K48-Ub2 ( 14% inhibition ) than wild-type ( 100% inhibition ) . The positioning of residues E71 and H74 on PLpro ( Figure 5D ) suggests that the N-terminal half of ISG15 ( equivalent to Ub2 ) and Ub2 of K48-Ub2 are interacting with a helical region ( residues D63 to H74 ) at the top of the thumb domain of PLpro . To test this hypothesis , three additional residues in this region were mutated to more accurately define the area of contact between PLpro and the two ligands ( Figure 6 , starred mutants ) . Two of these mutants , F70S and L76S , greatly affected both ISG15 and K48-Ub2 binding without causing detrimental effects to catalysis of Ub-AMC ( Figure 6 , starred mutants ) . F70 and L76 together comprise a solvent-exposed , hydrophobic patch on the thumb domain of PLpro that is neighbored by E71 and H74 , also shown to contribute to binding of ISG15 and K48-Ub2 ( Figure 6 ) . The retention of Ub-AMC hydrolysis activity and loss of recognition for the second ubiquitin or ubiquitin-like domain ( Ub2 ) is a significant result as the decoupling of these two sites has yet to be achieved in other USP enzyme systems and signifies that it is achievable . The biological significance of decoupling diubiquitin from monoubiquitin recognition was explored next . The mutant analysis allowed further refinement of our structural models of PLpro bound to ISG15 or K48-Ub2 , providing an established area of contact on PLpro . Both models were manually edited and refined to reflect a closer association on the thumb domain of PLpro ( Figure 7 ) . The N-terminus of ISG15 was shifted towards the thumb and rotated slightly about the flexible linker connecting the N- and C-terminal domains to present a small hydrophobic region on ISG15 ( V58-P59 ) to the complementary surface ( F70 , L76 ) on the PLpro thumb ( Figure 7B ) . Because there is little information on the conformational flexibility of ISG15 , only conservative manipulations of the structure were made . Conversely , K48-Ub2 has been shown to adopt numerous conformations [39] . Because the hydrophobic patch of Ub ( I44 , V70 , and L80 ) is ostensibly critical to the majority of interactions between ubiquitin and its binding partners [40] , an effort was made to align this patch in the more distal ubiquitin of K48-Ub2 with the identified hydrophobic patch on the PLpro thumb domain ( Figure 7A ) . The structural model is able to position and accommodate this interaction well , further supporting the results of the mutagenesis studies . To determine if PLpro mutants , particularly at residue F70 , affect PLpro deubiquitinating activity , we analyzed this activity in transfected cells ( Figure 8A ) . Wild-type PLpro cleaves ubiquitin from host cells in a dose dependent manner . In contrast , mutation of the catalytic cysteine ( C112A ) or residue F70 ( F70A and F70S ) within the ridge region of the SUb2 subsite show a reduction in the dose-dependent deconjugation of ubiquitin from host cell substrates . These data agree with the loss in affinity of PLpro for Ub2 and ISG15 by this and other ridge mutants in vitro ( Figure 6 ) . The ability of the F70S and F70A PLpro mutants to recognize and cleave the viral polyprotein was determined next by testing for their ability to cleave a substrate in a trans-cleavage assay [27] . Importantly , both the F70S and F70A ridge mutants retained protease activity at similar levels to WT PLpro , as demonstrated by the ability to process the nsp2-3 polyprotein substrate ( Figure 8B and Figure S3A ) . As expected , the control C112A mutant was unable to process the nsp2-3 polyprotein . Previous studies documented PLpro as an antagonist of the NFκB signaling pathway [27] , [41] . We sought to determine if mutations in PLpro that alter interactions with K48-Ub affect the K48-polyUb-mediated degradation of IκBα induced upon activation of the NFκB pathway by TNFα [42] . 293T cells were transfected with plasmid DNA encoding IκBα-HA and either PLpro WT or the F70S and F70A ridge mutants . At 16 hours post transfection , cells were treated with TNFα to induce ubiquitination and degradation of IκBα , and protein levels of IκBα and PLpro were monitored after 20 minutes by western blotting ( Figure 8C ) . As expected , IκBα is rapidly degraded in mock-infected cells . In contrast , IκBα is more abundant in cells expressing PLpro and there is no detectable degradation of IκBα after treatment with TNFα . Interestingly , the response of cells to TNFα treatment in the presence of either the C112A catalytic mutant or the F70S or F70A ridge mutants of PLpro were essentially the same as mock-infected cells . Similar data and results were obtained with each individual ridge mutant or with combinations of ridge mutants ( Figure S3 ) . These results are consistent with an inability of these mutants to either hydrolyze the isopeptide bonds ( C112A ) or bind ubiquitin chains ( F70S or F70A ) such as K48-Ub2 and ISG15 or in vitro ( Figure 5 ) and support a model whereby the deubiquitinating activity of WT PLpro is responsible for removing K48-linked Ub from IκBα thereby preventing signaling of NFκB . To further investigate the effect of mutation of the ridge domain on antagonism of the NFκB mediated activation of transcription , we evaluated PLpro WT , the C112A catalytic mutant , and both the F70A and F70S mutant PLpro enzymes in a dual-luciferase reporter assay containing an NFκB response element regulating transcription of firefly luciferase ( Figure 8D ) . In mock-transfected cells , the NFκB-dependent reporter is potently activated by the addition of media containing TNFα . This activation is antagonized in a dose-dependent manner when cells are transfected with different concentrations of wild-type PLpro ( Figure 8D ) . In contrast , the PLpro C112A catalytic mutant and F70S mutants are unable to effectively antagonize the NFκB pathway . These results , combined with the IκBα degradation assay described above , are consistent with a mechanism by which SARS-CoV PLpro blocks NFκB activation , and suggest that the ridge-region of the PLpro SUb2 subsite is important for the antagonism of the NFκB pathway .
Distinguishing the roles of multifunctional enzymes in viral replication is challenging . A critical first step is the proof-of-principle that an enzyme has distinct binding sites for different substrates . For SARS-CoV PLpro , an enzyme with peptidase and isopeptidase activities , the X-ray structure of PLpro in complex with ubiquitin aldehyde , and subsequent modeling of ubiquitin-chain and ISG15 interactions , suggested that a ridge region in PLpro is likely involved in binding ubiquitin-like modifier substrates . Analysis of PLpro mutants in protease and isopeptidase ( DUB and DeISGylase ) activity and competitive binding assays identified a hydrophobic portion of this ridge region in the SUb2 subsite that is essential for robust deISGylaing activity and interactions with ubiquitin chains , but is not required for protease or Ub-AMC hydrolysis activity . Thus , we were able to successfully decouple the viral protease processing activity of PLpro from its ubiquitin-chain and ISG15 hydrolysis activities . The use of co-crystal structures to identify ubiquitin and ISG15 interacting regions and subsequent re-engineering of enzymatic activity has also been applied to the study of cellular DUBs such as USP21 , and to another class of viral proteases that also have DUB activity such as the ovarian tumor ( OTU ) domains of Crimean Congo Hemorrhagic Fever Virus ( CCHFV ) [14] , [15] , [16] . Analysis of USP21 with linear diubiquitin revealed an ubiquitin-specific SUb2 ( S2 ) binding site that , when re-engineered , weakened interactions with diubiquitin and reduced the ability of the enzyme to remove ISG15 from cellular substrates [14] . Co-crystal structure studies of the viral OTU domain from CCHFV revealed a critical hydrostatic interaction with ISG15 . Mutagenesis of vOTU residues interacting with ISG15 , specifically Q16R [15] or E128A [16] , resulted in enzymes that were dramatically reduced in the ability to hydrolyze ISG15 while retaining the ability to hydrolyze polyubiquitin . Thus , co-crystal structure analysis of multifunctional enzymes can reveal novel sites that allow for separation of enzymatic activity and allow for targeted development of therapeutics . The structure of PLpro bound to ubiquitin-aldehyde illustrates how the multi-domain architecture of an enzyme can distinguish different ubiquitin chains and ubiquitin-like proteins by forming the support structure for two separate ubiquitin binding domains . For PLpro , these binding sites are in the palm domain ( SUb1 ) for the interaction with Ub1 ( red shaded region in Figure 4 ) and a hydrophobic patch , termed the “ridge” domain ( SUb2 ) for the interaction with Ub2 ( blue shaded region in Figure 3D ) . The canonical right-hand domain layout of ubiquitin-specific proteases ( USPs ) allows for a large , primary ubiquitin binding surface on the palm and finger domains , with the C-terminus of ubiquitin firmly coordinated in the active site through numerous intermolecular hydrogen bonds [8] . In this conformation , the hydrophobic patch of Ub1 is well positioned to interact with hydrophobic residues on the palm and fingers of PLpro ( Figure 1 ) . This general mode of binding is relatively conserved across known structures of USPs with [8] , though the tilt of Ub relative to its C-terminus , or size of the protein-protein interface , can be influenced by accessory loops , or by the relative sizes of the domains themselves . These factors also dictate the environment of different ubiquitin lysine residues , and thus the potential for a USP to bind a particular type of ubiquitin linkage internally . USP14 , a human deubiquitinating enzyme associated with the proteasome [43] , is specific for K48-linked Ub chains , which it removes from proteins immediately prior to their degradation in the proteasome . USP14 is comprised of a much larger fingers domain relative to that of PLpro , which may obstruct extended K63 Ub chains from binding and being cleaved by USP14 [44] . Conversely , CYLD , a human DUB that negatively regulates the NF-kB pathway , is specific for K63-based Ub chains . CYLD contains a truncated fingers domain , which is expected to accommodate extra ubiquitin moieties extending from K63 of the primary ubiquitin [45] . It has been shown that certain proteins capable of distinguishing different ubiquitin chains do so at the region around the linkage [4] , [46] . In the case of the DUBs , this recognition would occur at the scissile isopeptide bond in the active site , or rather how the proximal ubiquitin ( Ub1′ ) is oriented at the exit of the active site ( Figure 4 ) . Assuming equivalent binding of Ub1 at the SUb1 subsite , UB1′ would be presented differently to the SUb1′ subsite of a DUB depending on the linkage . AMSH-LP , a zinc-dependent DUB that regulates receptor trafficking , favors K63-Ub linkages by recognizing the particular ubiquitin surface around K63 of Ub1′ , allowing for correct alignment of the isopeptide bond in the active site [46] . Though it is not directly clear if SARS-CoV PLpro favors cleavage of one type of linkage over the other at the active site , our results suggest that the binding of ubiquitin species in at least two Ub subsites ( SUb1 and SUb2 ) located distal to the isopeptide bond in the active site are crucial for recognition . Due to the inhibitory nature of K48-Ub2 ( Figure 3C ) , we can infer that this species binds more tightly to the SUb2-SUb1 subsites instead of the SUb1-SUb1′ subsites since the concentration levels of K48-Ub2 increase during the hydrolysis of K48-linked ubiquitin chains causing product inhibition and hence the reduction of cleavage of K48-Ub2 to mono-ubiquitin . Otherwise , the K48-Ub2 species would be expected to be efficiently cleaved to monoubiquitin ( Figure 3A ) . Thus , K48-Ub2 strongly favors occupation of the SUb2-SUb1 subsites of SARS-CoV PLpro ( red and blue shaded regions in Figure 4 ) in favor of spanning the Ub1-Ub1′ binding sites . The activity of SARS-CoV PLpro towards Ub-AMC rivals those of the most active human DUBs that have been characterized so far . For example , UCH-L3 [47] has a kcat , Ub-AMC value of 9 . 1±0 . 1 s−1 , and USP8 [48] has a kcat , Ub-AMC value of 2 . 4 s−1 [14] , [15] . SARS-CoV PLpro is highly active towards the substrate ISG15-AMC ( Figure 3B ) with a kcat , ISG15-AMC value of 6 . 2±0 . 3 s−1 . This turnover rate is so far unmatched by any other human or viral DUB studied to date . The most active vOTU is from the Crimean-Congo hemorrhagic fever virus with a kcat , ISG15-AMC = 0 . 54±0 . 05 s−1 [49] , [50] and the most active USP so far is human USP21 with a kcat , ISG15-AMC = 5 . 9×10−3±1 . 3×10−3 s−1 . We attempted to determine the kinetic rates of hydrolysis of ISG15-AMC by the human deISGylase USP18 ( UBP43 ) , but the rate was at least 103 slower than that with PLpro , which is consistent with other studies . Though the exact outcome of PLpro's deISGylating activity during viral infection is currently unknown , ISG15 is an important activator of the immune response . Recent studies have shown that ISG15 is conjugated to newly synthesized proteins , a potential cellular strategy to disrupt the function of viral proteins , which are abundantly produced during infection [51] . SARS-CoV and other viruses , including nairo- and arteriviruses [52] , may have evolved deISGylating strategies to counteract such defenses . ISG15 has also been shown to positively regulate IRF3 activation by preventing an inhibitor , Pin1 , from binding to IRF3 [53] . SARS-CoV PLpro , which has been shown to prevent IRF3 activation [26] , may deISGylate IRF3 or bind to IRF3 through a conjugated ISG15 molecule to prevent its activation . It is intriguing to speculate that the DUB activity of SARS-CoV PLpro is important either for viral replication or pathogenesis . SARS-CoV PLpro is extremely efficient at deconjugating K48-linked chains and has measurable activity with K63-based chains . K48-based Ub conjugation has been demonstrated to be an important positive activator of the NFκB pathway [54] . Moreover , K63-based chains have been shown to be key components of the innate immune response as they mediate protein complex formation [1] . Three human DUBs , CYLD , A20 and DUBA , have been shown to negatively regulate the innate immune response by removing K63-based chains [55] , [56] , and USP15 inhibits the NFκB pathway by removing K48-Ub from IκBα , preventing its degradation [57] . We examined the effect of mutations within the ridge of PLpro on NFκB antagonism ( Figures 8 and S3 ) . IκBα is phosphorylated and degraded in response to TNFα , an activator of the NFκB pathway . The presence of PLpro prevents this degradation , and leads to an increase in basal levels of IκBα . These data indicate that PLpro is preventing the degradation and normal turnover of IκBα . The F70 ridge mutants ( F70A and F70S ) of SARS-CoV PLpro are unable to prevent the degradation of IκBα , suggesting that the K48-linked deubiquitinating activity of PLpro is required for PLpro's ability to antagonize the NFκB pathway . Importantly , Frieman et al . have shown that PLpro does not block TNFα stimulated IκBα phosphorylation [27] supporting a model whereby the stabilization of cellular IκBα is achieved through its K48-linked deubiquitination by PLpro . By separating the protease and isopeptidase activities of SARS-CoV PLpro , we have shown that K48-linked ubiquitin interaction is a mechanism behind the NFkB antagonism functions of SARS-CoV PLpro and that viral deubiquitinating enzymes might be critical determinants of pathogenesis . The recent outbreak of the new MERS-CoV [58] , a coronavirus that produces SARS-like symptoms in patients and has a high mortality , emphasizes the importance of continuing studies on the pathogenesis of deadly coronaviruses . MERS-CoV encodes a single papain-like protease ( PLpro ) similar to SARS-CoV , but the primary structure is more homologous to the bat coronavirus HKU5 ( Figure S4 ) . Comparing the amino acid sequences of SARS-CoV and MERS-CoV , specifically the amino acids that were mutated in this study and shown to have an effect on ubiquitin-like modifier recognition ( Figure 6 ) , reveals that there is little to no sequence conservation among these residues in the ridge region . This observation suggests that MERS-CoV PLpro is likely to recognize and process ubiquitin and ISG15 substrates differently than SARS-CoV PLpro . Since these residues are in the ridge region of the SUb2 subsite and are involved in binding Ub2 , the differences are most likely to be found in ubiquitin-like chain recognition . It will be interesting to compare the polyubiquitin chain recognition and cleavage properties of MERS-CoV to SARS-CoV PLpro to test this hypothesis . To date no FDA-approved vaccines or antivirals are available for human coronaviral infections , so information about SARS-CoV virulence factors are of critical importance [59] . Attenuation of SARS-CoV by affecting PLpro DUB activity provides a novel strategy in SARS-CoV and SARS-like coronavirus vaccine development . Directly targeting PLpro using antivirals would not only prevent polyprotein cleavage , but could also inhibit the innate immune antagonist functions of PLpro . Furthermore , understanding the molecular basis for ubiquitin chain interactions with cellular DUBs is an emerging strategy for developing specific human USP inhibitors that may be useful in the treatment of cancer , neurologic and infectious diseases [60] . In summary , we have determined the structure of a viral protease , SARS-CoV PLpro , in complex with ubiquitin aldehyde and have used this structure in conjunction with kinetic studies on the interaction of PLpro with Ub chains and ISG15 to guide molecular modeling . The resulting models of PLpro bound to two high affinity ligands , K48-Ub2 and ISG15 , suggest that contacts with these ligands extend beyond the mono-ubiquitin binding site ( SUb1 ) . Site-directed mutagenesis studies identified a region of the thumb domain of PLpro as a second ubiquitin binding site ( SUb2 ) for binding both K48-Ub2 and ISG15 , but not K63-Ub2 , a weaker ligand . Finally , these mutations are relevant to PLpro's function as an innate immune antagonist , as PLpro ridge mutants lose the ability to block NFκB-mediated signaling . Re-engineering of these PLpro mutants or others in the context of SARS-CoV may reveal a role for PLpro DUB activity in viral pathogenesis .
Ubiquitin aldehyde ( Ubal ) was synthesized using the intein-fusion method as previously described [61] with minor modifications . Expression of ubiquitin ( residues 1–75 ) fused to intein and chitin-binding domains ( plasmid pTYB2-Ub1–75 ) was carried out in BL21 ( DE3 ) Codon+ cells ( Novagen ) , using LB auto-induction media [62] supplemented with 50 µg/mL carbenicillin . Following 24 h of growth at 25°C , 3 L of culture were pelleted , resuspended in Buffer A ( 25 mM HEPES , pH 6 . 8 , 50 mM sodium acetate , 75 mM NaCl ) and lysed by addition of Triton X-100 to 0 . 16% followed by sonication . The lysate was clarified by centrifugation and loaded onto a gravity-flow column containing 50 mL of chitin beads ( New England Biolabs ) equilibrated with Buffer A . Following a series of column washes with Buffer A to remove unbound proteins , 25 mL of Buffer A containing freshly prepared 100 mM sodium 2-mercaptoethanesulfonate ( MESNA ) ( Sigma-Aldrich ) was added to the resin to create a slurry . The slurry was incubated overnight at 4°C with gentle rocking . The following day , the resin was separated from the effluent , which contained released ubiquitin thioester , and washed with 80 mL of Buffer A to fully elute any remaining released ubiquitin . Samples containing ubiquitin thioester were combined , concentrated to 20 mL ( 1–3 mg/mL ) , and dialyzed against 5 mM HCl overnight at 4°C . Ubiquitin acetal was generated by reacting 10 mL of ubiquitin thioester with 2 mL of 4M aminoacetaldehyde diethyl acetal , pH 8 . 5 ( Sigma-Aldrich ) and 50 µL of freshly prepared 2M N-hydroxy-succinimide ( Sigma-Aldrich ) and incubated at room temperature . Upon full conversion ( ∼1 hr ) , the ubiquitin acetal was dialyzed against 50 mM sodium acetate , pH 4 . 5 , overnight at 4°C . The acetal was deprotected with 0 . 15 M HCl to produce ubiquitin aldehyde , then quenched with 0 . 15 M Tris base , and subsequently desalted into 20 mM HEPES , pH 7 . 5 . Conversions from thiol ester to acetal and from acetal to aldehyde were monitored by HPLC on a C8 column as previously described [61] . The concentrations of the ubiquitin species throughout the purification were determined by HPLC , using unmodified ubiquitin ( Boston Biochem ) as a standard . The catalytic domain of SARS-CoV PLpro , polyprotein residues 1541–1855 , was expressed and purified as previously described [25] . To form the PLpro-Ubal complex , 30 mg of purified PLpro was added to 11 mg of Ubal in 20 mM HEPES , pH 7 . 5 , and the complex was incubated overnight at 4°C . The activity of PLpro was monitored to ensure full inactivation of the protease by Ubal . The resulting complex was purified from unreacted Ubal and hydrolyzed ubiquitin , a byproduct of aldehyde synthesis , using a 10/100 GL Mono-Q column ( GE Healthcare ) at pH 7 . 5 . Fractions containing PLpro-Ubal were pooled and concentrated to 10–15 mg/mL . PLpro-Ubal crystals were grown by vapor diffusion at room temperature from hanging drops containing 1 µL of protein complex ( 3–12 mg/mL PLpro-Ubal in 20 mM Tris , pH 7 . 5 ) and 3 µL of precipitant ( 0 . 1 M HEPES , pH 7 . 5 , 10% isopropanol , 20% PEG 4 , 000 ) . Crystals , which grew within two days , were flash-frozen in liquid nitrogen without the need for additional cryo-protectants , and then transferred to a SPINE puck system for X-ray data collection at the Advanced Photon Source of Argonne National Laboratory . X-ray data were collected at LS-CAT on beamline 21-ID-F to 2 . 76 Å and resulted in an overall Rmerge value of 6 . 3% with 98% completeness . Data were processed and scaled using the HKL2000 program suite [63] , and crystals of PLpro-Ubal were indexed to the P3121 space group with unit cell dimensions of a = b = 47 . 12 Å and c = 332 . 56 Å , with one monomer in the asymmetric unit . Phases for the complex were obtained from a molecular replacement solution using a monomer of the unbound PLpro structure ( PDB entry: 2FE8 ) and free ubiquitin ( PDB entry: 1UBQ ) as search models using Phaser [64] , which identified one unique molecular replacement solution of the complex . Rigid body refinement followed by iterative rounds of restrained refinement and modeling building using the programs Refmac5 [65] and WinCoot [66] reduced the R-factors to 18 . 7% ( Rcryst ) and 27 . 9% ( Rfree ) . The final structure and associated structure factors have been deposited in the PDB under PDB ID 4MM3 and RCSB ID RCSB082081 . Site-directed mutations were introduced into a pET15b-PLpro construct using synthetic primers purchased from Integrated DNA Technologies and the QuikChange Site-Directed Mutagenesis Kit ( Stratagene ) . All mutations were verified by sequencing . The pET15b-PLpro construct expresses an N-terminal His6-tag followed by the ubiquitin-like and catalytic domains of SARS-CoV PLpro ( polyprotein residues 1541–1855 ) . Wild-type and mutant versions of PLpro were expressed in BL21 ( DE3 ) cells and were purified using Ni-NTA agarose ( Qiagen ) . Enzyme concentration was determined using the Bradford protein assay . Proteolytic cleavage of homogeneous , K48-linked or K63-linked penta-ubiquitin ( Boston Biochem ) was carried out under the following conditions: 0 . 07 µg of purified PLpro ( 23 nM ) was incubated with 50 µg of K48-Ub5 or K63-Ub5 at 25°C in an 85 µL volume containing 50 mM HEPES , pH 7 . 5 , 0 . 1 mg/mL BSA , 100 mM NaCl , and 2 mM DTT . A control reaction was incubated under identical conditions with the exclusion of enzyme . At various time points , 10 µL were removed and quenched with the addition of SDS-PAGE sample loading dye to a 1× concentration ( 25 mM Tris , pH 6 . 8 , 280 mM β-mercaptoethanol , 4% glycerol , 0 . 8% SDS , 0 . 02% bromophenol blue ) , and heat treated at 95°C for 5 minutes . The samples were analyzed by electrophoresis on a 4–12% SDS-PAGE and stained with Coomassie dye . Wild-type PLpro inhibition by free Ub , ISG15 , K48-Ub2 , and K63-Ub2 was measured in a 96-well plate format in duplicate at 25°C . Assays contained 50 mM HEPES , pH 7 . 5 , 0 . 1 mg/mL BSA , 5 mM DTT , an unsaturating concentration of the substrate RLRGG-AMC ( 37 . 5 µM ) , 45 nM purified PLpro , and varying concentrations of Ub , ISG15 , K48-Ub2 , or K63-Ub2 ( Boston Biochem ) . Reaction progress curves were monitored continuously using a Tecan Genios Pro plate reader , where free AMC fluorescence was measured at λexcitation = 360 nm , λemission = 460 nm . Where applicable , the percent inhibition of the reaction was calculated relative to PLpro activity in the absence of the ligand . An IC50 value was calculated using the equation v = a/ ( 1+ ( [I]/IC50 ) where v is the velocity of the reaction in the presence of inhibitor , [I] is the concentration of the inhibitor , a is the velocity of the reaction in the absence of inhibitor , and IC50 is the concentration of inhibitor resulting in 50% inhibition of enzyme . Data were fit to the equation using the Enzyme Kinetics module of Sigmaplot ( Systat Software ) . All reactions were carried out at 25°C in 50 mM HEPES , pH 7 . 5 , 0 . 1 mg/mL BSA , and 5 mM DTT . AMC fluorescence was detected as above . Wild-type and mutant activities with 400 nM Ub-AMC and 400 nM ISG15-AMC were measured at 15 nM and 0 . 15 nM enzyme , respectively . Inhibition by K48-Ub2 was measured by comparing enzyme activity ( at 63 nM ) with 37 . 5 µM RLRGG-AMC in the absence and presence of 5 µM K48-Ub2 . The models of K48-Ub2 , K63-Ub2 , and ISG15 bound to PLpro were initially generated by superimposing the most proximal portions of each molecule onto the ubiquitin molecule in the PLpro-Ubal structure . The PDB files utilized for each model are as follows: ISG15 ( 1Z2M ) , K63-Ub2 ( 3H7S ) , K48-Ub2 ( 2KDF ) . Only the model of PLpro-K48-Ub2 required further editing , with the distal ubiquitin being manually manipulated about the isopeptide bond in Pymol to alleviate clashes with PLpro . HEK293T cells in 12-well Cell-Bind plate ( Corning ) were transfected ( LT1 , Mirus ) with 300 ng FLAG-Ub and either 125 ng , 250 ng , or 500 ng of pcDNA3 . 1-SARS-PLpro per well . Cells were incubated for 18 hours then lysed with IkBα lysis buffer ( 20 mM Tris ( pH 7 . 5 ) , 150 mM NaCl , 1 mM EGTA , 1 mM EDTA , 1% Triton X-100 , 2 . 5 mM Na pyro-phosphate , 1 mM Beta-glycerophosphate , 1 mM Na ortho-vanadate , 1 ug/ml Leupeptin ) and incubated for 20 min on ice . Lysates were subjected to centrifugation at 4C and the cytoplasmic contents added to 2× sample buffer and separated by SDS-PAGE on a 4–20% gradient gel ( BioRad ) . Gel was transferred to PVDF using semi-dry apparatus ( BioRad ) and immunoblotted with anti-FLAG ( Sigma ) , anti-V5 ( Invitrogen ) , and anti-calnexin ( BD ) . HEK293 cells were transfected with constructs expressing nsp2-3-GFP and SARS-CoV PLpro-V5 wild type , catalytic mutant ( C112A ) or ridge mutants ( F70A , F70S ) . Cells were incubated for 24 hours and then lysed with IkBα lysis buffer ( 20 mM Tris ( pH 7 . 5 ) , 150 mM NaCl , 1 mM EGTA , 1 mM EDTA , 1% Triton X-100 , 2 . 5 mM Na pyro-phosphate , 1 mM Beta-glycerophosphate , 1 mM Na ortho-vanadate , 1 ug/ml Leupeptin ) . Lysates were subjected to centrifugation at 4C and the cytoplasmic contents added to 2× sample buffer and separated by SDS-PAGE on a 10% PAGE gel ( BioRad ) . Samples were transferred to PVDF immunoblotted with anti-GFP ( Life Technologies ) and anti-V5 ( Invitrogen ) . HEK293 cells were plated in 60 mm dishes and transfected with 3 ug pcDNA3 . 1 – PLpro and 250 ng pIkBα-HA per dish ( Mirus LT-1 ) [42] . Cells were incubated for 24 hours , media , removed , and fresh media containing TNFα ( Roche ) at a final concentration of 20 ng/ml was added and incubated with cells . Cells were incubated with TNFα for the indicated time points . Cells were then lysed with an IkBα lysis buffer: 20 mM Tris ( pH 7 . 5 ) , 150 mM NaCl , 1 mM EGTA , 1 mM EDTA , 1% Triton X-100 , 2 . 5 mM Na pyro-phosphate , 1 mM Beta-glycerophosphate , 1 mM Na ortho-vanadate , 1 ug/ml Leupeptin . Lysates were kept on ice and centrifuged at 14000× g for 10 min at 4C . After centrifugation , supernatant was added to equal volume 2× sample buffer , boiled for 5 min , and then immediately run on 10% SDS-PAGE gel . Gel was transferred to PVDF using a semi-dry blot apparatus ( TransBlot Turbo , Bio-Rad ) and blocked in 5% milk overnight . Antibodies were used at a concentration of 1∶5000 ( mαHA , Covance; mαV5 , Invitrogen ) . Western blots were quantified using a Typhoon Image Scanner and ImageQuant5 . p< . 05 indicates a significant difference from WT PLpro transfected cells and was determined using a student t test with Systat . HEK293 cells were plated in 24-well plates ( Cell-Bind from Corning ) and transfected with 50 , 100 , or 150 ng of pcDNA3 . 1 – PLpro plasmid , 50 ng pNFkB-luc , and 25 ng pRL-TK ( Renilla control ) per well in triplicate ( Mirus LT-1 ) . Cells were incubated for 12 hours , media removed , and fresh media containing TNFα ( Roche ) at a final concentration of 10 ng/ml was added and incubated with cells for 4 hours . Cells were then lysed with passive lysis buffer and wells assayed for dual luciferase expression ( Promega ) by luminometer . p< . 05 indicates a significant difference from mock transfected cells and was determined using a student t test with Systat .
|
All coronaviruses such as the SARS virus and the recently identified Middle East Respiratory Syndrome ( MERS ) virus encode in their genomes at least one papain-like protease ( PLpro ) enzyme that has two distinct functions in viral pathogenesis . The first function is to process the viral polyprotein into individual proteins that are essential for viral replication . The second function is to remove ubiquitin and ISG15 proteins from host cell proteins , which likely helps coronaviruses short circuit the host's innate immune response . The 3-dimensional structure of SARS virus PLpro in complex with a human ubiquitin analog was determined and reveals how coronavirus PLpro enzymes strip ubiquitin and ISG15 from host cell proteins at the molecular level . A series of amino acid residues involved in interactions between PLpro and ubiquitin were mutated to identify which interactions are important only for the recognition of ubiquitin and ISG15 modified proteins by PLpro and not for recognition and cleaving of the viral polyprotein . The 3D structure of SARS PLpro with ubiquitin-aldehyde sheds significant new light into how PLpro interacts with ubiquitin-like molecules and provides a molecular road map for performing similar studies on other deadly coronaviruses such as MERS .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"biomacromolecule-ligand",
"interactions",
"biochemistry",
"enzyme",
"structure",
"proteins",
"enzymes",
"biology",
"and",
"life",
"sciences",
"enzymology",
"biophysics"
] |
2014
|
Structural Basis for the Ubiquitin-Linkage Specificity and deISGylating Activity of SARS-CoV Papain-Like Protease
|
In addition to modulating the function and stability of cellular mRNAs , microRNAs can profoundly affect the life cycles of viruses bearing sequence complementary targets , a finding recently exploited to ameliorate toxicities of vaccines and oncolytic viruses . To elucidate the mechanisms underlying microRNA-mediated antiviral activity , we modified the 3′ untranslated region ( 3′UTR ) of Coxsackievirus A21 to incorporate targets with varying degrees of homology to endogenous microRNAs . We show that microRNAs can interrupt the picornavirus life-cycle at multiple levels , including catalytic degradation of the viral RNA genome , suppression of cap-independent mRNA translation , and interference with genome encapsidation . In addition , we have examined the extent to which endogenous microRNAs can suppress viral replication in vivo and how viruses can overcome this inhibition by microRNA saturation in mouse cancer models .
MicroRNAs ( miRNAs ) are a class of small , ∼22 nt regulatory RNAs that modulate a diverse array of cellular activities . Through recognition of sequence complementary target elements found most often in the 3′UTR of cellular mRNAs , miRNAs post-transcriptionally regulate numerous cellular processes by way of mRNA translation inhibition or , less commonly , by catalytic mRNA degradation . It is thought that upwards of one-third of all human mRNAs are regulated by the over 700 human miRNAs that are currently known [1] , [2] . Many miRNAs can have tissue-specific localizations and , in addition , some are now known to have cancer-specific signatures . Cancer-specific miRNAs can be both oncogenic and oncosuppressive , and growing evidence now indicates that certain miRNAs are also involved in disease progression , through the promotion of metastasis [3] . The mechanisms by which a miRNA regulates a given mRNA are influenced by parameters such as the degree of sequence homology [4] and target site multiplicity [5] as well as by features of the mRNA itself , including target site secondary structure [6] and location [7] . In addition , the cellular machinery used to translate mRNAs is thought to profoundly affect miRNA regulation . While capped mRNAs are known to be amenable to both catalytic miRNA-induced cleavage and miRNA-mediated translational repression , it has been suggested that uncapped mRNAs that rely on an IRES ( Internal Ribosome Entry Site ) for translation initiation are not susceptible to translational repression [8][9] . In addition to their roles in the pathogenesis of human disease , endogenous cellular miRNAs can also play a role in viral infection , acting to suppress [10][11][12][13] or enhance [14] viral replication . Recently , miRNAs have been exploited to influence the tissue tropism and pathogenicity of viruses used as vaccines [10] , anticancer therapeutics [13] , and gene transfer vehicles [15] . MiRNAs of both cellular and viral origin are thought to be involved in regulating the host response to viral infection [16] and miRNAs have been shown to regulate viral antigen presentation [17] , the antiviral interferon response [18] , viral tissue tropism and antiviral immunity [19] . Engineering miRNA-responsiveness in viruses used as cancer therapeutics has been shown to be an effective way to generate tumor selectivity [19] . And although miRNAs clearly play a role in multiple aspects of both the viral replication cycle and the host response to viral pathogens , little is known about the mechanisms by which these regulatory molecules act to directly influence the level of virus replication . To this end , we engineered the oncolytic picornavirus Coxsackievirus A21 ( CVA21 ) , to encode artificial miRNA targets complementary to cellular miRNAs . Here , we look at the ability of these miRNA targets to restrict viral replication in cells expressing cognate miRNAs and the mechanisms by which they are able to do so . By utilizing CVA21 derivatives bearing different miRNA target elements ( miRTs ) , including targets with varying degrees of homology to endogenous cellular miRNAs and targets in orientations designed to target either the positive-strand RNA genome or the negative-strand antigenome , we were able to identify multiple , distinct steps in the picornaviral life cycle that are amenable to miRNA-mediated regulation . CVA21 is a picornavirus known to cause upper respiratory and inflammatory muscle infections in humans [20][21] . It replicates quickly and efficiently in both cell culture and in mouse models , and is a typical positive-strand RNA virus . The genome of CVA21 is a single-stranded , uncapped , positive-strand RNA , which is translated to yield a polyprotein that is then cleaved by viral proteases to form the viral capsid and nonstructural proteins . We have previously reported that CVA21 has potent oncolytic activity against a variety of human cancers , and can mediate complete tumor regression in mice bearing human melanoma or myeloma xenografts [13] . However , this potent and curative oncolytic activity is accompanied by a high-level viremia and rapid-onset lethal myositis that hampers the feasibility of employing this virus as an anticancer therapeutic in the clinic . Incorporation of miRNA target elements ( miRTs ) corresponding to two muscle-specific cellular miRNAs ( miR-133 and miR-206 ) was shown to mediate silencing of CVA21 gene expression in cells expressing muscle-specific miRNA mimics , in muscle culture lines and , most importantly , in vivo in mice . Cells expressing exogenous or endogenous miRNAs complementary to miRTs are intrinsically immune to infection by the engineered CVA21 and are protected from the cytolytic effects of this virus . Thus , tumor-bearing mice infected with a recombinant muscle-restricted virus ( CVA21 miRT ) were cured of established tumors whilst being protected from productive muscle infection and the resultant myositis . Our ability to engineer the oncolytic picornavirus CVA21 to contain diverse miRTs provided an opportunity to look at the mechanisms by which cellular miRNAs can act directly on a virus to silence gene expression and mediate tumor selectivity .
To investigate at what stage ( s ) of the viral life cycle cellular miRNAs are able to perturb viral replication , we utilized our previous recombinant miRT design whereby four tandem copies of a given target element corresponding to a cellular miRNA are incorporated into the 3′UTR of the viral genome ( Figure 1A ) . In order to identify a cellular miRNA that could effectively inhibit CVA21 , we engineered the CVA21 genome to encode tandem miRTs complementary to four different tissue-specific miRNAs . Target elements corresponding to muscle-specific ( miR-133 and miR-206 ) , hematopoetic-specific ( miR-142-3p ) or tumor-suppressor ( miR-145 ) miRNAs were incorporated in the 3′UTR of CVA21 and protection of HeLa cells transfected with sequence-complementary miRNA mimics ( synthetic dsRNAs corresponding to cellular miRNA duplex intermediates ) was analyzed ( Figure 1B ) . While miRNA-mimics were able to protect cells infected with viruses containing sequence complementarity from ∼60% to levels that were not significantly different from control , the hematopoetic-specific miR-142-3p target provided the most complete and most consistent protection , and hence provided the best candidate to study the different steps of virus replication at which cellular miRNAs could act to perturb viral replication . To identify miRNA-mediated antiviral activity acting at distinct steps in the viral life cycle , we designed a panel of recombinant viruses expressing variations of the miR-142-3p target . Four tandem fully complementary sites were designed to provide the best opportunity for direct catalytic cleavage of the viral RNA genome and mRNA ( CVA21 142T ) ; four copies in reverse orientation were designed to address the existence of miRNA-mediated antiviral activity acting on the negative-sense antigenome ( CVA21 142rT ) . In addition , two viruses were constructed to look at the potential for translational silencing of the viral mRNA: a virus containing 7 base pairs of mismatch between the target and miR-142-3p ( CVA21 mm7T ) and a virus that contained the entire miR-142-3p target site with a 6 base pair ( bp ) stuffer sequence inserted between bp 10/11 of the target site ( CVA21 b6T ) , each with four replicate copies ( Figure 1C ) . The latter , or ‘bulge’ , virus contained extra sequence at the catalytic cleavage site to provide a strong candidate for miRNA-mediated translational suppression: four perfect copies of the miRT to promote miRNA recognition , and a “stuffer sequence , ” at the site at which the RNA induced silencing complex ( RISC ) ‘slicer’ activity cleaves ( to block catalytic degradation of the viral RNA ) . These recombinant viruses were rescued in the absence of the corresponding miRNAs and found to replicate to high titer with growth kinetics analogous to the wild-type ( WT ) virus in cells lacking miR-142-3p ( Figure 1D ) . Cellular microRNAs are known to mediate post-transcriptional silencing of sequence-complementary mRNAs through translational repression and/or transcript degradation . However , in the absence of a 5′ cap , and in the presence of the IRESes utilized by many viruses , it has been hypothesized that miRNAs are unable to induce translational repression [22] . These studies have been performed primarily with reporter mRNAs containing one of several viral IRESes and generally in the presence of short interfering RNAs ( siRNAs ) , cell free extracts , or miRNA mimics supplemented in trans [8] , [23] [24] . Translation of the CVA21 mRNA into the viral polyprotein is mediated by a type I IRES in a cap-independent manner . We reasoned that the analysis of recombinant CVA21 variants targeted by miRNAs might be particularly useful in studying the translational silencing of cap-independent mRNAs because cells can be infected at very low multiplicities of infection ( MOI ) , thus circumventing the potentially confounding issue of miRNA saturation [25] , and enabling the quantitative analysis of virus expansion whilst simultaneously allowing precise measurement of cell death at specific times post-infection . Similar to experiments conducted by other groups [9] [23] , we found that in the presence of miRNA mimics added in trans , viruses containing multiple imperfectly complementary target sites were unresponsive to miRNA-mediated repression and cells were not significantly protected from viral cytopathic effects ( Figure 2A ) . Increasing miRNA mimic copy numbers did not cause any increase in protection from cytolysis by these viruses ( data not shown ) . However , hematopoetic cell lines and primary hematopoetic cells expressing high levels of endogenous miR-142-3p were significantly less susceptible to viruses containing mismatch ( mm7T ) or bulge ( b6T ) targets ( Figure 2B; p<0 . 01 and p<0 . 004 , respectively ) . In both cases , viral titers were reduced by up to 5 orders of magnitude ( Figure 2C ) . Although targeting the viral antigenome had no significant effect on cell viability , this did reduce viral titers by up to 100 fold ( Figures 2B and 2C ) . To determine whether the observed increase in cell viability was indeed as a result of translational silencing rather than transcript cleavage , we looked at CVA21 RNA copy numbers . A panel of hematopoetic cells , as well as H1-HeLa cells ( serving as a control cell line ) were supplemented with control or miR-142 mimics , and all cells were infected with the panel of recombinant CVA21 viruses . Total cellular RNA was harvested and viral RNA was quantified and normalized to cellular GAPDH mRNA levels . While inserting perfectly complementary miR-142-3p targets ( in CVA21 142T ) reduced CVA21 RNA levels by up to a million-fold , the mismatched CVA21 mm7T and bulged CVA21 b6T targets induced only a modest decrease in viral RNA ( Figure 2D ) . Translational silencing would be expected to indirectly cause a small drop in viral RNA levels through production of less viral polymerase , presumably accounting for the up to 40 fold decrease in CVA21 RNA in the presence of bulged targets ( Figure 2D ) . To test whether CVA21 142T RNA was reduced in the presence of cells bearing sequence-complementary miRNAs due to direct cleavage of the viral RNA , or rather because translational silencing had reduced the amount of viral polymerase to amplify the viral RNA , we performed a trans-complementation experiment . Hematopoetic MEC-1 cells were infected with two different CVA21s: miRT CVA21 ( containing muscle-targets ) , and CVA21 142T ( containing hematopoetic targets ) . In MEC-1 hematopoetic cells , CVA21 142T should be miRNA-targeted , while miRT CVA21 should replicate unencumbered . qRT PCR was then performed to ascertain the total amount of viral RNA in the cells , and what portion of it was miRT CVA21 ( the difference between total CVA21 RNA and miRT CVA21 RNA would therefore be RNA corresponding to CVA 142T in cells doubly infected with miRT CVA21 and CVA21 142T ) . Complementation by miRT CVA21 polymerase should be capable of synthesizing CVA21 142T if any remained that had not been catalytically destroyed , ( however the efficiency of trans-complementation in picornaviruses still remains under debate [26][27][28] ) . Quantitative PCR analysis revealed that all virus present in singly infected miRT CVA21 analysis was indeed from miRT CVA21 ( Figure 2E ) . MEC-1 cells infected with only CVA21 142T had a larger than 5 log decrease in total CVA21 RNA ( and , as expected did not contain any miRT CVA21 ) . In cells doubly infected with WT+ miRT virus , less than 50% of progeny genomes were found to be miRT genomes ( the remainder are therefore WT progeny genomes ) . However , in cells infected with CVA21 142T + miRT CVA21 , 100% of progeny genomes were found to be miRT genomes ( none were therefore 142T genomes ) . Hence , 142T genomes are destroyed and not amplified in cells containing the 142T miRNA , even when the cells are infected with another virus that provides all viral proteins in trans . While we observed that the CVA21 142T RNA genome was profoundly decreased in the presence of perfectly matched miR-142 mimics , there was significant variability in the degree of inhibition seen in different hematopoetic cell lines ( Figure 2D ) . To determine whether this variability correlated with the level of miR-142-3p expression in these cells , we analyzed the abundance of endogenous miR-142-3p in our panel of hematopoetic cell lines by qRT-PCR . The chronic lymphocytic leukemia ( CLL ) line MEC-1 and the multiple myeloma cell line Kas 6/1 expressed the highest levels of endogenous miR-142-3p ( Figure 2F ) , and showed the greatest inhibition of CVA21 142T RNA expression ( Figure 2D ) , followed by the CLL lines MEC-2 , MEC-1 , and WAC-3 . We observed an inverse correlation between miR-142-3p expression levels and the CVA21 142T RNA copy numbers detected , suggesting that the abundance of this particular miRNA contributes substantially to the efficiency with which it acts to silence complementary mRNAs . Overall these results show that perfectly matched miR-142 targets are recognized and catalytically degraded . In addition , while clearly not as efficient as mRNA cleavage , translational repression was seen to occur with cap-independent type I IRES-containing mRNAs and provided protection from cytopathic effects mediated by the virus , and the reduced the infectious virus released into the cellular supernatant . This translational silencing was only detected in the presence of high levels of endogenous miRNAs , however , and at a low multiplicity of viral infection . Having determined that viral replication was amenable to miRNA-mediated regulation at two of the earliest steps in viral infection ( genome cleavage after uncoating and initial mRNA translation ) , we next sought to look at later steps in the viral life cycle . While perturbing the virus life cycle at early steps after infection ( attacking genome after uncoating , initial mRNA expression ) would be most likely to slow or inhibit a productive infection , later steps such as mRNA accumulation , the bulk of mRNA and genome accumulation , and encapsidation of the viral genome could be equally amenable to miRNA-mediated regulation as exposed targets are still present on viral RNAs . In order to examine the ability of microRNAs to act on late steps of viral replication , we conducted a time-course experiment where miRNA mimics were added prior to , concomitant with , or at specific times after infection . Picornaviruses are known to inhibit host cell translation beginning at about 1 hour after infection [29] , with viral mRNA accumulation peaking by 5 hours post infection [30] , followed by encapsidation of the positive-strand viral RNA into the provirion and cytolytic release thereafter [31] . In order to present blocks at different stages of the picornaviral life cycle , we therefore infected cells either pre-treated with miRNA mimic , or mimics were added at 0 , 3 , or 6 hours post infection ( pi ) . Analysis of cell viability in the presence of miRNA mimics added at successive times after infection demonstrated that , while mimic presence prior to infection provides the greatest protection from viral cytolysis , miRNAs can negatively impact viral replication such that cell viability is protected by nearly 50% even when added 6 hours after infection ( Figure 3A ) . Moreover , production of viral progeny is greatly diminished in the presence of miRNAs added at varying time points up to 6 hours post-infection ( Figure 3B ) . To confirm that this was not an artifact of virus expansion ( ie cells merely being protected by virtue of not being infected until one round of viral replication had occurred ) , this experiment was repeated at an MOI = 3 or 10 and similar results observed ( Figure 3C–3F ) . In addition , we analyzed the quantity of CVA21 142T by qRT-PCR in the presence of control miRNA mimics as compared to CVA21 142T infection in the presence of miR-142-3p mimic at different MOI ( Figure 4A ) and found that virus abundance was similar to that observed when analyzed through infectious titer ( Figure 3B , 3D , 3F ) . Since the picornavirus life cycle is particularly rapid and the bulk of viral RNA amplification , protein expression and provirion production has already taken place by 6 hours after infection , we wondered whether miRNA mimics might be interfering with genome encapsidation . We therefore sought to look at the particle: infectivity ratio of the released progeny viruses in the presence of miRNA mimics added after viral infection . Infectivity was determined by titration of cell supernatant on H1-HeLa cells while capsid protein content was assayed after concentration through sucrose cushions of the supernatant of infected cells by both protein assay and Western analysis . Capsid proteins were not detected in sucrose-purified samples of cellular supernatant to which sequence-complementary mimics ( miR-142-3p ) were added before or simultaneously with CVA21-142T infection ( Figure 4B ) . However , when miR-142-3p mimic was added 3 or 6 hours after virus infection , total capsid protein was not significantly diminished as compared to cells transfected with a control miR ( p = 0 . 3 , p = 0 . 14 , Figure 4B ) . Infectious units , however , were decreased by 4 and 3 logs , respectively ( Figure 3B ) . Similarly , when these sucrose-purified fractions were subjected to Western analysis with antibodies raised against CVA21 , CVA21 protein was roughly equivalent when CVA 142T infection was performed with in the presence of control miRNA mimics ( Figure 4C ) . However , miR-142-3p mimic present 4 hours prior to infection or simultaneously with CVA 142T infection inhibited detectable expression of CVA21 capsid in the immunoblot . In contrast , when miR-142-3p mimic was added at either 3 or 6 hours after expression , accumulation of viral capsid was detected in Western blots of the sucrose-purified portion and corresponded with capsid quantified by protein assay . Together these results reveal that alternative steps in the picornaviral life cycle are amenable to miRNA-mediated inhibition: including ( but perhaps not limited to ) RNA cleavage after uncoating , reduced translation of the viral mRNAs , reduced viral mRNA accumulation , and diminished encapsidation of the viral genome into the procapsid ( Figure 5 ) . While the in vitro determination of the mechanisms by which miRNAs are able to inhibit virus replication can facilitate the design of better and more elegant ways to exploit cellular miRNAs for gene transfer and vaccine development for multiple virus families , in vivo evaluation is necessary to validate this targeting paradigm for translational purposes , and to demonstrate that it will indeed occur in model organisms . Therefore , we next sought to confirm that tumor cells expressing endogenous miRNAs were capable of targeting the CVA21 genome and would indeed protect from viral replication in vivo in the mouse model . To this end , 5×106 Mel-624 ( melanoma ) or MEC-1 ( chronic lymphocytic leukemia , CLL ) cells were implanted subcutaneously in immunodeficient mice . When tumors reached an average of 250 mm3 in size , the animals were treated with a single intratumoral dose of 106 TCID50 of a recombinant CVA21 , after which time tumor growth and survival were monitored ( Figure 6 ) . As expected , all CVA21 variants ( WT , miRT and142T ) replicated in the melanoma-xenografts ( which express no sequence-complementary miRNAs ) such that complete ( or near complete ) tumor regression took place ( Figure 6A ) . Animals receiving the WT or 142T viruses rapidly succumbed to myositis , necessitating euthanasia ( Figure 6E ) . Also as expected , animals treated with our previously described muscle-restricted virus ( miRT ) had complete tumor regressions and were protected from the development of myositis , while Opti-MEM injected control tumors grew unencumbered and mice had to be sacrificed when tumors reached ∼2 . 0 cm3 . In contrast to the melanoma xenografts , MEC-1 CLL xenografts ( which express abundant miR-142 ) were susceptible to the WT and muscle-restricted viruses but were highly resistant to oncolysis by CVA21 142T ( Figure 6B ) . Since all recombinant viruses behaved as expected in mice bearing one susceptible or one restricted tumor type , we next decided to examine their behavior in animals with a different tumor on each flank after intravenous virus administration . By using contralateral tumors we were able to create a more representative in vivo scenario , with multiple CVA21-permissive tissues having distinct ( permissive or restrictive ) miRNA expression profiles . Unexpectedly , the virus bearing tandem repeats of miR-142 ( CVA21 142T ) replicated in both the permissive melanoma xenograft ( Figure 6C ) and in the “nonpermissive” CLL xenograft ( Figure 6D ) in these mice , such that both tumors regressed just as rapidly as when they were infected with the nontargeted WT or miRT viruses . Once again , only the muscle restricted virus ( miRT CVA21 ) prolonged the survivals of tumor-bearing mice , ( Figure 6E p = 0 . 0018 , Figure 6F p = 0 . 0004 , Figure 6G p = 0 . 0062 ) . To confirm that CVA 142T was indeed replicating in both the melanoma and myeloma xenografts , both tumors were explanted from two separate mice at time of euthanasia . Tissue was homogenized and overlaid on H1-HeLa cells to look for viral recovery . In each tumor of both animals , virus was recovered and found to contain no alteration in the 142T sequence , indicating that there was indeed productive infection in both tumors . In order to determine why regression of mir-142-expressing hematopoetic tumors was observed in animals bearing contralateral melanoma xenografts when treated with the miR-142 restricted virus , but not in animals with a single hematopoetic tumor , we compared the viral titers in serum from animals treated with CVA21 142T versus WT CVA21 ( Figure 6H ) . Serum titers were typically high in mice treated with WT CVA21 , irrespective of whether they were implanted with melanoma , CLL or bilateral xenografts . In mice treated with CVA-142T , the serum titers were very high ( >108 ) in animals bearing melanoma xenografts , very low ( <102 ) in those with hematopoetic tumor xenografts , and were again very high in animals with the combination of melanoma and hematopoetic xenografts on either flank . Since serum titers were very high in mice with bilateral tumors ( >108 TCID50/ml ) , we investigated the possibility that endogenous miR-142 expressed in MEC-1 xenografts could have been saturated by this high-level viremia . To investigate this possibility , we conducted an in vitro assay looking at saturation of endogenous miR-142 in MEC-1 cells . While cells were completely protected against cytotoxicity at MOIs of up to 30 , increasing virus titer thereafter resulted in decreased cell viability such that by an MOI = 1 , 000 we saw near complete cytopathic effects in CVA21 142T infected cells ( Figure 6I ) , thus suggesting that the endogenous miR-142 may have been effectively saturated , a phenomenon which has been observed by a number of groups previously [32] [25] . While our in vitro data suggest that a sequence-specific saturation of endogenous miRNAs is possible , other possibilities include a non-specific shutdown of all RNAi machinery in infected cells , or possibly the destruction of tumor vasculature that serves as a necessary scaffolding for xenografts . Together these data show that miRNAs act in vivo to restrict virus tropism in cells permissive to viral replication . Perhaps not surprisingly , we found that persistent high-level viremia was able to overcome this miRNA-mediated restriction , possibly through allowing the saturation of endogenous miRNAs for which our virus bore tandem cognate targets . We speculate that in animals bearing bilateral tumors treated with intravenous CVA21 142T , virus trafficked to the susceptible melanoma tumor , which created a reservoir of viral amplification . This then resulted in a high-level viremia , which was able to overwhelm the miRNA-restricted CLL tumor by saturating the sequence-complementary endogenous miRNA . We detected no sequence alteration in the virally encoded miRT insert in animals treated with miRT or CVA21 142T , and therefore can dismiss the possibility that the viruses overcame miRNA restriction in this manner . However , we did identify significant sequence insert changes in viruses bearing the tumor suppressor miR-145T inserts when administered to this same panel of animals ( data not shown ) . We therefore surmise that miRNA targets are differentially effective and susceptible to different selective pressures to mutationally inactivate these targets , and that this may correspond to the specific tissues from which they are being excluded .
Increasing insights into the interactions between miRNAs and viruses have suggested new therapeutic targets for antiviral drugs and new techniques for the creation of vaccines , as well as for regulating the tissue specificity of gene transfer vehicles and oncolytic viruses . However , the ways in which cellular miRNAs can act to inhibit viral replication have remained ill-defined . In this report , we identify alternative steps in the viral life cycle that are amenable to miRNA-mediated inhibition in the context of a replication-competent oncolytic virus that acts as a fully curative therapy in xenograft-bearing mice . Here , we show that picornaviruses are most susceptible to miRNA-mediated attack immediately after infection , and that the viral RNA genome can be recognized and cleaved by fully sequence complementary miRNAs . However , contrary to the currently accepted paradigm , we also present evidence to suggest that miRNAs can translationally suppress uncapped , IRES-dependent mRNAs bearing imperfectly matched target sequences at low multiplicity of infection . Several previous studies have used the tumor-suppressor miRNA let-7a to look at miRNA regulation in cap dependent vs . IRES dependent mRNAs . However , despite very high let-7a abundance in a number of cell lines , target elements corresponding to let-7 do not functionally suppress sequence complementary targets as efficiently as does miR-142 ( or a number of other well-characterized miRNAs ) [5] . In addition , the interpretation of data generated using cell-free extracts and synthetic miRNAs supplemented in trans , as employed previously by several groups , may be made more difficult by miRNA saturation . We believe the use of a highly efficient miRNA target ( miR-142T ) , cell lines expressing high levels of endogenous , sequence-complementary miRNAs , and the ability to infect cells at low MOI , have enabled us to see the subtle translational suppression of a viral mRNA in a cap-independent context . The observation that the picornaviral life cycle can be efficiently interrupted by miRNAs even at very late stages after infection , by which time the viral genome copy number has amplified to a very high level , is truly remarkable . This late interruption appears to require perfect complementarity between the viral genome and the miRNA , and probably in large part reflects the rapid catalytic destruction of viral genomes . An interesting consequence of adding the miRNA late after infection to interrupt the viral life cycle is that the cell still releases abundant virus particles , but that they lack infectivity ( Figures 3 , 4 ) . The possibility exists that miRNAs could also block encapsidation of viral RNA genomes by steric hindrance ( RISC bound-RNAs being too large to be packaged ) , although our data do not address this possibility . While our data show that miRNAs can recognize and antagonize a virus , our observation that endogenous cellular miRNAs can actually overwhelm this regulation is also of potential importance . The phenomenon of miRNA saturation , whereby miRNA targets can act as ‘sponges’ to bind and sequester endogenous cellular miRNAs , has been reported previously , but has been generally thought to occur in the context of translational silencing [25] [32] . Here , we show that miRNA saturation can also occur when miRNA regulation is occurring primarily by catalytic degradation of a perfectly complementary RNA target ( Figure 6I ) , and that it can happen in the context of viral infection in vivo . We have shown that miRNAs can suppress virus propagation in vivo and can provide post-entry blocks to virus replication in otherwise permissive cells . However , we also show that miRNA defenses can be overcome at very high multiplicities of viral infection in vivo , allowing unencumbered virus replication . Therefore , clinicians considering therapeutic intervention utilizing miRNA targeted-viruses ( ie vaccines and cancer therapeutics ) must be particularly wary of the possibility that unencumbered virus replication in permissive cells , eg a tumor , may eventually lead to viral infection and pathogenesis in normal cells that express a restricting miRNA ( if that miRNA becomes saturated ) . MiRNAs are known for their ability to regulate numerous cellular functions , often at multiple steps within a single signal cascade . In this report , we have shown that miRNAs can also act to antagonize sequence complementary viruses at multiple , alternative steps in the virus life cycle such that tumor selectivity is generated , but also a mechanism by which this can be overcome . Although we show that picornaviruses are particularly susceptible to miRNA interference , the different replication steps affected by miRNAs are common to many viruses and it is therefore likely that multiple virus families will be amenable to miRNA-mediated regulation .
Mayo Clinic Institutional Care and Use Committee reviewed and approved protocol A7007 with investigators SJR , EJK , and EMH , entitled “MicroRNA-mediated targeting of an oncolytic enterovirus , Coxsackievirus A21” . All mice were housed in the BSL-2 facility at Mayo Clinic and experiments were performed in compliance with outlined and approved institutional guidelines . H1-HeLa , cells were obtained from American Type Culture Collection and were maintained in DMEM supplemented with 10% FBS in 5% CO2 . MEC-1 , MEC-2 , WAC-3 , and Kas 6/1 cells were obtained from Diane Jelenik , Dept . of Immunology , Mayo Clinic . MEC-1 and MEC-2 cells were maintained in IMDM 10% FBS in 5% CO2 . WAC-3 and Kas 6/1 cells were cultured in RPMI with 10% FBS in 5% CO2 . pGEM-CVA21 clone was kindly provided by Matthias Gromeier . miRNA sequences were obtained from the Sanger Institute miRBase ( http://microrna . sanger . ac . uk/sequences/ ) . The following sequences were cloned into the 3′UTR of pGEM-CVA21 in between bp 7344/7345 by overlap extension PCR . miR-142 3pT TCCATAAAGTAGGAAACACTACACGATTCCATAAAGTAGGAAACACTACAACCGGTTCCATAAAGTAGGAAACACTACATCACTCCATAAAGTAGGAAACACTACA miR-142revT TGTAGTGTTTCCTACTTTATGGAATCGTGTAGTGTTTCCTACTTTATGGAACCGGTTGTAGTGTTTCCTACTTTATGGAATCGTGTAGTGTTTCCTACTTTATGGA miR-142 mm7T TTAATGCAGTCATAAACACTACACGATTTAATGCAGTCATAAACACTACAACCGGTTTAATGCAGTCATAAACACTACATCACTTAATGCAGTCATAAACACTACA miR-142 b6T TCCATAAAGTAGGTCGATTAAACACTACACGATTCCATAAAGTAGGTCGATTAAACACTACAACCGGTTCCATAAAGTAGGTCGATTAAACACTACATCACTCCATAAAGTAGGTCGATTAAACACTACA miR-145T AGGGATTCCTGGGAAAACTGGACCGATAGGGATTCCTGGGAAAACTGGACACCGGTAGGGATTCCTGGGAAAACTGGACTCACAGGGATTCCTGGGAAAACTGGAC Viral RNA was produced using Ambion Megascript and Megaclear T7 polymerase kit according to manufacturers instructions . For rCVA21 rescue 1 ug RNA/well was transfected into H1-HeLa cells in 12 well plates using the Mirus RNA transfection reagent and at 24 hours post infection wells were scraped and cell pellets harvested . Cell pellets were subjected to 3 freeze/thaw cycles in liquid N2 , cell debris was cleared by centrifugation and cleared lysate was added to H1-HeLa cells in a T-75 flask . Titration of CVA21 was performed on H1-HeLa cells . Cells were plated in 96 well plates at 50 percent confluence . After 24 hours , serial ten-fold dilutions ( −2 to −10 ) were made of the virus; 100 uL of each dilution was added to each of eight duplicate wells . Following incubation at 37°C for 72 hours , wells were then assessed for CPE and TCID50 values were determined using the Spearman and Kärber equation . H1-HeLa cells were incubated with rCVA21 at a multiplicity of infection ( MOI , determined by TCID50 per cell ) of 3 . 0 for 2 hours at 37°C . Following this incubation , cells were washed and resuspended in fresh growth media at predetermined time-points ( 2 , 4 , 6 , 18 , 12 , 24 , hours ) , cells pellets were harvested and frozen at −80°C . At the completion of all time-points , they were thawed , and cell pellets were cleared from the samples by centrifugation providing a cleared cell lysate fraction . miRNA mimics were purchased from Dharmacon , Inc . Control miRNA mimic corresponded to a C . elegans miRNA with no predicted miRTs in mammalian cells according to manufacturer . miRNA mimics were transfected with Mirus ™ RNA transfection reagent at a 200 nM concentration . 4 hours post mimic transfection , cells were infected with recombinant CVA21 at MOI = 1 . 0 , 3 , or 10 , unless other time noted . After 24 hrs . post infection , cells were harvested for MTT viability assay and supernatant was harvested for titration . Total cellular RNA was harvested using the Qiagen RNeasy Kit , according to manufacturer instructions . Primers and probes corresponding to the 2A region or miRT insert region of CVA21 were used to quantitate CVA21 RNA and GAPDH primers and probes were used to normalize total cellular RNA . For each sample analyzed , 50 ng of total RNA was subjected to qRT-PCR in triplicate on Stratagene Mx4000 qPCR system using the Taqman One Step RT PCR master mix . Small RNAs were harvested from all cell lines with the Ambion miRVana microRNA isolation kit , according to manufacturer instructions . RNA was resuspended in 50 ul nuclease free water , and quantified by spectrophotometer . For analysis of miRNA expression , the Applied Biosystems Taqman microRNA Assay system was used . For each miRNA analyzed , 5 ng of small RNA was subjected to qPCR in triplicate on Stratagene Mx4000 qPCR system . Ten-centimeter dishes of H1-HeLa cells were transfected with miRNA mimics and infected with CVA21 142T at MOI = 1 . 0 at indicated time points . Cell supernatant was collected centrifuged for 5 mins at 10 , 000 g to clear cell debris . Infectious titer was calculated on cell supernatant , and thereafter . 5% SDS and 2 mM EDTA was added to supernatant , and then overlaid on a 5 ml sucrose cushion . Total virus particles were subjected to ultracentrifugation for 4 h at 28 , 000 rpm using an SW28 swinging bucket rotor . Supernatants were discarded and centrifuge tubes were rinsed with PBS and re-spun . After PBS wash , virus pellets were resuspended in PBS containing 0 . 2% SDS and 5 mM EDTA . Virus capsid protein was then quantified using Biorad protein assay kit and normalized to mock infected and concentrated samples . Twenty female Balb/C mice were inoculated intraperitoneally with 1e6 CVA21 three times over a period of six weeks . Two weeks following the final inoculation , mice were terminally bled , serum was collected and pooled and frozen at −20°C for use in immunoblotting . Supernatant from miRNA-mimic timecourse experiments were sucrose-purified as above and run on a 12 . 5% SDS page gel and transferred to a PVDF membrane using the Trans-Blot SD semi-dry transfer apparatus ( BioRad ) for 45 minutes at 15 Volts . Blots were blocked in TBS-10% milk for 1 hour at room temperature . Primary CVA21 antibody generated as described above was diluted 1∶500 in TBS-5% milk + . 05% Tween and blots were incubated overnight at 4°C . Membranes were washed in TBS + . 1% Tween and incubated in secondary Goat-Anti Mouse IgG ( Dako ) at a 1∶5000 dilution in TBS-5% milk+ . 05%Tween at room temperature for 1 hour . Membranes were washed in TBS + . 1% Tween and bound antibodies were detected using SuperSignal West Pico Chemluminescent reagent ( Pierce ) . All animal protocols were reviewed and approved by Mayo Clinic Institutional Care and Use Committee . CB17 ICR-SCID mice were obtained from Harlan . Mice were irradiated and implanted with 5e6 Kas 6/1 or Mel 624 cells in the right flank . When tumors reached an average of . 5× . 5 cm , tumors were treated with 1e6 CVA21 . Tumor volume was measured using a hand held caliper and blood was collected by retro-orbital bleeds .
|
Virus host range is shaped by cellular determinants such as transcription factors and receptor expression . In addition , we have previously shown that tissue-specific microRNAs can be utilized to direct the specificity of a replication competent picornavirus , Coxsackievirus A21 . In this report , we demonstrate the mechanism by which microRNAs are able to directly influence oncolytic viruses , an important class of anticancer agents . We show that microRNA expression is an important determinant of permissivity to oncolytic virus replication , but the actual abundance of that expression is far more important . In addition , we show that there are actually multiple different stages in the life cycle of a replication competent picornavirus that are amenable to regulation by cellular microRNAs . We proceed to illustrate that microRNAs can regulate virus tropism in vivo , but demonstrate that circulating high viral titers in the blood can overcome this mechanism of conferring tissue specificity . MicroRNAs are well known to have both oncogenic or oncosuppressive activities in human cancers . Here , we show that tissue-specific microRNA expression can also be used to modulate the efficacy of viral anticancer therapeutics , and the mechanism by which they are able to do so .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"virology/viral",
"replication",
"and",
"gene",
"regulation",
"molecular",
"biology/translation",
"mechanisms",
"molecular",
"biology/mrna",
"stability",
"oncology/myelomas",
"and",
"lymphoproliferative",
"diseases",
"virology/mechanisms",
"of",
"resistance",
"and",
"susceptibility,",
"including",
"host",
"genetics",
"infectious",
"diseases/viral",
"infections",
"virology/viruses",
"and",
"cancer",
"molecular",
"biology/translational",
"regulation",
"virology/host",
"antiviral",
"responses"
] |
2010
|
MicroRNA Antagonism of the Picornaviral Life Cycle: Alternative Mechanisms of Interference
|
Evidence is mounting that mutation rates are sufficiently high for deleterious alleles to be a major evolutionary force affecting the evolution of sex , the maintenance of genetic variation , and many other evolutionary phenomena . Though point estimates of mutation rates are improving , we remain largely ignorant of the biological factors affecting these rates at the individual level . Of special importance is the possibility that mutation rates are condition-dependent with low-condition individuals experiencing more mutation . Theory predicts that such condition dependence would dramatically increase the rate at which populations adapt to new environments and the extent to which populations suffer from mutation load . Despite its importance , there has been little study of this phenomenon in multicellular organisms . Here , we examine whether DNA repair processes are condition-dependent in Drosophila melanogaster . In this species , damaged DNA in sperm can be repaired by maternal repair processes after fertilization . We exposed high- and low-condition females to sperm containing damaged DNA and then assessed the frequency of lethal mutations on paternally derived X chromosomes transmitted by these females . The rate of lethal mutations transmitted by low-condition females was 30% greater than that of high-condition females , indicating reduced repair capacity of low-condition females . A separate experiment provided no support for an alternative hypothesis based on sperm selection .
Germ-line mutation is the ultimate source of heritable variation , but the vast majority of new mutations affecting fitness are deleterious . The unremitting presence of deleterious mutations causes a reduction in mean fitness , a phenomenon known as mutation load . Mutation load can be substantial even if individual mutations are of small effect and are held at low frequencies by natural selection . For example , classic theory [1] predicts that mutation load will reduce mean fitness by more than 60% if there is just one deleterious germ-line mutation per genome per generation . The constant influx of deleterious mutations may pose a serious challenge to natural populations [2] . Mutation load can accelerate the extinction of endangered species [3 , 4] and may be an important public health concern in humans [5] . Large mutation loads have been invoked as a possible explanation for a wide variety of other phenomena , such as the maintenance of genetic variation [6] , the evolution of specialization [7 , 8] , the evolution of outcrossing [9–11] , and the evolution of sexual reproduction [12–15] . Mutation rate ( U ) is the most important factor determining the magnitude of mutation load . However , estimates of the mutation rate vary over two to three orders of magnitude [16–20] . While much of this variance may be due to measurement error ( especially in earlier studies ) , some of this variance likely has real biological causes [17 , 21–23] . Of special interest is the possibility that variance in mutation rate arises from individual variation in condition , because individuals of low condition may have elevated rates of mutation . In a stable environment , condition dependence of the mutation rate is expected to alter the mutation load [24] because of the positive feedback loop it creates: individuals with an excess of deleterious alleles tend to be in low condition and so experience a high mutation rate . ( Interestingly , the mutation loads of sexual and asexual populations are affected very differently by condition dependence [24] . ) Condition dependence is also expected to accelerate adaptation to new stressful environments if the mutational input is elevated under poor condition [25 , 26] . Although mutation rate is often treated as though it is constant and nonplastic , there is no compelling reason to believe this should be true . Condition dependence is common among other traits , including recombination , another “genomic” trait [27–29] . Moreover , there is evidence that mutation rate varies across environments in some unicellular organisms [30–32] . However , there are reasons to question whether patterns observed in unicellular organisms apply to multicellular organisms . First , the unicellular organisms cited above are predominantly asexual , and theory [33 , 34] predicts that facultative elevation of mutation rate in response to stress is more likely to evolve for adaptive reasons in asexual species than in sexual species . Second , unicellular organisms may be particularly sensitive to environmental effects on DNA processes simply because they are unicellular . Nonetheless , it is reasonable to predict that mutation rates are also condition-dependent in multicellular organisms , though the reasons may be different from those in unicellular organisms . For instance , condition dependence may occur because maintaining DNA with perfect fidelity is a costly enterprise and low-condition individuals are less able to pay this cost . Despite the potentially important consequences of this phenomenon , there has been little effort to look for evidence of condition dependence in multicellular organisms . Mutation rate is a function of two factors: ( 1 ) the rate at which DNA damage occurs and ( 2 ) an organism's ability to repair that damage . Condition dependence in mutation rate is expected if either of these factors is condition-dependent . Using Drosophila melanogaster , we tested whether individuals in low condition are less able to repair DNA damage without inducing a mutation . When lesions in the DNA occur , these must be repaired for the cell cycle to continue properly . Some repair pathways are conservative and do not result in mutation; other repair pathways are error-prone so that mutations are generated in the process of removing DNA lesions . Conservative pathways are thought to be more costly than error-prone pathways [35] . The premise of our experiment was simple: expose high- and low-condition individuals to damaged DNA and assess their ability to repair the DNA without introducing error . One cannot simply expose flies to the same mutagen treatment because high- and low-condition individuals might respond differently ( e . g . , by eating or absorbing different amounts of the mutagen ) such that the level of damage differs between the treatments . To circumvent this difficulty , we took advantage of the maternal repair system in D . melanogaster . When males are mutagenized , DNA damage in sperm persists because there is little , if any , postmeiotic repair in males [36 , 37] . However , premutational DNA lesions can be repaired after fertilization by maternal repair proteins . For example , Vogel et al [37] mated standard males that had been mutagenized with methyl methanesulfonate ( MMS ) to either wild-type or repair-deficient ( Mei-9 mutant ) females . Repair-deficient females produced daughters carrying recessive lethal mutations on their paternally derived ( i . e . , mutagen-exposed ) X chromosomes almost eight times more frequently than did wild-type females . This result indicates repair-deficient females were less able to repair DNA damage on chromosomes coming from mutagenized sperm without producing a mutation . We used a similar design comparing high- and low-condition females rather than wild-type and repair-deficient females . Specifically , we used a larval diet manipulation to create high- and low-condition females that were genetically wild type with respect to DNA repair genes . These females were mated to standard mutagenized males; daughters were then screened for recessive lethals on the paternally inherited X chromosome ( Figure 1 ) . As reported below , low-condition females transmitted more of these sex-linked recessive lethals ( SLRLs ) than did high-condition females . An alternative interpretation to condition-dependent repair is condition-dependent sperm selection . In this scenario , heavily damaged sperm would be less likely to fertilize eggs in high-condition females than in low-condition females . We examined this possibility by doing a separate sperm competition experiment in which we measured selection against mutagenized sperm in both high- and low-condition females . There is no evidence that selection against mutagenized sperm is stronger in high-condition females .
A diet manipulation was used to produce flies of high and low condition . Females emerging from the low-condition treatment tended to be visibly smaller than females from the high-condition treatment , but all flies were well within the normal range of body sizes observed in typical fly cultures . Both high- and low-condition females were mated to males that had been reared under standard conditions and then mutagenized with alkylating agent MMS . As expected , the diet manipulation affected condition: females from the low-condition treatment produced approximately 32% fewer offspring than females from the high-condition treatment ( F1 , 646 = 82 . 0 , p < 0 . 0001 ) . Averaging across all of Experiment 1 , we found that approximately 15% of paternally derived ( i . e . , mutagenized ) X chromosomes carried lethal ( or near-lethal ) mutations . This frequency is consistent with other studies using a similar dose of this mutagen and is about two orders of magnitude greater than the spontaneous rate [38] . Our primary interest is whether a mutagenized X chromosome was more likely to eventually harbor a lethal mutation if it passed from a sperm into an egg in a low-condition female rather than a high-condition female . Such a pattern would be expected if low-condition females were more likely than high-condition females to employ error-prone pathways to repair damaged DNA from sperm . The frequency at which high- and low-condition females transmitted SLRL mutations from their mutagenized mates to their offspring is given in Table 1 . We observed a higher frequency of lethal-bearing X chromosomes being transmitted by low-condition females than by high-condition females . A randomization test revealed that we were unlikely to observe this large a difference by chance ( n = 552 , p = 0 . 04 ) . Averaging over both blocks of Experiment 1 , the rate at which low-condition females transmitted lethal-bearing X chromosomes ( 0 . 159 ) was approximately 28% higher than the rate at which high-condition females transmitted lethal-bearing X chromosomes ( 0 . 124 ) , i . e . , 0 . 159/0 . 124 = 1 . 28 . We performed a second experiment similar to that described above except that females were mated individually to mutagenized males to prevent any effects of pre-copulatory sexual selection . As in Experiment 1 , the diet manipulation in Experiment 2 affected condition: females from the low-quality diet treatment produced significantly fewer offspring than females from the high-quality diet treatment ( F1 , 851 = 10 . 7 , p < 0 . 001 ) . In Experiment 2 , the overall rate of SLRLs was approximately 0 . 10 , a lower frequency than in Experiment 1 but within the normal range expected for this type of mutagenesis [38] . As in Experiment 1 , there was a difference in the frequency of SLRLs transmitted by low- and high-condition females ( low: 0 . 109; high: 0 . 083 ) . Relative to the daughters of high-condition females , the daughters of low-condition females were approximately 31% more likely to harbor a lethal mutation on their paternally inherited X chromosome ( n = 595 , p = 0 . 03 ) . The data shown in Table 1 indicate that our results are consistent across both blocks of both experiments: low-condition females transmit lethal-bearing paternally derived X chromosomes at a higher rate than high-condition females . Considering the evidence from both experiments together by combining p-values [39] indicates this is a strongly significant effect ( weighted Z = −2 . 53 , p = 0 . 006 ) . It is possible that the results above could be due to sperm selection rather than DNA repair capacity . Sperm carrying more heavily damaged chromosomes might be less likely to successfully fertilize eggs in high-condition females than in low-condition females . In other words , it is possible that there is stronger selection against mutagenized sperm in high-condition females than in low-condition females . To test this possibility , we measured the siring success of mutagenized males and non-mutagenized males when mated to high- or low-condition females . Females were first mated to standard males and then mated to either mutagenized or non-mutagenized males . We measured the proportion of offspring sired by the second male ( P2 ) , thus allowing for estimates of the P2 abilities of mutagenized and non-mutagenized sperm against a standard competitor . Mean P2 scores are shown in Table 2 . Analysis of these data with a generalized linear mixed model revealed a significant negative effect of mutagenesis on P2 score ( F1 , 18894 = 9 . 92 , p = 0 . 002 ) , i . e . , mutagenized sperm was less successful than non-mutagenized sperm . There was no significant effect of female condition ( F1 , 18894 = 0 . 00 , p = 0 . 96 ) , and there was no significant interaction between female condition and whether sperm had been mutagenized ( F1 , 18894 = 0 . 31 , p = 0 . 58 ) . In addition to the analysis described above , we performed a different likelihood analysis that allowed us to model the strength of selection against mutagenized sperm in low- and high-condition females as separate parameters that are more easily interpreted . Consistent with the previous analysis , this likelihood analysis indicated selection against mutagenized sperm in low-condition females ( sL = 0 . 038 ) was considerably stronger than in high-condition females ( sH = 0 . 007 ) . Although sL was found to be significantly greater than sH in this analysis , these point estimates are primarily of heuristic value as this latter analysis ignores variation among individual females so that the model will tend to underestimate the uncertainty in the parameter estimates . Nonetheless , the results of both analyses show that whereas there is weak selection against mutagenized sperm , there is no indication that selection is stronger in high-condition females than in low-condition females—the evidence is in the opposite direction .
According to classic theory , deleterious alleles can have large effects on a population if the mutation rate is sufficiently large , i . e . , on the order of U = 1 . Though estimates of U have varied considerably , recent studies [16 , 17] employing modern techniques indicate that mutation rates are likely to be high enough to create large loads . As estimates of the mutation rate continue to improve , we can begin to acknowledge and study the variation in this genomic property . It is well known that transposable element activity increases in response to extrinsic stress [40 , 41] , but much less is known about variation in the rates of more traditional types of mutation . There is recent evidence that repair capacity and net mutation rate are temperature-dependent [23 , 42 , 43] , though this may not be too surprising since both DNA and protein stability are sensitive to temperature . Some recent mutation-accumulation studies have found evidence that mutation rates vary between closely related species and even among lines of the same species [17 , 22] . One study reported that the mutation rate accelerated within a line over the duration of a long mutation-accumulation experiment [21] . The reasons for this variation are unclear . In some cases , the variation in mutation rate can be attributed to genetic differences among lines because all lines accumulated mutations in the same environment [22] . Even so , it is unknown whether the important genetic differences occur at loci that are directly involved in DNA replication and/or repair , or alternatively , whether genetic differences affecting condition indirectly lead to differences in mutation . Despite the important consequences of condition dependence for mutation load [24] and adaptation to stressful environments [25 , 26] , there has been little effort to test for this type of mutational plasticity in multicellular organisms . We investigated whether DNA repair ability is condition-dependent by exploiting the maternal repair system in D . melanogaster . When fertilized with mutagenized sperm , low-condition females were approximately 30% more likely than high-condition females to produce daughters carrying paternally derived X chromosomes that harbored recessive lethals . This result was consistent across two separate experiments . The mutagen used in this experiment , MMS , causes lesions by alkylation of N atoms in the DNA [44] . These lesions can be repaired , without error , by excision repair . If a lesion is not repaired by excision prior to DNA replication , alternative error-prone repair mechanisms may be employed to remove the lesion , resulting in mutation [45] . Our data suggest that females in low condition are less able to efficiently repair DNA lesions without creating a mutation . This may be because low-condition females are more likely to employ error-prone repair pathways than are high-condition females or because low-condition females use error-free repair pathways less efficiently than do high-condition females . We considered an alternative hypothesis based on sperm selection though there are several reasons to doubt this possibility . First , it is unlikely that the DNA lesions that lead to mutations causing lethality are the direct targets of selection , because only a very small fraction of genes are expressed in sperm [46] . However , the mutagen may cause physiological effects on sperm performance such that sperm exposed to heavier doses of mutagen would have both reduced performance and a higher likelihood of DNA damage at potentially lethal sites . In other words , lethal mutations may be eliminated via a correlated response to selection against other effects of the mutagen . Effective removal of X-linked lethal recessives through a correlated response would require either strong selection or a large covariance between the true target of selection ( e . g . , physiological effects of the mutagen ) and the occurrence of X-linked lethal recessives . Most importantly , sperm selection alone is not sufficient to explain the observed pattern . Rather , selection against mutagenized sperm must be stronger in high-condition females than in low-condition females . We tested this possibility and found no evidence for it . In fact , our data indicated that selection against mutagenized sperm was stronger in low-condition females . The reasons for this latter result are unclear but need not be adaptive . It is possible that the reproductive tract of a low-condition female is simply a harsher environment for sperm and imposes stronger selection . Finally , it is worth noting that , although we did detect significant selection against mutagenized sperm in the sperm competition experiment , this selection was weak . Moreover , this weak selection represents the selective difference between two extremes: mutagenized and non-mutagenized sperm . In contrast , any sperm selection that might have occurred in Experiments 1 and 2 would have been among sperm that experienced varying degrees of exposure to the mutagen , i . e . , quantitative differences in exposure rather than qualitative differences as in the sperm competition experiment . Thus , any such selection in Experiments 1 and 2 would be expected to be even weaker than what we measured in the sperm competition experiment . In sum , we can infer that sperm selection was very weak in Experiments 1 and 2 , and most likely worked in a direction opposite to the observed pattern with respect to sex-linked lethals . Our data match the prediction expected under condition-dependent repair . Theory indicates that under most conditions , selection should favor reduced mutation rates in sexually reproducing organisms [47] . The direct costs of maintaining perfect DNA fidelity ( i . e . , the costs of perfect replication and error-free repair ) are thought to prevent mutation rates from evolving to extremely low levels . This implies that repair mechanisms are expected to operate at a level that is somewhat costly . As has been discussed in other contexts ( most notably for life history traits and secondary sexual characters ) , the expression of costly traits may often differ between individuals in high versus low condition [48–50] . Low-condition individuals may have higher mutation rates , not because selection favors more mutations , but simply because low-condition individuals cannot afford to invest as heavily in efficient repair . We do not know whether the condition dependence in repair capacity reported here reflects condition dependence in mutation rates under natural conditions . If there is a relationship , we can speculate on how this would affect mutation load . Let us assume , as recent data suggest [17] , that individuals in good condition have a mutation rate of Umin = 1 . In our study , females reared on the low-quality diet had 20%–30% lower fecundity and transmitted approximately 30% more recessive lethals than females reared on high-condition food . If we assume that a 30% reduction in condition translates into a 30% increase in mutation rate and that the relationship between condition and mutation rate is linear , we can calculate the mean fitness using previously developed theory [24] . Under these conditions , the mean fitness of a population at equilibrium is expected to be approximately 57% lower than expected if mutation was not condition-dependent ( i . e . , if U = 1 for all individuals regardless of condition ) . Obviously , this calculation is based on a number of untested assumptions , but nonetheless , it serves to illustrate that condition dependent mutation could have large effects on populations . Further work is required to explore these assumptions and evaluate the magnitude of condition dependence in mutation rate under natural conditions .
Experiment 1: females of high and low condition were produced through a larval diet manipulation . High-condition flies were created by rearing larvae on 7 . 5 ml of standard sugar-yeast-agar media at a low density ( 40 larvae per 8-dr [32-ml] vial ) . In the low-condition treatment , larvae were reared under identical conditions , but the media contained 25% of the standard concentration of sugar and yeast . Adult virgin females were collected within 8 h of eclosion . Both high- and low-condition adult females were held in vials containing standard media , but high-condition females were also given live yeast immediately . Females of both treatments received additional live yeast 3 d prior to mating . Females were mated to mutagenized males ( described below ) when they were 4 d ( block 2 ) or 5 d ( block 1 ) old . All of the females used were heterozygous for the balancer X chromosome Basc that is marked with the dominant eye mutation B ( bar eyes ) . The Basc chromosome had been crossed into our standard large outbred population ( Dah ) more than ten generations prior to the experiment; the balancer chromosome was maintained in this stock by selection . The Dah outbred population was originally collected in 1970 in Dahomey ( now Benin ) , West Africa . It has been maintained at large population size in various labs since that time and most recently in the current lab for over 3 y . Thus , the Basc heterozygous females used in this experiment had wild-type outbred genotypes other than the presence of the Basc chromosome . Wild-type males ( from the Dah population ) were kept without access to food or water for several hours then exposed to sugar water with 1 . 5 mM MMS . The following day , males were transferred to recovery bottles for 2 h before allowing them to mate with the high- and low-condition females described above . Matings occurred in vials containing approximately 20 flies of each sex . The next day , females were transferred to individual vials to lay eggs for 1 d . All vials contained standard medium and live yeast . We counted the number of offspring emerging from these vials , allowing us to examine how the diet manipulation of mothers affected their production of offspring . We used these F1 offspring to look for SLRL mutations . Our assay for SLRLs is very similar to traditional designs [38] and is shown in Figure 1 . From each of the original Basc heterozygous females , four to eight ( nonvirgin ) Basc/Xi daughters ( F1 ) , which had mated with their brothers prior to collection ( see Figure 1 for details ) , were placed in individual vials to lay eggs so that their offspring could be scored . The symbol “Xi” represents a paternally inherited ( i . e . , mutagen-exposed ) X chromosome . Regardless of her mate , a Basc/Xi female should produce two types of sons , Basc/Y and Xi/Y , in equal frequency . However , if a recessive lethal mutation has occurred on this chromosome , then the Basc/Xi daughter will be unable to produce wild-type sons ( Xi/Y ) . If there is no recessive lethal on Xi , then the expected frequency of wild-type males among the F2 progeny is 25% , assuming no viability differences among genotypes . If Xi contains a recessive lethal ( or near-lethal ) mutation , then the frequency of wild-type males among the F2 progeny should be much less than 25% . We determined whether a given Xi was likely to contain such a mutation by examining the observed frequency of wild-type males among a set of F2 progeny relative to the expectations if there was no mutation and if there was a mutation . Specifically , for each set of F2 offspring originating from a single F1 female , we calculated R = L1/40/L1/4 . L1/4 is the likelihood ( assuming a binomial distribution ) of the observed offspring array if the true frequency of wild-type males among all possible sets of viable progeny of the family is 25% , i . e . , the expected frequency if there is no mutation . Similarly , L1/40 is the likelihood of the observed offspring array if the true frequency of wild-type males among all possible sets of viable progeny of the family is 2 . 5% ( as expected if the viability of wild-type males was approximately 10% of normal ) . A low value of R indicates it is unlikely that the Xi in question contains a recessive lethal ( or near lethal ) , whereas a high value of R indicates the opposite . When R ≥ 10 , we classified the Xi as carrying a recessive lethal; when R ≤ 0 . 1 , we classified the Xi as not carrying a recessive lethal . For intermediate values , 0 . 1 < R < 10 , we were unable to clearly assign Xi to either category , and these data were excluded . Using this criteria , we calculated the frequency of lethal-bearing , paternally inherited X chromosomes transmitted by each of the original high- and low-condition females . Over 78 , 000 flies from 2 , 470 sets of F2 offspring were scored in Experiment 1 . From these 2 , 470 sets of F2 offspring , we were able to classify 1 , 461 Xi chromosomes as being unlikely to carry a recessive lethal ( R ≤ 0 . 1 ) and 236 Xi chromosomes as being likely to carry a recessive lethal ( R ≥ 10 ) ; the Xi chromosomes from the remaining 773 sets of F2 offspring were not classifiable ( 0 . 1 < R < 10 ) . For these 1 , 461 + 236 = 1 , 697 classifiable Xi chromosomes , the average number of F2 offspring upon which each classification had been based was 43 . 9 ( standard error [SE] = 0 . 75 ) . These 1 , 697 classifiable Xi chromosomes had been transmitted by 552 parental generation females ( 283 high condition + 269 low condition ) , giving an average of 3 . 1 classifiable Xi chromosomes per parental generation female . For each of these 552 parental generation females , we calculated the frequency of her classifiable Xi chromosomes that were likely to carry a recessive lethal . The average frequency of lethal transmission was compared between females from the two diet treatments . The true value of the difference in average frequency of lethal transmission was compared to a null distribution created by randomizing the data across treatments but within blocks; 10 , 000 randomizations were performed . The reported p-values for both Experiment 1 and 2 are for one-tailed tests because we had a clear a priori prediction about the direction of effect . Experiment 2: several months later , we performed a second experiment that was very similar to the one described above except that mutagenized males were mated individually to experimental females , thereby suppressing the opportunity for sexual selection . In this experiment , females were 4 d old ( block 1 ) or 3 d old ( block 2 ) at the time of mating . Up to ten Basc/Xi daughters were tested per female . Over 130 , 000 flies from 4 , 857 sets of F2 offspring were scored in Experiment 2 . From these 4 , 857 sets of F2 offspring , we were able to classify 2 , 648 Xi chromosomes as being unlikely to carry a recessive lethal ( R ≤ 0 . 1 ) and 291 Xi chromosomes as being likely to carry a recessive lethal ( R ≥ 10 ) ; the Xi chromosomes from the remaining 1 , 918 sets of F2 offspring were not classifiable ( 0 . 1 < R < 10 ) . For these 2 , 648 + 291 = 2 , 939 classifiable Xi chromosomes , the average number of F2 offspring upon which each classification had been based was 41 . 6 ( SE = 0 . 43 ) . These 2 , 939 classifiable Xi chromosomes had been transmitted by 595 parental generation females ( 309 high condition + 286 low condition ) , giving an average of 4 . 9 classifiable Xi chromosomes per female . For each of these 595 parental generation females , we calculated the frequency of her classifiable Xi chromosomes that were likely to carry a recessive lethal . As in Experiment 1 , we used a randomization test to compare the average frequency of lethal transmission between high- and low-condition females . To assess the total evidence for an effect of female condition on the rate of sex-linked lethal mutation , we used the weighted Z-transform method [39] to obtain a combined p-value from Experiments 1 and 2 . Each study was weighted by its sample size . The unweighted Z-transform method provided a very similar result . High- and low-condition females were created as described above . All of the females for this experiment were homozygous for a recessive bw− allele that causes brown eyes . Virgin females were mated to standard red-eyed males ( homozygous wild-type at the bw locus ) when they were 2 d ( block 1 ) , 3 d ( block 2 ) , or 4 d ( block 3 ) old . After 2–4 h , males were removed , and females were placed in individual vials to lay eggs for 4 d . Females that did not produce viable eggs during this period were discarded . Females were then mated to bw−/bw− males . Prior to mating , these males had either been mutagenized with 1 . 5 mM MMS as described above or put through an equivalent sham treatment without any mutagen . After 1 d , the males were discarded , and the females were transferred to new , individual vials to lay eggs for 2 d ( egg-laying vial 1 ) . Females were then flipped into new vials for another 2 d of egg laying ( egg-laying vial 2 ) . Offspring emerging from both egg-laying vials 1 and 2 were scored for eye color to determine paternity . Because of the strong last-male precedence in Drosophila , it is likely that females producing no brown-eyed offspring had not mated with the second male; these females were excluded from the analysis . We also excluded vials containing fewer than ten offspring . Data were analyzed with a generalized linear mixed model using PROC GLIMMIX in SAS with a logit link function and a binomial error structure where female condition and male mutagen treatment were included as fixed factors , and block and vial were included as random effects . In addition to the analysis above , we also used a likelihood framework to model the proportion of offspring sired by the second male , P2 , as a function of block , female condition , and selection against mutagenized sperm . Separate parameters modeled the strength of selection occurring in low-condition females and in high-condition females , though the model can be constrained so that selection is the same in both types of females . This analysis ignores variation among vials and so may underestimate the uncertainty in parameter estimates . Nonetheless , the parameters are easily interpreted biologically and thus have heuristic value . Specifically , the likelihood analysis worked as follows . Let X be an indicator variable specifying whether a female's second mate had been mutagenized ( X = 1 ) or not ( X = −1 ) . For low-condition females , the proportion of offspring sired by the second male is modeled as P2 , low ( X ) = ki ( 1 − f ) ( 1 − X tlow ) ; for high-condition females , P2 , high ( X ) = ki ( 1 + f ) ( 1 − X thigh ) . The parameters ki , f , tlow , and thigh describe the effect of different factors on P2: ki is approximately the average level of P2 in block i ( i ∈{1 , 2 , 3} ) , f is the effect of female condition on P2 , tlow is the disadvantage of mutagenized sperm in low females , and thigh is the disadvantage of mutagenized sperm in high females . Let mij be the observed number of offspring from female j in block i that were sired by the second male; let nij be the total observed number of offspring from this female . If the true expected siring success of the second male is p , the probability that m out of the n offspring produced by a female will be sired by the second male is given by the binomial distribution , Pr ( m|n , p ) = ( n ! / ( m ! ( n-m ) ! ) ) pm ( 1 − p ) n-m where p is P2 , low ( X ) or P2 , high ( X ) , as appropriate depending on the female's condition and the mutagen status of her second mate . Considering the data from all females , the negative log likelihood of parameter set x = {k1 , k2 , f , tlow , thigh} is given by where Ni is the total number of females in block i . A modified simulated annealing procedure , originating from 25 different random parameter combinations , was used to find the parameters that minimized the value of l ( x ) , i . e . , the maximum likelihood parameter estimates . We calculated the maximum likelihood of the unconstrained model and the maximum likelihood of a constrained model in which the disadvantage of mutagenized sperm was assumed to be the same in both high- and low-condition females , i . e . , tlow = thigh . The maximum likelihood parameters of the unconstrained model gave l ( xmax , unconstrained ) =1 , 587 . 8 , whereas for the constrained model , l ( xmax , constrained ) =1 , 596 . 4 . The unconstrained model had a significantly higher likelihood than the constrained model ( likelihood ratio test , χ2 = 17 . 4 , df = 1 , p = 3 × 10−5 ) indicating that strength of sperm selection differed significantly between high- and low-condition females . The maximum likelihood parameter estimates for the constrained model were k1 = 0 . 96 , k2 = 0 . 93 , k3 = 0 . 95 , f = −0 . 002 , tlow = 0 . 019 , and thigh = 0 . 003; the latter two values indicating selection against mutagenized sperm is stronger in low-condition females than in high-condition females . The standard population genetic parameterization of selection s comes from the reduction in fitness of the less fit type relative to the more fit type , i . e . , wless = ( 1 − s ) wmore . In this case , the less fit type are mutagenized males and the more fit type are non-mutagenized males . The parameter t is related to s by the equation s = 2t/ ( 1 + t ) . This relationship was used to produce the values of sL and sH given in the Results .
|
A variety of evolutionary phenomena are affected by the rate at which mutations enter a population and how those mutations are distributed amongst individuals . Although it is typically assumed that mutations occur randomly among individuals , this may not be the case . Individuals in poor condition may experience elevated mutation rates if they are more prone to experiencing DNA damage or are less able to repair such damage . Using the fruit fly Drosophila melanogaster , we tested whether individuals in poor condition had a reduced capacity to efficiently repair mutagen-induced DNA damage . Consistent with the prediction , we recovered approximately 30% more mutations from low-condition individuals than from high-condition individuals in two separate experiments . Such condition dependence in mutation rate may cause populations to carry considerably heavier loads of deleterious mutations than otherwise expected .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"evolutionary",
"biology"
] |
2008
|
Increased Transmission of Mutations by Low-Condition Females: Evidence for Condition-Dependent DNA Repair
|
Human T-lymphotropic virus type 1 ( HTLV-1 ) has been discovered in 1980 and has been linked to tropical spastic paraparesis ( HAM/TSP ) in 1985 in Martinique . There is no data on HAM/TSP incidence trends . We report , in the present work , the temporal trends incidence of HAM/TSP in Martinique over 25 years . Martinique is a Caribbean French West Indies island deserved by a unique Neurology Department involved in HAM/TSP diagnosis and management . A registry has been set up since 1986 and patients diagnosed for a HAM/TSP were prospectively registered . Only patients with a definite HAM/TSP onset between 1986 and 2010 were included in the present study . The 25-year study time was stratified in five-year periods . Crude incidence rates with 95% confidence interval ( 95%CI ) were calculated using Poisson distribution for each period . Age-standardized rates were calculated using the direct method and the Martinique population census of 1990 as reference . Standardized incidence rate ratios with 95% CIs and P trends were assessed from simple Poisson regression models . Number of HTLV-1 infection among first-time blood donors was retrospectively collected from the central computer data system of the Martinique blood bank . The HTLV-1 seroprevalence into this population has been calculated for four 5-year periods between 1996 and 2015 . Overall , 153 patients were identified ( mean age at onset , 53+/-13 . 1 years; female:male ratio , 4:1 ) . Crude HAM/TSP incidence rates per 100 , 000 per 5 years ( 95%CI ) in 1986–1990 , 1991–1995 , 1996–2000 , 2001–2005 and 2006–2010 periods were 10 . 01 ( 6 . 78–13 . 28 ) , 13 . 02 ( 9 . 34–16 . 70 ) , 11 . 54 ( 8 . 13–14 . 95 ) , 4 . 27 ( 2 . 24–6 . 28 ) and 2 . 03 ( 0 . 62–3 . 43 ) . Age-standardized 5-year incidence rates significantly decreased by 69% and 87% in 2001–2005 and 2006–2010 study periods . Patients characteristics did not differ regarding 1986–2000 and 2001–2010 onset periods . Between 1996–2000 and 2011–2015 study periods , the HTLV-1 seroprevalence significantly decreased by 63% . Martinique faces a sudden and rapid decline of HAM/TSP incidence from 2001 in comparison to 1986–2000 periods . Reduction of HTLV-1 seroprevalence , that may result from transmission prevention strategy , could account for HAM/TSP incidence decrease .
Human T-lymphotropic virus type 1 ( HTLV-1 ) is associated with many diseases including HTLV-1-associated myelopathy/tropical spastic paraparesis ( HAM/TSP ) . It is estimated that about 10–20 million people are infected with HTLV-1 throughout the world [1] . Whereas HTLV-1 seroprevalence is unknown for 80% of world population [2] , data are available in endemic regions and ranges from less than 1 per 10 , 000 people to more than 10% . The highest rates are found in Japan , Brazil , Colombia , the Caribbean islands , Equatorial Africa , Northeast Australia and Papua New Guinea [3] . Routes of infection include unscreened transfusion [4 , 5] and organ transplants [6] , sharing of needles or syringes with infected subjects , sexual contact [7] and breast-feeding [8 , 9] . The predominant HTLV-1 horizontal transmission through condom-less sex leads to dramatic seroprevalence regional variations even in high endemic area . Prevalence in population increases steadily with age and is higher in females [10] . The risk of developing HAM/TSP in HTLV-1 infected individuals has been assessed and deeply varies between studies and ethnic groups . In southern Japan the lifetime risk is 0 . 25% [11] while the 10-year risk reaches 5 . 3% in a Brazilian cohort [12] . Whereas , HTLV-1 seroprevalence rate is widely reported [3] , no data is available on HAM/TSP incidence in general population and in defining area . We report in the present study the incidence of HAM/TSP and its temporal trends , in the population of Martinique , French West Indies , over a 25-year period from 1986 to 2010 . Temporal trends in HTLV-1 infection prevalence among first-time blood donors between 1996 and 2015 have also been calculated .
Martinique is one of the most highly developed islands in the Caribbean , classified high ( 41st ) in terms of global human development at the world level . The population steadily increased between 1982 ( 328 , 566 ) and 2010 ( 394 , 171 ) . The great majority ( 95% ) of the population is Afro-Caribbean . Our Neurology department , located at the University Hospital of Martinique , constitutes the only neurological facility for patients in Martinique . It is highly involved in HAM/TSP diagnosis and management since the original description of the association between HTLV-1 and HAM/TSP [13] . The unique virology department is included in the University hospital . The island is also deserved by 3 neurological rehabilitation centers that work closely with the Neurology department . Since HAM/TSP diagnosis requires cerebrospinal fluid analysis , practitioners and other hospitals use to refer suspected cases to our department . Between 1986 and 2014 , the global health care networks implemented for HTLV-1 infection experienced only slight changes . The most significant changes related to care provided in the Neurology and Rehabilitation departments for HAM/TSP symptoms that have improved between 1986 and 2014 . To reduce HTLV-1 transmission among Martinique population , prevention programs were implemented in the early nineties . Blood-donor and organ-donor screening was systematically introduced . Blood and organ donations are invariably rejected if HTLV-1 is detected . Similarly , antenatal screening was implemented to prevent mother-to-child transmissions of HTLV-1 by breastfeeding avoidance . Information campaigns were regularly performed encouraging the use of condom to prevent sexually transmitted infection including HIV and HTLV-1 . The present study is based on a retrospective analysis on data prospectively collected between 1986 and 2015 in a dedicated registry . Patients with a HAM/TSP onset between 1986 and 2010 were considered . A registry has been set up since January 1986 and was still in progress in June 2015 . HAM/TSP characteristics were collected carefully and continuously . Physicians involved in HTLV-1 diseases including neurologists , rehabilitation physicians , ophthalmologists , hematologists and dermatologists reported patients included in the registry . On a yearly basis , a cross-analysis of virology and clinical database was performed . A standardized case report form has been elaborated and collected the demographic data , the medical history , the clinical symptoms , the initial symptoms ( gait impairment or urinary disturbances ) , the onset-to-diagnosis delay and the CSF analysis . The year of first symptoms , such as stiffness or weakness in the legs or urinary disturbances , defined the year of HAM/TSP onset . Peripheral blood mononuclear cells ( PBMCs ) were isolated from EDTA-enhanced blood by density gradient centrifugation . Real Time TaqMan PCR was performed on DNA extracted from dry pellets of 106 cells stored at -80°C after preparation from blood sample PBMCs . Forward and reverse primers for HTLV-I DNA quantitation: SK110 ( 5′-CCCTACAATCCAACCAGCTCAG-3′ , HTLV-I nucleotide 4758–4779; SK111 ( 5′-GTGGTGAAGCTGCCATCGGGTTTT-3′ , HTLV-I nucleotide 4943–4920 ) . Internal HTLV-I TaqMan probe: 5′CTTTACTGACAAACCCGACCTACCCATGGA-3′; Located between position 4829 and 4858 of the HTLV-I genome , and carried a 5′ reporter dye FAM ( 6-carboxy fluorescein ) and a 3′ quencher dye TAMRA ( 6-carboxy tetramethyl rhodamine ) . Forward and reverse primers for human albumin DNA quantification: Alb-S ( 5′-GCTGTCATCTCTTGTGGGCTGT-3′ ) and Alb-AS ( 5′-AAACTCATGGGAGCTGCT GGTT-3′ ) . Alb TaqMan probe ( 5′-FAM-CCTGTCATGCCCACACAAATCTC TCC-TAMRA-3′ ) were used as described previously [14] . Results were expressed as the number of HTLV-1 copies per 106 PBMCs . For the purpose of the present study , each chart was reviewed and HAM/TSP diagnosis was defined according to an updated staged approach to the World Health Organization criteria [15] . Only definite HAM/TSP was retained in the presence of the following: 1 ) A non-remitting progressive spastic paraparesis with sufficiently impaired gait to be perceived by the patient . Sensory symptoms or signs can be present . When present , they remain subtle and without a clear-cut sensory level . Urinary and anal sphincter signs or symptoms can be present; 2 ) Presence of HTLV-1 antibodies in serum and CSF confirmed by western blot analysis and/or a positive PCR for HTLV-1 in blood and/or CSF; 3 ) Systematic spinal cord imaging and extensive biological work-up ruled out differential diagnosis of progressive paraparesis . In Martinique , all blood donations are performed in the center for blood transfusion ( Etablissement Français du Sang ) . Demographic characteristics of all first-time blood donors tested positive for HTLV-1 were retrospectively collected from the central computer data system of the blood bank . Only donors tested positive with ELISA assay and confirmed with Western blot assay were retained . Data were available for the 20-year period from January 1 , 1996 and December 31 2015 . Characteristics of all donors were not available for the present study . The present research was conducted according to the principles of the declaration of Helsinki . In 1995 , a hospital ethic committee ( Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale ) was constituted and approved the HAM/TSP registry . Between 1986 and 1994 , only oral informed consents were obtained whereas all patients included in the registry after 1994 have given their written informed consent . After standardized information , the oral consent was obtained and collected in the patient’s chart . The hospital ethic committee allowed retrospectively oral consents for 1986–1994 period . For child inclusion participant , a parent has provided informed consent on the child’s behalf . Analysis was performed on anonymized data to ensure confidentiality . Due to the small number of events per year , the 25-year study time , from January 1 , 1986 to December 31 , 2010 , was stratified in five-year periods for HAM/TSP incidence analysis . Similarly , the 20-year study time , from January 1 , 1996 to December 31 , 2015 , for HTLV-1 seroprevalence in first-time blood donors was stratified in four 5-year periods . The numerator for calculation of incidence rate was the number of HAM/TSP patients with a disease onset in the defining period . The denominator was based on data provided by French National Institute for Statistical and Economic Studies ( www . insee . fr ) . By convention , the population of the last year of each defined period was considered for calculation . Crude incidence rates with 95% confidence intervals ( 95%CIs ) were calculated using Poisson distribution for each period . Age-standardized rates were calculated using the direct method and the Martinique population census of 1990 as reference . The Martinique population census of 1990 constituted the denominator to calculate HAM/TSP incidence for the first study period 1986–1990 . Then populations of 1995 , 2000 , 2005 and 2010 were used for 1991–1995 , 1996–2000 , 2001–2005 and 2006–2010 study periods . Standardized incidence rate ratios with 95% CIs and P trends were assessed from simple Poisson regression models . P for trend was used to either support or reject a linear trend in HAM/TSP standardized incidence rates from the first to the last study period . Seroprevalence of HTLV-1 within the four periods was expressed as number of seropositive HTLV-1 subjects per one hundred new blood donors . A binomial analysis was used for 95% CI of seroprevalence rates . The first study period ( 1996–2000 ) was used as reference to calculate the odds ratio of seroprevalence with 95%CI in the four study periods . The χ2 test or Fisher-exact test , as appropriate , and Student t test or Mann-Whitney test were used to examine differences in nominal and continuous values . All analyses were performed in SAS 9 . 3 ( SAS Institute , Inc , Cary , NC ) and MS Excel software .
The 5-year crude incidence rates ( 95%CI ) per 100 , 000 in 1986–1990 , 1991–1995 , 1996–2000 , 2001–2005 and 2006–2010 study periods are reported in Table 2 . Between 1986 and 2000 rates were stable and ranged from 10 . 01 ( 6 . 78–13 . 28 ) to 13 . 02 ( 9 . 34–16 . 7 ) whereas rates declined in the two earliest periods 2001–2005 and 2006–2010 , 4 . 27 ( 2 . 24–6 . 28 ) and 2 . 03 ( 0 . 62–3 . 43 ) . Age-standardized incidence rates on Martinique population census of 1990 ( used for 1986–1990 study period ) significantly decreased by 69% and 87% in 2001–2005 and 2006–2010 study periods ( Table 2 ) . Regarding HAM/TSP incidence rate decline from 2001 , we compared patient’s characteristics between the two groups whose disease onset was in 1986–2000 and 2001–2010 periods . Table 3 compares the characteristics of HAM/TSP patients with a disease onset in 1986–2000 and in 2001–2010 . Sex ratio , initial symptoms , onset-to-diagnosis delay and history of blood transfusion before 1990 were comparable between the two groups . We found a trend to significance for an older age at onset in 2001–2010 incident cases . Level of HTLV-1 in PBMC was available for 123 out of 153 patients . The mean level of HTLV-1 proviral load did not differed between the two study periods . Table 4 shows the prevalence of HTLV-1 infection in first-time blood donors within the four 5-year study periods . Seropositive male subjects were statistically younger in the first study period compared to the 3 earliest periods . Compared to the first study period ( 1996–2000 ) , HTLV-1 infection frequencies were significantly lower in the earliest study periods ( 2001–2005 , 2006–2010 , 2011–2015 ) as illustrated by the absence of 95%CI range limits overlap . Between 1996–2000 and 2011–2015 study periods , the HTLV-1 seroprevalence significantly decreased by 63% , 67% and 61% in the whole , female and male population of first-time blood donors respectively .
Over 25 years , 1986–2010 , our study reports a significant decrease of new definite HAM/TSP cases . Whereas the crude 5-year incidence rates remained stable between 1986 and 2000 , ranging from 10 . 1 to 12 . 2/100 , 000 , the rates rapidly decline between 2001 and 2010 , ranging from 3 . 2 and 2 . 9/100 , 000 . The age-standardized 5-year incidence rates exhibit a dramatic decrease by 70% from 2001 and reaching 87% for 2006–2010 periods . In accordance with previous studies , the age at onset of HAM/TSP patients was around 50 years old [16–18] . We found a trend to a significant older age at onset after 2000 ( 52 . 1 versus 57 . 5 , p = 0 . 06 ) that may reflect an age cohort effect and that could be indicative of a rapid decrease in HTLV-1 seroprevalence . As usually reported , females are highly predominant in Martinique HAM/TSP patients [19] and the proportion does not differ between 1986–2000 and 2001–2010 onset cases . Blood transfusion constitutes a usual infection route and a study from Japan showed a 16% decrease in the incidence of HAM/TSP 2 years after blood-donor screening implementation [20] . We did not observe any differences in blood transfusion history frequencies between patients whose onset was before or after 1990 . As previously reported , [19] initial neurological symptoms were predominantly gait impairment without any difference between the two onset groups . The median delay between onset and diagnosis was stable over the study time and was set at 3 years . In 1989 , HTLV-1 seroprevalence was appraised at 2 . 2% in the general population of Martinique [21] , while it was estimated to 1 . 93% among pregnant women [22] and to 0 . 4% among blood donors cohort in the mid-nineties [23] . Although no relevant data are available on HTLV-1 seroprevalence trends in our general population , infection rates in blood donors were analyzed over twenty years from 1996 to 2015 . Seroprevalence of HTLV-1 infection among first-time blood donors decreased significantly by 63% between 1996–2000 and 2011–2015 study periods . Seroprevalence decrease may partly account for HAM/TSP incidence decline and may result from several factors . These include the rapid westernization of life style that spread through the whole population from the eighties , the systematic screening for HTLV-1 antibodies in volunteer blood donor and pregnant women [16 , 17] that has been implemented in the early nineties in Martinique and the iterative information campaigns for HIV and HTLV-1 prevention that particularly encouraged the use of condoms in the population . Prevention strategy leading to refrain HTLV-1 infected mother to breast fed their children has demonstrated efficiency in seroprevalence decrease [9] . However , such a dramatic HAM/TSP incidence decline from 2001 is intriguing . Indeed , individuals aged around 50 years old between 2001 and 2010 became sexually active before implementation of HTLV-1 transmission prevention strategy . Additionally , they were all exposed to HTLV-1 contamination through breastfeeding and were at exposure risk through untested blood transfusion at least the first 20 years of life . HTLV-1 transmission prevention strategy could not be sufficient to account for such an early incidence decrease . This raises the question of a co-factor that could play a role in refraining the development of HAM/TSP disease among HTLV-1 infected population . Rapid and extensive westernization improved the socioeconomic level of the Martinique population and was followed by sanitary changes to the environment . Sanitation and hygienic behavior improvements resulted in a rapid reduction in parasite infection . Between 1978 and 1994 , the intestinal parasitism in children decreased from 70% to 8% [24] . Few studies , with divergent conclusions , focused on HAM/TSP development and parasite infection . Whereas a study assume that Schistosoma proteins may reduce inflammatory process associated with HTLV-1 infection through interferon-gamma level decrease [25] , another one report that treatment of helminthic infection does not affect the risk of HAM/TSP development [26] . Gillet et al . [27] suggest that co-infection with Strongyloides is associated with formation of new HTLV-1-infected T-cell clones and with oligoclonal proliferation of certain HTLV-1 clones , increasing thereby the risk of HTLV-1 disease expression . On the other hand , whereas HTLV-1 and Strongyloides infect 43% and 35% of tested indigenous in Central Australia [28] , the HAM/TSP prevalence appears very low in these communities with extremely poor access to sanitation and hygiene facilities [29] . Studies are still needed to investigate the role of a potential co-factor associated with sanitary change and potentially leading to decline in number of HAM/TSP diagnosis . The present study does not allow reliable conclusions about the synergistic effects of HTLV-1 transmission prevention programs and improvement of sanitary conditions in HAM/TSP incidence rates decline from 2001 in Martinique . Whereas meta-analyses show no clear evidence that structural interventions at community level , to increase condom use , prevent the transmission of HIV and other sexually transmitted infections [30] , no reliable study has focused on HTLV-1 infection . We assume that HTLV-1 sexual transmission is probably the next challenge in Martinique to continue HTLV-1 seroprevalence decrease . Prevention strategies focusing on sexual transmission have to be intensified and should lead to increase acceptability of condom in the general population using mass media campaigns , public opinion , meeting’s school information and social marketing advertisement; improve condom accessibility particularly for low-income population; improve female access to condom; extend the availability of condoms such as condom machine installation in public and private spaces . Despite the large study period of 25 years , the number of new HAM/TSP cases remains limited with low statistical power particularly in comparing the characteristics of the two groups of patients with a disease onset before and after 2001 . Between 1986 and 2015 , technological improvement of imaging and laboratory investigations may have influenced HAM/TSP diagnosis accuracy . However , HAM/TSP diagnosis criteria did not change during the study time and imaging methods that are crucial to rule out alternative myelopathy and especially spinal cord compression , were available through the whole study time . A reduction of HAM/TSP patients referred to our network by the primary health setting over the study period might be raised . However , only slight changes occurred to the global health care network for HTLV-1 infected patients . Moreover , we assume that improvement of medical management of HAM/TSP patients in Neurology and Rehabilitation centers may have prompted general practitioners and primary care centers to refer more systematically patients to our network . We showed that HTLV-1 seroprevalence have decreased over 20 years among first-time blood donors , but the absence of demographic characteristics of this population limits our findings interpretation . In summary , Martinique faces a rapid and sustained decline of HAM/TSP incidence from 2001 estimated by 70% in comparison to 1986–2000 periods . Implementation of HTLV-1 transmission prevention strategy may result in seroprevalence reduction and in HAM/TSP incidence decrease observed 15 years later . The monitoring of new HAM/TSP cases has to be continued and HTLV-1 seroprevalence in blood donors and pregnant women must be sequentially assessed .
|
Human T-lymphotropic virus type 1 ( HTLV-1 ) was discovered in 1980 and HTLV-1-associated myelopathy/tropical spastic paraparesis ( HAM/TSP ) was described five years later in 1985 . HAM/TSP is a progressive disabling disorder characterized by spastic paraparesis with bladder and bowel dysfunction that constitutes a significant public health problem in endemic areas . Up to date , there is no efficiency treatment of HAM/TSP and prevention of HTLV-1 transmission is critical to limit the disease spreading throughout communities . In the present 25-year-study time , we report a significant decrease of HAM/TSP incidence estimated more than 70% in early 2000 compared to 1986–2000 period in Martinique a French West Indies Island . We found a trend to a significant older age at onset after 2000 ( 52 . 1 years versus 57 . 5 years , p = 0 . 06 ) that may reflect an age cohort effect and that could be indicative of a rapid decrease in HTLV-1 seroprevalence . We showed a significant decline in HTLV-1 infection among first-time blood donors between 1996–2000 and 2011–2015 study periods . Thus , probable HTLV-1 seroprevalence decrease secondary to HTLV-1 antibodies screening in blood donors and pregnant women and to iterative information campaigns could partly account for HAM/TSP incidence decline . This study emphasizes the importance of prevention strategies to control HAM/TSP development in HTLV-1 endemic areas .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"geomorphology",
"medicine",
"and",
"health",
"sciences",
"clinical",
"laboratory",
"sciences",
"pathology",
"and",
"laboratory",
"medicine",
"landforms",
"pathogens",
"topography",
"geographical",
"locations",
"microbiology",
"neurorehabilitation",
"census",
"health",
"care",
"retroviruses",
"viruses",
"north",
"america",
"research",
"design",
"rna",
"viruses",
"sexually",
"transmitted",
"diseases",
"caribbean",
"transfusion",
"medicine",
"islands",
"research",
"and",
"analysis",
"methods",
"infectious",
"diseases",
"medical",
"microbiology",
"htlv-1",
"microbial",
"pathogens",
"rehabilitation",
"medicine",
"hematology",
"people",
"and",
"places",
"diagnostic",
"medicine",
"survey",
"research",
"viral",
"pathogens",
"earth",
"sciences",
"neurology",
"biology",
"and",
"life",
"sciences",
"blood",
"donors",
"blood",
"transfusion",
"martinique",
"organisms"
] |
2018
|
Temporal trends in Human T-Lymphotropic virus 1 (HTLV-1) associated myelopathy/tropical spastic paraparesis (HAM/TSP) incidence in Martinique over 25 years (1986-2010)
|
Pulmonary adenoma susceptibility 1 ( Pas1 ) is the major locus responsible for lung tumor susceptibility in mice; among the six genes mapping in this locus , Kras is considered the best candidate for Pas1 function although how it determines tumor susceptibility remains unknown . In an ( A/J×C57BL/6 ) F4 intercross population treated with urethane to induce lung tumors , Pas1 not only modulated tumor susceptibility ( LOD score = 48 , 69% of phenotypic variance explained ) but also acted , in lung tumor tissue , as an expression quantitative trait locus ( QTL ) for Kras-4A , one of two alternatively spliced Kras transcripts , but not Kras-4B . Additionally , Kras-4A showed differential allelic expression in lung tumor tissue of ( A/J×C57BL/6 ) F4 heterozygous mice , with significantly higher expression from the A/J-derived allele; these results suggest that cis-acting elements control Kras-4A expression . In normal lung tissue from untreated mice of the same cross , Kras-4A levels were also highly linked to the Pas1 locus ( LOD score = 23 . 2 , 62% of phenotypic variance explained ) and preferentially generated from the A/J-derived allele , indicating that Pas1 is an expression QTL in normal lung tissue as well . Overall , the present findings shed new light on the genetic mechanism by which Pas1 modulates the susceptibility to lung tumorigenesis , through the fine control of Kras isoform levels .
The Pulmonary adenoma susceptibility 1 ( Pas1 ) locus , mapping in the distal region of chromosome 6 , is the major modulator of lung tumor susceptibility in mice [1] , [2] . In inbred mouse strains , the Pas1 locus has a conserved haplotype consisting of six genes clustered in a ∼450-kb region [3] . From proximal to distal , these genes are branched chain aminotransferase 1 ( Bcat1 ) , lymphoid-restricted membrane protein ( Lrmp ) , cancer susceptibility candidate 1 ( Casc1 ) , LYR motif containing 5 ( Lyrm5 ) , v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog ( Kras ) and intermediate filament tail domain containing 1 ( Ifltd1 ) . Four of these genes , namely Lrmp , Casc1 , Kras and Ifltd1 , have been singled out as candidate genes for Pas1 locus functions in fine-mapping studies [4]–[6] . Although for none of these genes is there a clear demonstration of a role in lung tumor susceptibility , there is substantial evidence supporting the involvement of Kras as the key effector of the Pas1 locus . Kras encodes a small GTPase that functions as a molecular switch in signal transduction , influencing cell proliferation [7] . Permanently activating mutations , at codons 12 , 13 and 61 , are frequently found in both spontaneous and chemically induced lung tumors in mice [8] , [9] . Moreover , in mouse models in which mutant Kras can be activated by somatic recombination in the lung , animals are highly susceptible to lung tumorigenesis and develop multiple lung tumors at 100% incidence with a short latency [10] , [11] . Nonetheless , as heterozygous Kras knockout mice have higher susceptibility to chemically induced lung tumorigenesis than wild-type mice , the wild-type Kras allele may have a tumor suppression function [12] . The double role of Kras in lung carcinogenesis—tumor suppressor when wild-type and oncogene when mutated—led to the hypothesis that lung cancer susceptibility could result from the subtle balance between expression levels of wild-type and mutated Kras [13] . Given the frequent occurrence of activating Kras mutations in mouse lung tumors , an increase in Kras gene expression in lung tumor tissue could be expected to raise the level of active Kras protein , thereby providing the growth advantage characteristic of neoplasms . Evidence supporting this mechanism was provided by the observation that , in mice , Kras mRNA levels in normal lung tissue were ∼2-fold higher in strains susceptible to lung tumorigenesis ( both highly susceptible A/J mice and intermediate-susceptible FVB/N and 129/Sv mice ) than in a resistant strain ( C57BL/6 ) [13] . Our understanding of the mode of action of Kras in lung tumorigenesis is further complicated by the existence of two main transcripts , namely Kras-4A and -4B , generated by alternative splicing of its fourth coding exon . The two transcripts differ in their 3′-termini and give rise to two proteins ( of 189 and 188 residues , respectively ) with different C-terminal sequences . Because the C-termini of Ras proteins function in plasma membrane binding ( through both electrostatic binding of basic residues and hydrophobic binding of fatty acylated residues ) [14] , sequence variations in this region could affect the biological functions of the isoforms . Indeed , in transfected cells , human Kras-4A , but not Kras-4B , efficiently induced transformed foci and enabled anchorage-independent growth [15] , [16] . Additional differences between the two isoforms regard their expression levels , which were found to be modulated during mouse embryogenesis and to vary in different tissues [17] , [18] . Moreover , the ratio between Kras-4A and Kras-4B mRNA levels in lung tissue was higher in mouse strains susceptible to lung cancer than in resistant ones [17] . More recent evidence implicating the Kras-4A oncoprotein in lung tumorigenesis was provided by two studies in which mice expressing only the Kras-4B isoform had greater resistance to chemical carcinogenesis than did wild-type mice also expressing Kras-4A [19] , [20] . Despite these observations , several aspects of the candidacy of Kras as the major effector of the Pas1 locus in lung tumor susceptibility remain to be clarified . Wild-type alleles from susceptible and resistant mice code for the same protein; consequently , the mechanism by which Kras determines genetic susceptibility to lung tumorigenesis could , instead , depend on its overall expression level or on the ratio of its isoforms in normal lung tissue . To address these points and to clarify the involvement of the two Kras isoforms , we studied Kras mRNA levels and germline variants using an established model of urethane-induced lung tumorigenesis in an advanced intercross population between A/J mice ( tumor susceptible ) and C57BL/6 mice ( resistant ) . This model was previously instrumental in defining the Pas1 locus based on the pattern of tumorigenesis in F2 intercross mice [1] . In the present study , we used F4 mice ( of a new pedigree ) in order to have a greater resolution power . After confirming Pas1 as a quantitative trait locus ( QTL ) modulating lung tumor multiplicity in this pedigree , we performed a genome-wide association study to look for expression QTL modulating the levels of Kras-4A and Kras-4B transcripts in lung tumors and normal lung tissue . Our findings indicate that Kras-4A and , to some extent , also Kras-4B are modulated by cis-acting elements within the Pas1 locus itself . These results underpin a possible causal relationship between germline variations and mRNA levels of the Kras-4A isoform , consequently influencing lung tumor susceptibility .
From 80 of the 183 ABF4 urethane-treated mice used in the genetic linkage analysis , we were able to resect a tumor specimen from lung tissue . This subgroup included 37 mice homozygous at rs6265387 for the A/J-derived allele ( associated with tumor susceptibility ) , 34 heterozygous mice , and 9 mice homozygous for the C57BL/6 resistant allele that nevertheless had developed a lung tumor . Indeed , the intrinsic low susceptibility to lung tumorigenesis of the 42 animals homozygous for the C57BL/6-derived allele ( only 9 of these mice developed a lung tumor ) did not allow us to analyze a subgroup of the original population with a genotype ratio typical of such intercrosses ( i . e . , 1∶2∶1 ) . RNA was extracted from each specimen and used in quantitative PCR to measure the levels of Kras-4A and -4B mRNA . Genome-wide linkage analysis for Kras-4A found a single expression QTL on chromosome 6 ( peak LOD = 4 . 5 ) , corresponding to the Pas1 locus ( Figure 2A ) . In contrast , no expression QTL was found for Kras-4B on any chromosome , including chromosome 6 where LOD scores remained below the threshold for significance . When the 80 mice were grouped according to the genotype at rs6265387 , we found that Kras-4A mRNA was highest in mice homozygous for the A/J-derived susceptible allele ( GG ) , intermediate in heterozygous ( AG ) animals , and lowest in mice homozygous for the C57BL/6 resistant allele ( AA ) ; these differences were significant ( GG vs . AG , P = 3 . 2×10−3; GG vs . AA , P = 8 . 9×10−5; ANOVA followed by Tukey's test for multiple comparisons , Figure 2B ) . These results implicate Kras-4A in the mechanism of Pas1-dependent tumor formation . The genetic susceptibility to cancer is an intrinsic feature of normal tissue that , in urethane-treated mice , influences both the probability of tumor initiation and the number of tumors that develop . Consequently , it is likely that the fundamental elements underlying this phenomenon are present not only in tumors but also in normal tissue . To test this hypothesis , we attempted to validate in normal lung tissue the genetic linkage we observed in lung tumors . Therefore , we genotyped 111 untreated male ABF4 mice for nine markers on chromosome 6 spanning from 96 . 7 Mb to 148 . 3 Mb . These markers include 8 SNPs and one 37-bp sequence variation that in C57BL/6 mice occurs as a tandem repeat; this insertion mutation is located in the second intron of Kras [21] . Additionally , we assayed lung mRNA from these mice for the two Kras isoforms , and examined the association of mRNA levels with genotype . A strong linkage was found between the level of Kras-4A mRNA and genotyped markers in the Pas1 locus , describing a LOD curve with a peak of 23 . 2 ( Figure 3A ) ; this expression QTL explained 62% of the phenotypic variance . A much weaker , yet statistically significant linkage for the Kras-4B transcript was observed ( maximum LOD score = 4 . 1 ) . These results confirm and strengthen the major role of Kras-4A isoform found in tumors ( Figure 2 ) . We then grouped the mice according to genotype at the Kras 37-bp insertion to examine the relationship between Kras mRNA levels and the number of inherited A/J-derived alleles of the Pas1 locus . A strong association was found for Kras-4A levels ( P = 9 . 7×10−14 , ANOVA; square-root-transformed values are shown in Figure 3B ) : mice with two A/J-derived alleles ( i . e . no insertion , -/- ) had on average 1 . 9-times more transcript than did mice receiving two C57BL/6-derived alleles ( ins/ins ) , and heterozygous mice had intermediate levels . In contrast , the linkage between Pas1 genotype and Kras-4B transcript levels was weaker ( P = 9 . 7×10−5 , ANOVA , Figure 3C ) . We then examined whether this specific genetic control exerted by the Pas1 locus was also present in normal lung of the parental inbred strains A/J and C57BL/6 ( Figure 4 ) . The median level of the Kras-4A transcript in A/J ( tumor-susceptible ) mice was 2 . 3-fold higher than that of C57BL/6 ( resistant ) mice ( P = 6 . 0×10−9 , ANOVA; square-root-transformed values are shown in Figure 4A ) . In contrast , the Kras-4B levels were not significantly different between mouse strains ( Figure 4B ) . Altogether , these results indicate that the A/J-derived allele of the Pas1 locus , which confers susceptibility to lung tumorigenesis , is also associated with higher steady-state levels of the Kras-4A splice variant . These observations suggest that subtle modulation of Kras-4A mRNA production or stability may be the key effector of Pas1-controlled lung tumor susceptibility , possibly via a cis-acting element within Pas1 itself . To test whether the allele-specific levels of Kras isoforms are attributable to variations in cis-regulatory elements also mapping in the Pas1 locus , we examined the possibility of differential allelic expression . For this analysis , we considered two SNPs ( rs29968550 and rs30022167 , mapping in a region of Kras mRNA common to the two isoforms ) whose genotypes permitted us to distinguish the A/J- from the C57BL/6J-derived allele . By pyrosequencing , we determined the frequencies of the two alleles in Kras-4A and -4B cDNA from normal lung tissue of 20 heterozygous untreated ABF4 mice and from lung tumor specimens of 15 heterozygous treated mice and in genomic DNA from the same animals , used as control for the amplification efficiency of the two alleles ( Figure S1 ) . Indeed , when allelic imbalance is tested , uneven amplification unrelated to differential transcription may occur in some of the assays . Therefore , the experimentally obtained allelic ratio for genomic DNA , which in principle is equal to 1 in heterozygotes , is used as the baseline to evaluate the ratios observed for cDNAs . Based on the results for both SNPs , comparing the mean allelic ratios for Kras-4A cDNA with those of the corresponding genomic DNA , we observed a ∼2-fold higher expression of the A/J-derived allele in Kras-4A cDNA in both tumor and normal tissue indicating a higher steady state of this isoform associated with the A/J-derived allele , possibly due to either a higher production rate or a lower degradation rate ( Figure 5 ) . For Kras-4B , the mean allelic ratios for cDNA from tumors were similar to those for genomic DNA , meaning equal production from both alleles; in normal tissue , however , a slightly higher ratio for Kras-4B cDNA relative to genomic DNA was observed ( Figure 5 ) . In the statistical analysis , allelic ratios were log10-transformed to approximate a normal distribution before applying Welch's t test ( Table 1 ) . This analysis showed that the high allelic ratios for Kras-4A cDNA , observed at two SNPs and in both normal and tumoral lung tissue , are significantly different from those of genomic DNA ( P<0 . 001 , see Table 1 ) . In the case of Kras-4B cDNA , the allelic ratios were significantly different from that of genomic DNA only for normal lung tissue ( P<0 . 001 ) . These results provide evidence for the differential allelic expression of Kras transcripts , with higher levels being produced from the A/J-derived allele in ABF4 intercross mice . These findings are compatible with the presence of allele-specific germline variations in cis-acting elements that mainly influence the selection or stability of the Kras-4A isoform .
In this study , we verified the major role of the Pas1 locus in modulating lung tumorigenesis in an ABF4 advanced intercross population , where the lung tumor multiplicity ( Nlung ) phenotype behaved as a monogenic trait under the control of the Pas1 locus , explaining almost 70% of the phenotypic variance . In addition , we found that Pas1 is an expression QTL because it controls the level of the Kras-4A isoform in urethane-induced lung tumors . This genetic control is an inherited trait , in that it was already present in normal lung tissue of untreated mice . Finally , we observed the preferential expression of Kras-4A from the A/J-derived allele in heterozygous mice , suggesting the existence of allele-specific germline variations in cis-acting elements that influence splicing bias , selection or stability of this isoform . We carried out this genetic study in an ABF4 intercross since it was an effective way to accumulate , in a population of a relatively small size , 3-fold more recombination events than those possible in a conventional F2 intercross population of the same size [22] . Indeed , each ABF4 mouse underwent six informative meioses compared to the two informative meioses of an F2 mouse , thus providing improved resolution for mapping loci affecting strain-related phenotypes [22] . This methodological choice allowed us to obtain a sharp peak of linkage between Pas1 and Nlung , with very high association ( LOD score = 48 in 183 ABF4 mice ) . The high rate of recombination in this intercross was also useful for testing our hypothesis that Kras isoform expression is modulated by Pas1 . Animals homozygous for the C57BL/6-derived allele had low susceptibility to lung tumorigenesis and , therefore , rarely developed lung tumors after urethane treatment . Hence , the genetic class with the resistant phenotype was under-represented in the experiment designed to identify expression QTLs in tumors . Another difficulty that presented in our research was the known dysregulation of gene expression architecture in cancer , including mouse lung tumors [23] , [24] , which could have reduced the detectability of expression QTLs . Despite these limitations , we observed a significant linkage of the Pas1 locus with mRNA levels of Kras-4A ( LOD score = 4 . 5 in 80 mice ) but not of Kras-4B . In normal lung tissues from untreated mice , where all three genotypes were present at roughly the expected ratio and where no pathologic alterations could have played a confounding role , we observed that the Kras-4A isoform levels were again significantly linked with the Pas1 locus ( LOD score = 23 . 2 in 111 mice ) . In these untreated animals , the Kras-4B isoform also had significant linkage with Pas1 , although the LOD score of 4 . 1 was only ∼17% that obtained for Kras-4A . The differential allelic expression of Kras-4A and , to some extent , Kras-4B , in normal lung tissue and in lung tumor specimens , indicated the existence of regulatory polymorphisms located within the Pas1 locus . These cis-acting elements have not yet been identified . However , since the Kras gene promoter has been shown to have similar activity in susceptible and resistant mice [25] , it is conceivable that these cis-acting elements are located in other regions such as introns or the 3′-untranslated region of the Kras transcript . Regulatory elements in these regions can be expected to impact more on mRNA splicing and stability than on promoter activity . Several polymorphisms in non-coding regions of Kras have already been found in various mouse inbred strains and the functional roles of some of them have been investigated [21] , [25] , [26]; these studies , however , reached contrasting conclusions . Future investigations could focus on the targeted re-sequencing of the A/J-derived Kras gene and flanking regions , to search for not yet identified polymorphisms that affect splicing selection or mRNA stability of Kras isoforms . The candidate functional polymorphisms could then be tested by cloning into reporter vectors for transfection and assay of cell lines . Overall , these results shed new light on the role of expression QTLs in mouse lung tumorigenesis , since they indicate that the genetic control exerted by Pas1 mostly affects the expression of the Kras-4A transcript . These results point to the 4A isoform as the functional element of Pas1 , the major locus modulating lung tumorigenesis in mice .
All animals received humane care according to the criteria outlined in a protocol approved during the meeting board of December 21 , 2006 , by the institutional ethical committee for animal use ( CESA ) at the Fondazione IRCCS Istituto Nazionale dei Tumori . A pedigree of intercross mice was generated by mating lung tumor-susceptible A/J mice ( A; 5 females ) with lung tumor-resistant C57BL/6 mice ( B; 5 males ) . From generation ABF2 to ABF4 , male and female mice were individually labeled , randomly selected using a random number generator and bred , avoiding the pairing of siblings ( 20 female and 20 male ABF2 mice , and 55 female and 55 male ABF3 mice were mated ) . After weaning , ABF4 mice were sexed and ear-tagged , and a section of tail was collected for DNA extraction . Male ABF4 mice were then randomly assigned to two groups: untreated mice ( n = 131 ) were raised under standard conditions , while 188 male mice were treated with a single intraperitoneal injection of urethane ( 1 g/kg body weight ) at 4 weeks of age to induce the development of lung tumors [27] . To obtain normal lung tissue , 111 untreated AFB4 mice , 11 A/J and 11 C57BL/6 mice ( all males ) were anesthetized and killed at 16 weeks of age; lung lobes were isolated and frozen in liquid nitrogen . To obtain lung tumor tissue , the 183 urethane-treated mice were killed at 40 weeks of age; the chest was opened , the trachea was dissected to permit needle access , and the lungs were filled with 0 . 5 ml RNALater solution . Then , the lung lobes were removed , placed in RNALater , and kept overnight at 4°C degrees . The next day , lungs were examined for surface tumors . For each mouse , we recorded the total number of tumors ( Nlung ) and , whenever possible , resected one tumor of about 1–1 . 5 mm under a stereo-microscope . These tumor specimens were frozen and stored at −80°C degrees . Genomic DNA was extracted from tail samples using the DNeasy Blood & Tissue Kit ( Qiagen , Valencia , CA , USA ) and quantified using Picogreen dsDNA Quantitation Kit ( Invitrogen , Life Technologies , Paisley , UK ) . Total RNA was extracted from normal lung tissue and from lung tumor specimens using RNeasy Midi Kit ( Qiagen ) , purified with RNeasy MinElute Cleanup ( Qiagen ) , and quantified by spectrophotometry ( ND-1000 spectrophotometer , NanoDrop , Wilmington , DE , USA ) . RNA integrity was verified using the RNA 6000 Nano Assay Kit ( Agilent Technologies , Palo Alto , CA , USA ) ; the mean RIN value was 8 . 9 ( SD = 0 . 8; range , 6 . 7 to 9 . 9 ) . Genomic DNA from urethane-treated ABF4 mice was used for genome-wide SNP genotyping carried out with the GoldenGate Genotyping Assay according to the manufacturer's protocol ( Illumina , San Diego , CA , USA ) , using the Mouse MD Linkage Panel representing 1449 mouse loci . Additionally , the Pas1 locus in untreated ABF4 mice was genotyped for nine genetic markers , including eight SNPs ( rs6349084 , rs33863668 , rs31000839 , rs31005929 , rs13479063 , rs33893742 , rs13479082 , rs3711088 ) and a 37-bp tandem repeat in Kras [21] . Briefly , a genomic fragment surrounding each marker was PCR-amplified in reactions containing 30 ng genomic DNA , 1 U AmpliTaq Gold ( Applied Biosystems , Life Technologies ) , 1× AmpliTaq Gold buffer ( Applied Biosystems , Life Technologies ) , 1 . 5 mM MgCl2 , 200 µM dNTPs and 5 pmol of each of a pair of specific primers ( Table S1 ) in a total volume of 25 µl . To genotype the eight SNPs , PCR products were pyrosequenced on a PyroMark Q96 ID system running PyroMark Q96 ID Software ( Qiagen ) . To genotype the Kras 37-bp repeat , PCR amplicons were analyzed by 3% agarose gel electrophoresis for fragment size . RNA from normal lung ( 1 µg ) or from lung tumor specimens ( 0 . 5 µg ) was used to synthesize cDNA by reverse transcription using the Transcriptor First Strand cDNA Synthesis Kit ( Roche , Basel , Switzerland ) . Kras-4A and -4B transcript levels in the cDNA were measured in quantitative PCR ( qPCR ) assays using Fast SYBR Green PCR Master Mix ( Applied Biosystems ) and 300 nM intron-spanning primers ( Table S1 ) . Hypoxanthine phosphoribosyltransferase 1 ( Hprt1 , Table S1 ) was used as reference gene . Relative expression levels were calculated using the comparative Ct method using one of the synthesized cDNA samples as calibrator . The allelic expression of Kras-4A and Kras-4B isoforms was analyzed taking into consideration two SNPs in the 3′-UTR region of the Kras gene ( Figure S1 ) : rs30022167 ( A or C , for the A/J- or C57BL/6-derived allele , respectively ) and rs29968550 ( A or G , for the A/J- or C57BL/6-derived allele , respectively ) . Among the animals heterozygous for markers at the Pas1 locus , we selected 20 mice from the untreated group and 15 from the urethane-treated group , from which we already had genomic DNA . In addition , from these animals , we also had cDNA from normal lung ( n = 20 ) or from tumor specimens ( n = 15 ) , having been reverse-transcribed for the qPCR experiments . For each Kras isoform , we carried out a first amplification step on the cDNA samples using a forward primer located in exon 4A ( Kras-4A ) or in the junction between exon 3 and exon 4B ( Kras-4B ) and a common reverse primer located in the 3′-UTR region of Kras gene downstream of rs30022167 and rs29968550 ( Table S1 and Figure S1 ) . The resulting PCR product as well as a sample of genomic DNA from the same animal were PCR-amplified using SNP-specific primers . These PCR products were pyrosequenced on a PyroMark Q96 ID system running PyroMark Q96 ID Software ( Qiagen ) . For each SNP , the proportions of the two alleles present in each sample ( genomic DNA , Kras-4A cDNA and Kras-4B cDNA from untreated and tumor-bearing mice ) were determined from pyrogram peak heights , and allelic ratios ( A/J-derived allele/C57BL/6-derived allele ) were calculated . Nlung values and mRNA levels were square-root transformed to improve the normality of distribution . The transformed values and genotype data were analyzed by simple interval mapping using R/qtl [28] to identify QTLs . LOD scores were considered significant if greater than the 95% LOD threshold ( α = 0 . 05 ) , calculated by 10 , 000 permutations . The percent phenotypic variance explained by a given QTL was calculated from the LOD score by the following formula: R2 = [1−10 ( −2LOD/n ) ] , where n is the sample size [28] . Differences between genotype groups were analyzed by one-way ANOVA followed by Tukey's test for multiple comparisons . Log10-transformed allelic ratios were compared between cDNA and genomic DNA using a two-tailed Welch's t test . Data were considered significant when P<0 . 01 .
|
A person's risk of developing cancer depends on both genetic and environmental factors . To study the genetic predisposition to cancer without the influence of environmental variables , scientists study mice treated with urethane , a chemical carcinogen that induces lung tumors . By crossing inbred ( genetically identical ) strains of mice that are either resistant or susceptible to urethane-induced cancer , researchers can search for genes associated with tumor formation in the offspring . From previous work of this type using second-generation mice , it was already known that a region on chromosome 6 was associated with tumor formation . Now , a new study , carried out in a fourth-generation mouse population , focused to a single gene of chromosome 6 called Kras . This gene forms two different messenger RNA transcripts , called Kras-4A and Kras-4B , that produce two proteins with slightly different structure and , perhaps , function . The study found that mice susceptible to lung tumors have relatively more Kras-4A messenger RNA than resistant mice and that this difference may be due to small variations in the DNA near or within this gene .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"animal",
"models",
"cancer",
"genetics",
"model",
"organisms",
"genetic",
"polymorphism",
"gene",
"expression",
"genetics",
"gene",
"regulation",
"biology",
"and",
"life",
"sciences",
"population",
"genetics",
"molecular",
"genetics",
"genomics",
"evolutionary",
"biology",
"mouse",
"models",
"genetics",
"of",
"disease",
"dna",
"transcription",
"research",
"and",
"analysis",
"methods"
] |
2014
|
Mouse Pulmonary Adenoma Susceptibility 1 Locus Is an Expression QTL Modulating Kras-4A
|
Yeast two-hybrid screens are an important method for mapping pairwise physical interactions between proteins . The fraction of interactions detected in independent screens can be very small , and an outstanding challenge is to determine the reason for the low overlap . Low overlap can arise from either a high false-discovery rate ( interaction sets have low overlap because each set is contaminated by a large number of stochastic false-positive interactions ) or a high false-negative rate ( interaction sets have low overlap because each misses many true interactions ) . We extend capture–recapture theory to provide the first unified model for false-positive and false-negative rates for two-hybrid screens . Analysis of yeast , worm , and fly data indicates that 25% to 45% of the reported interactions are likely false positives . Membrane proteins have higher false-discovery rates on average , and signal transduction proteins have lower rates . The overall false-negative rate ranges from 75% for worm to 90% for fly , which arises from a roughly 50% false-negative rate due to statistical undersampling and a 55% to 85% false-negative rate due to proteins that appear to be systematically lost from the assays . Finally , statistical model selection conclusively rejects the Erdös-Rényi network model in favor of the power law model for yeast and the truncated power law for worm and fly degree distributions . Much as genome sequencing coverage estimates were essential for planning the human genome sequencing project , the coverage estimates developed here will be valuable for guiding future proteomic screens . All software and datasets are available in Datasets S1 and S2 , Figures S1–S5 , and Tables S1−S6 , and are also available from our Web site , http://www . baderzone . org .
Maps of pairwise protein–protein interactions are being generated in increasing numbers by the two-hybrid method [1] . Genome-scale two-hybrid screens have now been conducted for Saccharomyces cerevisiae ( yeast ) [2 , 3] , Caenorhabditis elegans ( worm ) [4] , and Drosophila melanogaster ( fly ) [5] . More recently , screens have been reported for herpesviruses and human [6–8] . These datasets have stimulated large-scale analysis of the topology of protein interaction networks . Limitations in the data , both false positives ( spurious interactions reported from high-throughput screens ) and false negatives ( true interactions missing from the screens ) , continue to make it difficult to infer network properties [9–11] , including distinctions as basic as the difference between Erdös-Rényi ( ER ) , power law [12–14] , and other network degree distributions [15] . A recent review points out the challenges in estimating false-positive rates , false-negative rates , and completion to full coverage of protein interaction networks [16] . Virtually every published method falls back to an estimate based on intersections of datasets . For false-positive rates , these methods have large variance when assays have little overlap , and indeed could not be used to analyze the existing large-scale maps for worm and fly . Estimates for false-negative rates based on overlap of datasets may have even larger uncertainty . Finally , global estimates of false-positive and false-negative rates say little about protein-specific properties , including whether certain classes of proteins behave well or badly in two-hybrid screens . The goal of this work is to develop and apply a statistical model for two-hybrid pairwise interaction screens . Previous methods typically summarize the presence or absence of an interaction as a 1/0 binary variable , and possibly split off a high-confidence core dataset . The method we describe reaches back to the raw counts of observed bait–prey clones . This frees the statistical method from the need for an external gold standard of true-positive and true-negative interactions , or even a second dataset . It permits protein-specific predictions that for the first time permit tests of hypotheses that some classes of proteins are more or less likely to have nonspecific interactions . Finally , estimates of false-negative rates permit statistically grounded confidence intervals for the total number of pairwise interactions present in model organism proteomes . A flowchart of a two-hybrid screen orients the discussion by showing where true-positive interaction partners can be lost and where false-positive , spurious interactions may arise ( Figure 1 ) . In a two-hybrid assay , one protein is fused to the binding domain ( bait construct ) of a yeast transcription factor , and a second protein is fused to the activation domain ( prey construct ) . Physical interactions between bait and prey proteins reconstitute transcription factor activity . Due to the expense of the assay , not every protein may be selected to be made into a bait or prey construct . Furthermore , some constructs may not be functional at all due to improper folding or incompatibility with the two-hybrid system . These missing interactions are important to consider when estimating the total number of interactions in a proteome . High-throughput two-hybrid screens have used multiplexed pairwise tests , either by testing a single bait versus a pool of preys [4 , 5] , or by pooling both baits and preys [3] . Unnormalized prey pools can be generated from mRNA extracted from growing cells . With access to clone collections , pools can be normalized by designing baits and preys individually for each protein or protein domain , then mixing preys in equal proportion . The yeast screen considered here [3] tested 62 normalized bait pools versus 62 normalized prey pools , each pool having approximately 96 genes . The fly screen and worm screen each tested one bait in turn versus both normalized and unnormalized pools . The testing occurs by using mating or transformation to express both the bait and prey construct in a single yeast cell . True-positive interactions drive reporter genes that permit the yeast cell to grow in selective media . Yeast cells whose bait–prey constructs do not interact are expected to drop out during the population expansion . True positives may also be lost during the population expansion for at least two reasons . First , the mating or transformation may lack enough cells to ensure that every combination is tested . Second , a particular construct may have domain-specific misfolding , making it functional for some interactions but nonfunctional for others . True interactions that are not represented in the cells following the population expansion are systematic false negatives for a particular screen . False negatives due to insufficient mating/transformation and due to nonfunctional domains could in principle be discriminated by repeating the mating or transformation step and the selective population expansion . Without this additional step , however , losses during the population expansion combine to yield a systematic false-negative rate termed 1 − psyst , with psyst representing the true-positive rate for an interacting pair to survive the population expansion . Some cells expressing noninteracting proteins may also survive the population expansion , and the final population of cells will be a mixture of true positives and false positives . In Figure 1 , the mass fraction of true-positive cells is 1 − α , and of false-positive cells is α . The ratio of false positives to the total number of true negatives is the false-positive rate . Usually , however , the ratio is with respect to the total number of observed interactions ( Equation 31 ) , defined as the false-discovery rate and synonymous with the parameter α . An ongoing point of contention in two-hybrid screens is the possibility that two proteins that never interact in vivo in the host organism might have a strong , reproducible interaction in vitro in the engineered two-hybrid system . Conversely , proteins with a strong two-hybrid interaction might nevertheless fail to interact in vivo . For the purposes of this work , we assume that such cases are rare and we classify any pair of proteins with a reproducible two-hybrid interaction as a true positive . While the total false-positive fraction α may be large , it represents a sum over many different false-positive pairs . Most models , including ours , assume that any particular false positive is rare , with vanishing probability of observing a specific false-positive interaction more than once . Interactions detected in pooled screens often require sequencing to identify the interacting partners , although advanced pooling designs may improve deconvolution efficiency [17] . Cost constraints limit the number of interactions that can be sampled for sequencing . If the number of clones selected for sequencing is smaller than the number of true interaction partners of a bait , some true partners will certainly be lost . Limited sampling depth also truncates the observed degree distribution for baits . The false-negative rate due to undersampling is termed 1 − psamp in Figure 1 . False-discovery rates have typically been estimated by comparing datasets [18–20] , suggesting up to 50% false positives , but these analyses can confound false-positive and false-negative error sources . Estimated error rates have large uncertainty because few interactions are observed in multiple datasets . For example , comparing the Uetz and Ito two-hybrid datasets for yeast reveals only 9 . 1% of the total interactions in common [3] , and comparing the two-hybrid interactions with mass spectrometry interactions reveals only 0 . 6% in common [20] . Similarly , comparison of two fly screens reveals few interactions in common [5 , 21] . Cross-species comparisons have also revealed little overlap in the reported interactions [4] , although protein and network evolution are additional confounding factors . Efforts to estimate the true number of interaction partners of a protein have used contingency tables for observing an interaction in multiple screens . These methods require that all the interactions be true positives , for example by excluding singleton observations [22] , which can reduce the estimated interaction count . A notable exception is previous work in the context of mass spectrometry of protein complexes [23] , which used a Bayesian model to infer global parameters for screen-specific false-positive and false-negative rates . These parameters then provided posterior estimates for the probability of a true interaction given results of one or more screens . This work is important in using the number of trials and successes , rather than a single summary yes/no observation , in its probability model; it serves as motivation for developing similar models for the more complicated two-hybrid sampling process involving strong protein-specific effects . Quantitative predictions of the amount of work required to identify some fraction of true interactions would be analogous to formulas for genome sequencing [24] and would be useful for planning new experiments [25] . The new work presented here uses the raw screening data to estimate the false-negative rate from undersampling , together with the false-positive rate . A schematic illustrates the sampling process ( Figure 2 ) . Interactions are sampled with replacement from two sets , one representing true positives and the other true negatives . The observations are the number of times that each interaction is sampled , which we summarize with three variables: n , the total number of samples drawn; w , the number of unique interactions within the n samples; and s , the number of interactions observed exactly once . From these observations we are to estimate the unknown ( hidden ) values of k , the total number of true interaction partners , and f , the number of false positives within the sample n . We also estimate the parameter α representing the fraction of false positives in the mixture ( the false-discovery rate ) , as well as parameters representing the probability distribution for k . For simplicity , the illustration suggests sampling interactions in the entire network; in reality , this sampling process occurs separately for each bait , and the estimation of k and f is performed separately for each bait . This estimation problem is akin to estimating population sizes or species counts from capture–recapture experiments , estimating vocabulary size from word counts , estimating the number of distinct alleles at a particular locus , and estimating the number of facts in the scientific literature [26–33] . Classic capture–recapture theory permits heterogeneous capturability rates , here analogous to different probabilities of observing each true interaction partner of a bait . The canonical estimator has a simple form: k̂ = w + s2/2k2 [34–36] , where k2 is the number of partners observed exactly twice . The classic estimator fails in the two-hybrid setting because it does not account for false positives . To our knowledge , false positives have never been discussed in the capture–recapture setting . False positives will vastly inflate the interaction count by adding to the number of singleton observations , s , and to the total observed count , w . The standard estimator has high variance when the number of observations is small , yielding a small value for the denominator k2 . The estimator fails to converge when each partner is observed only once , yielding n = w = s , k2 = 0 , and k̂ → ∞ . We present a front-to-back statistical model for both false-positive and false-negative error rates in two-hybrid screens . A glossary of model terms is provided ( Table 1 ) . The overall approach is to start by estimating the parameters of a mixture model for true positives and false positives following the population expansion . This permits us to estimate bait-specific false-discovery rates and false-negative rates due to undersampling . We can then back-calculate the false-negative rate due to systematic effects . Putting the results together yields an overall estimate for the false-negative rate of a screen and a basis for comparing interaction lists produced by different efforts . Along the way we examine issues that our model is able to address quantitatively: selecting the best model for the protein degree distribution; correlating false-discovery rates with bait properties such as “sticky” or “promiscuous” domains or hydrophobic regions; and determining the relative performance of prey libraries generated from cDNA libraries or ORFeome collections .
We applied our methods to experimental data from two-hybrid screens conducted in the model organisms yeast , worm , and fly . The key parts of the datasets are the numbers of times that a specific bait identifies each prey , from which all other required values may be calculated . Yeast data was taken from ITO FULL , with clone counts from the IST HIT column [3] . Worm data was from WI5 with clone counts in the NumHitADcDNA and NumHitADORF columns [4] . The worm interactions were from the CORE_1 , CORE_2 , and NON_CORE sets; interactions annotated as SCAFFOLD ( previous screens by the same group ) , LITERATURE ( interactions reported in the scientific literature ) , and INTEROLOG ( interactions inferred cross-species ) were excluded . Fly data was from the CuraGen screen with clone counts in the baitprey and preybait columns [5] . A summary of the data sources is provided ( Table 2 ) , and a compendium of the data sources is available ( Dataset S1 ) . In collecting these datasets , we noted that many two-hybrid screening publications do not report the clone counts that are required for capture–recapture analysis . This includes one of the two major yeast high-throughput screens [2] , a screen for Helicobacter pylori interactions [37] , and important recent screens for human protein–protein interactions [6 , 7] . Part of the motivation of this work is to demonstrate the value of making this type of raw data available for analysis . The relevant variables describing a two-hybrid screen are listed in Table 1 and summarized here . Each of N baits is screened against a prey library . For bait i , ni clones from a two-hybrid screen are sampled and the preys are identified . The number of times that prey j occurs within bait i's sample is termed nij . The number of unique preys within the ni clones is termed wi . The number of preys observed exactly once ( singletons ) is si . The ni clones comprise a mixture of false positives and true positives , but it is not known a priori which are the false positives , or even the total number fi of false positives . The goal of our analysis is to estimate the number of false positives , fi , and the number of true positives that were left unsampled for each bait . Our statistical model makes the following assumptions . Prey constructs are either functional ( with probability psyst ) or systematically lost ( with probability 1 − psyst ) with respect to a particular true interaction partner bait construct . Due to possible differences in binding sites , a prey may be functional for one bait and nonfunctional for a different bait . The total number of true positives for a particular bait i is termed κi , of which ki ≡ psystκi are functional . The parameter psyst is estimated from the observed probabilities of bidirectional interactions . Prey libraries are normalized , with each prey present at equal concentration . True-positive interaction partners are sampled with equal probability with replacement from the ki functional preys . False-positive preys occur stochastically , not systematically , with a low probability per prey and negligible probability that any single true negative is sampled twice for a given prey . Thus , clones observed once are a mixture of false positives and true positives; clones observed two or more times are assumed to be true positives . The cumulative probability that a particular clone is a false positive may be large because it sums over all the possible true negatives . This cumulative false-positive rate is the false-discovery rate per clone , termed αi for bait i , and may be different from bait to bait . These assumptions are justified in the Materials and Methods section . Even if restrictive , they still provide a necessary starting point for building more complicated models . Given these assumptions , we show in Materials and Methods how false-discovery rates and corrected counts of interaction partners can be determined for each bait . The posterior estimates for false-discovery rates and interaction counts depend on the functional forms selected for the bait-to-bait heterogeneity in the false-positive rate and the protein interaction degree distribution . We used a variety of model selection criteria , also described in Materials and Methods , that had perfect performance on simulated data . While false positives are a recognized byproduct of two-hybrid screens , there has been little work to investigate bait-to-bait variation in the false-discovery rate . We investigated three models for bait-specific false-discovery rates , described in words here and mathematically in Materials and Methods , Equation 2 . The false-positive rate in the model is expressed per sampled clone , rather than per prey in the library ( which would be a much smaller error rate ) or per unique interaction ( which would be a larger error rate ) . We selected three representative functional forms as possible models for the probability that a bait protein has k functional interaction partners in the prey library , described in text here and more formally in the section Theory . The summary results , Table 3 , extrapolate the number of interaction partners from the estimated number of true positives within the preys screened to the total number in the proteome . The results suggest about 40 pairwise interaction partners per protein in yeast , and roughly 100 pairwise interaction partners per protein in worm and fly . These numbers , however , are based on the estimated means . For long-tailed degree distributions , the median values may provide greater intuition , and may in fact be more robust by discounting outliers with high interaction counts . Median numbers of interaction partners obtained from parametric degree distributions ( see the section Total Interaction Counts ) , are provided at the bottom of Table 3 . The final values obtained are roughly ten partners per yeast protein , 61 for worm , and 46 for fly . The 1 . 5-fold difference between worm and fly might point to built-in biases in the screens ( different baits and preys , different selection thresholds , etc . ) rather than any fundamental biological differences . Using the median and mean estimates as brackets , our results suggest between 30 , 000 and 140 , 000 pairwise interactions in yeast; 600 , 000 to 1 , 200 , 000 in worm; and 300 , 000 to 600 , 000 in fly . Other work , using a contingency table approach similar to k∩ , has suggested a 95% confidence interval of about 40 , 000 to 75 , 000 interactions in yeast , and 150 , 000 to 370 , 000 in human [16] . This previous work was unable to make predictions for worm or fly , however , due to the lack of multiple datasets for comparison .
The methods introduced here provide a new model for false-positive and false-negative rates for two-hybrid screens . To our knowledge , this is the first model that considers the number of observations of each prey , as opposed to a binary interaction / no-interaction summary statistic , to calculate these rates . We have validated the model thoroughly using simulated data and using published biological datasets . The applications to published data demonstrate the crucial ability to predict how many new interactions will be observed as more preys are collected , together with the true-positive and false-positive fractions . One of the major criticisms of the two-hybrid method has been a high false-positive rate . Unlike previous methods that produce average false-discovery rates over an entire screen , our method provides bait-by-bait estimates . False-discovery rates are heterogeneous: some baits perform better than others . As others have suggested [38] , this permits the possibility of correlating false-positive rates with hydrophobicity and related protein properties . We find strong evidence for higher false-discovery rates for membrane proteins , but not for hydrophobic proteins in general . Two-hybrid screens such as the split-ubiquitin system [55] have been developed to detect interactions between membrane proteins . These assays could very well show a correlation of false-positive rates with other classes of proteins . Classification of proteins according to enzymatic function reveals that those in signaling pathways have lower false-discovery rates than those in metabolic pathways . This suggests greater evolutionary pressure to maintain specificity of information-carrying networks . One suggested mechanism of network evolution is that recent paralogs may continue to share interaction partners . This would imply that proteins within a single family should show cross-reactivity with each other's binding partners , eventually leading to false positives due to weak remnants of ancestral interactions . We rejected this hypothesis by finding no significant correlation between false-discovery rate and family size . Analysis of false-positive rates also provides a quantitative estimate of the value of using constructs from a sequence-verified ORF collection rather than from cDNA libraries . When we classify bait constructs as “good” or “bad , ” we find that the “good” category is 90% for an ORF collection and 45% for a cDNA library . On the prey side , using an ORF library produces one-third fewer false positives than a cDNA library . This model yields estimates of false-negative rates from screening statistics , and to our knowledge is the first attempt to discriminate between false negatives due to undersampling and false negatives due to biological and systematic effects . We find that sampling and systematic factors are both important contributors to false negatives , with undersampling yielding a roughly 2× reduction in interactions , and systematic effects yielding an additional 2× to 6× reduction . False-negative rates estimated from the statistical model are in general agreement with those estimated from comparisons between datasets or to a gold standard . The statistical framework provides a convenient route to assessing the likelihood of different population-level functional forms for the protein degree distribution and the false-discovery rate . We provide conclusive evidence that , among one-parameter degree distribution models , a PL model is far superior to an ER model . We find evidence for exponential truncation of the degree distribution in worm and fly , but not in yeast . The number of interactions per protein is predicted to increase from about ten partners for yeast to about 50 partners for the metazoans worm and fly . These results suggest that more complex organisms have more interactions per protein component , as well as more components overall . This model will have value in application to ongoing pool-based assays for protein–protein interactions in model organisms and human . An immediate demonstration is the ability to predict the total number of pairwise protein–protein interactions based on two-hybrid data . We suggest that the total number of pairwise interactions observable by the two-hybrid system is roughly 140 , 000 in yeast , and 600 , 000 to 1 , 300 , 000 in worm and fly , with about 95% remaining to be discovered . An attractive extension of the model presented here is to include unequal capture probabilities for true interaction partners . The current model represents true-positive preys as a two-component mixture: a fraction 1 − psyst of true-positive preys are considered absent from the pool , with capture probability 0; the remaining k true-positive preys have uniform capture probability . It would be possible to include more components , or even a continuous variable representing an inhomogeneous capture probability of a prey . This is important for libraries generated directly from mRNAs with varying abundances , and could still be important for libraries generated from normalized clone collections due to varying effective nuclear concentrations and binding constants . Including heterogeneous capture rates for baits could be accomplished by extending the model to represent the true-positive rate psyst for each bait as an additional hidden variable to be optimized within the Expectation–Maximization ( EM ) framework . In this work , psyst is a global parameter calculated after the sampling-based parameters have been estimated . Both of the above extensions would involve a probability model that considers interactions in both directions , bait–prey and prey–bait , and would necessarily add complication to what is already a mathematically detailed model . While a more complicated model would seem unlikely to lead to different conclusions from those presented here , it could answer questions relating capture probabilities to protein physical properties , protein abundances in libraries , and transient versus stable protein interactions . Developing related statistical models for other types of protein interaction screens will also be important . A constant proviso attached to interaction screens is the suspicion that methods such as two-hybrid screens , affinity pull-downs [56 , 57] , and protein binding chips [58] will identify different subsets of interactions . Quantitative comparisons are difficult , however , because systematic assay-specific differences are confounded with random loss of interactions due to incomplete sampling . Methods such as the one presented here will contribute to understanding what different screening technologies tell us about the proteome .
An overview of notation is provided ( Table 1 ) . Consider a particular protein j used as one of N baits in a two-hybrid screen against a pool of Γ species of preys of which κj are true interaction partners . We assume that κj = Γ so that Γ − κj ≈ Γ is a good approximation for the number of true negatives for each bait . We model the first stage of a two-hybrid screen as an all-or-none process reflecting whether a bait mates successfully with a prey and yields progeny that survive selection . For simplicity , and to reduce the number of free parameters , we assume an identical systematic true-positive rate psyst for each bait j with each of its κj true interaction partners . The parameter psyst includes systematic biological effects , such as generating functional fusion proteins in the two-hybrid system . The number of surviving true positives is kj with binomial distribution We further assume that the Γ true negatives continue to grow slowly in the selective media with a population expansion that is only p* times the population expansion of surviving true positives . We assume a stochastic , not systematic , model for false positives , with p* = 1 and identical for each prey . Also for simplicity , and in accord with prey libraries constructed from normalized ORF collections , we assume that each prey is present initially at equal concentrations . The final mass fraction of false positives is denoted αj = p*Γ[kj + p*Γ] , yielding a scaled error model that depends on a single constant p*Γ ≡ a . More generally , a variety of error models are possible: The MIXTURE model introduces an index zj ∈ {1 , 2… , m} to one of m possible values of α and prior probabilities π ( 1 ) + π ( 2 ) + …+ π ( m ) = 1 for the m components . With m = 2 , this permits “good” baits ( zj = 1 ) and “bad” baits ( zj = 2 ) with α ( 1 ) ≤ α ( 2 ) . The second stage of screening bait j is to sequence nj clones from the mixture of true positives and false positives . We assume that each of the kj true positives is sampled with uniform probability ( 1 − αj ) /kj and each of the Γ false positives is sampled stochastically with uniform probability αj/Γ . The number of times that prey species m , either a true positive or a false positive , is sampled within the nj clones is njm , with 0 ≤ njm ≤ nj and ∑mnjm = nj . The probability of the observed counts njm is a multinomial , [nj ! /Πmnjm ! ]Πm , with θm = ( 1 − αj ) /kj or αj/Γ . As is typical in a capture–recapture setting , it is more convenient to work in the context of abundance classes . Let for i ≥ 1 represent the observed data as the number of preys observed exactly i times within the nj samples . For convenience , we introduce sj as a synonym for , the number of singleton preys observed only once . The total number of distinct preys observed is wj , where δ ( arg ) is 1 if its argument is true and 0 if false . The standard generalized multinomial distribution is obtained by summing over the {S} permutations that yield wj distinct species , identical to Equation 3 of [34] . A rough motivation for this formula is that nj/Πi≥1 is the number of distinct permutations of the nj clones , |S|/Πi≥1 is the number of distinct permutations of the observed species , and Πm is the probability of selecting the species in specified order . Our final approximation is that each true negative occurs at most once as a false positive , njm = 0 or 1 when m is a true negative . The expected number of false-positive clones within the nj clones is αnj . The probability that each of these , selected at random from the Γ total possibilities , is distinct is [1 − ( m − 1 ) /Γ] , or approximately exp[− ( m − 1 ) /Γ] = exp[−αjnj ( αjnj − 1 ) 2Γ] , analogous to the Birthday Paradox ( the probability that two people in a large random group share a birthday ) . An appropriate constraint ensuring distinct false positives is that nj ≤ /αj . With genome-size prey libraries , Γ ≥ 5 , 000 , and we anticipate that αj ≤ 0 . 5 , making this approximation valid for nj ≤ 140 . For yeast , ten baits ( 0 . 67% ) violate this constraint; for worm , 18 baits ( 2 . 5% ) ; for fly , ten baits ( 0 . 27% ) . Denote fj as the number of false-positive observations within the sample nj . By the above assumption , the false positives must be within the sj singletons , and 0 ≤ fj ≤ sj . Using the uniform capture probabilities , Πm = . The number of permutations |S| can be calculated under the above assumption of singleton false positives as The first factor is the number of ways that false positives can be assigned to a subset of fj of the sj singleton species . We have used Γ ! / ( Γ − fj ) ! ≈ , which is valid because Γ ≫ nj ≥ fj . The second factor is the number of permutations that select the wj − fj observed true positives out of kj . Combining results yields with an additional factor of π ( zj ) depending on the hidden variable zj that indicates the component for the MIXTURE error model , Equation 2 . The probability distribution for the hidden variables fj and kj are obtained through the Bayesian relation For the MIXTURE model , the analogous equation includes the hidden variable zj . When kj is independent of nj and αj , Pr ( kj | nj , αj ) = Pr ( kj ) ≡ Pr ( kj | Φ ) , where Φ comprises one or more global parameters describing the interaction degree distribution . The simplified Bayesian result is or Pr ( yj | xj , Q ) where the hidden variables yj = { kj , fj} , and possibly zj; the observed variables xj are the counts of singletons ( sj ) , distinct preys ( wj ) , and total samples ( nj ) ; and the parameters Q are the global parameters for the error model ( a , α , or {α ( 1 ) … α ( m ) ;π ( 1 ) …π ( m ) } ) and the protein degree distribution . The three summary statistics {sj , wj , nj} are sufficient statistics for the observed data due to the assumption of homogeneous probabilities for observing each true-positive and true-negative species . The sum over k formally starts at wj − f , which may equal 0 when each of the nj observations is a singleton . In the results , however , we restrict attention to probability distributions for which Pr ( k = 0 | Φ ) = 0 and start the summation at k = 1 . Three distributions are considered: The normalization of the Poisson distribution by 1 − e−λ in Equation 11 reflects that the summation begins at k = 1 rather than at 0 . In keeping with the definition of 1 − psyst as the systematic false-negative rate , it may be more appropriate to use parametric distributions for Pr ( κ ) , than to obtain Pr ( k ) as the convolution Pr ( k ) = ∑κ≥k Pr ( k | κ ) Pr ( κ ) . We are in a sense replacing Pr ( k | κ ) of Equation 1 by a delta function near the mean value κpsyst in order to retain the form of a simpler parametric distribution . Estimates for {kj , f} for each bait could in principle be obtained using Equations 10–11 . This requires , however , estimates for the global parameters Q . Furthermore , the asymptotic form of the summand in Equation 10 is Pr ( k | Φ ) . Writing the asymptotic form of Pr ( k | Φ ) as k−ɛ , existence of a maximum a posteriori estimator requires nj + ɛ − wj > 0; convergence of the sum requires nj + ɛ − wj > 1; and convergence of the mean of kj requires nj + ɛ − wj > 2 . For the ER prior , λ > 0 guarantees convergence of all powers of kj; for the TPL prior , c > 0 guarantees convergence . Convergence could be achieved by normalization of k in Equation 11 to a finite cutoff Γ rather than to ∞ . In practice , however , results for the PL model are sensitive to the cutoff value when ɛ < 2 . The TPL and ER models are not sensitive to a cutoff , as both provide a natural cutoff as part of the model parameters . If a cutoff is appropriate , we anticipate that these models will provide improved descriptions of a degree distribution . To overcome both these difficulties , we use EM to obtain parameter estimates Q̂ that maximize the probability of the observed data [59 , 60] , assuming a uniform prior Pr ( Q ) . We introduce the notation for the mean of a generic function F ( x , y ) of the hidden and observed variables . The sum over the hidden variables expands to for the scaled α and single α error models , and to for the m-component mixture . As mentioned before , while some models permit kj = 0 , the power law model does not; for consistency , we start all degree distributions at kj = 1 and the lower limit of the sum over kj is effectively max ( 1 , wj − fj ) . The standard equations giving a new parameter estimate Q in terms of a previous estimate Q′ are where the simplification holds because Pr ( xj , yj | Q ) = Pr ( xj | yj , Q ) Pr ( yj | Q ) and Pr ( xj | yj , Q ) = Pr ( xj | yj ) is independent of Q . Update equations for the error models are as follows: Update equations for the degree distribution are as follows: An interesting and unfortunately common boundary case occurs when only a single clone is sequenced for a bait , nj = 1 . In these cases , sj and wj must also be 1 , and Pr ( xj | Q ) = 1 regardless of Q . Thus , baits with n = 1 do not affect the final model parameters because the partial derivatives of their contributions to the log-likelihood are always 0 . The appearance of the expectation of log k rather than k in the EM equations for the power law parameter ɛ in the PL and TPL models suggests the use of the posterior mean of log k as a route to estimating the hidden variable decay . We define this estimator as k̂ , The three error models and the three degree distribution models yield a total space of nine possible models , with varying degrees of freedom ( df ) : 1 df for the scaled and single error models; 2m − 1 df for the m-component mixture error model; 1 df for the ER and PL degree distributions; 2 df for the TPL distribution . We used three separate criteria to assess which model provides the best fit: log-likelihood cross-validation ( CV ) ; full data Bayesian information criterion ( BIC ) ; and bootstrap BIC . The CV method with F-fold cross-validation divides the full data into F subsets . For subset f , model parameters Qf are estimated using the remaining F − 1 subsets , and the log-likelihood of subset f is calculated as loglikf = log Pr ( {xf}|Qf ) . This procedure is repeated F times , once for each subset . Thus , each subset is used F − 1 times to obtain model parameters and 1 time to obtain an unbiased log-likelihood . The final log-likelihoods , ∑f loglikf , can be compared directly . The statistical significance of a difference in log-likelihoods for two models can be assessed using a paired test , such as the nonparametric Wilcoxon rank signed test , for the differences − for pairs of models M and M′ . The BIC is an appropriate heuristic for performing model selection in the context of maximum likelihood parameter estimation for Q and a uniform prior over model classes M: where d is the number of df in the model and N is the number of sets of observations , here baits . A smaller BIC indicates a more likely class of models , and the term d log N penalizes more complex models . Overfitting is unlikely for our models: the typical number of sets of observations N ∼ 1 , 000 , while the models have only two to four free parameters . Sometimes , the BIC heuristic may indicate a small preference for one model over another . Bootstrap replicates may be used to assess the stability of the BIC results . Bootstrap replicates are constructed by selecting N examples from the full data of N examples uniformly and with replacement . Thus the number of times n that an example occurs in a bootstrap replicate is approximately Poisson with Pr ( n ) = 1/ ( n ! e ) . The BIC heuristic for each model is then calculated for each bootstrap replicate , and the number of times that each model has the best BIC score is recorded . We calculated the cumulative number of clones sampled for a domain , ndom , and the cumulative posterior estimate for the number of false positives , f̂dom , by summing over the counts for each protein annotated as having that domain: p-Values for the upper and lower tail , p> and p< , were calculated assuming a binomial distribution with ndom trials and a success rate equal to the overall false-discovery rate α̂ for each organism ( 0 . 093 for yeast , 0 . 122 for worm , 0 . 157 for fly ) . To ensure a conservative test , fractional values of f̂dom were rounded down for the upper-tail test and rounded up for the lower-tail test , Finally , the single-value p-values were adjusted for the number of domains observed among baits in each species ( 783 for yeast , 473 for worm , 1 , 310 for fly ) . When two domains refer to an identical subset of proteins , results for only a single domain are displayed . Parameter estimates . Simulations were performed separately for each of the three protein degree distributions ( ER , PL , TPL ) combined with each of the three error models ( scaled , single , mixture ) , yielding nine total models . Simulations over a range of parameter values used N = 1 , 000 baits and n = 10 clones sampled per bait , and were usually repeated three times at each parameter value ( Figures S1–S3 ) . The agreement between known and estimated parameters is very good over parameter values that span the estimated values for the published datasets ( Table S1 ) . Agreement is quantified by the root-mean-square ( RMS ) difference between the known and estimated parameters and the R2 goodness of fit measure , defined as 1 − ∑t ( θt − θ̂t ) 2/∑t ( θt − $theta;̄ ) 2 with sum t over trials , θt the true parameter value for trial t , $theta;̄ the mean of θ over the trials , and $theta;̄t the estimated value for trial t . The RMS values are generally less than 0 . 05 in absolute units . The R2 values for the TPL-mixture range from 0 . 65 for the false-discovery rate parameter to 0 . 98 for the power law parameter . The R2 values depicted for the TPL-mixture model are not typical but rather the worst results obtained over all combinations of degree distribution and error model ( Figure S1–S3 ) . Other models with fewer parameters are more accurately estimated . For example , simulations using the PL-mixture model yield an R2 of 0 . 99 for ɛ , 0 . 97 for π ( 1 ) , and 0 . 88 for α ( 1 ) and α ( 2 ) . Simulations using the PL-single model and PL-scaled model yield R2 ≥ 0 . 99 for all parameters ( Table S1 ) . Hidden variable estimates . Simulations also assessed the ability to predict hidden variables ( Figure S4 ) . We simulated a dataset using parameters obtained by fitting the fly experimental data with a TPL degree distribution and a two-component error model . The simulated dataset had N = 1 , 000 baits and n = 10 baits per prey . Next , the EM algorithm was used to estimate the model parameters ɛ̂ , α̂ ( 1 ) , α̂ ( 2 ) , and π̂ ( 1 ) . The estimated parameter values , rather than the true known values that generated the data , were used to avoid introducing a favorable bias in the hidden variable predictions . The converged model parameters were then used to obtain the posterior estimates 〈log k〉 , 〈f〉 , and 〈δ ( z = 1 ) 〉 for each bait , with the notation 〈…〉 defined by Equation 13 ( Figure S4 ) . Using the log-transform of k is more natural than using k due to the long tail of the power law distribution . It is also motivated by the EM equations for the power law exponent ɛ , which depends on 〈log k〉 rather than 〈k〉 as shown in Equation 18 . The hidden variable 〈log k〉 is estimated with R2 = 0 . 84 , indicating good correlation between true and estimated values . The RMS error in the estimate is 0 . 42 , indicating the ability to predict the true value of k within a factor of exp ( ±0 . 42 ) , or 1 . 5-fold . The RMS for the number of false positives is 1 . 1 , which means that the estimate for the number of false positives for a bait is usually within 1 of the true count . Estimates of the error component of a bait are accurate for low values ( low error rate ) and high values ( high error rate ) of ẑ . Baits with intermediate estimated values , 1 . 2 ≤ ẑ ≤ 1 . 6 , may come from either error component . For these baits , all but one or two of the preys are singletons , and it is difficult to determine whether this is due to a large bait degree or a large error rate . Model selection for simulated data . We next validated the BIC heuristic for model selection ( Table S2 ) . In this test , we used each of the nine possible models ( three degree distributions × three error models ) to generate datasets with N = 1 , 000 baits and n = 10 preys per bait , then calculated the log-likelihood for each of the nine models . A total of 81 fits were performed ( nine generative models × nine fitting models ) . The parameters of the generative models were deliberately selected to yield similar data by using the values obtained by fitting the experimental worm data ( Table 2 ) . In each case , the BIC identified the model accurately ( Table S2 ) . The probability to obtain a perfect result by chance is approximately ( 1/9 ) 9 , or 2 . 6 × 10−9 . The BIC results indicate the TPL model can provide a good fit for data generated by all three models: ER , PL , and TPL . The TPL model includes the PL model as a special case with the exponential decay constant c = 0 . A large value of c , which truncates the degree distribution , permits the TPL model to mimic the ER model . The BIC score adds a penalty of log 1 , 000 = 6 . 91 to the TPL fit to account for the extra parameter . In several of the entries of Table S2 , this penalty is essential to select the true generative model over the TPL model . Properties of the experimental datasets for yeast , worm , and fly are summarized at the top of Table 2: total number of baits , N; number of preys sampled per bait , n; number of unique observed interaction partners per bait , w; and number of interaction partners observed a single time , s . Each dataset was fit using three separate degree distribution models ( ER , PL , TPL ) and three error models ( SCALED , SINGLE , MIXTURE ) , a total of nine possible combinations of degree distribution and error model . Model selection . The BIC heuristic selects the PL-MIXTURE model for yeast and the TPL-MIXTURE model for worm and fly ( Table 2 ) . In general , the PL and TPL models are much better than the ER models . The MIXTURE error model is somewhat better than the SINGLE error model , which in turn is much superior to the SCALED error model . To explore the robustness of this conclusion , we used 10-fold cross-validation ( CV ) to compute the log-likelihood of the data under each of the models . The CV method identifies the TPL-MIXTURE model as the best for worm and fly , and finds no significant difference between PL-MIXTURE and TPL-MIXTURE for yeast ( p-value = 0 . 35 ) . Finally , we generated 100 bootstrap replicates each of the yeast , worm , and fly datasets , calculated the BIC scores for each of the nine models , and tabulated the number of times that each model had the best score . The PL-MIXTURE model won 94/100 times for yeast , and the TPL-MIXTURE won 98/100 times for worm and 100/100 times for fly . Model parameters . The PL models yield power law parameters ɛ that are robust to the choice of error model: ɛ = 1 . 67–1 . 72 for yeast , 1 . 48–1 . 53 for worm , and 1 . 50–1 . 58 for fly . In contrast , we can mimic a conventional fit by estimating the PL ɛ using Equation 18 but with k ( the corrected number of interaction partners ) replaced with w ( the observed number of unique partners , which may include false positives ) . The conventional fit introduces two sources of error: it inflates bait degrees by including false positives , and it deflates bait degrees by excluding false negatives . The conventional estimates for yeast , worm , and fly yield exponents of 2 . 22 , 1 . 66 , and 1 . 61 , which are larger ( have steeper decay of the degree distribution ) than the model results . Thus , the error due to false negatives may dominate the error due to false positives when PL parameters are estimated . Parameter estimates based on prey degree rather than bait degree might be less sensitive to these sources of error . Work by others connects the inverse of c to the typical domain size in a network [15] . The best-fitting TPL models for worm and fly have c ≈ 0 . 04 to 0 . 07 , suggesting a domain size of ten to 30 proteins in a subnetwork . These estimates may not be overly precise as the TPL parameters are sensitive to the error model . The worm data , for example , yields ( ɛ , c ) = ( 1 . 25 , 0 . 012 ) for scaled , ( 0 . 46 . 0 . 035 ) for single , and ( 0 . 95 , 0 . 040 ) for mixture error models . The extra variability arises because a larger value of ɛ can compensate for a smaller value of c . The yeast network , which is best fit by a PL-MIXTURE model , yields a very small value of c when fit by any TPL model . This suggests that the yeast network may show less modular structure than either the worm or fly networks . Interaction networks from viruses , parasites , and simpler organisms have been shown to be less clustered than interaction networks from more complex organisms [52] . The fraction of false positives is estimated consistently by the SCALED , SINGLE , and MIXTURE error models regardless of the choice of degree distribution . The false-positive fraction is calculated as ∑if̂i/∑ini from the estimated false-positive count f̂i and number of preys ni for each bait i . This fraction is estimated as 0 . 08–0 . 09 for yeast , 0 . 12 for worm , and 0 . 16 for fly . The false-positive fraction is , of course , larger when defined relative to the number of unique interactions identified rather than the number of preys . Although the generative model , the model parameters , and the hidden variables are unknown for the experimental yeast , worm , and fly data , cross-validation is still possible . The cross-validation used half of the dataset to predict the number of new and single interactions in the remainder of the dataset . First , for each bait , we extracted a random half of the preys to serve as a training set . For baits with an odd number n of preys , we selected ( n − 1 ) / 2 and ( n + 1 ) / 2 for the training set with equal probability . Next , we used the training half of the data to estimate model parameters for each organism . For simplicity , we restricted attention to TPL–MIXTURE model . With the maximum likelihood parameter estimates , we then obtained posterior estimates for the false-positive rate and true interaction count of each bait protein . The false-positive rate was determined directly from the mixture model . Based on experience with simulated data , we used exp ( 〈log k〉 ) for the posterior estimate of the bait degree . As noted before , the EM equations for PL-like networks , Equation 18 , suggest that the logarithm of the degree is a more natural variable than the degree itself ( which is not even guaranteed to converge ) . Finally , we use the statistical model to predict how many unique interaction partners and singleton interaction partners are expected to be observed as the remaining test half of the data is added back . The predictions of the model based on the training half can then be compared with the actual results for the number of unique partners , w , and the number of singleton partners , s . Since the posterior estimate of the bait degree may be fractional , we replaced the factorial function with the Gamma function . Also , rather than starting with the values for s and w for the training set and calculating the marginal increase , we performed a more demanding comparison by using the model parameters to estimate s and w for the observed training set as well . Fitting error was observed only for the count of yeast singletons . Our model can distinguish between random or stochastic false negatives due to undersampling , which could be detected by sampling additional clones , and systematic false negatives that cannot be corrected by deeper sampling of clones from a two-hybrid screen . As described in the section Theory , the overall true-positive rate is the product of the random and systematic rates , psamp × psyst . The true-positive rate from sampling for bait i is ( wi − fi ) / ki , the number of true positives observed out of the ki true positives represented in the pool . The values of fi and ki are hidden , however , and psamp must be estimated . An appropriate estimator , weighting each true positive equally , is using Equation 19 to define k̂i . The false-negative rate due to systematic losses may be estimated from interactions observed in both directions . We first restrict attention to proteins that were found to have at least one interaction as a bait and as a prey . Using the notation that nij indicates the number of times that bait i recovers prey j , we extract the subset with nij ≥ 2 , ensuring that the interaction between i and j is a true positive . These N2 cases are then subdivided into N0 not observed in the reverse orientation , with j as bait and i as prey , and N+ observed at least once in the reverse orientation: The indicator function δ ( arg ) is 1 for a true argument and 0 for a false argument . The expected value of δ ( nji ≥ 1 ) is the true-positive rate for bait j , equal to the product of the true-positive rates accounting for systematic losses and for random undersampling , Similarly , the expectation of δ ( nji ≥ 0 ) is 1 − psyst ( wj − f̂j ) /k̂j . The estimated true-positive rate from just the sampling step for the interactions contributing to N2 , denoted , is In practice , we find that is somewhat larger than the value of ( w − f̂ ) /k̂ averaged over all baits . For yeast , worm , and fly , the values for are ( 0 . 81 , 0 . 84 , 0 . 81 ) , and the values for averaged over all baits are ( 0 . 58 , 0 . 60 , 0 . 71 ) . Note that , the average of the ratios , is distinct from p̂samp , the ratio of the averages of w − f̂ and k̂ ( Table 10 ) . An estimator for N+ in terms of the unknown psyst and other quantities that are known is An obvious route to estimating psyst is to assume that N+ follows a binomial distribution for N2 trials with success rate . The corresponding maximum likelihood estimate p̂syst and its standard error are The estimated number of true interaction partners corrected for sampling losses and systematic losses is then κ̂i , with An alternative estimate for the false-negative rate of a high-throughput screen may be obtained by comparison to literature data . To accomplish this , we downloaded protein interaction datasets from the Database of Interacting Proteins ( DIP ) [54] . As is common in studies such as this , we required a “gold-standard” dataset that did not use information from the two-hybrid screens we are studying and also was unlikely to be contaminated by false positives . We therefore filtered the entire DIP database to include interactions from only small-scale experiments . Because there is no firm definition of small scale , we used cutoffs of 10 and 100 for the number of interactions reported . For each unique gold-standard interaction , denote the pair of proteins ( i , j ) . We defined bi = 1/0 if protein i was / was not used as a bait in the high-throughput screen , and pi = 1/0 if protein i was / was not used as a prey . We defined similar terms for bj and pj . Also define I ( i , j ) as 1/0 if the high-throughput screen detected or missed the interaction between i as bait and j as prey . We then calculated the following values: These summary statistics consider the gold-standard interaction in both orientations . For gold-standard interactions between identical proteins , only one of the two identical terms was included in the sum . The bait fraction screened and the prey fraction screened indicate possible correlation in the choice of baits and preys in the small-scale experiments and the high-throughput screens . The summary statistic for the false-negative rate should correct for this bias by only considering interactions that were within the space considered by the high-throughput experiment . This false-negative rate is interpreted as including both the undersampling loss and the systematic loss . The mean number of true interaction partners per bait , corrected for false positives and for undersampling , is Using the definition of the false-negative rate from undersampling , psamp from Equation 25 , the mean may also be written as Correcting the mean for systematic losses , psyst from Equation 30 , yields This value requires a final correction for the actual search space relative to the entire genome size . The correction we use is the number of preys with at least one interaction , Nprey , relative to the number of proteins annotated in the entire genome , Nproteome . The final proteome-wide mean interaction count is ( Nproteome/Nprey ) 〈κ̂〉 . The number of pairwise interactions in the entire proteome is then An alternative method for estimating interaction counts is to use the inferred probability distribution directly: Pr ( k | λ ) for Poisson; Pr ( k | ɛ ) for PL; and Pr ( k | ɛ , c ) for TPL , Equation 11 . For brevity , denote each of these Pr ( k | θ̂ ) for the appropriate parameter estimate θ̂ . An overall value for k denoted k̂ may be obtained using an analog of the bait-specific estimate Equation 19 , A more typical value for the interaction count corrected for undersampling may be obtained using the median defined by the parametric distribution . We use linear interpolation to estimate a non-integer median , where k+ is the smallest value of k such that CumPr ( k ) ≥ 0 . 5 and CumPr ( k ) is the cumulative probability Pr ( k′ | θ̂ ) The median interaction count corrected for systematic losses , κ̂med , is then and the median number of interaction partners per protein , corrected for the space screened , is ( Nproteome/Nprey ) κ̂med .
|
The genome sequence of an organism provides a parts list of proteins , but not an instruction manual for assembling the parts into a cell . Assembly instructions now come from experiments such as two-hybrid screens that detect physical interactions between pairs of proteins . Defining the resources required for generating a full interaction map requires accurate estimates of the false-negative and false-positive rates of genome-scale screens . Two-hybrid screens often select a query protein and sample its interaction partners . True partners may be missed , and false partners may be spuriously identified . This sampling process resembles a capture–recapture experiment , except that classical capture–recapture theory assumes no false positives . Novel extensions to capture–recapture theory permit its application to proteomic screens . This new theory provides statistically grounded answers to long-standing questions: false-discovery rates of high-throughput screens ( possibly over 50% per unique interaction , but probably no more than 15% per clone ) ; the quality of different screening libraries; protein properties leading to “sticky” or “promiscuous” interactions; the global network topology; and , most importantly , the coverage of existing two-hybrid maps . Models estimate roughly 30 , 000 total pairwise interactions in yeast and 500 , 000 to 1 , 000 , 000 in metazoans . The majority of these interactions remain to be discovered .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"mathematics",
"caenorhabditis",
"computational",
"biology",
"drosophila",
"saccharomyces"
] |
2007
|
Where Have All the Interactions Gone? Estimating the Coverage of Two-Hybrid Protein Interaction Maps
|
Data-driven models of functional magnetic resonance imaging ( fMRI ) activity can elucidate dependencies that involve the combination of multiple brain regions . Activity in some regions during resting-state fMRI can be predicted with high accuracy from the activities of other regions . However , it remains unclear in which regions activity depends on unique integration of multiple predictor regions . To address this question , sparse ( parsimonious ) models could serve to better determine key interregional dependencies by reducing false positives . We used resting-state fMRI data from 46 subjects , and for each region of interest ( ROI ) per subject we performed whole-brain recursive feature elimination ( RFE ) to select the minimal set of ROIs that best predicted activity in the modeled ROI . We quantified the dependence of activity on multiple predictor ROIs , by measuring the gain in prediction accuracy of models that incorporated multiple predictor ROIs compared to models that used a single predictor ROI . We identified regions that showed considerable evidence of multiregional integration and determined the key regions that contributed to their observed activity . Our models reveal fronto-parietal integration networks , little integration in primary sensory regions , as well as redundancy between some regions . Our study demonstrates the utility of whole-brain RFE to generate data-driven models with minimal sets of ROIs that predict activity with high accuracy . By determining the extent to which activity in each ROI depended on integration of signals from multiple ROIs , we find cortical integration networks during resting-state activity .
Data-driven models constrained with functional magnetic resonance imaging ( fMRI ) data can elucidate some of the dependencies that involve a combination of multiple brain regions underlying the observed activity . Previous studies show that the activity in some brain regions during resting-state fMRI can be predicted with high accuracy using the activities in other regions [1] . The activity in each region of interest ( ROI ) can be modeled as a weighted sum of the activities of predictor ROIs ( “features” ) , and the weights are found by optimization methods . Weights estimate the contribution of activity in predictor ROIs to activity in the modeled ROI , and can give clues about the connectivity between the regions ( either indirect or direct ) with respect to signal integration . While previous models of ROI resting-state activity achieved high prediction accuracy [1] , it remains unclear in which ROIs resting-state activity depends on multiregional integration . Some degree of prediction accuracy can be achieved using the predictor ROI most correlated with the modeled ROI . Therefore , it is important to identify ROIs whose activity exhibits unique integration of multiple varied signals ( in contrast to activity that is redundant with other ROIs and/or integrates fewer varied signals ) . Estimating dependencies between ROIs is challenging due to the low temporal resolution and correlations between regions that can increase false positives . Previous studies used support vector regression to find ROI dependencies , in a manner that was inclusive in terms of predictors [1] . Finding a minimal set of ROIs that predict activity in a modeled ROI with high accuracy could serve to reduce false positives and thus better estimate the key ROIs involved and their contribution to the observed activity in the given ROI . To address this question , methods of feature selection , which generate sparser models and relate prediction accuracy to the number of predictors in the model , are particularly useful . One commonly-used method is recursive feature elimination ( RFE ) , a backward feature selection method that generates models iteratively , eliminating one or more predictors in each iteration–commonly the predictor with smallest weight , which least influences the model [2 , 3] . By examining the effect of elimination on validation error , RFE can be used to exclude redundant and spurious predictors , to arrive at a model with the minimal set of predictors that yields the most accurate prediction . RFE was originally used to classify cancer genes [3] , and more recently was used successfully to classify brain states and conditions ( e . g . , resting vs . task or disease vs . control ) from fMRI activity in predictor ROIs [4–9] . However , the method has not been used thus far to fit and predict activity in modeled ROIs from the activity in predictor ROIs . Another method of feature selection , “least absolute shrinkage and selection operator” , or Lasso , uses L1 norm regularization of the regression . Regularization constrains properties of the predictor weights , and generally yields models with better prediction accuracy than ordinary least-squares regression [1 , 10 , 11] . L1 norm regularization minimizes the magnitude of the weights , and therefore yields sparse models with fewer redundant and spurious predictors [12 , 13] . As such , Lasso has been successfully applied to fMRI data for feature selection , disease prediction and classification [14–18] , though it has not been used thus far to fit and predict activity in modeled ROIs from the activity in predictor ROIs . Whereas RFE examines directly the relationship between model prediction accuracy and the number of predictors in the model , Lasso relates prediction accuracy to the magnitude of predictor weights . While feature selection methods such as RFE and Lasso are useful in addressing the challenging issue of spurious predictors ( reduce false positives ) , other methods such as elastic net use additional L2 norm regularization to reduce missing true predictors ( reduce false negatives ) , although at some cost to the reduction of false positives [11] . The choice of which modeling method to use depends on the question at hand . In this study we were interested in finding the predictors on which activity strongly depended ( reduce false positives ) , and accordingly chose RFE and Lasso as the primary modeling methods , and used elastic net models to corroborate our results . It is noteworthy that while the covariance matrix or partial correlations provide information about potential dependencies between pairs of ROIs [19 , 20] , models derived from the activity data by ROI selection infer the number and combination of multiple ROIs that are jointly necessary for explaining/predicting the activity in a given ROI . This information cannot be derived just from looking at the correlations . In this study , we first compared different methods for modeling ROI activity , in terms of prediction accuracy and feature selection . We used resting-state fMRI activity data from 46 subjects , and for each ROI per subject we generated models using whole-brain RFE , Lasso or elastic net to select the minimal set of ROIs that best predicted activity in the ROI . We then determined the degree to which ROI activity depended on multiple predictor ROIs , by comparing prediction accuracy of models incorporating multiple predictor ROIs to models that used a single predictor ROI . We identified ROIs that showed evidence of multiregional integration , and determined their key predictor ROIs , thereby identifying integration networks during resting state .
We used resting-state fMRI activity data of 46 subjects ( ~20 minutes in length per subject ) , and divided each brain volume to 128 ROIs ( Fig 1A , see Methods ) . We first compared different methods for feature selection and modeling of ROIs ( for each ROI per subject ) , in terms of model prediction error and the number of predictor ROIs selected . Two of the methods used RFE for feature selection , the third used Lasso , and the fourth used elastic net ( see Methods ) . In the RFE approach ( Fig 1B and 1C ) , the validation error of models generally decreased as predictor ROIs were removed , reaching a minimum beyond which elimination of predictors generally increased the validation error . One method to generate the final model for the ROI ( which we termed RFE2 ) , was to use the model that had the minimal validation error during the elimination process ( Fig 1C , the minimum of the curve ) . However , due to fluctuations in the validation error curve , even after the minimum there were decreases in validation error that followed some of the increases , indicating that not all the predictor ROIs of the model at the minimum were necessary . Therefore , in the primary RFE method that we used , the final model for the ROI was generated using only predictor ROIs whose elimination increased the smallest validation error across the rest of the iterations that followed ( Fig 1C , magenta dots ) . After generating models for ROIs using the different methods mentioned above , we assessed their prediction error using the test data segment , which was not used in any stage of model generation . The average prediction error across 108/112 cortical ROIs ( excluding temporal pole and entorhinal cortex that had particularly large errors in all models , see below ) was similar for the different modeling methods ( 24 . 8 ± 6 . 4% of the activity variance for ROIs modeled using RFE , 24 . 4 ± 6 . 5% for RFE2 , 21 . 4 ± 5 . 4% for Lasso , and 21 . 2 ± 5 . 4% for elastic net , Fig 2A ) . Prediction error of null models ( see Methods ) was significantly larger ( 111 . 5 ± 17 . 4% , chance level = 100% , p < 10−79 ) . For all ROIs , Lasso and elastic net models had comparable prediction error ( difference < 1% ) , which was also slightly smaller than the prediction error of RFE models . However , for most ( 95/108 ) ROIs the difference between RFE and Lasso/elastic net ranged between 0 . 1–5 . 3% , and was therefore negligible in terms of effect size . The 13 ROIs with slightly larger difference ( 5–8% ) had large prediction error in Lasso/elastic net models ( 25–36% ) , and were therefore not well-modeled by any of the methods . Thus , performance of models generated by the different methods was similar . In contrast to the similarity in prediction error between the different modeling methods , RFE selected considerably fewer predictor ROIs ( 9 ± 1 ) compared to RFE2 ( 22 ± 4 , p < 10−64 ) , Lasso ( 49 ± 5 , p < 10−104 ) or elastic net ( 59 ± 7 , p < 10−101 , Fig 2B ) . Thus , RFE yielded models with similar prediction accuracy as the other methods , and had fewer redundant predictor ROIs . In light of our aim at reducing false positives to better elucidate key multiregional dependencies , we conducted the subsequent analyses using the parsimonious models derived with RFE . We also note that RFE had shorter runtime compared to Lasso and elastic net . Fitting models for each subject took ~10 minutes with RFE vs . ~1 hour with Lasso/elastic net ( and for the entire dataset the difference in runtime was ~8 hours vs . ~50 hours , respectively ) . Models generated using simple regression had larger prediction error than models generated using the other methods ( 30 . 1 ± 8 . 2% ) , a significant difference both in terms of effect size and statistics ( overall average difference of 6–9% , p < 10−39 ) . Furthermore , simple regression models for all ROIs included the entire set of predictor ROIs and thus provided no selection of predictors . There was no correlation between prediction error of ROIs and their connectivity in terms of number of predictors ( r = -0 . 03 , p = 0 . 75 ) . An example of model ROI with high prediction accuracy is shown in Fig 3A . Among models of the 112 cortical ROIs , small prediction errors ( 13–23% ) were observed in 49 ROIs–in occipital cortex , superior and inferior parietal cortex , precuneus and cuneus cortex , isthmus and anterior cingulate , lingual gyrus , superior and rostral/caudal middle frontal gyrus ( Fig 3B ) . Larger prediction errors ( 30–40% ) were seen in models of anterior ROIs in temporal lobe ( superior temporal gyrus1 , middle temporal gyrus1 , inferior temporal gyrus1 ) , banks of superior temporal sulcus , and caudal/rostral anterior cingulate cortex . Otherwise , particularly large prediction errors ( > 40% ) were observed only for 4 ROIs ( in each hemisphere ) –temporal pole , parahippocampal gyrus , entorhinal cortex and transverse temporal cortex . Among subcortical ROIs , models of thalamus and cerebellum had small prediction errors ( 16–22% ) , whereas models of the pallidum , amygdala , and accumbens had large errors ( > 40% ) . Prediction error was weakly inversely-correlated with the temporal signal to noise ratio ( tSNR , see Methods ) of the ROI ( r = -0 . 33 , p < 0 . 001 , n = 128 ROIs ) . Importantly , the ROIs with particularly large prediction errors , such as temporal pole and entorhinal cortex , also had a small tSNR ( 0 . 92 and 0 . 71 , respectively , where the tSNR across ROIs ranged from 0 . 71 to 1 . 99 ) . In contrast , the anterior temporal and cingulate ROIs with prediction error between 30–40% mentioned above had tSNR of average magnitude ( ~1 . 3 ) , except for the inferior temporal gyrus1 , suggesting that resting-state activity in these ROIs depended on other regions in a more variable manner compared to other ROIs . To examine the dependence of activity on multiple predictor ROIs , we compared modeled ROIs in terms of a novel measure that we termed “multiregional prediction gain”–the improvement in prediction accuracy of a model that used multiple predictor ROIs compared to a model generated using the single predictor ROI most correlated with the modeled ROI ( see Methods ) . For example , the model for activity in right inferior parietal cortex2 of an example subject shown in Fig 4A had prediction error of 19% using 5 predictor ROIs selected by RFE . An alternative model for the same ROI , generated using only the single predictor ROI most correlated with the modeled ROI in the training data , had a larger prediction error of 68% ( Fig 4B ) . The multiregional prediction gain for this ROI was therefore 49% ( p < 10−15 , determined by t-test between the square errors of the two models ) , and thus there was evidence in the data of considerable dependence of ROI activity on the activities in multiple regions . Among the 49 cortical ROIs that were well-modeled ( average prediction error 13–23% ) , 20 ROIs , primarily parietal and frontal , had large prediction gain ( 13–23% , Fig 4C ) . These integrator ROIs included right superior frontal gyrus1-4 , left superior frontal gyrus2 , left/right rostral-middle frontal gyrus2 , 3 , left/right caudal-middle frontal gyrus , left inferior parietal cortex2 , right inferior parietal cortex1 , 2 , right superior parietal cortex2 , right supramarginal gyrus2 , 3 , right precentral gyrus2 , and right/left lateral orbital cortex . The large multiregional prediction gain in these ROIs reflected a significant difference between the prediction errors of models that had multiple predictor ROIs and the prediction errors of models that used only a single predictor ROI ( p < 10−6 ) , as determined by t-test corrected for multiple comparisons ( see Methods ) . ROIs with moderate multiregional prediction gain ( 8%– 13% ) included left/right precuneus cortex2 , left/right fusiform gyrus2 , left cuneus cortex , and left/right lateral occipital cortex1-3 . ROIs with small multiregional prediction gain ( < 8% ) , and thus no dependence of activity on unique multiregional integration , included pericalcarine cortex , isthmus cingulate cortex , precuneus cortex1 , and lingual gyrus . For these ROIs , the activity was strongly correlated with a single predictor ROI , indicating a strong coupling between the two ROIs , which masked evidence of dependence of the activity on other ROIs . For example , the small gain of left isthmus cingulate cortex was due to strong coupling with right isthmus cingulate cortex in 85% of the cases ( with average activity correlation of 0 . 84 between the two ROIs ) , left precuneus cortex1 in 10% of the cases ( activity correlation = 0 . 8 ) , and left precuneus cortex2 in the rest of the cases ( activity correlation = 0 . 77 ) . For these ROIs , the data therefore indicated redundancy in terms of signal integration during resting state . We note that whereas ROIs whose activity is highly correlated with other ROIs would have small prediction gain , ROIs that are well-modeled and whose activity depends on multiregional integration ( large prediction gain ) cannot be found just by looking at the activity correlation matrix . This is because ROIs that are only moderately correlated with other ROIs could differ in terms of how well their activity can be predicted from multiple regions , and therefore differ in their multiregional prediction gain . For example , the maximal activity correlation of right inferior parietal cortex2 with other ROIs was 0 . 78 and its average activity correlation was 0 . 42 , its average prediction error was 17% , and its average prediction gain was 23 . 5% . In contrast , whereas the maximal and average activity correlations of left supramarginal gyrus3 with other ROIs were 0 . 78 and 0 . 42 as well , its average prediction error was higher ( 27% ) and its prediction gain was lower ( 12% ) . Thus , the level of activity correlation did not in itself indicate the extent of multiregional integration or prediction error . We also note that the magnitude of multiregional prediction gain need not necessarily be correlated with the number of predictor ROIs . For example , a multiregional prediction gain of 20% could theoretically depend on 2 predictor ROIs just as it could depend on 20 predictor ROIs . In the first case , each predictor ROI would be responsible on average for a larger portion of the prediction gain compared to the latter case ( 10% vs . 1% , respectively ) . Thus , a difference in prediction gain between ROIs could either be related or not to the number of predictor ROIs . We therefore examined the relationship between the multiregional prediction gain and the number of predictor ROIs in the models , and found that ROIs with larger multiregional prediction gain tended to include a larger number of predictor ROIs ( r = 0 . 78 , p < 10−26 ) . Therefore , the different gain between ROIs corresponded to a difference in the number of predictor ROIs on which the activity depended , rather than a difference in the average contribution of predictor ROIs to the prediction gain . Multiregional prediction gain for an ROI was not correlated with the tSNR of its time series ( r = 0 . 16 , p > 0 . 05 , n = 128 ROIs ) . We found no correlation between multiregional prediction gain and subject age in any of the ROIs ( p > 0 . 05 ) . For each of the 20 integrator ROIs ( with large multiregional prediction gain ) mentioned above , we looked for key predictor ROIs on which the activity depended most ( see Methods ) . In the analysis of predictor ROI weights , the weights were normalized as % of the total weight onto the modeled ROI ( see Methods ) . For example , in models of right inferior parietal cortex2 , key predictor ROIs included both neighbouring ROIs ( right inferior parietal cortex1 , 3 , right precuneus cortex1 , 2 , right middle temporal gyrus3 ) and more distal ROIs ( > 50 mm apart , right caudal-middle frontal gyrus , right superior frontal gyrus1 ) , as well as the homotopic left inferior parietal cortex2 ( Fig 5A ) . The set of integrator ROIs and their key predictor ROIs formed interconnected integration networks ( Fig 5B ) that involved primarily parietal and frontal ROIs , with long-distance ( > 50 mm ) connections between inferior parietal and caudal-middle/superior frontal ROIs . Key predictor ROIs were primarily ipsilateral to the integrator ROIs , and included also the homotopic ROI , but did not include subcortical ROIs . The weights of predictor ROIs were moderately correlated with the activity correlations between the predictors and modeled ROI ( r = 0 . 76 ± 0 . 05 , p > 10−53 , determined by shuffling the activity correlations ) , further supporting the meaningfulness of the estimated weight . We did not find any correlation between subject age and predictor weights ( p > 0 . 05 for all ROIs ) . To corroborate our results with modeling methods that do not prune predictors as heavily as RFE or even Lasso , we compared predictor ROI weights as determined by RFE to weights determined by elastic net . Predictor ROI weights were strongly correlated between the two types of models , with the difference mainly in that RFE pruned predictors that had small weights in models derived by elastic net ( Fig 6A ) . For the 20 integrator ROIs highlighted above , the correlations between model weights derived using the two methods were 0 . 77 ± 0 . 03 on average , and ranged from 0 . 69 ± 0 . 13 to 0 . 79 ± 0 . 11 ( Fig 6B ) . Thus , key predictor ROIs were identified to a similar degree by the two methods . Prediction error of models for cortical ROIs ( excluding the 4 ROIs in each hemisphere with large prediction error mentioned above ) was only weakly correlated with ROI size , either for single subjects ( r = -0 . 19 , Fig 7A ) or across subjects ( r = -0 . 2 , Fig 7B ) . Some correlation is to be expected by the de-noising effect of averaging more voxels in larger ROIs compared to smaller ROIs . While the estimated head motion was small ( 0 . 12 ± 0 . 07 mm , see Methods ) but correlated with subject age ( r = 0 . 71 , p < 10−7 ) , as is commonly the case [21] , the average prediction error was not correlated with subject head motion ( r = -0 . 09 , p = 0 . 56 , Fig 7C ) or subject age ( r = 0 . 1 , p = 0 . 49 , Fig 7D ) .
We used whole-brain recursive feature elimination to select minimal sets of ROIs that best predicted the resting-state fMRI activity in ROIs of each subject . We used these data-driven models to highlight ROIs whose activity depended considerably on multiple predictor ROIs , reflected in a substantial gain in prediction accuracy of models that used multiple predictor ROIs compared to single predictor ROI . We determined the key predictor ROIs that contributed to the activity in each integrator ROI , and showed that the integrator and predictor regions formed interconnected integration networks during resting state . Multiregional prediction gain measures both the variety of signals that a region integrates and the uniqueness of this integration compared to other regions . Thus , a region that is well-modeled and whose activity integrates different streams of information more uniquely compared to other regions ( less correlated with other regions ) will have a larger multiregional prediction gain . In contrast , a region that integrates similar streams of information or whose activity is coupled ( strongly correlated ) to the activity of other regions will have a small multiregional prediction gain , so far as is evident in the data of activity during the particular brain state tested . Importantly , whereas high activity correlations are associated with small multiregional prediction gain , data-driven models are necessary to find ROIs with large multiregional prediction gain , and to infer the number and combination of multiple predictor regions on which ROI activity depends . ROIs whose resting-state activity depended on multiple predictor ROIs were seen within superior frontal gyrus , rostral-middle frontal gyrus , caudal middle frontal gyrus , inferior parietal cortex , and lateral orbital cortex in both hemispheres , as well as right superior parietal cortex , right supramarginal gyrus , and right precentral gyrus . The large dependence on multiregional integration that we found in the frontal/parietal regions agrees with their involvement in higher cognitive functions such as reflection during resting state [22] , and supports previous indications based on correlations that these regions are functional hubs during resting state [23 , 24] . The high prediction accuracy in these regions agrees also with previous modeling studies [1] and with their high activity during resting-state [25] . The integrator ROIs and their key predictor ROIs formed interconnected resting-state integration networks , consisting mostly of ROIs in the parietal and frontal lobes , in line with previous findings [23] . Our models indicate long-distance dependencies connecting the frontal and parietal networks between parietal cortex and middle/superior frontal gyrus . The activity in the integrator ROIs did not depend significantly on subcortical ROIs , indicating that during resting-state they integrated primarily cortical activity . For some ROIs , the models predicted activity with high accuracy but did not depend on multiple predictor ROIs , due to strong and stable correlation ( coupling ) between the modeled ROI and a single predictor ROI . The size of ROIs was sufficiently large ( e . g . 55 ± 14 voxels for pericalcarine or 74 ± 21 for isthmus cingulate ) to be spatially well-separated to reduce the effect of spatial correlations between nearby voxels . Thus , the small multiregional prediction gain in these regions indicates a degree of redundancy among some regions during resting state in terms of informative activity ( as in the case of isthmus cingulate cortex ) , and/or that the regions do not integrate numerous and varied cortical signals ( as in the case of primary sensory regions such as pericalcarine cortex ) . While it is not possible to distinguish between these two possibilities solely based on multiregional prediction gain , the small gain can be informative along with other pieces of information . E . g . , as isthmus cingulate cortex was previously shown to be a hub during resting state [26] , the small multiregional prediction gain in our study indicates that its signal integration during resting state is redundant with that of homotopic and precuneus ROIs . We have shown that RFE models predict resting-state activity with similar accuracy as Lasso or elastic net models , but with considerably fewer predictor ROIs ( on average less than 10 ) . RFE is therefore a useful whole-brain method for generating parsimonious models that predict activity with high accuracy , and for discovering key predictor ROIs in high-dimensional brain activity data . Future studies that focus on minimizing false negatives rather than false positives can use our framework of analyzing multiregional prediction gain and predictor ROIs with other types of models such as elastic net . Our method is somewhat related to other multivariate methods such as partial correlations [19 , 20] and partial least squares analysis [27] , with the advantage that our modeling method grounds the estimated interregional dependencies with their joint prediction of ROI activity . Future studies should systematically compare our method to other common multivariate methods of fMRI analysis , for the purpose of properly ascertaining the manner in which these methods complement one another . In addition , data-driven RFE models of ROI activity could complement connectivity analysis of pairwise correlations , by pruning ROIs that are correlated with the modeled ROI but are not necessary for predicting its activity and therefore are likely to be spurious . The high prediction accuracy of RFE models indicates stability of integration of signals during resting state , which agrees with the previously reported stability in activity correlations between structurally connected regions [28] . High prediction accuracy was associated with errors of ~15%– 25% of the activity variance , indicating that 75%– 85% of the activity in these regions was driven by stable interactions . The prediction error could result from measurement noise as well as from activity driven by variable interactions on top of a stable core . Given that prediction error was only weakly inversely-correlated with tSNR of the ROI time series , it is likely that the error was more influenced by the variability in interactions between regions . Thus , for ROIs such as anterior superior/middle temporal gyrus , and rostral/caudal anterior cingulate cortex , where the error was larger ( 30–40% ) and tSNR was average , it is likely that the resting-state activity depended on other ROIs more variably . Considerably large prediction errors ( 50–60% ) were seen only in temporal pole , parahippocampal gyrus , entorhinal cortex and transverse temporal cortex , which are prone to susceptibility artifacts [29] and had small tSNR in our study . We did not observe significant relationships between subject age and prediction accuracy or multiregional prediction gain , indicating that the portion of stable interactions between regions was similar across a wide age span . While care must be taken in interpreting predictor weights [30] , in our analysis of predictor ROI weights we have used normalized weights ( % of total weight onto the modeled ROI ) rather than raw coefficients , to allow a meaningful comparison of weights within and between models . Furthermore , there are several indications that the predictor weights we estimated signify their importance in influencing the modeled ROI activity . First , the models predicted new data with high accuracy , indicating that the estimated weights capture true dependencies underlying the observed activity . Second , the weights of predictor ROIs were strongly correlated with the activity correlations between the predictor ROIs and modeled ROI , further supporting their meaningfulness . Similarly to previous modeling studies [1] , we show that despite the low temporal resolution of fMRI and the correlations between regions , training data segments that are not too long ( ~300 points , or ~10 minutes of scanning ) can be used to generate models with high prediction accuracy for most regions . Nevertheless , future models will benefit from utilizing longer datasets ( e . g . , the Human Connectome Project [31] ) , which we expect will better constrain the models and improve determining the key dependencies between regions . Such datasets could also be used to corroborate the findings of this study . Other validations could be made by comparing the robustness of models and results using different parcellations [32 , 33] , or testing the methods using ground-truth artificial data generated by brain simulation tools [34] . Future studies should utilize our data-driven modeling and analysis method to data acquired during other brain states ( e . g . during task ) , to determine the dependence of activity on multiple predictor ROIs and the key regions involved . The interconnected sets of integrator ROIs and their predictor ROIs could also be analyzed in terms of network theoretic measures [35–37] , relationship with structural connectivity and activity correlations [23 , 38–41] , and temporal dynamics [42–44] .
We used resting-state fMRI blood-oxygen-level dependent ( BOLD ) data from 46 healthy subjects ( 18 males , 28 females ) of ages between 18 and 77 ( 41 ± 18 years , median = 35 ) , acquired at Berlin Center for Advanced Neuroimaging , Charité University Medicine Berlin [34 , 45] . Subjects were recruited as volunteers and gave written informed consent before the study , which was approved by the local ethics committee in accordance with the institutional guidelines at Charité Hospital , Berlin , and performed in compliance with the Code of Ethics of the World Medical Association ( Declaration of Helsinki ) , the relevant laws and institutional guidelines . Briefly , subjects were lying relaxed in a supine position inside the MR scanner . They were instructed to relax , lie still , and minimize movement ( especially head movement ) . The subject’s head was immobilized with removable cushions ( Siemens equipment ) , and earplugs were used to prevent adverse effects due to scanner noise . Subjects were instructed to close their eyes but not to fall asleep . Functional MRI ( BOLD-sensitive , T2*-weighted echo planar imaging , TR 1940 ms , TE 30 ms , FA 78° , 32 transversal slices ( 3 mm ) , voxel size 3 × 3 × 3 mm , FoV 192 mm , 64 matrix ) was recorded using a 3 Tesla Siemens Tim Trio MR scanner simultaneously with 64 channels EEG [46] . Each scan spanned ~20 min ( 661 TR ) . Subjects also underwent anatomical T1-weighted scans ( magnetization-prepared rapid gradient-echo , TR 1900 ms , TE 2 . 25 ms , 192 sagittal slices ( 1 . 0 mm ) , voxel size 1 × 1 × 1 mm , FA 9° , FoV 256 mm , 256 matrix ) as well as T2-weighted scans ( 2D turbo spin echo , TR 2640 ms , TE1 11 ms , TE2 89 ms , 48 slices ( 3 . 0 mm ) , voxel size 0 . 9 × 0 . 9 × 3 mm , refocusing FA 150° , FoV 220 mm , 256 matrix ) . Preprocessing was performed using FEAT ( fMRI Expert Analysis Tool ) Version 6 . 0 from the FMRIB ( Functional MRI of the Brain ) Software Library ( FSL FEAT , 2014 ) . It included correction for head motion using MCFLIRT [47] , fieldmap correction to reduce spatial distortion of EPI images [48] , BET brain extraction to remove non-brain tissue [49] , and high-pass filtering ( cutoff at 100s ) to adjust for baseline drift of the signal . Functional data was registered to the individual high-resolution anatomical T1 images using FreeSurfer . The BOLD signal was not smoothed nor corrected for slice-timing differences , in line with suggestions by previous studies [50] . We divided the brain to 128 gray matter ROIs ( 64 in each hemisphere , Fig 1A ) , 112 cortical and 16 subcortical ( cerebellum , thalamus , caudate , putamen , pallidum , hippocampus , amygdala , and accumbens ) . The parcellation was conducted in two stages: the brain was first divided to 84 ROIs ( 68 cortical and 16 subcortical ) according to anatomical features detected in the anatomical MRI scan , using an automated method described previously [51] . Next , 26 large cortical ROIs ( 13 in each hemisphere ) were subdivided into several contiguous ROIs of equal size along the region’s main axis: Superior frontal gyrus ( 4 ROIs , anterior-posterior axis ) , inferior parietal cortex ( 3 , superior-inferior ) , lateral occipital cortex ( 3 , posterior-anterior ) , middle temporal gyrus ( 3 , anterior-posterior ) , precentral gyrus ( 3 , superior-inferior ) , rostral-middle frontal gyrus ( 3 , anterior-posterior ) , superior parietal cortex ( 3 , posterior-anterior ) , superior temporal gyrus ( 3 , anterior-posterior ) , supramarginal gyrus ( 3 , posterior-anterior ) , inferior temporal cortex ( 2 , anterior-posterior ) , fusiform gyrus ( 2 , anterior-posterior ) , postcentral gyrus ( 2 , superior-inferior ) , and precuneus cortex ( 2 , posterior-anterior ) . Subdivided ROIs of the larger regions were referred to in the text as regionx , where x is the number of the ROI along the region’s main axis . For example , left superior frontal gyrus2 was the second-most anterior ROI of left superior frontal gyrus . Subdividing the large regions resulted in more similar sizes across ROIs ( 124 ± 40 vs . 211 ± 161 voxels ) . The activity in each ROI was the average over its voxels , converted to percent signal change ( from the average of the time-series for that ROI ) . For each subject we split the time-series into three segments ( Fig 1B ) : training data ( first half ) , validation data ( third quarter ) , and test data ( last quarter ) . The model for each ROI expressed the activity as a weighted sum of the activities in other ROIs ( “features” ) , or: yi=∑j≠iωj , ixj ( 1 ) Where xj is the observed activity in predictor ROIj , and ωj , i is the weight of predictor xj . We used RFE to select the minimal set of ROIs that best predicted the validation data ( Fig 1 ) . The feature selection algorithm involved the following steps: The prediction error ( see below ) of each model was assessed using the validation data segment , and referred to as the validation error . The algorithm yielded a set of models ( each having one less predictor ROI than the previous ) and their corresponding validation errors ( Fig 1C ) . At this point we considered two methods to yield the final model for the ROI . In one method ( which we termed RFE2 ) , the final model was the model with smallest validation error during the elimination process ( the minimum of the validation error curve , Fig 1C ) . Because the validation error fluctuated , with decreases in error following increases , we also considered another method ( which we termed RFE ) to further eliminate redundant predictor ROIs , whereby we selected only the predictor ROIs whose elimination resulted in increase of the smallest validation error across the rest of the iterations that followed , and we used these predictor ROIs to generate a new final model for the ROI , fitting the training data by regression . We also generated for each ROI a third model using Lasso ( in Matlab ) , a different feature selection method , to compare with RFE . Lasso adds a regularization term to the regression , the L1 norm of the feature weights , which minimizes the size of weights and reduces the number of predictors [12] , and by avoiding extreme parameter values reduces overfitting and increases prediction accuracy [1 , 10 , 11] . Lasso minimizes: 〈 ( yi−xi ) 2〉+c|ω| ( 2 ) where yi is the model activity of ROIi ( Eq 1 ) , xi is the observed activity of ROIi , c is the magnitude of regularization , and |ω| is the L1 norm of the weight vector . For each ROI , we generated models using 200 different magnitudes of the regularization and consequently a different number of predictor ROIs in the models . We tested the candidate models on the validation data segment , and selected as the final model the one with smallest validation error . We corroborated our results using a fourth modeling method , elastic net ( in Matlab ) , that prunes predictors less strongly than RFE or Lasso and thus can reduce false negatives [11] . Elastic net uses both L1 and L2 norms of the predictor weights vector to regularize the regression , selecting groups of correlated predictors together instead of discarding redundant predictors . Elastic net therefore has an additional parameter compared to Lasso , namely the relative contribution of L1 vs . L2 norms ( ranging from 0 to 1 ) . To keep the runtime sensible , for each ROI we generated models similarly to Lasso , but searched 50 magnitudes of the regularization for each of 5 magnitudes of the L1 vs . L2 contribution ( 0 . 01 , 0 . 05 , 0 . 1 , 0 . 5 , and 1 ) , or 250 different combinations overall . We measured prediction error in % of the activity variance , the mean-squares error between the observed and predicted time-series , normalized by the total mean squares of the observed time series , or: 〈 ( yi−xi ) 2〉/〈 ( xi−xi¯ ) 2〉 ( 3 ) where xi is the observed activity in the ROIi and yi is the model activity in the ROIi ( Eq 1 ) . Prediction error of candidate models during model generation was determined using the validation data segment and referred to as validation error ( see above ) , whereas prediction error of the final model was determined using the test data segment , which was not used during model generation . The significance of prediction error was assessed using null models , by calculating the prediction error when randomly reordering the weights of predictor ROIs in the model for the ROI ( 500 permutations per ROI ) . For each ROI , we calculated the difference between the prediction error of its multiple-predictors model and the prediction error of a model generated using only the predictor ROI most correlated with the modeled ROI in the training data . We termed the difference as “multiregional prediction gain” , since it measured the improvement in prediction accuracy using the multiple-predictors model . The multiregional prediction gain was therefore a measure of the effect size for the difference in prediction errors of the two types of models , and was reported in this study together with its p-value significance which was determined using t-test on the square errors of the two models . Because effect size is an important factor in determining significance , we focused in this study on highlighting cases where the gain was both large and significant . To establish robust estimates of some measures across subjects ( e . g . , predictor ROI weights ) , we estimated the average across subjects using bootstrapping in Matlab , by sampling 500 times with replacement to estimate the measure’s mean and the 95% confidence intervals . For each modeled ROI , we determined key predictor ROIs that had significantly large weights , as established by t-test ( see below ) vs . weights of null models ( see above ) . In this analysis of predictor ROI weights , to compare predictor weights within and across models , we normalized the predictor coefficients by taking their absolute value and dividing by the sum of absolute weights onto the modeled ROI . The weights were therefore expressed in % of the total weight . Previous studies show correlation between head motion and age [21] , which could have residual effects even after head-motion correction was applied [52 , 53] . Hence , when we examined correlations between subject age and prediction error , multiregional prediction gain , or predictor weight , we controlled for the correlations between head motion and age , by calculating the partial correlation ( in Matlab ) between the measure of interest and subject age while controlling for correlations between age and head motion estimated by the motion correction procedure . We used the estimated mean displacement between consecutive fMRI volumes scanned as an estimate for head motion [52] . All significance tests reported in this study for the difference in measures between models were determined pair-wise , using t-test in Matlab . When we conducted multiple comparisons , we corrected the p value using Bonferroni correction ( multiplying the p value by the number of comparisons ) . Correlation between variables and its significance were determined using Pearson correlation coefficient in Matlab . It is not trivial to estimate SNR in resting-state fMRI , and the noise also differs spatially . For the purpose of this study we used the temporal SNR [54] in each ROI , quantified by the average magnitude of % signal change divided by the variance of the time series , or: tSNR=〈|x|〉/Var ( x ) ( 4 ) Where x is the time series of the ROI ( in % signal change from average ) .
|
Models of fMRI activity can elucidate underlying dependencies that involve the combination of multiple brain regions . However , it remains unclear in which regions activity depends on unique integration of multiple predictor regions . To address this question , sparse ( parsimonious ) models could serve to better determine key interregional dependencies by reducing false positives . We used resting-state fMRI data , and for each brain region we performed whole-brain recursive feature elimination to select the minimal set of regions that best predicted activity in the region . We identified integrator regions by quantifying the gain in prediction accuracy of models that incorporated multiple predictor regions compared to single predictor region . Our study provides data-driven models that use minimal sets of regions to predict activity with high accuracy . By determining the extent to which activity in each region depended on integration of signals from multiple sources , we find cortical integration networks during resting-state activity .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"cingulate",
"cortex",
"medicine",
"and",
"health",
"sciences",
"diagnostic",
"radiology",
"functional",
"magnetic",
"resonance",
"imaging",
"data",
"acquisition",
"engineering",
"and",
"technology",
"signal",
"processing",
"brain",
"neuroscience",
"magnetic",
"resonance",
"imaging",
"brain",
"mapping",
"neuroimaging",
"research",
"and",
"analysis",
"methods",
"computer",
"and",
"information",
"sciences",
"imaging",
"techniques",
"parietal",
"lobe",
"radiology",
"and",
"imaging",
"entorhinal",
"cortex",
"diagnostic",
"medicine",
"anatomy",
"signal",
"to",
"noise",
"ratio",
"biology",
"and",
"life",
"sciences",
"cerebral",
"cortex"
] |
2017
|
Multiregional integration in the brain during resting-state fMRI activity
|
The lion Panthera leo is one of the world's most charismatic carnivores and is one of Africa's key predators . Here , we used a large dataset from 357 lions comprehending 1 . 13 megabases of sequence data and genotypes from 22 microsatellite loci to characterize its recent evolutionary history . Patterns of molecular genetic variation in multiple maternal ( mtDNA ) , paternal ( Y-chromosome ) , and biparental nuclear ( nDNA ) genetic markers were compared with patterns of sequence and subtype variation of the lion feline immunodeficiency virus ( FIVPle ) , a lentivirus analogous to human immunodeficiency virus ( HIV ) . In spite of the ability of lions to disperse long distances , patterns of lion genetic diversity suggest substantial population subdivision ( mtDNA ΦST = 0 . 92; nDNA FST = 0 . 18 ) , and reduced gene flow , which , along with large differences in sero-prevalence of six distinct FIVPle subtypes among lion populations , refute the hypothesis that African lions consist of a single panmictic population . Our results suggest that extant lion populations derive from several Pleistocene refugia in East and Southern Africa ( ∼324 , 000–169 , 000 years ago ) , which expanded during the Late Pleistocene ( ∼100 , 000 years ago ) into Central and North Africa and into Asia . During the Pleistocene/Holocene transition ( ∼14 , 000–7 , 000 years ) , another expansion occurred from southern refugia northwards towards East Africa , causing population interbreeding . In particular , lion and FIVPle variation affirms that the large , well-studied lion population occupying the greater Serengeti Ecosystem is derived from three distinct populations that admixed recently .
Lion fossils trace to the Late Pliocene in Eastern Africa and the Early Pleistocene in Eastern and Southern Africa coincident with the flourishing of grasslands ∼2–1 . 5 million years ago [1] , [2] . By Mid Pleistocene ( ∼500 , 000 years ago ) , lions occupied Europe and by the Late Pleistocene ( ∼130 , 000–10 , 000 years ago ) lions had the greatest intercontinental distribution for a large land mammal ( excluding man ) , ranging from Africa into Eurasia and the Americas [3] . Lions were extirpated from Europe 2 , 000 years ago and within the last 150 years from the Middle East and North Africa . Today , there are less than 50 , 000 free-ranging lions [4] that occur only in sub-Saharan Africa and the Gir Forest , India ( Figure 1A ) . Understanding the broader aspects of lion evolutionary history has been hindered by a lack of comprehensive sampling and appropriately informative genetic markers [5]–[9] , which in species of modern felids requires large , multigenic data sets due to its generally rapid and very recent speciation [10] , [11] . Nevertheless , the unique social ecology of lions [12]–[14] and the fact that lions have experienced well-documented infectious disease outbreaks , including canine distemper virus , feline parvovirus , calicivirus , coronavirus , and lion feline immunodeficiency virus ( FIVPle ) [15]–[18] provide a good opportunity to study lion evolutionary history using both host and virus genetic information . Indeed , population genetics of transmitted pathogens can accurately reflect the demographic history of their hosts [19] , [20] . Unlike other of the 36 cat species , lions have a cooperative social system ( prides of 2–18 adult females and 1–9 males ) and their populations can have high frequencies of FIVPle , a lentivirus analogous to human immunodeficiency virus ( HIV ) , which causes AIDS-like immunodeficiency disease in domestic cats . FIVPle is a retrovirus that integrates into the host genome and is transmitted by cell-to-cell contact , which in felids occurs during mating , fighting and mother-to-offspring interactions . Thus , viral dissemination is a function in part of the frequency of contact between infected and naïve lions within and among populations . The virus is quite genetically diverse in lions [15] , [18] , offering a unique marker for assessing ongoing lion demographic processes . To unravel lion population demographic history we used a large multigenic dataset . Distinct sets of markers may not necessarily yield similar inferences of population history , as coalescent times vary as a function of their pattern of inheritance [21] . There is also a large variance in coalescent times across loci sharing a common pattern of inheritance especially in complex demographical histories ( Table 1 ) . However , the accurate interpretation of the differences among loci can provide a more resolved and coherent population history , affording more-nuanced insights on past demographic processes , levels of admixture , taxonomic issues , and on the most appropriate steps for effective conservation and management of remaining populations . The goal of this study was to assess the evolutionary history of lion by ( 1 ) characterizing lion population structure relative to patterns of FIVPle genetic variation , ( 2 ) detect signatures of migration using both host and viral population genomics , and ( 3 ) reconstruct lion demographic history and discuss its implication for lion conservation . We assess genetic variation from 357 lions from most of its current distribution , including mitochondrial ( mtDNA; 12S–16S , 1 , 882 bp ) , nuclear ( nDNA ) Y-chromosome ( SRY , 1 , 322 bp ) and autosomal ( ADA , 427 bp; TF , 169 bp ) sequences , and 22 microsatellites markers . We further document patterns of FIVPle variation in lions ( FIVPle pol-RT gene , up to 520 bp ) .
Genetic analyses of 357 lions from throughout the extant species range showed that paternally inherited nDNA ( SRY ) and maternal inherited ( mtDNA ) sequence variation was generally low ( only one paternal SRY-haplotype and 12 mtDNA haplotypes; π = 0 . 0066 ) ( Figure 1; Figure S1; Tables S1 and S2 ) . The most common mtDNA haplotype H11 was ubiquitous in Uganda/Tanzania and parts of Botswana/South Africa , H1 was common in Southern Africa , and H7 and H8 were unique to Asian lions . The autosomal nDNA sequences showed fairly distinct patterns of variation ( Figure 1; Figure S1 ) . Of the five ADA haplotypes , A2 was the most common and most-widely distributed . The other four haplotypes , which are derived and much less common , included one ( A5 ) that was fixed in Asian lions . The three TF haplotypes were more widely and evenly distributed . Levels of population subdivision among lions were assessed using microsatellite and sequencing data . Eleven groups were identified using Bayesian analyses [22] and three-dimensional factorial correspondence analyses [23] ( Figure 2; Table S3 ) . Most clusters represented geographically circumscribed populations: Namibia ( Nam ) , Kruger National Park ( Kru ) , Ngorongoro Crater ( Ngc ) , Kenya ( Ken ) , Uganda ( Uga ) , and Gir ( Gir ) . Two distinct clusters were found in Botswana , Bot-I that included lions from southern Botswana and Kalahari ( South Africa ) ( Fk = 0 . 24 ) and Bot-II found exclusively in northern Botswana ( Fk = 0 . 18 ) . Surprisingly , three distinct clusters were found in a single geographical locale ( approximately 60×40 km square ) in the large panmyctic population of the Serengeti National Park ( Ser-I/Ser-II/Ser-III ) ( Fk = 0 . 18 , 0 . 21 , and 0 . 15 , respectively ) . Two captive lions from Angola ( Ang ) , one from Zimbabwe ( Zbw ) and four Morocco Zoo Atlas lions ( Atl; presently extinct from the wild ) ( Figure 1A ) were included in the analyses to explore the relationship of lions from more isolated , endangered , or depleted areas . Ang and Zbw lions were assigned to Bot-II ( q = 0 . 90 and 0 . 87; 90%CI: 0 . 47–1 . 00 ) and Kru ( q = 0 . 85; 90%CI: 0 . 52–1 . 00 ) ( Bayesian analyses [22] ) populations , respectively , as expected based on their geographical proximity . However , these lions differed from Bot-II and Kru by up to 8 mtDNA mutations , sharing haplotypes with the Atl lions ( H5 in Ang and H6 in Zbw ) ( Figure 1B and 1C ) . The Atl lions did not group in a unique cluster . Both nDNA and mtDNA pairwise genetic distances among the 11 lion populations showed a significant relationship with geographic distance ( R2 = 0 . 75; Mantel's test , P = 0 . 0097; and R2 = 0 . 15; Mantel's test , P = 0 . 0369; respectively ) ( Figure 3 ) . The significant positive and monotonic correlation across all the scatterplot pairwise comparisons for the nDNA markers ( bi-parental ) was consistent with isolation-by-distance across the sampled region . However , the correlation between nDNA FST and geographic distance considerably decreased when the Asian Gir population was removed ( R2 = 0 . 19; Mantel's test , P = 0 . 0065 ) suggesting that caution should be taken in interpreting the pattern of isolation-by-distance in lions . We further compared linearized FST estimates [24] plotted both against the geographic distance ( model assuming habitat to be arrayed in an infinite one-dimensional lattice ) and the log geographic distance ( model assuming an infinite two-dimensional lattice ) . The broad distribution of lions might suggest a priori that a two-dimensional isolation-by-distance model would provide the best fit for the nDNA data ( R2 = 0 . 25; Mantel's test , P = 0 . 0022 ) , but instead the one-dimensional isolation-by-distance model performed better ( R2 = 0 . 71; Mantel's test , P = 0 . 0476 ) ( Figure S2 ) . The pattern observed for the mtDNA ( maternal ) was more complex . While there was a significant relationship between mtDNA FST and geographic distance , there was an inconsistent pattern across broader geographic distances ( Figure 3 ) . This is partly due to the fixation or near fixation of haplotype H11 in six populations and the fixation of a very divergent haplotype H4 in Ken population ( Figure 1B and 1C ) . The removal of the Ken population considerably increased the correlation between mtDNA FST and geographic distance ( R2 = 0 . 27; Mantel's test , P = 0 . 0035 ) . Thus , the null hypothesis of regional equilibrium for mtDNA across the entire sampled region is rejected despite the possibility that isolation-by-distance may occur regionally . These contrasting nDNA and mtDNA results may be indicative of differences in dispersal patterns between males and females , which would be consistent with evidence that females are more phylopatric than males . Alternatively , selection for matrilineally transmitted traits upon which neutral mtDNA alleles hitchhike is possible , given the low values of nucleotide diversity of the mtDNA ( π = 0 . 0066 ) . A similar process has been suggested in whales ( π = 0 . 0007 ) [25] and African savannah elephants ( π = 0 . 0200 ) [26] , where both species have female phylopatry and like lions , a matriarchal social structure . However , genetic drift tends to overwhelm selection in small isolated populations , predominantly affecting haploid elements due to its lower effective population size ( Table 1 ) . Therefore , we suggest that the contrasting results obtained for nDNA and mtDNA are more likely further evidence that lion populations underwent severe bottlenecks . The highly structured lion matrilines comprise four monophyletic mtDNA haplo-groups ( Figure 4A; Figure S3 ) . Lineage I consisted of a divergent haplotype H4 from Ken , lineage II was observed in most Southern Africa populations , lineage III was widely distributed from Central and Northern Africa to Asia , and lineage IV occurred in Southern and Eastern Africa . Seroprevalence studies indicate that FIVPle is endemic in eight of the 11 populations but absent from the Asian Gir lions in India and in Namibia and southern Botswana/Kalahari regions ( Nam/Bot-I ) in Southwest Africa ( Figure 1B ) . Phylogenetic analysis of the conserved pol-RT region in 117 FIV-infected lions depicted monophyletic lineages [15] , [18] that affirm six distinct subtypes ( A–F ) that are distributed geographically in three distinct patterns ( Figure 4B; Figure S4 ) . First , multiple subtypes may circulate within the same population as exemplified by subtypes A , B and C all ubiquitous within the three Serengeti lion populations ( Ser-I , Ser-II and Ser-III ) and subtypes A and E within lions of Botswana ( Bot-II ) ( Figure 4B and 4C and Figure S4 ) . Second , unique FIVPle subtypes may be restricted to one location as subtype F in Kenya , subtype E in Botswana ( Bot-II ) , subtype C in Serengeti , and subtype D in Krugar Park ( Figure 4B and 4C and Figure S4 ) . Third , intra-subtype strains cluster geographically , as shown by distinct clades within subtype A that were restricted to lions within Krugar Park , Botswana and Serengeti and within subtype B that corresponded to Uganda , Serengeti and Ngorongoro Crater lions ( Figure 4B and 4C and Figure S4 ) . Not unexpectedly , FIVPle pairwise genetic distances , represented as population FST among the eight lion FIV-infected populations , were not significantly correlated with geographic distance ( R2 = 0 . 08; Mantel's test , P = 0 . 165 ) ( Figure 3 ) , and affirms that patterns of viral dissemination do not conform to a strict isolation-by-distance model . Rather , the two distinct clusters observed ( Figure 3 ) reflect the complex distribution of FIVPle among African lions . Indeed , despite the low geographic distance within East-African lion populations , the FIVPle genetic divergence showed a broader range in FST ( 0 . 03 to 0 . 79 for most of first cluster; Figure 3 ) . By contrast , approximately half of the range in FST ( 0 . 26 to 0 . 69 for the second cluster; Figure 3 ) was observed among East and Southern Africa in spite of its large geographic separation . In contrast with the patterns observed in lions , linearized FST estimates [24] for FIVPle were better correlated with log geographic distance ( two-dimensional lattice model ) ( R2 = 0 . 15 ) than with geographic distance ( one-dimensional model ) ( R2 = 0 . 02 ) , although in both cases the Mantel's test was not significant ( P>0 . 2474 ) ( Figure S2 ) . The mtDNA coalescence dating suggested that the East African lineage I ( Ken haplotype H4 ) had an old origin of ∼324 , 000 years ( 95% CI: 145 , 000–502 , 000 ) . Extant East African populations ( Ken/Ngc/Ser-I/Ser-II/Ser-III ) also showed a slightly significant higher nDNA allelic richness and genetic diversity ( Table S4 ) relative to populations to the south ( Kru/Nam/Bot-I/Bot-II ) and north ( Uga/Gir ) ( A = 2 . 43 , 2 . 39 , and 1 . 62 , P = 0 . 021; HO = 0 . 64 , 0 . 62 , and 0 . 34 , P = 0 . 019; respectively ) . Moreover , the FIVPle subtype diversity was higher in East African clades ( exhibiting four out of the six known viral-strains ) , including the most divergent FIVPle subtype C ( Figure 4B and 4C ) . These genetic data from lions and FIVPle is consistent with the older origin of extant East African lions , which is further supported by the oldest lion fossils discovered in East Africa [1] . Relative to East Africa , Southern lions have a slightly more recent mtDNA coalescence . Lineage II , found in Nam , Bot-II and Kru has an estimated coalescence of 169 , 000 years ( 95% CI: 34 , 000–304 , 000 ) and the more widespread lineage IV found in the Southern populations of Bot-I , Bot-II and Kru as well as the Eastern populations of Ser ( I , II , and III ) , Ngc and Uga , coalesces ∼101 , 000 years ago ( 95% CI: 11 , 000–191 , 000 ) . However , the similar levels of nDNA genetic diversity , the occurrence of an exclusively Southern mtDNA lineage II and highly divergent FIVPle subtypes , FIVPle subtype D found only in Kru and subtype E exclusive to Bot-II , suggests that both East and Southern Africa were important refugia for lions during the Pleistocene . Therefore , the co-occurrence of divergent mtDNA haplotypes ( 6 to 10 mutations; Figure 1B and 1C ) in southern populations may be the consequence of further isolation within refugia during colder climatic periods . Contemporary fragmentation of lion populations could further explain the results of nested-clade phylogeographical analysis ( NCPA [27] ) ( Figure S5 ) , which inferred restricted gene flow with isolation-by-distance between mtDNA haplotypes H9 ( Bot-II ) and H10 ( Kru ) ( χ2 = 10 . 00 , P = 0 . 0200 ) , between haplotypes H1 ( Bot-II/Nam ) and H2 ( Kru ) ( χ2 = 71 . 00 , P≤0 . 0001 ) , and between haplotypes H9–H10 ( Bot-II/Kru ) and haplotypes H11–H12 ( Bot-I/Kru/Ser/Ngc/Uga ) ( χ2 = 187 . 83 , P≤0 . 0001 ) . Further isolation within refugia ( sub-refugia ) may also have occurred in East Africa . This is suggested by the distinctive mtDNA haplotype H4 and the unique FIVPle subtype F found in the Kenya population , which may have resulted from reduced gene flow across the Rift valley , a scenario that has been suggested for several bovid and carnivore populations ( see [28] and references therein ) . The best example of concordance between host genome markers and viral transmission patterns is observed in the Serengeti National Park in Tanzania . Our previous findings described markedly high levels of FIVPle subtype A , B and C circulating within the Serengeti lion population to such an extent that 43% of the lions sampled were multiply-infected with two or three subtypes [15] , [18] and were hypothesized to represent recent admixture of three formerly separated populations . Such result is confirmed here by lion genomic markers ( Figure 2 ) . Further , although lions within the Serengeti can be assigned to one of three populations ( Ser-I , Ser-II or Ser-III ) by host genomic markers , FIVPle subtypes are distributed ubiquitously in all three , characteristic of rapid horizontal retroviral transmission subsequent to host population admixture . The possible isolating mechanism remains to be elucidated as there is no apparent barrier to gene flow in this ecosystem . Based on patterns of genetic diversity and phylogenetic analysis of lion nDNA/mtDNA and FIVPle markers , we propose a scenario of a period of refugia/isolation in the Late Pleistocene followed by two major lion expansions across Africa and Asia . The first expansion , supported by the mtDNA NCPA [27] ( χ2 = 690 . 00 , P≤0 . 0001; Figure S5 ) , was a long-distance colonization of mtDNA lineage-III ( Gir/Atl/Ang/Zbw ) around 118 , 000 years ago ( 95% CI: 28 , 000–208 , 000 ) , with subsequent fragmentation of haplotypes H5–H6 into Central and North Africa and haplotypes H7–H8 into West Asia ( M1- Figure 1A ) . Support for this initial expansion is also found in nDNA . The ADA haplotype A5 fixed in Gir in also present in Ken , Ser-II , and Ser-III , suggesting that lions likely colonized West Asia from the East Africa refugia ( Figure 1B ) . Such an expansion may have been favored by the start of a warmer and less arid period in Africa 130 , 000–70 , 000 years ago [29] . This “out-of-Africa event” would have occurred much later than the initial lion expansion through Eurasia based on fossils ( ∼500 , 000 years ago ) [3] . It is likely that multiple lion expansions occurred in the Pleistocene , as occurred with humans [21] . A second , more recent lion expansion probably occurred at the Pleistocene/Holocene transition , this one from Southern Africa toward East Africa ( M2- Figure 1A , Figure 3 ) . This is reflected in the mtDNA linage IV , where haplotypes present in Southern lions are basal ( older ) to those found in the East . Overall , mtDNA population nucleotide diversity decreases from Southern to East Africa ( Figure 1B and 1C ) , a finding supported by pairwise mismatch analysis [30] ( raggedness , r = 0 . 086; P<0 . 001 ) . The fixation of mtDNA haplotype H11 in Bot-I ( otherwise fixed only in East Africa populations ) suggests that the colonizing lions expanded northwards from the Kalahari Desert , which included bush , woodland and savannah habitats during the climatic fluctuations of the Pleistocene [31] . This expansion would have occurred relatively recently as the single rare tip mtDNA haplotype H12 , found only in Ser-I , is derived from the interior widespread haplotype H11 ( ∼14 , 000–7 , 000 years; given one mtDNA substitution every 7 , 000 years; Table 1 ) . This expansion is also supported by FIVPle subtype A where haplotypes present in Southern lions ( Kru and Bot-I ) are basal to those found in the East ( Ser-I , Ser-II and Ser-III ) and a decrease of nucleotide-diversity of this FIVPle subtype is observed from Southern ( π = 0 . 15 ) to Eastern Africa ( π = 0 . 03 ) ( Figure 3B ) . Interestingly , a similar northward colonization process from Southern Africa has been suggested for some of the lion preys , namely the impala , greater kudu , and wildebeest [32] , [33] . If we had restricted our inferences to mtDNA , we might have concluded that East African lion populations , which are fixed or nearly fixed for haplotype H11 , went extinct during the Pleistocene/Holocene transition ( similar to the well known mega-fauna extinctions of the Late Pleistocene [34] ) and were then colonized by Southern populations . However , our population genomics data better fit a scenario of lion population expansion and interbreeding rather than simple replacement . First , genetic diversity and allelic richness at nDNA are slightly higher in East Africa populations relatively to those in Southern Africa . This is contrary to the expected pattern of population expansion in which there is usually a progressive decline in genetic diversity and allelic richness . Second , Ser lions carry two diverse FIVPle subtypes found only in East Africa ( B–C ) , and not only FIVPle subtype A , which was presumably introduced in East Africa coincidently with the mtDNA expansion event northwards from South . Third , the East African FIVPle subtype B found in Uga/Ser-I/Ser-II/Ser-III/Ngc showed evidence of a population expansion ( raggedness , r = 0 . 004; P<0 . 01; Fs = −20 . 37; P<0 . 00001 ) and the highest nucleotide diversity observed within FIVPle subtypes ( π = 0 . 09 ) . Four , the FIVPle subtype diversity is higher in East African clades ( four out of the six viral strains ) . The utility of FIVPle pol-RT as a marker of lion population structure and natural history is that it can be informative on a contemporaneous time scale , though it may be less useful at capturing more ancient demographic events . The extreme divergence among FIVPle subtypes , considered with high sero-prevalence in eight of the 11 lion populations , and combined with patterns of geographic concordance , support the hypothesis that FIVPle is not a recent emergence within modern lions [35] . Populations that harbor one private FIVPle subtype such Ken ( subtype F ) , Bot-II ( subtype E ) , and Kru ( subtype D ) must have been sufficiently isolated for enough time for the virus to evolve into unique subtypes , a result corroborated by the high nDNA and mtDNA genetic structure present in these lion populations ( Figure 4 ) . Thus , it is possible that the initial emergence of FIVPle pre-dates the Late-Pleistocene expansions of contemporary lion populations [36] , but present day distributions are more useful indicators of very recent host population dynamics , a result also observed with FIVPco in a panmictic population of pumas in western North America [19] . Accurate interpretation of past and contemporary population demographic scenarios is a primary goal for the effective conservation of endangered species . In this study , we found substantial population subdivision , reduced gene flow , and large differences in FIVPle sequence and sero-prevalence among lion populations , as well as evidence of historic secondary contact between populations ( Figure 3C; Table S4 to S9 ) . The very low population level of mtDNA nucleotide diversity , the number of haplotypes private to a single population ( Figure 1 ) , and probably also the lack of SRY genetic variation across all male lions ( haplotype S1 , n = 183 ) suggests that lion numbers diminished considerably following the Late Pleistocene . The last century reduction in lion distribution further eroded its genomic diversity , and microsatellite variation suggested recent population bottlenecks in seven out of the 11 populations ( standardized differences test , P<0 . 05; Table S5 ) [37] . Although we did not explicitly try to address the adequacy of lion subspecies designations ( currently only one African subspecies is widely recognized ) [38] , [39] , we provided strong evidence that there is no evidence of substantial genetic exchange of matrilines among existing populations as the AMOVA [40] within-population component was uniformly high in all distinct subdivision scenarios ( ΦST≈0 . 920; P<0 . 0001; three-six groups; Table S6 ) . Similarly , significant population structure was detected from nDNA ( FST = 0 . 18 ) , with low levels of admixture evident from Bayesian analysis [22] ( α = 0 . 033 ) . Therefore , employing a bottom-up perspective that prioritizes populations , rather than large-scale units ( e . g . all African lions ) , might preserve and maintain lion diversity and evolutionary processes most efficiently [41] .
A total of 357 individuals were obtained across most of the lion range in Africa and Asia ( Figure 1A; Table S1 ) . Genetic variation among lion specimens was assessed using maternal ( 12S and 16S genes ) , paternal ( SRY gene ) and bi-parental autosomal ( 22 microsatelite loci , and the ADA and TF genes ) markers ( GenBank accession numbers: FJ151632–FJ151652 ) . Analyses of mtDNA in Panthera species are complicated by the presence of a 12 . 5 kb mtDNA integration into chromosome F2 [42] . Accordingly , mtDNA specific primers were designed for the 12S and 16S genes ( Table S2 ) and we used long-range PCR amplification . We designed primers to amplify segments of the ADA ( exon 10 and intron 10 ) and the TF ( intron 3 ) genes ( Table S2 ) , two of the most variable protein loci in lion populations [5] , localized on the domestic cat Felis catus chromosome A3p and C2q , respectively . The Y-chromosome SRY-3′UTR gene was also amplified [43] . PCR products were amplified from 50 ng of genomic DNA in a 25 µL reaction system containing 1 . 5 mM MgCl2 , 1 . 0 mM dNTPs , 0 . 25 units of AmpliTaq Gold DNA polymerase ( Applied Biosystems ) , and 1× PCR buffer II; the amplification protocol was: denaturation 10 min at 95°C , a touch-down cycle of 95°C for 30 s , 52°C for 60 s decreased by 1°C in the next cycle for 10 cycles , 72°C for 120 s , then 35 amplification cycles of 95°C for 30 s , 52°C for 60 s , and 72°C for 120 s , followed by an extension of 10 min at 72°C . PCR products were sequenced on an ABI 377 . Sequences were aligned and cleaned using SEQUENCHER ( Gene Codes ) . Twenty two polymorphic microsatellite loci ( 20 dinucleotide repeats: FCA006 , FCA008 , FCA014 , FCA069 , FCA077 , FCA085 , FCA091 , FCA098 , FCA105 , FCA126 , FCA129 , FCA139 , FCA205 , FCA208 , FCA211 , FCA224 , FCA229 , FCA230 , FCA247 , and FCA281; and two tetranucleotide repeats: FCA391 and FCA441 ) were amplified [44] . Microsatellites were scored in an ABI 377 and analyzed using Genescan 2 . 1 and Genotyper 2 . 5 . These loci are located on 11 of the 19 F . catus chromosomes , occurring in different linkage groups or at least 12 centimorgans apart [44] , [45] . Western blots using domestic cat and lion FIV as antigen were performed as previously described [46] , [47] . The supernatant from virus-infected cells was centrifuged at 200 g for 10 min at 5°C . The resultant supernatant was centrifuged at 150 , 000 g at 4°C for 2 hours . Pelleted viral proteins were resuspended in 1/20th of the original volume and total protein content was assayed using the Biorad Protein Assay . Twenty mg of viral protein were run on 4–20% Tris-Glycine gels and transferred to PDVF membranes ( BioRad ) . Membrane strips were exposed 2–12 h to a 1∶25 or 1∶200 dilution of serum or plasma . After washing , samples were labeled with goat anti-cat HRP or phosphate conjugated antibody ( KPL laboratories ) at a 1∶2000 dilution , washed , and incubated in ECL Western Blotting detection reagents ( Amersham Biosciences ) for 2 min , then exposed to Lumi-Film Chemiluminescent Detection Film ( Boehringer Mannheim ) or incubated in BCIP/NBP phosphatase substrate ( KPL laboratories ) for 15 min [46]–[48] . Results were visualized and scored manually based on the presence and intensity of antibody binding to the p24 gag capsid protein . Nested PCR amplification of partial FIVPle pol-RT was performed [18] , [46] . Briefly , first round PCR reactions used 100 ng of genomic DNA , 2 . 5 mM MgCl2 and an annealing temperature of 52°C . Second round PCR reactions used identical conditions with 1–5 µl of first-round product as template . All PCR products were sequenced as described above for lion genetic analyses ( GenBank accession numbers: AY549217–AY552683; AY878208–AY878235; FJ225347–FJ225382 ) . We used the Genetix 4 . 02 [49] , Genepop 3 . 3 [50] , Microsat [51] , and DnaSP 4 . 10 [52] to calculate the following descriptive statistics: ( i ) percentage of polymorphic loci ( P95 ) , number of alleles per locus ( A ) , observed and expected heterozygosity ( HE and HO ) , and number of unique alleles ( AU ) ; ( ii ) assess deviations from HWE; ( iii ) estimate the coefficient of differentiation ( FST ) , and ( iv ) nucleotide ( π ) and haplotype ( h ) diversity . We tested the hypothesis that all loci are evolving under neutrality for both the lion and the FIVPle data . For frequency data , we used the method described by Beaumont and Nichols [53] and implemented in Fdist ( http://www . rubic . rdg . ac . uk/~mab/software . html ) . The FST values estimated from microsatellite loci plotted against heterozygosity showed that all values fall within the expected 95% confidence limit and consequently no outlier locus were identified . For sequence data ( lion nDNA/mtDNA and FIVPle pol-RT ) , we ruled out any significant evidence for genetic hitchhiking and background selection by assessing Fu and Li's D* and F* tests [54] and Fu's FS statistics [55] . A Bayesian clustering method implemented in the program Structure [22] was used to infer number of populations and assign individual lions to populations based on multilocus genotype ( microsatellites ) and sequence data ( ADA , TF , and mtDNA genes ) and without incorporating sample origin . For haploid mtDNA data , each observed haplotype was coded with a unique integer ( e . g . 100 , 110 ) for the first allele and missing data for the second ( Structure [22] analyses with or without the mtDNA data were essentially identical ) . For K population clusters , the program estimates the probability of the data , Pr ( X|K ) , and the probability of individual membership in each cluster using a Markov chain Monte Carlo method under the assumption of Hardy-Weinberg equilibrium ( HWE ) within each cluster . Initial testing of the HWE in each of the populations defined by the geographic origin of sampling revealed no significant deviation from HW expectations with the exception of Ser and Bot population ( later subdivided by Structure [22] in 3 and 2 clusters , respectively; such deviations from HW expectations were interpreted as evidence of further population structuring ) . We conducted six independent runs with K = 1–20 to guide an empirical estimate of the number of identifiable populations , assuming an admixture model with correlated allele frequencies and with burn-in and replication values set at 30 , 000 and 106 , respectively . Structure also estimates allele frequencies of a hypothetical ancestral population and an alpha value that measures admixed individuals in the data set . The assignment of admixed individuals to populations using Structure [22] has been considered in subsequent population analyses . For each population cluster k , the program estimates Fk , a quantity analogous to Wright's FST , but describing the degree of genetic differentiation of population k from the ancestral population . Patterns of gene flow and divergence among populations were described using a variety of tests . First , to visualize subtle relationships among individual autosomal genotypes , three-dimensional factorial correspondence analyses [23] ( FCA ) were performed in Genetix [49] , which graphically projects the individuals on the factor space defined by the similarity of their allelic states . Second , neighbor-joining ( NJ ) analyses implementing the Cavalli-Sforza & Edwards' chord genetic distance [56] ( DCE ) were estimated in Phylip 3 . 6 [57] , and the tree topology support was assessed by 100 bootstraps . Third , the difference in average HO and A was compared among population groups using a two-sided test in Fstat 2 . 9 . 3 . 2 [58] , which allows to assess the significance of the statistic OSx using 1 , 000 randomizations . Four , the equilibrium between drift and gene flow was tested using a regression of pairwise FST on geographic distance matrix among all populations for host nDNA ( microsatellites ) /mtDNA and FIVPle data . A Mantel test [59] was used to estimate the 95% upper probability for each matrix correlation . Assuming a stepping stone model of migration where gene flow is more likely between adjacent populations , one can reject the null hypothesis that populations in a region are at equilibrium if ( 1 ) there is a non-significant association between genetic and geographic distances , and/or ( 2 ) a scatterplot of the genetic and geographic distances fails to reveal a positive and monotonic relationship over all distance values of a region [60] . We also evaluated linearized FST [i . e . FST/ ( 1−FST ) ] [24] among populations . We tested two competing models of isolation-by-distance , one assuming the habitat to be arrayed in an infinite one-dimensional lattice and another assuming an infinite two-dimensional lattice . Both models showed that genetic differentiation increased with raw and log-transformed Euclidean distances , respectively [24] . We determined the confidence interval value of the slope of the regression for the nDNA data using a non parametric ABC bootstrap [61] in Genepop 4 . 0 [62] . The demographic history of populations was compared using a variety of estimators based on the coalescence theory . First , signatures of old demographic population expansion were investigated for mtDNA and FIVPle pol-RT haplotypes using pairwise mismatch distributions [63] in DnaSP [52] . The goodness-of-fit of the observed data to a simulated model of expansion was tested with the raggedness ( r ) index [64] . Second , the occurrence of recent bottlenecks was evaluated for microsatellite data using the method of Cornuet & Luikart [37] in Bottleneck [65] and using 10 , 000 iterations . This approach , which exploits the fact that rare alleles are generally lost first through genetic drift after reduction in population size , employs the standardize differences test , which is the most appropriate and powerful when using 20 or more polymorphic loci [37] . Tests were carried out using the stepwise mutation model ( SMM ) , which is a conservative mutation model for the detection of bottleneck signatures with microsatellites [66] . Third , to discriminate between recurrent gene flow and historical events we used the nested-clade phylogeographical analysis [27] , [67] ( NCPA ) for the mtDNA data . When the null-hypothesis of no correlation between genealogy and geography is rejected , biological inferences are drawn using a priori criteria . The NCPA started with the estimation of a 95% statistical parsimony [68] mtDNA network in Tcs 1 . 20 [69] . Tree ambiguities were further resolved using a coalescence criteria [70] . The network was converted into a series of nested branches ( clades ) [71] , [72] , which were then tested against their geographical locations through a permutational contingency analysis in GeoDis 2 . 2 [73] . The inferences obtained were also corroborated with the automated implementation of the NCPA in ANeCA [74] . To address potential weaknesses in some aspects of the NCPA analysis [75] , [76] , we further validated the NCPA inferences with independent methods for detecting restricted gene-flow/isolation-by-distance ( using matrix correlation of pairwise FST and geographic distance ) and population expansion ( using pairwise mismatch distributions ) . Four , to test the significance of the total mtDNA genetic variance , we conducted hierarchical analyses of molecular variance [40] ( AMOVA ) using Arlequin 2 . 0 [77] . Total genetic variation was partitioned to compare the amount of difference among population groups , among populations within each groups , and within populations . Phylogenetic relationships among mtDNA and FIVPle pol-RT sequences were assessed using Minimum evolution ( ME ) , Maximum parsimony ( MP ) , and Maximum likelihood ( ML ) approaches implemented in Paup [78] . The ME analysis for mtDNA consisted of NJ trees constructed from Kimura two-parameter distances followed by a branch-swapping procedure and for FIVPle data employed the same parameter estimates as were used in the ML analysis . The MP analysis was conducted using a heuristic search , with random additions of taxa and tree-bisection-reconnection branch swapping . The ML analysis was done after selecting the best evolutionary model fitting the data using Modeltest 3 . 7 [79] . Tree topologies reliability was assessed by 100 bootstraps . For the FIVPle data , the reliability of the tree topology was further assessed through additional analyses using 520 bp of FIVPle pol-RT sequences in a representative subset of individuals . The time to the most recent common ancestor ( TMRCA ) for the ADA and TF haplotypes was estimated following Takahata et al . [80] , where we calculate the ratio of the average nucleotide differences within the lion sample to one-half the average nucleotide difference between leopards ( P . pardus ) and lions and multiplying the ratio by an estimate of the divergence time between lions and leopards ( 2 million years based on undisputed lion fossils in Africa ) [81] , [82] . The mtDNA TMRCA was estimated with a linearized tree method in Lintree [83] and using the equation H = 2μT , where H was the branch height ( correlated to the average pairwise distance among haplotypes ) , μ the substitution rate , and T the divergence time . Leopard and snow leopard ( P . uncia ) sequences were used as outgroups . Inference of the TMRCA for microsatellite loci followed Driscoll et al . [6] where the estimate of microsatellite variance in average allele repeat-size was used as a surrogate for evolutionary time based on the rate of allele range reconstitution subsequent to a severe founder effect . Microsatellite allele variance has been shown to be a reliable estimator for microsatellite evolution and demographic inference in felid species [6] .
|
The lion Panthera leo , a formidable carnivore with a complex cooperative social system , has fascinated humanity since pre-historical times , inspiring hundreds of religious and cultural allusions . Here , we use a comprehensive sample of 357 individuals from most of the major lion populations in Africa and Asia . We assayed appropriately informative autosomal , Y-chromosome , and mitochondrial genetic markers , and assessed the prevalence and genetic variation of the lion-specific feline immunodeficiency virus ( FIVPle ) , a lentivirus analogous to human immunodeficiency virus ( HIV ) that causes AIDS-like immunodeficiency disease in domestic cats . We compare the large multigenic dataset from lions with patterns of genetic variation of the FIVPle to characterize the population-genomic legacy of lions . We refute the hypothesis that African lions consist of a single panmictic population , highlighting the importance of preserving populations in decline rather than prioritizing larger-scale conservation efforts . Interestingly , lion and FIVPle variation revealed evidence of unsuspected genetic diversity even in the well-studied lion population of the Serengeti Ecosystem , which consists of recently admixed animals derived from three distinct genetic groups .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Material",
"and",
"Methods"
] |
[
"evolutionary",
"biology",
"genetics",
"and",
"genomics/population",
"genetics"
] |
2008
|
The Evolutionary Dynamics of the Lion Panthera leo Revealed by Host and Viral Population Genomics
|
The level of susceptibility to tuberculosis ( TB ) infection depends upon allelic variations in numerous interacting genes . In our mouse model system , the whole-genome quantitative trait loci ( QTLs ) scan revealed three QTLs involved in TB control on chromosomes 3 , 9 , and in the vicinity of the H2 complex on chromosome 17 . For the present study , we have established a panel of new congenic , MHC-recombinant mouse strains bearing differential small segments of chromosome 17 transferred from the TB-susceptible I/St ( H2j ) strain onto the genetic background of TB-resistant C57BL/6 ( B6 ) mice ( H2b ) . This allowed narrowing the QTL interval to 17Ch: 33 , 77–34 , 34 Mb , containing 36 protein-encoding genes . Cloning and sequencing of the H2j allelic variants of these genes demonstrated profound polymorphic variations compare to the H2b haplotype . In two recombinant strains , B6 . I-249 . 1 . 15 . 100 and B6 . I-249 . 1 . 15 . 139 , recombination breakpoints occurred in different sites of the H2-Aβ 1 gene ( beta-chain of the Class II heterodimer H2-A ) , providing polymorphic variations in the domain β1 of the Aβ-chain . These variations were sufficient to produce different TB-relevant phenotypes: the more susceptible B6 . I-249 . 1 . 15 . 100 strain demonstrated shorter survival time , more rapid body weight loss , higher mycobacterial loads in the lungs and more severe lung histopathology compared to the more resistant B6 . I-249 . 1 . 15 . 139 strain . CD4+ T cells recognized mycobacterial antigens exclusively in the context of the H2-A Class II molecule , and the level of IFN-γ-producing CD4+ T cells in the lungs was significantly higher in the resistant strain . Thus , we directly demonstrated for the first time that the classical H2- Ab1 Class II gene is involved in TB control . Molecular modeling of the H2-Aj product predicts that amino acid ( AA ) substitutions in the Aβ-chain modify the motif of the peptide–MHC binding groove . Moreover , unique AA substitutions in both α- and β-chains of the H2-Aj molecule might affect its interactions with the T-cell receptor ( TCR ) .
Tuberculosis ( TB ) remains a significant public health problem: one-third of the human population is infected with Mycobacterium tuberculosis ( MTB ) and 10% of those are at a risk of developing overt TB during their lifetime [1 , 2] . Although there is growing body of evidence that the outcome of infection is modulated both by bacterial and host genetics [3 , 4] , genetic factors regulating susceptibility to infection , transition from latency to reactivation and severity of the disease remain largely unknown . The important role of host genetic factors in TB disease control in humans has been clearly demonstrated in numerous studies , including adoption [5] , twin [6–8] , genome-wide association ( GWAS ) [9–12] , and case-control population [13–16] studies . Apart from rare cases of Mendelian susceptibility to mycobacterial diseases ( MSMD ) due to nonsense and missense mutations in key genes involved in protective immunity against intracellular pathogens ( reviewed in [17] ) , the complex patterns of TB susceptibility and disease manifestations clearly correspond to a polygenic type of genetic control with numerous epistatic interactions ( reviewed in [18] ) . Naturally , identification of TB control-relevant genes and alleles in humans remains a very difficult task which is complicated by the environmental and strain diversity , as well as by the lack of consensus in the definition of , and distinction between , clinical phenotypes . TB infection can be readily induced in mice , and some refined mouse TB models reproduce human-like pulmonary infection with appreciable accuracy ( see [19 , 20] for the review ) . In a few independent studies employing different inbred mouse strains , the whole genome scan approach has been applied for genetic mapping of quantitative trait loci ( QTL ) involved in TB susceptibility and disease control [21–25] . Since different inbred strains were selected as susceptible and resistant parental prototypes , and different phenotypes ( survival time post-infection , multiplication of mycobacteria in organs , dynamics of cachexia ) were analyzed , it is not surprising that the genomic locations of most of the QTL reported in different studies did not coincide . In our TB models , we use I/St TB-susceptible and A/Sn or C57BL/6 TB-resistant mice as prototypes . TB-infected I/St mice differ profoundly from their more resistant counterparts by early onset of mortality , rapid body weight loss , increased mycobacterial multiplication in lungs and spleens , and exacerbated lung histopathology [26] . Whole genome scans performed in F2 and N2 generations identified three QTL on chromosomes 3 , 9 and 17 whose allelic variation affected TB susceptibility [22 , 23] . The QTL on chromosome 17 , peaking at the D17Mit175 marker , overlaps with the location of the mouse major histocompatibility complex Н2 . Remarkably , this locus remains the only known case of co-localization of TB-controlling QTL reported in previous studies: Kramnik and colleagues mapped a QTL within the H2 region using a different combination of parental strains [21] . Associations of TB susceptibility/severity with the MHC polymorphic haplotypes have been previously reported both in humans [27–30] and mice [31–35] . In mice , allelic variations within the H2 complex were shown to affect survival time following challenge , the level of T-lymphocyte-mediated delayed type hypersensitivity ( DTH ) response , T-cell proliferation after stimulation with mycobacterial antigens and the efficacy of BCG vaccination against tuberculosis , production of IFN-γ by mycobacteria-specific T-cells , and production of mycobacteria-specific antibodies [31 , 36–39] . However , these early studies provided no information about any particular gene within the H2 complex affecting TB immunity . Progression from a defined QTL region to a particular gene remains a major challenge: about 3 , 000 QTLs have been mapped in mice and rats but less than 1% of the genes have been identified at the molecular level [40] . This is especially true for the ~3 . 5 Mb MHC region , which provides the highest density of coding genes in the genome . Furthermore , many of these genes , which display a very high level of allelic variation and extensive linkage disequilibrium , have fundamental roles in immunity . Not surprisingly , numerous associations with several diseases for this part of the genome have been reported [41–44] . To begin identification of the gene , we have started to narrow the interval for the chromosome 17 QTL using a classical homologous recombination approach and have developed a panel of recombinant congenic mouse strains bearing different intra-H2 segments from TB-susceptible I/St mice on the resistant B6 genetic background . Given the previous demonstration that an allelic variant of chromosome 17 QTL inherited from B6 mice determined resistance to infection [21] , we decided to use these common and genetically well-characterized mice as the TB-resistant prototype strain . At the initial stage of this study , we succeeded in narrowing the region on chromosome 17 which determines the level of TB-susceptibility from 8–65Mb to 33 , 77–34 , 34 Mb [45] . Since gene sequencing data for the I/St inbred strain are unavailable from the databases , in the present study we cloned and sequenced coding parts of all genes annotated for this region using I/St cDNA . As expected , the region displayed a very high level of genetic polymorphism and only a few out of 36 genes demonstrated identical sequences for B6 and I/St . In addition , the region under study contains many genes of importance for immunity and cell biology , thus being realistic candidates for the infection control . Thus , we searched for recombination events inside the TB-controlling region and established new mouse strains narrowing the region to the 34 , 24–34 , 33Mb interval . This interval contains only five coding genes , all belonging to classical and non-classical Class II: H2-Ob , H2-Aa , H2-Ab1 , H2-Eb1 and H2-Eb2 . Two recombinant strains with substantial differences in response to TB infection displayed recombination events in different parts of a single H2-Ab gene , which was critical for gene identification .
We transferred genomic regions covering the vicinity of the H2 complex from TB-susceptible parental I/St ( H2j ) mice onto the B6 ( H2b ) genetic background in successive backcross generations . Starting with the BC1 ( N2 ) generation , we applied simultaneous selection for the presence of two traits: TB-susceptible phenotype and Chromosome 17 markers of the I/St origin . At the generation N10-11 , more than forty B6 . I-H2 recombinant congenic strains on the B6 background carrying different , partly-overlapping genomic regions of the extended H2j- haplotype ( 17 Chr: 8 , 44–65 , 34 Mb ) were generated . Fig 1 displays the most informative B6 . I strains whose pheno- and genotyping allowed us to narrow the region of interest to the interval 33 , 77–34 , 34Mb ( a total of 0 , 57 Mb ) . Mice of all strains which inherited this region from I/St ancestors were significantly more susceptible to infection than those bearing B6 alleles as indicated by survival curves ( Fig 2 ) and the dynamics of cachexia ( S1 Fig ) . Fine mapping within this region was achieved by superposition: the resistant strain B6 . I-249 . 1 . 16 carries H2j alleles more proximal than the SNPs rs13482956 ( 17:33 , 773331 ) whereas the strain B6 . I-9 . 3 . 19 . 8 is susceptible although all distal genes starting with and including H2-Ea are of B6 origin ( Fig 1 ) . Being more susceptible than B6 , all recombinant strains carrying the region 33 , 77–34 , 34Mb inherited from I/St were more resistant than their I/St ancestors , indicating the influence of B6 background genes on survival . This is further supported by the fact that the level of resistance was identical in parental B6 mice and in recombinant mice which inherited the identified H2 segment from B6 ( Fig 2 ) . According to http://www . ensembl . org , the identified region contains 36 protein-coding genes , most of which may have important immunological and regulatory functions . No information was available about the genome sequence of I/St mice , so it was impossible to start searching for candidate genes by direct sequence comparison . Therefore , we cloned and sequenced the protein-coding regions of all 36 I/St-originated genes in the region ( GenBank , accession numbers KJ650201-KJ650234 ) . Table 1 displays all amino acid ( AA ) substitutions between H2b and H2j haplotypes , as judged from the cDNA sequencing data . As expected , the region appeared to be highly polymorphic: only seven genes ( Zbtb22 , H2-Ke2 , B3galt4 , Slc39a7 , Brd2 , H2-DMb2 , H2-DMb1 ) displayed no allelic polymorphism for the two haplotypes . The H2 segment under study contains numerous genes generally involved in immune response control , and for many of these genes evidence is available indicating their possible involvement in the control of TB infection . S1 Table briefly summarizes the data illustrating this point . In addition , alternative splicing isoforms for many I/St alleles in this region not annotated previously were revealed ( see: GenBank , accession numbers KJ663713- KJ663725 ) , making the general picture of genetic diversity even more complex . The rich deposit of polymorphic genes potentially influencing susceptibility to , and severity of , TB infection , as well as the potential contributions of both polymorphism and expression regulation of other genes in the region , justified further narrowing the interval by genetic recombination . To search for new recombination events inside the region 33 , 77–34 , 34 Mb , we performed several crosses between novel recombinant and B6 mice . In particular , the F2 progeny of ( B6 . I-249 . 1 . 15 x B6 ) F1 mice was used to develop a new set of congenic strains . In two new recombinant strains , B6 . I-249 . 1 . 15 . 100 ( hereafter–B6 . I-100 ) and B6 . I-249 . 1 . 15 . 139 ( hereafter–B6 . I-139 ) , standard genotyping identified the point of recombination between markers D17Mit21 and D17Mit22 ( Fig 3A ) . Surprisingly , these strains demonstrated sharply contrasting TB phenotypes ( Fig 3B ) . After aerosol challenge , B6 . I-139 mice did not differ by survival time from parental B6 mice ( mean survival time ( MST ) = 238 . 9 ± 13 . 41 and 249 ± 10 . 21 days , respectively , P > 0 . 5 ) . B6 . I-100 mice did not differ from the B6 . I -249 . 1 . 15 . 46 strain ( MST 152 ± 13 . 3 and 153 ± 10 . 97 days , respectively , P > 0 . 5 ) , but did differ significantly from the B6 . I-139 strain ( P < 0 . 001 ) . Phenotypic differences were confirmed by evaluation of cachexia dynamics ( S2 Fig ) , and by assessment of mycobacterial loads in the lungs at weeks 4 and 10 post-challenge ( Fig 3C ) . Annotation in the http://www . ensembl . org database provides the length of 98 , 588bp for the genomic region between D17Mit21 and D17Mit22 , which contains only 5 protein-coding genes , 2 lincRNA genes and no genes for micro-RNAs ( Fig 3D ) . The chromosomal segment sufficient to determine the contrasting TB phenotypes appeared to be very small , and we identified genetic material inside the segment by gene sequencing . Both strains carried the b allele of H2-Ob and the j allele of H2-Aa , but differed at the H2-Ab1 gene ( Fig 4 ) . The H2-Ab1 gene in both strains originated from recombination events between b and j haplotypes , but the crossing-over occurred at different sites . In B6 . I-139 mice , the whole polymorphic part of the H2-Ab1 gene encoding the extracellular functional domain of the molecule was of the H2b origin: only substitutions W222R in the connecting peptide and H249Y in the cytoplasmic domain were inherited from the H2j haplotype . In contrast , in B6 . I-100 mice this polymorphic part of the H2-Ab1 gene was identical with that of the H2j haplotype , except for a single substitution N29D ( Fig 4 ) . As far as both recombination events occurred in the translated part of the gene , we assume that the promoter region of B6 origin was identical for both strains and played no role in infection response . The fact that B6 . I-139 mice displayed the resistant phenotype similar to parental B6 mice suggests that two AA substitutions of I/St origin in the connecting peptide and the cytoplasmic domain are not major players in TB susceptibility . Analogously , the presence of the H2b-encoded aspartic acid in the H2-Ab1 of the B6 . I-100 strain is unlikely to influence the level of TB susceptibility , since B6-I . 100 mice display a phenotype identical to that of B6 . I-249 . 1 . 15 . 46 mice , whose entire H2-Ab1 gene was inherited from I/St mice . Taken together , these results demonstrate that the differences in TB susceptibility/severity between these two recombinant mouse strains were determined by allelic polymorphisms in a single β1 domain of the H2-Aβ molecule . Thus , independent recombination events within a single gene created genetic variation sufficient to markedly alter the response to TB infection . The newly identified H2-Ab1 alleles were designated as j* in the B6 . I-100 strain and b* in the B6 . I-139 strain . We performed fine genetic mapping using the most integrative TB characteristics–survival curves , mycobacterial multiplication in the lungs , and body weight loss . Differences in the regulation of lung tissue inflammation after infection in B6 and I/St mice are of critical importance for TB pathogenesis [46 , 47] . To characterize the influence of the H2-Ab1 polymorphism on TB-induced inflammation , we compared lung pathology 35 days post-infection in mice of both parental and new recombinant strains . As shown in Fig 5A , in B6 and B6 . I-139 mice , lung pathology was represented by granulomatous areas well-delimited from the breathing tissue , whereas I/St and B6 . I-100 mice developed diffuse TB pneumonia that was more severe in I/St mice . In good agreement with the histological results , the levels of key Type 1 inflammatory cytokines , IL-6 and TNF-α , after TB challenge were significantly lower in the lungs of resistant B6 and B6 . I-139 mice compared to susceptible I/St and B6 . I-100 mice ( Fig 5B ) . No difference in the levels of the TB-irrelevant Type 2 cytokine IL-5 between all four strains was found . Importantly , production of the key TB-protective Type 1 cytokine , IFN-γ , by CD4+ T-lymphocytes isolated from the lungs of infected mice and stimulated in vitro with a mixture of mycobacterial antigens followed the H2-Ab1-determined pattern . As shown in Fig 5C , the 5-fold difference in the numbers of IFN-γ-producing CD4+ T cells between parental B6 and I/St strains was reduced to a 2-fold difference between B6 . I-139 and B6 . I-100 mice but remained highly significant ( P < 0 . 01 ) . Calculation of the total numbers of IFN-γ-producing CD4 cells per lung provided the results consistent with the percentile evaluation ( S2 Table ) . These results establish connections between anti-TB protective immune responses , CD4+ T-cell function and allelic heterogeneity of the classical Class II antigen-presenting molecule , providing a mechanistic explanation for the differences in the severity of disease determined by a single MHC gene . IFN-γ production by the CD4+ T cell in response to MTB is generally considered the major mechanism of host defense [48] . TB-resistant B6 mice express only one MHC Class II molecule - H2-Ab on their antigen-presenting cells ( APC ) , whereas I/St mice express both MHC Class II molecules - H2-Aj and H2-Ej . A priori it was impossible to judge whether the defect in TB defense in mice bearing the H2j haplotype was determined by sub-optimal antigen presentation by the H2-Aj compared to the H2-Ab molecule , or by the parallel presentation of mycobacterial antigens by two Class II molecules which somehow interfered with the development of protective immunity . To resolve this issue , we assessed the presentation of mycobacterial antigens by the APC derived from mice with different Class II allelic composition . A mycobacteria-specific CD4+ T cell line derived from I/St mice [49] readily proliferated in the presence of mycobacterial antigens if the APC were derived from mice expressing the H2-Aj molecule , even if the H2-E molecule was not expressed ( Fig 6A ) . Moreover , the presence or absence of H2-E did not change the level of response , suggesting that the H2-A molecule presents mycobacterial antigens to the vast majority of T-cell clones . To prove that the H2-E-recognizing T-cell clones have not been lost due to repeated stimulation during T-cell line development , we repeated the experiment with highly purified CD4+ T cells from TB-immune lymph nodes of I/St mice and obtained similar results ( Fig 6B ) . Recombinant B6 . I-100 and B6 . I-139 mice express the H2-E molecule due to the presence of the H2-Ej α-chain . These mice possess identical H2-Aj α-chains but differ in their H2-Aj β-chains ( Fig 4 ) . To determine in which context mycobacterial antigens were presented to T cells by newly originated H2 haplotypes and to evaluate the efficacy of antigen presentation by recombinant H2-Aβ chains , we assessed the response of T-cells from recombinant mice in the presence of different APC . As shown in Fig 6C and 6D , B6 . I-100 and B6 . I-139 T cells recognized mycobacterial antigens in the context of the H2-A , but not the H2-E , molecule . Interestingly , B6 . I . 100 T-cells did not distinguish fully syngenic Aj* and progenitor Aj , whereas B6 . I-139 T cells recognized only Ab* β-chain , most likely because the hybrid H2-Aβb/Aαj molecule was formed . Taken together , these results indicate that the H2-A molecule plays a pivotal role in the presentation of mycobacterial antigens and the generation of TB-specific CD4+ T cell responses . The Aj and Aj* allelic variants are not intrinsically defective in the antigen-presenting function and elicit a level of T cell proliferation in response to soluble mycobacterial antigens similar to the Ab and Ab* alleles . These observations provided an opportunity to functionally test whether or not the result of T cell interaction with infected macrophages depends upon H2-A alleles . To this end , we performed co-culture experiments with macrophages from B6 and congenic B6 . I-9 . 3 . 19 . 8 ( see Fig 1 ) mice and immune effector CD4 T-cells obtained from ( B6 x B6 . I-9 . 3 . 19 . 8 ) F1 mice , which are able to interact with both allelic forms of H2-A . The H2-E-negative strain B6 . I-9 . 3 . 19 . 8 was used instead of B6 . I-100 to further exclude possible influence of the H2-E expression . As shown in Fig 6E , T-cells profoundly increased the ability of B6 and F1 peritoneal macrophages to inhibit mycobacterial growth , whereas only marginal effect was seen in B6 . I-9 . 3 . 19 . 8 macrophages , suggesting that recognition of “protective” H2-Ab vs . “non-protective” H2-Aj molecules by CD4+ T-cells leads to profound differences in macrophage activation . Remarkably , moderate capacity to inhibit mycobacterial growth in the absence of T-cells was similar in all macrophages , regardless their genetic origin . This is in full agreement with theoretical expectations: Class II alleles do not regulate the level of innate protective response . These results provide additional independent conformation in support of the conclusion that the H2-A allelic variation is sufficient to determine prominent variations in acquired anti-mycobacterial immunity . In contrast with experiments on genetic restriction of antigen-specific response described above , reciprocal functional experiment ( activation of F1 macrophages by H2-Ab and H2-Aj T-cells ) would not be informative , since allogenic Class II recognition provides unpredictable effects . A BLAST search in the Protein Data Bank ( PDB ) for α- and β-chains of the H2-Aj molecule revealed the highest score of sequence similarity ( 88% and 85% identity , respectively ) with the protein 2P24 [50] , with deletions or insertions lacking . Comparison between H2-Aj and H2-Ab ( PDB id 1MUJ ) provided sequence identity of 93% and 89% for α- and β-chains , respectively , with the 2-AA deletion ( P65 and E66 ) in the former allele . Two molecular models of the H2-Aj protein based upon atomic coordinates of 1MUJ and 2P24 provided a high level of similarity around the deletion point ( S1 Fig ) , which justified the use of the 1MUJ model for further comparisons . Comparison between the H2-Aj and H2-Ab molecules suggests that the most prominent structural dissimilarities occur in two protein backbone regions: the 310 helical fragment of the α-chain and the P65E66 deletion in the H2-Aβj chain ( Fig 7A ) . These deviations are not unique and are present in other H2 haplotypes ( reviewed in Ref [51] and displayed in S3 Fig ) . Analysis of the hydrogen bond network between H-2A Class II molecules and , as a model , invariant CLIP peptide stabilizing the complex before an antigenic peptide is loaded , demonstrated that the conserved H-bond interactions and the total number of H-bonds are identical for the H2-Ab and H2-Aj products despite two b → j substitutions , T71K and E74A in the Aβ-chain ( Fig 7B and 7C ) . The Aβ-position E74 is highly conserved among all known mouse H2-A haplotypes ( S4 Fig ) and the majority of human HLA-DQ molecules [52] . In the H2-Aj molecule , the A74-provided H-bond is lacking; however , it might be functionally substituted by the H-bond from the Aβ K71 . However , prominent differences in the structure and size of the peptide-anchoring pockets between the two allelic forms of the H2-A molecule were observed . Structural data for the peptide-anchoring pockets were either available from the 3D structure of the H2-Ab [53] or deduced for the H2-Aj from our model . AA substitutions distinguishing the H2-Aj protein from the prototypic H2-Ab should have profoundly changed the structure of the peptide-binding groove in binding pockets P1 , P4 , P6 , P7 and P9 . Due to the substitution L31W and A52T in the α-chain , the volume of the H2-Aj P1 pocket should be appreciably smaller compared to that of H2-Ab ( Fig 7D ) and , therefore , may accept smaller side chains capable of forming a hydrogen bond with the Aβ T52 hydroxyl group . Changes in other binding pockets are due to substitutions in the β-chain ( Fig 7E ) . The differences between H2-Aj and H2-Ab in the P4 binding pocket include substitutions Y26L , T28S , G13S and the most prominent one—E74A , allowing occupation of the pocket by a neutral residue in H2-Aj instead of a positively charged residue in H2-Ab . Changes in the P6 , ( substitutions Y30N and T71K ) , make this pocket in the H2-Aj molecule more permissive for negatively charged residues . Due to the H47F substitution , the pocket P7 is neutral in H2-Aj but positively charged in H2-Ab , while a minor substitution Y37F makes pocket P9 more permissive for lipophilic residues . We consider the most important differences between H2-Aj and H2-Ab to result from a unique substitution Q61K in the α-chain and substitution R70Q in the β-chain , since these substitutions alter the polarity of interactions between two chains ( S5 Fig ) . Two positively charged residues occupying opposite positions in the hybrid H2-A ( αj+βb* ) ( S5 Fig ) could well affect the Class II-TCR interactions .
An extremely high level of genetic polymorphism in the H2 chromosomal region and its unique saturation with immunity-relevant genes complicate the identification of allelic variants in a particular gene influencing the disease severity and outcome . This is particularly true for the classical Class I and II genes , for which genetic silencing approaches are hardly applicable since they lead to a severe abrogation of overall functions of acquired immunity . It took us about 7 years to develop the panel of more than 40 H2-recombinant inbred strains on the B6 genetic background sufficient for identification of the H2-Ab1 gene as the TB severity determinant using the forward genetic approach . The key point was establishing the fact that in the B6 . I-100 and B6 . I-139 strains distinct recombination breakpoints between the H2j and H2b haplotypes were located within the same H2-Ab1 gene . This resulted in a H2-Ab1b-like allele in TB-resistant B6 . I-139 and in a H2-Ab1j-like allele in TB-susceptible B6 . I-100 mice , which , being compared with previously characterized phenotypes and genotypes in other strains from the panel , identified the H2-Ab1 as the gene underlining the Chromosome 17 TB-controlling QTL . This is the first direct demonstration of the differences in TB infection susceptibility/severity depending on allelic polymorphisms in a single Class II MHC gene . Associations of TB susceptibility/severity with the MHC polymorphic haplotypes have been previously reported both in humans [27–30] and mice [31–35] , but direct evidence that the alleles of H2-A1 or its human orthologous gene HLA-DQ differentially regulate TB control by the host was lacking . Importantly , in our system both allelic variants apparently retain normal functional activity and are not defective in mycobacterial antigen presentation/recognition ( Fig 6 ) . This may reflect the situation which exists in natural populations: fine genetic differences may lead to pronounced shifts in adaptively important responses providing the subject for slowly operating natural selection on quantitative basis , whereas loss-of-function mutations in key immunologically active genes are eliminated rapidly . Allelic differences in a single H2-A1 gene influenced all major phenotypes characterizing severity of TB infection ( bacterial loads in affected organ , histopathology , cachexia , survival time post-challenge ) , suggesting that H2-Ab1 is one of major players in the TB control in mice . Naturally , this does not exclude the presence of other genes within an extended H2 region involved in TB control , but their possible contribution is likely to be weaker . The established panel of recombinant mouse strains may serve a useful tool for dissecting genetic control of susceptibility/severity in other models of TB and in other experimental infections . Indeed , one TB-susceptibility QTL , Sst5 , has been mapped to the H2 region in the B6 –C3HeB/FeJ mouse strain combination [21] . An extended H2 region contains QTL involved in genetic control of susceptibility/severity of several protozoal and metazoal pathogens: Lmr1 for Leishmania major [54] , Char3 for Plasmodium chabaudi [55] , Belr1 for Plasmodium berghei [56] , Tir1 for Trypanosoma congolense [57] and Sm2 for Schistosoma mansoni [58] . It is very unlikely that co-localization of all these QTL is coincidental , and our new panel of mouse strains may shed light on the architecture of the H2-driven genetics of host-parasite interactions in these other disease models . Regarding the molecular mechanisms underlining differences in TB susceptibility in the absence of overt gene dysfunctions , several possibilities are being considered . The most evident is differences in mycobacterial antigenic peptide repertoire presented to T-cells by structurally different Class II molecules . Alignment of the AA sequence of the polymorphic H2-Ab1j domain with annotated H2 haplotypes ( S4 Fig ) demonstrates that it shares the common deletion P65-E66 with haplotypes k , g7 , u , s , f , s2 , retains conserved AA residues at positions determining the basic 3D structure of the protein , and is not unique with this regard . However , there are non-conservative and potentially important substitutions located in four β-strands and in the α-helix that form the peptide-binding groove . Molecular modeling predicts that the motif for peptide binding should differ between H2-Aj and H2-Ab due to substitutions in the pockets P4 , P6 , P7 and P9 ( Fig 7 ) . However , at present no information is available concerning sets of mycobacterial peptides providing more vs . less protective anti-TB responses . We also considered the level and stability of the cell surface expression of the H2-A allelic forms as a factor potentially influencing the level and quality of T-cell activation . Since antibodies reacting with the H2-Ab and H2-Aj molecules with equal affinity are lacking , a direct quantitative comparison of expression levels was impossible . Thus we applied an antibody dilution approach described earlier [59 , 60] and found no differences in the expression levels for the H2-A molecule between B6 , B6 . I-100 and B6 . I-139 mice ( S6 Fig ) , most likely excluding this explanation . Yet another possible reason for the differences in anti-mycobacterial immunity between the carriers of H2-Ab1j and H2-Ab1b alleles could be selection of CD4 T cells in thymus and/or their maintenance in the periphery . Our preliminary studies demonstrated a significant difference in the CD4: CD8 ratio between B6 and I/St mice , as well as between some of the novel recombinant mice . More data and , possibly , new recombinant mouse strains expressing no H2-E molecule will be needed to precisely evaluate the importance of this MHC-dependent pathway of immune response regulation .
To evaluate severity of the disease , mice were infected with ~5 x 102 colony-forming units ( CFU ) of standard virulent M . tuberculosis strain H37RV ( sub-strain Pasteur ) using an Inhalation Exposure System ( Glas-Col , Terre Haute , IN ) exactly as described earlier [49] . Mortality was monitored daily starting at week 5 post-infection . To assess CFU counts , lungs from individual mice were homogenized in 2 . 0 ml of sterile saline , and 10-fold serial dilutions were plated on Dubos agar ( Difco ) and incubated at 37°C for 20–22 days . Pathology of the lung tissue was assessed as described [50] . Briefly , mice were euthanized by a thiopental ( Biochemie GmbH , Vienna , Austria ) overdose . Lung tissue ( the middle right lobe ) was frozen in a –60°C to –20°C temperature gradient in an electronic Cryotome ( ThermoShandon , UK ) , 6–8μm-thick sections were cut across the widest area of the lobe , fixed with acetone , stained with hematoxylin-eosin and mounted . To prepare T-cell lines , cells from the popliteal lymph nodes of I/St and B6 mice , immunized into rear footpads with 10μg/mouse of mycobacterial sonicate mixed 1:1 with incomplete Freund’s adjuvant , were cultured as described previously [62] . Briefly , 2 x 106/ml immune cells isolated on day 21 post-immunization were cultured in 24-well plates ( Costar , Netherlands ) in RPMI-1640 containing 10% FCS , 10 mM HEPES , 4 mM L-glutamine , 5 x 10−5 M 2-ME , vitamins , pyruvate , non-essential amino acids and antibiotics ( all components–HiClone , Logan , UT , USA ) for 14–16 days in the presence of 10μg/ml mycobacterial sonicate . Live immune cells ( >93% viability by trypan blue exclusion ) were isolated by centrifugation at 2500 g for 20 min at 20°C , on the Lympholyte M gradient ( Cedarlane Labs , Ontario , Canada ) , washed twice and counted . The next stimulation cycle was accomplished by co-culturing 2 x 105 isolated cells with mitomycin C-treated 1 . 5 x 106 splenic APC in the presence of sonicate for another 14–16 days . These cycles were repeated 4 times and resulted in stable antigen-specific CD4+ ( >99% purity by flow cytometry ) T-cell lines . To obtain fresh immune CD4+ T cells , at day 21 following immunization lymph node cells were purified by negative selection using magnetic beads ( CD4+ T-cell Isolation kit II , Miltenyi Biotec ) according to the manufacturer’s recommendations . To assess antigen-specific proliferation , either 105 purified CD4+ T cells , or 104 T-line cells were co-cultured with 2 x 105 mitomycin C-treated splenic APC in a 96-well flat-bottom plate ( Costar ) , at 37°C , 5% CO2 , in supplemented RPMI-1640 containing 10 μg/ml of H37Rv sonicate . Non-stimulated wells served as controls . Triplicate cultures were pulsed with 0 . 5 μCi [3H]-thymidine for the last 18 h of a 40 h incubation . The label uptake was measured in a liquid scintillation counter ( Wallac , Finland ) after harvesting the well’s contents onto fiberglass filters using a semi-automatic cell harvester ( Scatron , Norway ) . Peritoneal macrophages were obtained after stimulation with peptone as described previously [63] . 50 x 103 macrophages per well of 96-well plates in RPMI-1640 supplemented with 2% FCS and containing no antibiotics were infected with M . tuberculosis H37Rv at MOI 5:1 for 1 . 5 h . CD4+ T cells ( ~97% purity ) were obtained from spleens of ( B6 x B6 . I-9 . 3 . 19 . 8 ) F1 mice at day 21 after i . v . infection with 5 x 105 CFU of M . tuberculosis H37Rv using magnetic separation ( see above ) . T cells were added to infected macrophages at 1:1 ratio , and co-cultures kept for 72 h at 37°C in CO2 incubator . To assess mycobacterial viability , [3H]-uracil label was added for last 18 h of incubation , and the uptake assessed exactly as described in [63] . This method provides >99% correlation with CFU counting [63] . Infected B6 and I/St mice were euthanized by thiopental overdose , and lung cell suspensions were prepared using the methods described earlier [64] . Briefly , blood was washed out by repeated broncho-alveolar lavage with 0 . 02% EDTA-PBS with antibiotics , lungs removed , sliced into 1–2 mm3 pieces and incubated at 37°C for 90 min in supplemented RPMI-1640 containing 200 U/ml collagenase and 50 U/ml DNase-I ( Sigma , MO ) . Single cell suspensions obtained by vigorous pipetting were washed twice in HBSS containing 2% FCS and antibiotics . Suspensions of spleen and lymph node cells were obtained using routine procedures . Cells were incubated 5 min at 4°C with an anti-CD16/CD32 mAb ( BD Biosciences ) for blocking Fc-receptors and stained with FITC-anti-CD3 , APC-anti-CD8 and PerCP-anti-CD4 antibodies ( BD Biosciences ) . For intracellular IFN-γ staining , 1 . 5 × 106 cells were cultured in 24-well plates in the presence of 10 μg/ml mycobacterial sonicate for 48 h; GolgiPlug block ( 1 μl/ml; BD Biosciences ) was added for the last 10 hours . Cells were then stained with anti-IFN-γ mAb XMG1 . 2 ( BD Biosciences ) using the Cytofix/Cytoperm kit ( BD Biosciences ) . The expression of the Н2-Еα molecules on cell surface , discriminating Н2b ( Н2-Еα-negative ) and Н2j ( Н2-Еα-positive ) haplotypes was assessed using the PE-14-4-4S mAb ( BD Biosciences ) Cells were analyzed on BD Biosciences FACSCalibur flow cytometer using CellQuest and FlowJo software . The levels of cytokines in the lung tissue was measured individually in infected animals using whole-lung homogenates in 2 ml of sterile saline stored at –70°С before assessment . After thawing , debris was removed from the samples by centrifugation at 800 g , and cytokine levels in supernatants were assessed in an ELISA format using mouse OptEIA TNF-α Set , OptEIA IL-6 Set and OptEIA IL-5 Set ( BD Biosciences ) and mouse INF-γ Set ( Biolegend ) according to the manufacturer’s instructions . RNA was extracted from spleens using the SV Total RNA Isolation System ( Promega , USA ) and treated with DNase I ( AMPD1 , Sigma ) . Complementary DNA ( cDNA ) was synthesized with oligo-dT18 primers ( Thermo Scientific , Lithuania ) and M-MLV reverse transcriptase ( Promega , USA ) . Primer sequences for cloning were obtained from the Ensembl database ( version GRCm38 . p2 ) for the C57BL/6 strain . 5' ( forward ) primers ended at the start codon ( ATG ) ; reverse primers started at the ( TGA ) stop codon . Coding DNA was amplified with Advantage GC Genomic LA Polymerase ( Clontech , USA ) , PCR products were purified by gel extraction with Cleanup Mini Set ( Evrogen , Russia ) and cloned into the PCR-Script Amp Cloning vector using the PCR-Script Amp Cloning Kit ( Stratagene , USA ) or in pAL-TA ( Evrogen , Russia ) with preliminary 3 cycles of amplification of PCR products with Taq polymerase ( Helicon , Russia ) . The 4–6 positive clones were sequenced for each gene . Nucleotide sequences have been submitted to the GenBank ( http://www . ncbi . nlm . hih . gov/genbank ) under accession numbers KJ650201-KJ650234 , KJ663-713-KJ663725 . Molecular modeling was performed using an Octane2 workstation ( Silicon Graphics , USA ) equipped with the programs Insight II/Discover ( Accelrys , USA ) . Atomic coordinates of the mouse Class II MHC H2-Au MBP125-135 ( PDB id 2P24 ) [50] and Class II MHC H2-Ab in complex with the human CLIP peptide ( PDB id 1MUJ ) [54] were used for homology modeling . AA substitutions were deduced using the Biopolymer program . In order to minimize inter-atomic clashes , all individual AA conformations were chosen automatically , using the criteria of the lowest energy . To deduce the structure of the H2-Aj molecule using the H2-Ab template , the deletion P65-E66 was introduced manually using the Biopolymer program . Atomic coordinates available from two models , 1MUJ and 2P24 , for the Class II-CLIP structures were subjected to further refinement for the I-Aj using the Discover program and AMBER force field . Refinement stages included short energy minimization ( the steepest descent algorithm ) , followed by 1 ps molecular dynamics simulations at 298K and by the final energy minimization ( the conjugate gradient algorithm ) . Results were visualized using the Insight II and Discovery Studio software ( Accelrys , USA ) . All analyses were done using Graphpad Prism version 4 . Mortality was assessed using Kaplan-Meier survival analysis and the log-rank tests , CFU counts using Student’s t-test . P < 0 . 05 was considered statistically significant .
|
Many genes of the host regulate interactions with Mycobacterium tuberculosis and determine the level of susceptibility to , and severity of , tuberculosis ( TB ) . Identification of these genes and their alleles is continuing and contributes new knowledge about the host-pathogen interactions . So far , forward genetic approaches ( from phenotype to gene ) have identified several chromosomal segments involved in genetic control of TB in mice ( quantitative trait loci—QTL ) , but only one particular gene , Ipr1 , has been identified . Here , we report the identification of a second TB-controlling gene . On the basis of a pair of mouse inbred strains with polar susceptibility to TB infection ( susceptible I/St and more resistant C57BL/6 ) we established a panel of recombinant strains carrying small segments of Chromosome 17 from I/St on the genetic background of C57BL/6 . A combination of genetic mapping , gene sequencing , TB phenotypes assessment and immunological approaches demonstrates that the H2-Ab1 gene encoding the beta-chain of the Class II heterodimer H2-A determines susceptibility to TB infection . The importance of allelic polymorphisms in Class II genes encoding antigen-presenting molecules in susceptibility to infection has been suspected . This is the first prove of this role obtained by the methods of classical forward genetics .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
The QTL within the H2 Complex Involved in the Control of Tuberculosis Infection in Mice Is the Classical Class II H2-Ab1 Gene
|
Chinese Erhualian is the most prolific pig breed in the world . The breed exhibits exceptionally large and floppy ears . To identify genes underlying this typical feature , we previously performed a genome scan in a large scale White Duroc × Erhualian cross and mapped a major QTL for ear size to a 2-cM region on chromosome 7 . We herein performed an identical-by-descent analysis that defined the QTL within a 750-kb region . Historically , the large-ear feature has been selected for the ancient sacrificial culture in Erhualian pigs . By using a selective sweep analysis , we then refined the critical region to a 630-kb interval containing 9 annotated genes . Four of the 9 genes are expressed in ear tissues of piglets . Of the 4 genes , PPARD stood out as the strongest candidate gene for its established role in skin homeostasis , cartilage development , and fat metabolism . No differential expression of PPARD was found in ear tissues at different growth stages between large-eared Erhualian and small-eared Duroc pigs . We further screened coding sequence variants in the PPARD gene and identified only one missense mutation ( G32E ) in a conserved functionally important domain . The protein-altering mutation showed perfect concordance ( 100% ) with the QTL genotypes of all 19 founder animals segregating in the White Duroc × Erhualian cross and occurred at high frequencies exclusively in Chinese large-eared breeds . Moreover , the mutation is of functional significance; it mediates down-regulation of β-catenin and its target gene expression that is crucial for fat deposition in skin . Furthermore , the mutation was significantly associated with ear size across the experimental cross and diverse outbred populations . A worldwide survey of haplotype diversity revealed that the mutation event is of Chinese origin , likely after domestication . Taken together , we provide evidence that PPARD G32E is the variation underlying this major QTL .
The external ear is part of the auditory system and plays a vital role in collecting sound as the first step in hearing . Multiple congenital anomalies have been documented for human external ears . For instance , microtia , characterized by a small and abnormally shaped outer ear , occurs in approximately one in 8 , 000–10 , 000 births . However , only in a minority of cases has a genetic or environmental cause been found [1] . The domestic pig services as not only an agriculturally important animal for meat production but also an important large-animal model for human medicine [2] . Thousands of years of selective breeding has created diversity of phenotypes in pigs , such as ear size in Erhualian and White Duroc breeds . Erhualian is the most prolific pig breed and exhibits unusually large and floppy ears as breed character ( Figure 1 ) . Historically , the large-ear feature of Erhualian pigs had been favored by owners for the traditional sacrificial culture [3] . White Duroc is one of worldwide-popular boar line and has small and erect ears ( Figure 1 ) . We have created a four-generation White Duroc × Erhualian resource population , in which phenotypic traits related to ear size have been recorded in 1 , 027 adult F2 animals and 560 adult F3 individuals ( Table S1 ) . We mapped a major QTL for ear size around 58 cM on SSC7 ( Figure S1 ) using a genome scan on the White Duroc × Erhualian cross [4] , which confirmed the previously reported QTL affecting ear size in a Large White × Meishan F2 resource population [5] . The significant QTL had a small confidence interval of 2 cM and explained more than 40% of phenotypic variance . The aim of this study was to identify the genetic determinant underlying this major QTL .
To fine map the QTL , we genotyped 1 , 027 adult F2 animals and their 68 parents and 19 grandparents in the White Duroc × Erhualian cross using additional 17 SNP markers and 11 microsatellite markers in the QTL region . A final set of 33 markers covering the QTL region were then explored to deduce the QTL genotypes of F1 sires by the marker-assisted segregation analysis as proposed previously [6] . We determined QTL genotypes of all 9 F1 sires ( Figure S2 ) . All 9 Q-bearing chromosomes for increased ear size shared a haplotype of ∼1 . 2 Mb flanked by markers HMGA1 – TULP1 . The shared haplotype was distinct from q-bearing chromosomes ( Figure 2 ) . These observations strongly suggest that the QTL is located in the 1 . 2-Mb interval . Given the extremely divergent ear size phenotypes between Erhualian and White Duroc animals , we assumed that Q and q alleles were alternatively fixed in Erhualian and White Duroc founder animals; hence all Erhualian founder sows could share a chromosomal segment carrying the Q allele for increased ear size . To test this assumption , we reconstructed haplotypes of all 19 founder animals ( 2 sires and 17 dams ) using 50 markers ( 15 microsatellites and 35 SNPs ) in the QTL region . Almost all Erhualian founder sows shared a haplotype of ∼750 kb within the refined 1 . 2-Mb interval ( Figure 2 ) . As predicted , this shared haplotype was associated with increased ear size and presumably Q-bearing chromosomes . Two Erhualian founder sows carried a distinct haplotype ( denoted as Eq ) , which was unexpected because it was contrast with our initial assumption . We then conducted a statistical analysis of F2 animals in the White Duroc × Erhualian cross . The results revealed that the Eq chromosome had an effect on decreased ear size similar to the White Duroc chromosome ( Dq ) and significantly different from the Erhualian Q-bearing chromosome ( EQ ) . The least-squares means ( ± s . e . ) of ear weight were 323 . 07±4 . 55 for EQEQ and 266 . 66±18 . 9 for EQEq ( P = 0 . 04 ) ; 264 . 71±3 . 52 for DqEQ and 236 . 98±17 . 12 for DqEq ( P = 0 . 06 , Table 1 ) . The shared EQ chromosome allowed us to refine the location of the major QTL to the 750-kb interval between markers UHRF1BP1 and TULP1 ( Figure 2 ) . Historically , Erhualian pigs had undergone selection for ear size because pigs with extraordinary large and floppy ears were favored for the ancient sacrificial culture in the Taihu region of East China [3] . Reduced genetic variation in the critical region containing the QTL was therefore predicted . To define the region of reduced genetic variation , we collected 211 animals representing all lineages in 3 Erhualian nucleus populations , 216 animals from 6 Chinese indigenous breeds and 119 independent animals from 3 Western worldwide-popular commercial breeds . Using these samples , we genotyped 6 microsatellite and 32 SNP markers in the 750-kb region . We found that 18 adjacent markers in a 630-kb region between markers UHRF1BP1 and FANCE showed dramatically reduced polymorphisms in all Erhualian pigs with nearly all major allele frequencies of more than 0 . 90 . Notably , the 18 markers in the 630-kb region are monomorphic in the Erhualian nucleus population from Xishan county ( n = 72 ) . In comparison , the genetic polymorphisms of these markers were maintained in other Chinese , Western breeds , and wild boars ( Figure 3 ) . The 630-kb region showing strong selective-sweep effects on Erhualian pigs was therefore predicted to contain the responsible locus . We further genotyped the 18 markers in the 630-kb region on 188 adult animals of Sutai pigs . This breed was developed after 18-genereation selection from a Duroc ( 50% ) × Erhualian ( 50% ) cross in 1986 [7] , meaning that the breed has undergone 18 generations of meiosis reducing the extent of linkage disequilibrium between QTL and linked markers . The Erhualian-originated haplotype of 630 kb showed significant ( P = 0 . 009 ) association with increased ear size compared with other chromosomes in Sutai pigs ( Figure S3 ) , thereby supporting the conclusion that this region harbors the causative gene . The 630-kb region encompasses 9 annotated genes ( ANKS1A , DEF6 , FANCE , PPARD , SCUBE3 , TAF11 , TCP11 , UHR1BP1 and ZNF76 ) in the human homologous region . RT-PCR was performed to detect expression levels of these genes in ear tissues of piglets . Four genes including PPARD , FANCE , TAF11 and ZNF76 were highly expressed , whereas transcripts of other genes were almost absent in ear tissues ( data not shown ) . Of the 4 genes , PPARD ( peroxisome proliferator-activated receptor delta ) is a ligand-modulated transcription factor belonging to the nuclear receptor superfamily and plays crucial roles in diverse biologically important processes [8] . For instance , PPARD play a pivotal role in modulating cell differentiation in both keratinocytes and sebocyte of skin [9] . PPARD also serves as a key regulator in fat metabolism; it triggers fat burning and enhances energy uncoupling in adipose tissues and skeletal muscle [10]–[12] . Moreover , PPARD is a key player in Wnt/β-catenin pathway [13] , which has essential roles in diverse cellular activities including chondrocyte proliferation and differentiation [14] . The external ear is composed of skin , cartilage , connective tissues and fat . Given its crucial role in skin homeostasis , cartilage development and fat metabolism , PPARD stood out as a prime positional candidate for the major QTL . We monitored the relative mRNA expression of PPARD in ear tissues of Erhualian and Duroc pigs at four different ages by real-time RT-PCR . The expression levels were higher in samples at early ages ( days 0 , 45 and 90 ) compared with adult samples ( day 300 ) . However , no significant difference of expression levels was found in ear tissues between large-eared Erhualian and small-eared Duroc pigs ( Figure S4 ) . To search for causative mutations , we first sequenced the entire coding region of the PPARD gene using ear mRNA of two White Duroc and two Erhualian animals and identified only one nonsynonymous mutation . The G to A mutation caused a glycine to glutamic acid substitution at codon 32 ( GU565977 ) in the conserved intrinsically disordered domain of the PPARD protein predicted by SMART ( http://smart . embl-heidelberg . de/ ) . The intrinsically disordered domain is a distinctive and common characteristic of eukaryotic hub proteins like multifunctional nuclear receptors and serves as a determinant of protein interactivity [15] . Comparison of amino acids of this protein domain across mammals revealed that glycine is well conserved in mammalian PPARDs ( Figure 4 ) , while the derived glutamic acid occurs only in alleles increasing ear size in pigs . We thus speculated that the nonconservative substitution probably changes the PPARD interactivity with other protein partners and consequently affects the gene's regulation function . Genotypes of F1 sires ( 9 heterozygotes ) and F0 animals ( 17 homozygotes and 2 heterozygotes ) at the mutation site were 100% concordance with their QTL genotypes . The potentially altered function and QTL concordance of PPARD G32E corresponded to the hypothesis that this SNP may be the causative mutation underlying the major QTL . PPARD is involved in the Wnt/β-catenin signaling pathway that regulates diverse cellular functions . In the nucleus , PPARD interacts with β-catenin binding to TCF/LEF transcription factors that stimulate transcription of target genes important for multiple cellular activities including cartilage development and organogenesis [14] . To demonstrate functional significance of PPARD G32E , we cotransfected the 293T cells with the lentiviral expression vectors of wild-type or mutant PPARD and a TCF/LEF-driven luciferase reporter construct . A Renilla luciferase expression vector was used for the normalization of transfection efficiency . Overexpression of mutant PPARD led to a 40% decrease ( P<0 . 05 compared with the wild-type treatment ) in TCF/LEF reporter activity ( Figure 5A ) , indicating the G32E mutation mediates down-regulation of β-catenin downstream genes . To examine a direct functional role of PPARD G32E in target genes of β-catenin , we treated pig ear-derived primary fibroblast cells with the lentiviral PPRAD expression vectors and monitored the mRNA levels of β-catenin and its known downstream ( c-myc ) [16] and upstream ( Sox9 ) [17] genes along with GAPDH as a loading control by real time quantitative RT-PCR . The mRNA levels of β-catenin and c-myc were reduced respectively by 4 . 1-fold and 11 . 5-fold ( P<0 . 001 ) in mutant PPARD transfectants compared with the cells transfected with wild-type PPARD . Western blot analysis showed that both β-catenin and c-myc protein levels were decreased by the mutant PPARD treatment ( Figure 5B ) , thereby confirming the results of mRNA and luciferase reporter analyses . Sox9 mRNA expression in mutant PPARD transfectants was only slightly decreased to 1 . 1-fold of the wild-type PPARD treatment; the result was validated by Western blot ( Figure 5B ) . GAPDH was used as a protein loading control for total cell lysate , which was not affected by both wild-type and mutant PPARD treatments ( Figure 5B ) . Altogether , we conclude that PPARD G32E is a functional variant that mediate down-regulation of β-catenin and its target gene expression in the Wnt/β-catenin signaling pathway . Wnt/β-catenin signaling has been firmly demonstrated to suppress adipogensis [18]–[19] . The fact that PPARD is a key modulator of lipid production in the skin [9] and that PPARD G32E inhibits β-catenin expression led us to assume that the mutation stimulates lipid production and storage that are required for enlarged ear size . To confirm the effect of PPARD G32E on ear size , we performed a standard association test , a marker-assisted association test and an F-drop test [20] in the White Duroc × Erhualian cross . The SNP showed greatly significant ( P<0 . 0001 ) association with ear weight and ear size in the standard association test . In the marker-assisted association test , the SNP was more significant ( P<0 . 001 ) for these traits compared with the QTL effect . After fitting this polymorphism in the QTL model , the great QTL effect disappeared with F-value drop rations of less than 0 . 03 ( Table S2 ) . These results were in agreement with the hypothesis that the SNP is the causative mutation for the major QTL affecting ear size . Nevertheless , we cautioned the results because variants closely linked with a causative mutation also lead to strong association in F2 resource populations due to the high level of linkage disequilibrium between founder breeds [20] . To obtain additional supporting evidence , we further genotyped the G32E mutation on 667 mature pigs from 4 Chinese local breeds ( Erhualian , Hang , Yushan Black and Bama Xiang ) and 3 synthetic commercial lines ( Sutai , Suzhong , Sujiang ) with phenotypic data of ear size . These populations show a wide range of ear size and segregate for the mutation . The association analyses confirmed the effect of PPARD G32E on ear size . The 32E allele was significantly associated with increased ear size across the tested breeds ( P<0 . 05; Table 2 ) . Chinese local pig breeds have low levels of linkage disequilibrium extending up to only 0 . 05 cM [21] . The concordantly significant association across Chinese breeds thereby strengthened the hypothesis that PPARD G32E is the responsible locus for ear size . The effects of PPARD G32E differ in their magnitude in the tested breeds; one reason is that the effects are context-dependent and are influenced by different genetic backgrounds and environments . Another possibility is that PPARD G32E is only responsible for part of the effect on ear size in Erhualian pigs . To reveal the ancestral state and allele frequency of PPARD G32E in diverse pig breeds , we genotyped the mutation in a panel of 1 , 166 animals representing 31 domestic breeds and Chinese and European wild boars . Overall , the derived 32E allele for increased ear size occurred at high frequencies ( >0 . 80 ) in Chinese breeds with large and floppy ears . In contrast , the 32G allele for normal ear size was fixed in all wild boars , European local and commercial breeds , and occurred at low frequencies ( <0 . 30 ) in Chinese indigenous breeds having small and erect ears . These results indicated that PPARD G32E may occur in Chinese pigs after domestication . We detected only one heterozygote in European local breeds ( Table 3 ) . The animal was from Large Black pigs that exhibit large and floppy ears and have been influenced by Chinese breeds brought into England in the late 1800′s [22] . We further analyzed the genetic variability and haplotype structure around the G32E mutation in a worldwide pig panel . A total of 868 animals representing 34 breeds were genotyped for 32 SNPs in a 77-kb region of PPARD . Again , the Erhualian breed showed a selective sweep signal as it had negative classical selection statistics Tajimas D and much smaller nucleotide variability ( πN ) compared with other Chinese local breeds and Western commercial breeds ( Table S3 ) . Especially , the genetic variability at the 32 loci was wiped out in the Erhualian population from Xishan . Moreover , we plotted a distribution of the frequency of the derived 32E allele ( PA ) against Tajimas D index to elucidate the existence of directional selection for the G32E mutation . When PA = 0 , Tajimas D was highly variable across breeds , likely due to demographic and/or sampling effects . In stark contrast , Tajima's D took highly negative values when PA >0 . 8 in Erhualian and other Chinese large-eared breeds as expected in a classical directional selection ( Figure S5 ) . We reconstructed 16 major haplotypes with frequencies larger than 0 . 01 from the 32 SNPs genotyped . Of the 16 haplotypes , only one carried the derived 32E allele; it was at high frequencies in Erhualian pigs and intermediate frequencies in some floppy-eared Chinese breeds whereas absent in Western pigs and wild boars ( Table 4 ) . The NJ phylogenetic tree illustrated that the typical haplotype of Erhualian pigs was generally divergent from other haplotypes ( Figure S6 ) . These observations supported the assumption that the G32E mutation has a unique origin in Chinese breeds likely after domestication and has undergone selection in Erhualian pigs . We calculated linkage disequilibrium measures ( r2 ) between all pairs of loci and inferred haplotype blocks . Three and two haplotype blocks were identified in the PPARD region for Chinese indigenous pigs and Western commercial breeds , respectively . Only a single nevertheless larger block that spanned 53 kb and contained the G32E SNP was found in Erhualian pigs , reflecting a selection hitching effect ( Figure S7 ) . The G32E SNP was in high disequilibrium with very few of the SNPs analyzed ( two with r2>0 . 8 ) , and there was no observable trend between physical distance and disequilibrium measures for the G32E SNP and the rest of loci ( Figure S8 ) . The elucidation of the genetic basis of multifactorial traits in domestic animals is still a big challenge , and few successful examples have been reported [23]–[27] . In this study , a battery of genetic and functional assays obtained diverse pieces of supporting evidence that the PPARD G32E substitution underlies the major QTL effect on ear size on SSC7 . ( 1 ) The shared haloptypes of 9 F1 sires segregating for the QTL spanned a region of ∼1 . 2 Mb containing PPARD . ( 2 ) All Erhualian founder chromosomes shared a ∼750 kb segment spanning PPARD that were associated with the Q allele for increased ear size . ( 3 ) Erhualian pigs showed an obvious selective sweep signal in a 630-kb region encompassing PPARD; the signal was concordant with the breeding history of the breed . ( 4 ) The 630-kb haplotype showed similar QTL effect on increased ear size in Sutai pigs that were developed after 18-generation selection in the Erhualian × Duroc cross . ( 5 ) Of the 4 genes expressed in ear tissues within the critical region , PPARD stood out a prime candidate for its established essential roles in skin homeostasis , cartilage development and fat metabolism . ( 6 ) Only one missense mutation ( G32E ) was identified in PPARD using White Duroc and Erhualian founder animals . The mutation caused a nonconservative amino acid change at the conserved intrinsically disordered domain and was of functional significance . ( 7 ) The G32E SNP was concordant with QTL genotypes of F0 and F1 animals in the White Duroc × Erhualian cross . ( 8 ) The G32E SNP showed strikingly significant association with ear size across the experimental cross and diverse outbred populations . ( 9 ) The derived allele for increased ear size occurred at high frequencies only in Chinese floppy-eared breeds . Altogether , these data led us to conclude that G32E in the PPARD gene has an important contribution to ear size in pigs . The results establish , for the first time , a direct and novel role of PPARD in ear development and may be of relevance for the pathogenesis of external ear abnormalities in humans . The genomic region harboring PPARD G32E is of great interest in pig genetics , because significant QTL for diverse traits related to growth , carcass length , skeletal morphology and fat deposition have been consistently evidenced in the region using the current resource population and different crosses between Chinese Meishan and commercial breeds [28]–[33] . The overlapping QTL for multiple traits in the region led us to assume that there might be a single critical gene having pleiotrophic effects on these traits . We herein showed the causality of PPARD G32E for the QTL affecting ear size in the critical region . Given that PPARD serve as a crucial and multifaceted determinant of diverse biological functions including fat metabolism , cartilage development , chondrocyte proliferation and differentiation [8] , [10]–[14] , we thus speculate that PPARD is a strong candidate of the multiple significant QTL on SSC7 and that PPARD G32E might have pleiotropic effects on growth , carcass and fatness traits in pigs . Further investigations will be performed to validate the hypothesis in the future .
All animal work was conducted according to the guidelines for the care and use of experimental animals established by the Ministry of Agriculture of China . Microsatellite markers in the mapped interval were mined from the pig genome assembly ( Build 9 . 2 ) at http://www . ensembl . org/Sus_scrofa/Info/Index and were genotyped using standard procedures . Primers for amplification of microsatellite markers are given in Table S4 . QTL genotypes of F1 boars in the White Duroc × Erhualian intercross were determined by marker-assisted segregation analysis as described previously [6] . Briefly , a Z-score was calculated for each F1 sire; the score is the log10 of the H1/H0 likelihood ratio where H1 assumes that the boar is heterozygous at the QTL ( Qq ) , while H0 postulates that the boar is homozygous QQ or qq . Boars were considered to be Qq when Z >2 , QQ or qq when Z <−2 , and of undetermined genotype if −2<Z<2 . The pedigree and management of the intercross population with phenotypic data of ear size have been described elsewhere [4] . Haplotypes of founder animals were reconstructed with the SimWalk2 program . To detect the effects of a putative selection sweep on the genetic variation in Erhualian pigs compared with control animals , we analyzed the microsatellite and SNP genotypes of 211 Erhualian pigs and 335 control animals representing 10 different breeds ( Hetao Large-Ear: 56; Laiwu: 32; Yushan Black: 31; Wuzhishan , 32; Dianan Small-Ear: 31; Tibetan: 34; White Duroc: 12; Duroc: 29; Large White: 39; Landrace: 39 ) . SNP markers were genotyped using the ABI SNapshot protocol or PCR-RFLP assays . All primers are given in Table S4 . Total RNA was extracted from pig tissues using the Rneasy Fibrous Tissue Mini Kit ( Qiagen ) . To analyze expression of candidate genes in ears , products from the first strand-complementary DNA synthesis ( TaKaRa ) were amplified with primers given in Table S5 . The quantification of the PPARD transcripts was performed by the comparative Ct method ( 2−ΔΔCt ) using the primers and TaqMan probes shown in Table S5 . Real-time PCR was done with the Universal PCR Master Mix using an ABI7900 instrument ( Applied Biosystem ) . All samples were analyzed in triplicate . The β-actin gene was used as the internal reference gene . The entire coding region of porcine PPRAD was re-sequenced using ear mRNA of two White Duroc and two Erhualian animals . Primer pairs listed in Table S6 were used to generate overlapping PCR amplicons . All PCR products were purified using the NucleoSpin Extract II kit ( Macherey-Nagel ) and sequenced using the same primers . The sequence traces were assembled and analyzed for polymorphisms using the SeqMan program ( DNASTAR ) . The PPARD G32E mutation was genotyped using the ABI SNapshot protocol . A 385-bp DNA fragment was amplified with the F2/R2 primer pairs ( F2: 5′-CGG CTG TTT TAC AGG AAG GA-3′; R2: 5′- CTG CAC TCA GAC CCA GAT GA-3′ ) . SNapshot reactions were performed with Multiplex Ready Reaction Mix ( Applied Biosystem ) and an extension primer ( 5′-TTT TTT TTT TGC TGG AGG GAA GCG AGT GCT CTG GT -3′ ) using an ABI 3130XL DNA Analyzer ( Applied Biosystem ) . The coding region of pig PPARD was amplified with primers PPARD-Age-I-F ( 5′- GAG GAT CCC CGG GTA CCG GTC GCC ACC ATG GAG CAG CCG CCG GAG-3′ ) and PPARD-Age-I-R ( 5′- TCA TCC TTG TAG TCG CTA GCG TAC ATG TCC TTG TAG-3′ ) . The amplified cDNA was gel-purified and digested with AgeI and NheI ( NEB ) . The restricted fragments were cloned to pGC-FU-EGFP-3FLAG lentiviral expression vector ( Genechem ) . The sequence and orientation of the insert were verified by DNA sequencing . The expression of His-tagged PPARD in cultured cells was confirmed by Western blot analysis with anti-His antibody . The human 293T cells were infected with the lentiviral expression constructs of pig wild-type and mutant PPARD . The infected cells were seeded at a concentration achieving 80% confluence in 96-well plates 18 h before transfection . The cells were transiently transfected with TCF/LEF-Luc reporter vector ( Cignal , SAB ) along with a control Relina luciferase vector using Lipofectamine plus reagent . The cell lysates were obtained with 1× reporter lysis buffer ( Promega ) 48 h after transfection . The luciferase activity was assayed in a Berthold Auto Lumat LB953 luminometer ( Nashua , NH ) by using the luciferase assay system from Promega . The relative luciferase activity was normalized to the Relina luciferase activity in each sample . The pig ear-derived fibroblast cells were transfected with pGC-FU-EGFP-3FLAG lentiviral expression vector ( Genechem ) . Five days post-transfection , 1×106 cells were harvested for qPCR and Western blot analysis . Total RNA was extracted from harvested cells using Trizol ( Invitrogen ) . Two µg of total RNA was synthesized into cDNA with M-MLV reverse transcriptase ( Promega ) and oligo d ( T ) . Real time PCR was performed on the cDNA using the SYBR Premix Ex Taq ( TaKaRa ) and primers listed in Table S7 in a TP800 Real Time System ( TaKaRa ) . The quantification of transcripts was performed by the comparative Ct ( 2−ΔΔCt ) method . All values were reported as mean ± S . D . of triplicate assays of each cDNA sample . Rabbit anti-PPARD ( Sigma ) , mouse anti-β-catenin ( Abcam ) , rabbit anti-c-myc ( Cellsignaling ) , mouse anti-Sox9 ( Abcam ) and mouse anti-GAPDH ( Santa Cruz ) antibodies were used in Western blots in a routine way . The specific immunoreactive bands were visualized using an ECL plus kit ( GE Healthcare ) and quantified with the Molecular Imaging Software ( Kodak ) . The entire White Duroc × Erhualian resource population was genotyped for the PPARD G32E mutation . Association of the mutation with ear size and weight was evaluated using standard association , marker-assisted association and F-drop test as described previously [20] . Association analyses were also performed on 667 animals representing 7 different breeds . Photographs were taken for one ear of each animal after the ear was fixed and covered with a ruler as an internal reference of the size . Ear size was calculated using the Qwin software ( Laica ) . Significance was evaluated by the t-test in the GLM procedure of SAS 9 . 0 . Genomic DNA pools of White Duroc ( n = 2 ) and Erhualian ( n = 2 ) animals were amplified with primers given in Table S6 . All PCR products were purified with the Qiagen protocol and sequenced using the same PCR primers , revealing a subset of SNP markers in the genomic region of porcine PPARD . SNP markers were genotyped by iPLEX SEQUENOM MassARRAY platform . SNP genotype calls were filtered and checked manually , and aggressive calls were omitted from the dataset . Population genetics parameters including the mean number of pairwise differences across loci ( πN ) , Tajimas D , Fu and Li's D were estimated with DnaSP v5 [34] . Haplotypes were reconstructed with PHASE v2 [35] . Haplotype phylogenetic tree based on p-distance were drawn using MEGA4 [36] . The Haploview v4 . 1 program [37] was used to calculate linkage disequilibrium measures ( r2 and D' ) and to identify haplotype blocks .
|
A central but challenging objective in current biology is to dissect the genetic basis of quantitative traits . Numerous quantitative trait loci ( QTL ) have been uncovered in model and farm animals , providing unexpected insights into the biology of complex traits . However , only a few causal variants underlying the QTL have been explicitly identified . By using a battery of genetic and functional assays , we herein show that a major QTL effect on pig ear size is most likely caused by a single base substitution in an evolutionary conserved region of the PPARD gene . The protein-altered mutation is of functional significance and explains a proportion of variation in ear size across diverse pig breeds . A worldwide survey showed that the mutant allele for increased ear size was derived from a common ancestor in Chinese pigs , likely after domestication . These findings establish , for the first time , an essential role of PPARD in ear development and highlight the great potential of naturally occurring mutations in farm animals to gain insights into mammalian biology . Moreover , the knowledge of the PPARD causal mutation adds to the limited list of quantitative trait genes and quantitative trait nucleotides characterized in domesticated animals .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Methods"
] |
[
"animal",
"genetics",
"heredity",
"genetics",
"animal",
"management",
"biology",
"quantitative",
"traits",
"genetic",
"determinism",
"genetics",
"and",
"genomics",
"agriculture"
] |
2011
|
A Missense Mutation in PPARD Causes a Major QTL Effect on Ear Size in Pigs
|
Among animal models of schistosomiasis , the rhesus macaque is unique in that an infection establishes but egg excretion rapidly diminishes , potentially due to loss of adult worms from the portal system via shunts or death by immune attack . To investigate this , six rhesus macaques were exposed to Schistosoma mansoni cercariae and the infection monitored until portal perfusion at 18 weeks . Despite a wide variation in worm numbers recovered , fecal egg output and circulating antigen levels indicated that a substantial population had established in all animals . Half the macaques had portal hypertension but only one had portacaval shunts , ruling out translocation to the lungs as the reason for loss of adult burden . Many worms had a shrunken and pallid appearance , with degenerative changes in intestines and reproductive organs . Tegument , gut epithelia and muscles appeared cytologically intact but the parenchyma was virtually devoid of content . An early and intense IgG production correlated with low worm burden at perfusion , and blood-feeding worms cultured in the presence of serum from these animals had stunted growth . Using immunoproteomics , gut digestive enzymes , tegument surface hydrolases and antioxidant enzymes were identified as targets of IgG in the high responder animals . It appears that worms starve to death after cessation of blood feeding , as a result of antibody-mediated processes . We suggest that proteins in the three categories above , formulated to trigger the appropriate mechanisms operating in rhesus macaques , would have both prophylactic and therapeutic potential as a human vaccine .
Schistosomiasis remains a major public health problem in the Tropics , with tens of millions infected and many more at risk [1] . It has been estimated that greater than 250 , 000 deaths per annum are directly attributable to the disease [2] , and the subtle morbidities associated with chronic infection have a more serious impact than hitherto credited [3] . Treatment relies on a single drug ( praziquantel ) to eliminate the adult worms but , as this has no prophylactic properties and is ineffective against larval schistosomes [4] , a vaccine would augment efforts to control and ultimately eradicate the disease . Once established in the human portal tract adult Schistosoma mansoni are long-lived [5] , revealing their ability to deploy effective immune evasion strategies . In pre-pubertal children there is little evidence for immune-mediated prevention of worm recruitment , as a result of which the prevalence and intensity of infection rise gradually with age [6] . Even in those adults who are apparently resistant to reinfection , suggesting the development of acquired immunity , no mechanisms have been defined on which a vaccine might be based [7] . The difficulties inherent in research on human schistosomiasis have entailed the use of laboratory animal models , with some early studies being undertaken in the rhesus macaque ( Macaca mulatta ) [8]–[10] . In this species , exposure to a moderate number of cercariae elicited protection against a challenge given four to five months later while the adult worms that engendered the immune response were apparently unaffected [9] . By analogy with tumor transplantation , the term concomitant immunity was proposed as an explanation [11] . Resistance to challenge was also demonstrated in mice with a chronic S . mansoni infection [12] but was subsequently shown to be an artefact of pathology , not immune-mediated killing [13] . The porta-caval shunts that developed in mice as a result of egg-induced hepatic pathology prevented challenge larvae from establishing by providing them with an escape route from the portal to the pulmonary vasculature , and even permitted adult worms from the primary infection to exit and pass to the lungs [13] . A salient feature of the rhesus macaque host is that an infection becomes patent but , above a threshold worm burden , egg output declines over the ensuing weeks to months [10] , [14] . As in the mouse , such a decline might be explained by the escape of worms through developing porta-caval shunts , meaning that they could no longer deposit eggs in the intestinal wall . Conversely , if shunts do not develop in the rhesus macaque , the observed decline in egg output with time could reflect immune elimination of the primary worm burden . We sought to establish which of these two contending hypotheses provided the most likely explanation . A clear demonstration of anti-adult worm immunity would open a new route towards the elusive goal of a schistosome vaccine , potentially one with therapeutic properties .
The study used six adult female rhesus macaques ( mean age 15 . 8±7 . 1 years , mean weight 5 . 0±1 . 2 kg ) from the colony at the Biomedical Primate Research Centre ( BPRC ) , Rijswijk , The Netherlands . The experimental protocol was approved by the Institutional Animal Care and Use Committee at BPRC and the Biology Department Ethics Committee , University of York . Animals were exposed to 1000 S . mansoni cercariae ( Puerto Rican isolate; Department of Parasitology , Leiden University Medical Centre , The Netherlands ) via the shaved abdominal skin for 30 minutes , under tiletamine-zolazepam ( Zoletil; Virbac , Barneveld , NL ) anaesthesia supplemented with ketamine hydrochloride ( Alfasan International , Woerden , NL ) . Serum was obtained from finger prick bleeds at 2-wk intervals between weeks 6 and 14 , and by intravenous sampling prior to infection and at perfusion ( wk 18 ) . Fecal samples were collected overnight at 1 or 2 wk intervals from wk 6 . Eggs per gram of feces was determined from three individual samples/animal/time point , using the Percoll technique [15] . Soluble circulating anodic antigen ( CAA ) , released into the bloodstream from the parasite's gut , was detected by ELISA using specific monoclonal antibodies [16] . Animals were anaesthetised as above and laparotomized to measure portal blood pressure via the superior mesenteric vein , followed by the injection of 15×106 carbonised inert microspheres ( 15 µm diameter ) to measure the extent of porta caval shunting , both as previously described for rodents [17] . Animals were maintained under deep anaesthesia for 15 minutes before perfusion , after which the lungs and liver were weighed and samples retained . The fractional distribution of microspheres in lungs and liver+perfusate was then estimated by counting aliquots of tissues digested overnight in 5% KOH at 60°C [18] . Rhesus macaques received an intravenous injection of heparin followed by an overdose of anaesthetic before ligaturing the aorta and vena cava to isolate the portal vasculature . Portal perfusion was performed as described for the olive baboon [19] , and the worms were fixed in formal saline after counting under a stereomicroscope . Reference worms were recovered from C57BL/6 mice ( University of York Animal Facility ) seven weeks after exposure to 200 cercariae , and fixed in formal saline . Worms were prepared for confocal microscopy by post-fixation in AFA ( alcohol/40% formaldehyde/glacial acetic acid , in the ratio 85∶10∶5 ) , staining for 30 min in Langeron's Carmine [20] , differentiation in 70% acid alcohol , clearing , and mounting in DPX ( VWR International Ltd ) . Optical slices were obtained with a Bio-Rad MRC-1000 confocal microscope , with excitation at 514 nm from a 25 mW argon ion laser and a 585 nm long pass emission filter . Kalman averaging was carried out over 12 frames . For electron microscopy , worms were post-fixed in 2 . 5% glutaraldehyde/4% paraformaldehyde in 100 mM PBS , pH 7 . 2 , at 4°C overnight [21] . Mechanically transformed schistosomula [22] were cultured in M169 medium ( Invitrogen , Paisley , Scotland ) supplemented with 5% normal rhesus serum , 1% glutamine and 1% penicillin/streptomycin in 24-well plates , 95% O2∶5%CO2 , 37°C for 4 days . Medium was then replaced and 1% rhesus erythrocytes added to cause transformation to blood-feeding liver-stage worms . On day 8 , they were pooled and re-plated in 50% M169∶50% test or control serum plus 1% erythrocytes , with a medium change on day 16 . Test sera were pooled from the two animals with highest and the two with lowest worm burdens at perfusion . Control serum and erythrocytes from an uninfected rhesus macaque were the gift of Covance Laboratories , Harrogate , UK . On day 18 , the three worm populations were photographed on an Optiphot-2 microscope ( Nikon , Kingston , UK ) at ×100 with a TK-1070E camera ( JVC , London , UK ) , and the mean cross-sectional surface area of individuals ( n = 28 to 48 ) measured as an index of growth . Significance was determined by paired sample t test against the uninfected serum control . Levels of IgM and IgG antibodies against soluble adult worm proteins ( SWAP ) were determined by ELISA , as described previously [23] . Wells were probed with alkaline phosphatase-conjugated rabbit anti-monkey IgG ( Sigma-Aldrich , Poole , UK ) or goat F ( Ab' ) 2 anti-human IgM ( Biosource International , Nivelles , Belgium ) , diluted 1∶2000 , colour developed using p-Nitrophenyl phosphate substrate ( Sigma ) and absorbance read at 405 nm . Total IgE concentrations were determined by ELISA [24] on plates coated with 1∶2000 rabbit anti-human IgE ( DakoCytomation , Ely , UK ) and IgE binding detected with a 1∶1000 dilution of peroxidase-conjugated rabbit anti-human IgE ( DakoCytomation ) . A rhesus macaque serum was calibrated against a human serum immunoglobulin standard ( The Binding Site , Birmingham , UK ) . A stimulated adult worm secretory preparation ( SASP ) , derived predominantly from the worm gut , was generated as described [25] , apart from the final concentration step to <1 ml volume by centrifugation at 3000 rpm , 4°C using Vivaspin 20 µl 5 K MWCO filter units ( VWR International Ltd , Lutterworth , UK ) ; a 20 µl aliquot of protease inhibitor cocktail ( Sigma ) was added before storage at −20°C . A tegument surface preparation ( TSP ) was generated by incubating adult worms for 1 h in RPMI 1640 ( Invitrogen ) plus 10 mM HEPES containing phosphatidyl inositol phospholipase C ( Sigma ) at 1 . 25 units/ml . The culture supernatant was concentrated and protease inhibitor cocktail added before storage at −20°C . The protein composition of these two preparations , analysed by tandem mass spectrometry ( MS ) , will be reported elsewhere ( Hall , S . L . et al . , Borges , W . C . et al . , manuscripts in preparation ) . Concentrated supernatants were buffer-exchanged using 25 mM Tris , pH 7 . 5 before separation of 30 µg of protein on pre-cast mini 2D gels [26] . For TSP , 1D separations were performed on 1 . 0 mm 4–12% NuPAGE Bis-Tris gels ( Invitrogen ) [26] . ID and 2D separations were transferred to PVDF membranes [26] , for probing with a 1∶500 dilution of rhesus macaque sera , before incubation with 1∶1000 alkaline phosphatase-conjugated rabbit anti-rhesus IgG antibody ( Sigma ) , both for 1 . 5 h at RT , and development using 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium ( Sigma ) . Antibody targets were identified by matching Western blots to gels separated under identical conditions , by visual inspection or Phoretics Evolution Software ( Non-linear Dynamics , Newcastle , UK ) .
CAA and fecal egg output were monitored , as indirect indicators of the status of the schistosome infection ( Figure 1 ) . The production of CAA rose steadily as worms began blood feeding and reached maturity , with mean values peaking at week 8 before declining to a plateau from week 12 onwards . Eggs were first detected in the feces at week 7 , with mean numbers peaking at week 9 before declining towards the baseline; in two cases egg excretion reached zero by the end of the study . Although data from individual animals showed a wide variation for the two indirect indicators ( Table 1 ) , their peak values revealed that appreciable numbers of parasites established in all rhesus macaques . However , there was a large discrepancy in the number of worms recovered at perfusion ( range 12–708; Table 1 ) , signifying that many had been lost from the portal system in some animals . The declining values of the indirect indicators might reflect worm death or a more subtle deterioration in physiological function . In fact , when worms were counted a large proportion , particularly the females , had a pallid appearance as they lacked hematin pigment in their intestines . Many females were also shrunken compared to the mature equivalent from mice , with bodies approximately two-thirds the normal width . Confocal microscopy of the pallid female worms ( n = 15 ) revealed that the intestinal epithelium was thinner and its luminal surface lacked the numerous lamellar extensions seen in normal female worms from mice ( n = 9; Figure 2A cf . B ) . A decline in reproductive function of the pallid females was reflected by the greatly shrunken appearance of the ovary ( Figure 2C ) that contained very few oocytes compared to normal worms ( Figure 2D ) . The vitelline lobules were strikingly absent in the pallid worms ( Figure 2A cf . normal worms Figure 2B ) , as were forming eggs in the ootype ( data not shown ) . The degeneration of the reproductive system was not confined to females . Pallid males ( n = 7 ) had reduced numbers of spermatocytes in their testicular lobules , giving the contents a shrunken appearance with lacunae ( Figure 2E ) versus the tight packing of spermatocytes and maturing spermatozoa in the normal testis ( n = 4; Figure 2F ) . A few females of normal appearance had an ovary with healthy oocytes , a forming egg in the ootype and sperm present in the proximal part of the oviduct that acts as a receptaculum seminis; this last is indicative of functionally active male worms . These degenerative changes were further highlighted by an ultrastructural comparison . The cellular organization in the posterior half of the female body , comprising numerous individual vitelline cells packed with peripheral egg-shell precursor granules in worms from mice ( Figure 3B ) , was absent in pallid worms from rhesus macaques ( Figure 3A ) . The parenchyma appeared almost completely devoid of content , including glycogen and lipid stores , although individual cell membranes and isolated nuclei were evident . The absence of food in the gut lumen contrasted with the abundance of SEi-digested blood in the normal worm gut ( Figure 3A cf . B ) . Electron microscopy confirmed the degenerative state of the intestinal epithelium in pallid worms . Although plasma membranes were intact , the cytoplasm was less dense , nuclei were more rounded , and endoplasmic reticulum and Golgi apparatus were lacking , implying little or no protein synthesis ( Figure S1A ) . In contrast , the surface syncitial tegument of the parasite had a normal pitted appearance with the usual cytoplasmic inclusions ( Figure S1B ) . The subtegumental circular and longitudinal muscles that make up the body wall , together with the muscles surrounding the gut epithelium , also appeared relatively intact . Egg-induced changes to the portal vasculature were investigated as a potential cause of worm loss , using hepatic portal blood pressure and the extent of porta-caval shunting as indicators . Values for the former ranged from 8 . 1 to 32 . 3 cm H2O ( Table 1 ) , and portal shunting was negligible with the exception of one animal where 96% of the microspheres was recovered from the lungs ( Table 1 ) . There was no correlation between the worm burden , portal pressure and the extent of shunting . Collectively , these data indicate that the integrity of the portal system had not been compromised by egg deposition . We investigated whether the declining physiological status of worms was correlated with the humoral immune response . IgM , probed with SWAP , peaked at week 8 before gradually declining over the period of worm deterioration ( Figure 4A ) whereas IgG levels rose to a sustained plateau , with considerable variation between individual animals ( Figure 4B ) . The values for IgG level at weeks 8 and 18 , plotted against final worm burden , fell into 3 groups ( Figure 4C ) . In the two animals with lowest worm recoveries , levels were already high at week 8 and remained so at week 18 . Conversely , in the two with highest burdens , levels were low at week 8 with only small increments apparent by week 18 . The animals with intermediate burdens had the largest increments in IgG levels between the two sampling times . Levels of total IgE showed only a small perturbation from normal ( 4–5 IU ) around week 8 but had returned to background by week 12 ( data not shown ) . When blood-feeding worms were cultured in vitro with different serum pools to determine any deleterious effect on their physiology , a differential effect on growth but not on viability was observed over the 18-day test period . Thus , the mean cross-sectional surface area ( mm2±SE ) of worms in medium containing serum from the low burden pool ( 0 . 139±0 . 013 ) was approximately 50% less ( P<0 . 001 ) than those supplemented with normal ( 0 . 233±0 . 013 ) and high burden ( 0 . 269±0 . 022 ) sera , which did not differ significantly from each other ( n = 46 , 37 and 28 worms , respectively ) . To determine whether the IgG response was directed against secreted and/or accessible surface proteins we developed two corresponding novel antigen preparations . Separation of SASP by 2D electrophoresis revealed a complex mixture of proteins , many of which could be identified by tandem MS ( Figure S2 ) . When blots were probed with individual rhesus sera , only a minority of constituents was immunoreactive , with largely quantitative rather than qualitative variations observed between the six animals . The same basic pattern was present but at a higher intensity in the animals with the low burden and high titre ( Figure 5A ) , and vice versa ( Figure 5B ) . Targets identified by tandem MS on gels matched to the higher intensity blot are annotated in the figure . Two groups of highly reactive spots on the blots were either not detected on the gels by Sypro Ruby or Coomassie staining ( Figure 5 Region X ) , strongly indicative of glycan epitopes , or were present only in trace amounts that made tandem MS identification problematic ( Figure 5 Region Y ) . The former group is likely to comprise high molecular weight mucins while the two identities obtained for the latter ( Sm200 and α2 macroglobulin ) are indicative of tegument surface components . A relatively simple mixture of proteins was revealed by 2D separation of TSP , some sufficiently abundant for identification by tandem MS ( Figure S3 ) . TSP is a more scarce resource than SASP so probing with individual sera was restricted to Western blots of 1D separations ( Figure 6A ) . The complexity of serum reactivity increased in inverse proportion to the final worm burden . Both high ( 280 kDa ) and low ( 14 , 10 , 8 kDa ) molecular weight bands were strongly recognized by all animals whilst more numerous bands were evident in the four with medium to low burdens , especially R6 . Sufficient material was accumulated to perform a second 2D separation , under identical conditions to the gel in Figure S3 , for blotting and probing with R6 serum ( Figure 6B ) . The principal reactive targets identified by tandem MS were Sm200 , alkaline phosphatase , α2 macroglobulin , LMWP and Sm22 . 6 .
In this study we have attempted to discover why the fate of a primary S . mansoni infection in the rhesus macaque differs from that in other permissive primate hosts [27] . A notable feature of our data was the wide range of worm recoveries at 18 weeks but the peak values for fecal egg output and level of circulating parasite antigens indicate that a substantial population established in all animals . Independent validation was provided by five adult rhesus macaques , infected under identical conditions by one of us and perfused at 8 weeks , where a mean of 43% parasite maturation was observed ( range 152 to 718 worms ) [28] . Perfusion data from earlier studies revealed 49% maturation at 6–7 weeks [29] and reduced primary burdens between 12 and 27 weeks relative to 8 weeks [10] . These data on actual worm numbers , together with the two indirect estimates of infection intensity , point to a gradual but appreciable loss of the established worm burden in some animals . This is contrary to the impression given by advocates of concomitant immunity that adult worms of a primary infection are unaffected by the response they provoke [11] , [30] . To determine whether worms were able to escape from the portal vasculature via porta-caval shunts , as occurs in chronically infected mice [13] , [17] , we measured portal pressure and any associated loss of vascular integrity in the rhesus macaques . Three animals exhibited portal hypertension on the basis that normal mean values for portal pressure in this species are around 7 . 4 to 9 . 8 cm H2O ( 0 . 73 to 0 . 96 kPa ) [10] , [31] and values >14 cm H2O ( 1 . 37 kPa ) are considered hypertensive [32] . These data concur with an earlier study in which a small number of rhesus macaques had portal hypertension after schistosome infection [10] . However , as we observed only a single animal with ( substantial ) portal shunting , translocation of worms from the portal system can be ruled out as the reason for loss of adult burden . The absence of a pathophysiological explanation strengthens the case for an immunological mechanism . Direct evidence for inhibitory factors in the circulation is provided by the retarded growth of blood-feeding worms when cultured with a serum pool from low burden animals , compared to sera from high burden and normal animals . IgG is the most likely candidate since the mean profile of this isotype rises over the period when the indicators of worm number are declining , unlike those for IgM and total IgE that show an early rise and fall . The low levels of total IgE are unusual for infection with helminths but may reflect diminished immunostimulation . This could be due to the early cessation of egg laying , which normally drives Th2 polarization [33] , and/or the reduced worms . The kinetics of IgG production is consistent with the pattern of worm elimination , with the rapid responder animals having the lowest burdens by 18 weeks and vice versa . Indeed , IgG had reached high levels in the rapid responder animals at 8 weeks , before worm elimination began , providing strong evidence that it was the cause rather than an effect of worm death . Given that the high burden animals still had low IgG levels at week 18 , it appears that a threshold of specific antibody must be reached before deterioration in worm viability commences . Thereafter , we propose that sustained immunological pressure over weeks to months , not an acute lethal hit , is the cause of worm death . Support for the idea of a threshold is provided by earlier studies with animals given a low dose of cercariae where egg output continued undiminished over a prolonged period [14] , [34] . Our interpretation is that the antigenic stimulus in this situation was insufficient to promote high antibody levels . The appearance of worms at perfusion is a key pointer to the reason for their demise . The lack of blood and hematin in the gut , especially of the females that have a higher nutrient/energy requirement [35] , indicates cessation of feeding that would contribute to a reduction in gut-derived circulating antigens . The capacity to acquire nutrients across the tegument surface [36] would partly compensate for loss of gut function but the vacuolated parenchyma denuded of glycogen and lipid stores reveals its inadequacy for long-term worm maintenance . The associated atrophy of vitellaria and ovaries explains the declining egg production and hence fecal egg output . Parenthetically , a reduced fecundity and shorter length of female S . japonicum worms , recovered from rhesus macaques at 19 versus 6 weeks , has been reported but without commentary [37] . The structural integrity of the tegument , gut epithelium and muscles we observed suggests that these tissues are crucial to survival , although it is notable that the gut epithelium showed a significant loss of biosynthetic machinery associated with blood feeding . In this context it has recently been demonstrated by RNA interference that gut Cathepsin D is essential for blood digestion and worm maturation [38] . Overall , the degenerative changes point to a process of worm starvation leading to death from organ failure . We infer from the egg output and circulating antigen data that in the rapid responder animals the process begins approximately 10 to 12 weeks after infection , with worms taking several weeks to expire . Antigens released from or exposed by the live parasite provide a potential stimulus for the production of effector antibodies . TSP contains some of the most “external” proteins from the tegument , hence the ones likely to be accessible to antibodies while SASP contains the secreted proteins of the gut epithelium plus some tegument constituents released during short-term culture . The quantitative rather than qualitative differences in the reactivity of individual rhesus sera to TSP and SASP , suggest that the gradual recognition of specific targets coincides with an increasing ability to eliminate the worm population . We were able to identify digestive enzymes in gut secretions , tegument surface hydrolases and a number of antioxidant enzymes , plus proteins of unknown function especially at the tegument surface , as antibody targets . We believe that the unusual ability of the rhesus macaque to eliminate adult worms is mediated by IgG but the mechanism is unclear . Our data do not support rapid opsonisation and/or complement fixation leading to lysis and/or leukocyte attack , since surface-adherent cells were not evident on starving worms . Similarly , a lack of leukocyte adherence [39] has been reported on adult worms from mice , in spite of IgM , IgG and Complement proteins at the tegument surface [40] , [41]; in this host the complement cascade appears to be arrested at C4 [42] . An alternative mechanism , by analogy with well-characterized autoimmune responses , could be antibodies operating in a blocking ( as in Myasthenia gravis [43] ) or stimulatory ( as in Grave's disease [44] ) capacity . We envisage blocking antibodies impacting on nutrient uptake or the ability of worms to combat oxidative stress , while antibodies stimulating receptors could trigger hyperactivity . Due to genetic variability , some worms will be able to resist immune attack for longer than others; similarly , the speed and intensity with which a rhesus macaque responds to the schistosome infection will dictate the timing of its worm elimination . In reappraising the rhesus macaque model we have shifted the emphasis from the more common perception of a resident worm population protecting against further parasite recruitment [11] to anti-adult immunity against a primary infection in a self-cure process [27] . This interpretation provides a novel rationale for the development of a schistosome vaccine that would be both therapeutic and prophylactic , thereby providing the ultimate control tool . Tegument surface proteins and gut secretions appear to be the source of key antigens . Indeed , two recent studies have demonstrated the protective potential in rodents of a tegument surface tetraspanin [45] and an anti-oxidant enzyme [46] , the latter involving anti-adult worm immunity . The inventories of gut and tegument proteins that we have obtained by proteomic analysis [40] , [41] provide the basis for a reverse vaccinology approach [47] to identify the most promising candidates .
|
Infection with blood-dwelling schistosome worms is a major cause of human disease in many tropical countries . Despite intensive efforts a vaccine has proved elusive , not least because the chronic nature of the infection provides few pointers for vaccine development . The rhesus macaque appears unique among animal models in that adult worms establish but are eventually lost . We investigated whether this was due to pathological or immunological causes by monitoring the fate of a schistosome infection , and were able to rule out escape of worms from the portal system as a result of egg-induced vascular shunts . A substantial worm population established in all animals but there was a wide variation in the numbers recovered at 18 weeks . We observed a strong inverse association between the rapidity and intensity of the IgG response and worm burden . Rather than an acute lethal attack , immune-mediated elimination of worms appeared to be a prolonged process directed against vital components of exposed surfaces , causing worms to starve to death . We suggest that if the mechanisms deployed by the rhesus macaque could be replicated in humans by administration of key recombinant antigens , they would form the basis for a vaccine with both prophylactic and therapeutic properties .
|
[
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] |
[
"infectious",
"diseases/helminth",
"infections",
"immunology/immunity",
"to",
"infections",
"immunology/immune",
"response",
"pathology/histopathology"
] |
2008
|
Elimination of Schistosoma mansoni Adult Worms by Rhesus Macaques: Basis for a Therapeutic Vaccine?
|
Postzygotic reproductive isolation is characterized by two striking empirical patterns . The first is Haldane's rule—the preferential inviability or sterility of species hybrids of the heterogametic ( XY ) sex . The second is the so-called large X effect—substitution of one species's X chromosome for another's has a disproportionately large effect on hybrid fitness compared to similar substitution of an autosome . Although the first rule has been well-established , the second rule remains controversial . Here , we dissect the genetic causes of these two rules using a genome-wide introgression analysis of Drosophila mauritiana chromosome segments in an otherwise D . sechellia genetic background . We find that recessive hybrid incompatibilities outnumber dominant ones and that hybrid male steriles outnumber all other types of incompatibility , consistent with the dominance and faster-male theories of Haldane's rule , respectively . We also find that , although X-linked and autosomal introgressions are of similar size , most X-linked introgressions cause hybrid male sterility ( 60% ) whereas few autosomal introgressions do ( 18% ) . Our results thus confirm the large X effect and identify its proximate cause: incompatibilities causing hybrid male sterility have a higher density on the X chromosome than on the autosomes . We evaluate several hypotheses for the evolutionary cause of this excess of X-linked hybrid male sterility .
Speciation occurs when two populations become reproductively isolated from each other through the evolution of one or more barriers to gene flow [1 , 2] . One of the most intensively studied forms of reproductive isolation is intrinsic postzygotic isolation , the inviability or sterility of species hybrids . A model describing the evolution of hybrid inviability and hybrid sterility was proposed independently by Dobzhansky [1] and Muller [3] . The essence of their idea is that divergence at interacting loci between species can cause deleterious , incompatible epistatic interactions in interspecific hybrids . Genetic studies have now amassed abundant evidence that these hybrid incompatibilities are a common cause of intrinsic hybrid fitness problems [4] . In 1989 , Coyne and Orr suggested that “two rules of speciation” further characterize the genetics of postzygotic isolation [5] . The first is Haldane's rule , which states that when hybrids of just one sex are dead or sterile , it is usually the heterogametic ( XY ) sex [6] . This rule is widely obeyed in both male-heterogametic ( XY; e . g . , Drosophila and mammals ) and female-heterogametic ( ZW; e . g . , Lepidoptera and birds ) taxa [7–12] . After two decades of intensive study , most speciation geneticists now agree that Haldane's rule is caused by a combination of at least two phenomena [4] . First , the dominance theory posits that the alleles causing hybrid incompatibilities are , on average , partially recessive for their effects on hybrid fitness [3 , 13–16] . Thus , heterogametic F1 hybrids ( hereafter XY males ) experience the full effects of all recessive X-linked hybrid incompatibilities , whereas homogametic F1 hybrids experience few or none . The main prediction of the dominance theory is supported by both genetic [17–27] and comparative studies [28 , 29] . Second , the faster-male theory posits that incompatibilities causing hybrid male sterility accumulate faster than those causing hybrid inviability or hybrid female sterility [8 , 30] . Two processes might give rise to such faster-male evolution . First , sexual selection on male-specific genes could increase divergence at these loci , increasing the chance of hybrid male sterility . Second , spermatogenesis itself might be an inherently sensitive developmental process that is easily perturbed in hybrids . Regardless of its underlying causes , evidence of faster-male evolution has been obtained from genetic studies [21 , 22 , 26 , 31] , hybrid gene misexpression studies [32–34] , and comparative analyses [29] . Two other phenomena , faster evolution of X-linked loci ( the faster-X theory [35]; but see [36 , 37] ) and some forms of genetic conflict [31 , 38–42] , have also been suggested as causes of Haldane's rule , but their general importance remains unclear . The second rule of speciation is the so-called large X effect [5 , 7 , 35] . In backcross analyses of species hybrids , substitution of one species's X chromosome for the other's has a disproportionately large effect on hybrid fitness relative to similar substitution of an autosome . The large X effect ( recently dubbed “Coyne's rule” [43] ) has been observed in genetic analyses of hybrid inviability [44–46] and hybrid sterility [47–58] , and also has been inferred from patterns of gene flow across natural hybrid zones: X-linked loci often pass across hybrid zone boundaries less freely than do autosomal loci [59–64] . Despite these observations , the causes , and indeed the existence , of the large X effect remain controversial . Wu and Davis [8] pointed out that backcross analyses of hybrid males compare the hemizygous effects of X chromosomes with the heterozygous effects of autosomes: hybrid males suffer from all recessive incompatibilities on the X while those on the autosomes remain mostly masked . They argued that the large X effect , although expected under the dominance theory [10 , 16] , arises as a consequence of the backcross design and signifies nothing special about the X chromosome per se . However , the large X effect could also result from a higher density of hybrid incompatibility loci on the X [5 , 8 , 9 , 35] . Therefore , to distinguish the relative contributions of dominance versus density to the large X effect , hemizygous X chromosome segments must be compared to homozygous autosomal segments [8] . Several studies have performed this test with mixed results . Hollocher and Wu [22] introgressed three regions from both the D . sechellia and the D . mauritiana second chromosome into a largely D . simulans genetic background and compared the homozygous effects of the autosomal introgressions on postzygotic reproductive isolation to those of hemizygous X-linked introgressions . Their results showed that homozygous autosomal introgressions do , in fact , have effects on hybrid inviability and hybrid sterility similar in magnitude to X-linked introgressions . They concluded that , to their level of resolution , the large X effect is indeed a methodological consequence of dominance . Two other studies in Drosophila have , however , found tentative evidence for a higher density of hybrid male sterility on the X . True et al . [21] performed a genome-wide screen for hybrid incompatibilities between D . mauritiana and D . simulans . Using a collection of D . mauritiana lines bearing selectable P-element markers at 87 known cytological positions , they generated 355 homozygous introgressions after backcrossing to D . simulans for 15 generations . Their results showed that D . mauritiana introgressions on the X cause hybrid male sterility 50% more often than those on the autosomes . More recently , Tao et al . [26] used a fine-scale mapping approach in the same hybridization to study the distribution of hybrid incompatibilities by generating 218 overlapping D . mauritiana introgressions covering the third chromosome . Although they did not create a similar set of D . mauritiana introgressions on the X , Tao et al . compared the effects of their third chromosome introgressions with previously published data from 265 X-linked introgressions used to map hybrid male sterility between these two species . These comparisons suggested that the X carries ∼2 . 5 times more hybrid sterility loci than the autosomes . Thus , both True et al . and Tao et al . tentatively concluded that the X chromosome has a higher density of hybrid male sterility factors . There are two possible explanations for the conflicting results between these three studies . The first involves introgression size . If larger introgressions are more likely to cause hybrid incompatibilities [65] , then large introgressions would upwardly bias estimates of the number of hybrid incompatibilities regardless of their location . For example , Hollocher and Wu compared autosomal introgressions roughly the size of a chromosome arm ( ∼20–30 Mb ) with X-linked introgressions that were as small as one-third of the size of their autosomal introgressions ( and collected from a different experiment ) . This difference in introgression size might have led them to overestimate the relative effects of the second chromosome introgressions . Although True et al . 's data suggested that no obvious difference in size existed between their X-linked and autosomal introgressions , the authors urged caution in the interpretation of their results , as they were unable to systematically obtain introgression size estimates . The second possible explanation involves sampling bias . Tao et al . used X-linked introgressions in their comparison that were significantly smaller than their third chromosome introgressions . Although this appears to argue in favor of the large X effect , these X-linked introgressions were not a random sample: they came from regions known from previously published work to have large effects on hybrid male sterility ( see [66–69] ) . Thus , to date , there has not been an analysis that simultaneously phenotypes and genotypes many X-linked and autosomal introgressions from the same experiment . The large X effect hypothesis makes a clear prediction: introgressions of a given size will be incompatible more often when they reside on the X chromosome than when they reside on an autosome . Because introgression size affects hybrid fitness , it is important to compare introgressions of modest size on the X versus the autosomes . Here , we take an approach similar to that used by True et al . , but we study the genome-wide distribution of hybrid incompatibilities between D . mauritiana and D . sechellia , two island endemic species of fruit flies that diverged ∼300 , 000 y ago [70] . Using a collection of D . mauritiana P-element insertion lines , we generated 142 independent introgressions that collectively cover ∼70% of the genome . We use these data to test theories about the genetic causes of the two rules of speciation: Haldane's rule and the large X effect .
Crosses between D . mauritiana and D . sechellia produce fertile F1 hybrid females and sterile F1 hybrid males in both directions of the cross . To place our D . mauritiana introgressions in an otherwise D . sechellia genetic and cytoplasmic background , all introgressions were initiated by crossing D . sechellia white ( w ) females with D . mauritiana P[w+] males . Each D . mauritiana P[w+] insertion line carries a mini-white gene , which acts as a semi-dominant visible eye-color marker , allowing us to select heterozygous P[w+] females in a D . sechellia w background for each backcross . Following the introgression scheme shown in Figure 1 ( the same crossing scheme used in [21] ) , we introgressed 66 different P[w+]-marked chromosomal regions from D . mauritiana into a D . sechellia w genetic background via 15 generations of repeated backcrossing . On average we constructed 2 . 2 independent replicates per P[w+] insert , for a total of 142 sublines . From each subline , we scored the viability and fertility of homozygous ( hemizygous ) introgressions ( see Materials and Methods ) . An average of 156 flies were scored per subline , for a total of 22 , 128 flies . Of the 142 sublines , 55 ( 39% ) show some form of postzygotic reproductive isolation ( Table S1 ) . Haldane's rule in male-heterogametic taxa is thought to result from the general recessivity of hybrid incompatibility alleles ( the dominance theory ) and the more rapid accumulation of incompatibilities that cause hybrid male sterility ( the faster-male theory ) . Our introgression data support both theories . The crossing scheme used to create the introgressions required backcrossing through hybrid females that were heterozygous for D . mauritiana genetic material . Although a few sublines were lost during backcrossing to D . sechellia w , at least one subline per P[w+] insert survived the introgression procedure . This suggests that there are few ( if any ) incompatibilities sufficiently dominant to cause strong sterility or inviability in heterozygous introgression females . Similarly , after 15 generations of backcrossing , hybrid males also appear unaffected by dominantly acting incompatibilities: all autosomal sublines produced viable heterozygous P[w+] males ( H1 males; see Figure 1B for notation ) , and 94% ( 526 of 562 H1 males tested ) of these were fertile . Indeed , every autosomal subline produced at least two fertile H1 males of five tested . Thus , virtually all of the hybrid incompatibilities we detect act fairly recessively , consistent with the dominance theory . Second , we find a dramatic excess of hybrid male sterility over hybrid inviability or hybrid female sterility ( Figure 2 ) . Of 142 sublines , only 9 ( 6% ) cause hybrid inviability . The fertility of most introgressions could therefore be tested for both sexes in 133 sublines . Surprisingly , all viable sublines produce fertile hybrid females . This result differs from that of True et al . [21] who found that ∼6% of their D . mauritiana–D . simulans sublines were hybrid female-sterile . Most striking , however , is the large number of sublines that cause hybrid male sterility: 44 of 133 viable sublines ( 33% ) produce completely sterile hybrid males . This excess of hybrid male-sterile introgressions over hybrid inviable or hybrid female-sterile introgressions supports the faster-male theory of Haldane's rule , and is consistent with other genetic studies in Drosophila [4 , 21 , 22 , 26] . The incompatibilities we observe could , however , be artifacts in two ways . First , deleterious mutations might be segregating within each parental stock . However , we observe inviability or sterility only in some introgression lines—not in the parental lines—which indicates that the incompatibilities detected result from incompatible interactions between genes from D . mauritiana and D . sechellia . Second , some of the inviability and sterility we observe might be caused by spontaneous mutations that arose during the introgression procedure . Previous estimates show that mutations account for 1%–5% of lethal sublines and ∼0 . 2%–2 . 5% of sterile sublines seen in D . mauritiana–D . simulans hybrids [21 , 26] . Although our study was performed in D . mauritiana–D . sechellia hybrids , there is no evidence of a difference in mutation rates between hybrids of these three species [71] . Spontaneous mutation cannot , therefore , account for the relative abundances of the different classes of hybrid incompatibilities . Two additional facts militate against spontaneous mutation: ( 1 ) the rate of spontaneous mutation to lethality is several-fold higher than that to male sterility in Drosophila [72 , 73] and thus cannot account for the excess of hybrid male-sterile introgressions compared to hybrid inviable or hybrid female-sterile introgressions , and ( 2 ) hybrid incompatibilities were typically confirmed with replicate sublines ( 20 of 32 P[w+] inserts ) . Even if we exclude singly incompatible sublines from our data , our results remain qualitatively the same: a large proportion of D . mauritiana introgressions cause hybrid male sterility in a D . sechellia genetic background . Introgressions causing hybrid male sterility are not randomly distributed throughout the genome ( Figure 2 ) . Instead , we find that the X chromosome possesses a significant excess of hybrid male steriles compared to the autosomes: 60% ( 27 of 45 ) of X-linked introgressions are hybrid male-sterile , whereas only 18% ( 17 of 97 ) of autosomal introgressions are hybrid male-sterile ( χ2 = 25 . 9 , p < 0 . 0001 ) . Although this pattern is consistent with that predicted by the large X effect , it could have several trivial causes . Two of these can be ruled out . First , we can exclude a clustering of P[w+] inserts in male-sterile regions on the X as the cause of this pattern , as the collection of P[w+] inserts was shown previously to have a random distribution within D . mauritiana chromosome arms [21 , 74] . Second , we can exclude the possibility that P[w+] inserts in hybrid male-sterile regions on the X are represented by more sublines than those in fertile regions ( which would inflate the fraction of sterile X-linked introgressions ) , as the number of sublines scored for sterile and fertile P[w+] inserts is the same on both the X ( χ2 = 0 . 01 , p = 0 . 91 ) and the autosomes ( χ2 = 0 . 18 , p = 0 . 67 ) . Last , the apparent excess of hybrid male-sterile sublines on the X chromosome could result from systematically larger introgressions on the X versus the autosomes . To test this possibility , we estimated the size of our introgressions by genotyping three microsatellite markers on each side of the P[w+] insert in 108 sublines from 55 P[w+] inserts . The three markers on each side were spaced ∼50 kb , ∼500 kb , and ∼1 Mb away from the P[w+] insert ( see Materials and Methods ) . For 67 of these sublines , we could reliably score or infer the genotype at all six markers , hereafter referred to as “complete” sublines . For the remaining 41 sublines , we were able to obtain only partial genotypes because of repeated PCR failure or a lack of diagnostic markers . Our results show that X-linked and autosomal introgressions appear similar in size ( Figure 3 ) . We nevertheless tested for potential size differences in two ways . First , we compared the distribution of recombination events on either side of each P[w+] insert using the 67 complete sublines . If X-linked and autosomal introgressions are similar in size , we expect them to possess similar fractions of sublines with one or more recombination events within the ∼2-Mb window around the insert ( 1 Mb to the left and 1 Mb to the right ) . Consistent with this expectation , we find that X-linked and autosomal introgressions have a similar distribution of recombination events ( χ2 = 4 . 32 , df = 2 , p = 0 . 12 ) . Second , we compared estimated introgression sizes between X-linked and autosomal introgressions using unpaired t-tests with null distributions generated from 1 , 000 randomizations of the data . We find no significant size difference between X-linked ( mean ± one standard error = 1 . 33 ± 0 . 10 Mb; median = 1 . 46 Mb ) and autosomal ( 1 . 49 ± 0 . 06 Mb; median = 1 . 47 Mb ) introgressions using the 67 complete sublines ( p = 0 . 15 ) or using all 108 sublines ( p = 0 . 12 ) ; similar results hold when using nonparametric Mann-Whitney tests ( pcomplete = 0 . 33; ptotal = 0 . 36 ) . Thus , the finding that X-linked introgressions cause hybrid male sterility more often than autosomal ones cannot be explained by a systematic difference in introgression size between the X chromosome and the autosomes . The probability that a D . mauritiana introgression will cause hybrid male sterility in D . sechellia is greater when the introgression resides on the X chromosome . Our results thus provide a proximate explanation for the large X effect: there is a higher density of hybrid male steriles on the X chromosome . Interestingly , our data also show that hybrid male-sterile introgressions appear to be slightly larger than fertile introgressions . When we compare the distribution of recombination events among the complete sublines , we find no significant difference between fertile and sterile introgressions either within the X chromosome ( χ2 = 4 . 15 , df = 2 , p = 0 . 13 ) or within the autosomes ( χ2 = 1 . 41 , df = 2 , p = 0 . 49 ) . However , in both comparisons we see a trend towards sterile introgressions being larger than fertile ones: 76% of sterile introgressions show one or more recombination events compared to 82% of fertile introgressions ( Figure 4 ) . Comparing mean size between fertile and sterile introgressions , we find no significant difference using the complete sublines ( p = 0 . 11 ) , but a marginally significant difference using all 108 sublines ( p = 0 . 02 ) . However , this marginally significant result does not affect our interpretation of the large X effect . Even if fertile and sterile introgressions differ slightly in size , introgression sizes on the X and the autosomes are the same . Our data allow us to map the locations of 108 D . mauritiana introgressions , and thus obtain a rough estimate of the minimum number of hybrid inviable and hybrid male-sterile regions in the genome ( Figure 2 ) . We estimate that a minimum of four hybrid inviable regions separate D . sechellia and D . mauritiana: three on chromosome arm 3L and one on the X chromosome . We also estimate a minimum of eight hybrid male-sterile regions distributed roughly uniformly across the autosomal genome: three on chromosome arm 2L , zero on 2R , two on 3L , and three on 3R . Previous studies have shown that there are no hybrid inviables or hybrid male steriles on the small dot fourth chromosome between these species [75] . The average autosomal arm thus carries approximately two hybrid male-sterile regions . In contrast , a minimum of nine hybrid male-sterile regions are distributed over the length of the X chromosome—more than four times the number on an average ( and similarly sized ) autosomal arm . Thus , we find that at least one hybrid male sterility locus resides on each major chromosome , consistent with previous studies of postzygotic isolation in this species pair [76] . These numbers are , of course , minimum estimates for three reasons . First , each of our hybrid inviable or hybrid male-sterile introgressions might contain more than one hybrid incompatibility gene . Second , hybrid inviable introgressions may mask tightly linked hybrid male steriles . Third , we were unable to screen some regions of the genome and so may have missed some hybrid incompatible regions . However , because our coverage of the genome is fairly good , we do not expect that the qualitative difference—especially for hybrid male sterility—between the X and the autosomes would be much affected by higher resolution mapping . Although dominance may contribute to the large X effect [10 , 16] , our analysis distinguishes the dominance of hybrid male steriles from their relative density in two ways . First , the excess of hybrid male steriles on the X chromosome cannot be attributed to a methodological consequence of dominance [8 , 22] as we compare hemizygous X and homozygous autosomal effects . Second , we introgress more D . mauritiana material on the autosomes than on the X ( 97 versus 45 introgressions ) , thus exposing more potential recessive hybrid incompatibilities on the autosomes . Under the dominance theory ( which assumes equal densities of hybrid incompatibilities on the X and on the autosomes [15 , 16] ) , we would expect to uncover more hybrid male steriles on the autosomes than on the X , in contrast to our findings . Our introgression data demonstrate a higher density of hybrid male steriles on the X chromosome , but they do not explain why there are more on the X . We can exclude two explanations . First , there is not a higher concentration of male fertility-essential genes on the X chromosome . If anything , the opposite appears to be true in Drosophila: genes mutable to male sterility appear to be randomly distributed throughout the genome [77] and genes with male-biased expression are underrepresented on the X [78] . Second , the faster molecular evolution of X-linked loci does not appear to contribute to the large X effect . Charlesworth et al . [35] ( see also [5] ) showed theoretically that X-linked loci experience faster rates of substitution than autosomal loci when beneficial mutations are , on average , partially recessive . If true , we might expect the X chromosome to accumulate hybrid incompatibilities faster than equivalently sized autosomes . This theory predicts that X-linked loci will show greater sequence divergence between species than do autosomal loci . Population genetic tests for faster X evolution , however , show that the substitution rates of X-linked and autosomal loci in Drosophila are similar [36 , 37 , 79] . It is worth noting , however , that the effect of even a slightly elevated substitution rate on the X chromosome would be amplified , as the number of hybrid incompatibilities increases at least as fast as the square of divergence [16 , 80] . Three plausible explanations of the large X effect remain . First , recent discoveries of sex-ratio distortion in weakly fertile hybrid males [41 , 81 , 82] have renewed interest in the idea that genetic conflict might drive the evolution of X-linked hybrid male steriles [38 , 39] . Tao and Hartl [31] hypothesize that recurrent bouts of invasion by sex-chromosome meiotic drive loci can increase the density of hybrid incompatibilities on the sex chromosomes . They argue that because sex-ratio distorters affect gametogenesis , genes involved in conflict over sex ratio could have pleiotropic effects on fertility . As sex-ratio distorters usually reside on the X chromosome [39 , 83] , this might give rise to a higher density of hybrid male steriles on the X versus the autosomes . One way to test for histories of sex-ratio conflict is to screen for sex-ratio distortion in species hybrids: distorters that are masked by suppressors in one species can be unmasked on the naïve genetic background of another species [41 , 82] . We therefore scored the sex ratio of 54 fertile introgression lines . We found only one subline , an autosomal introgression , that consistently produces a moderately male-biased sex ratio ( Table S1 ) . Further work , however , showed that this sex-ratio bias is not a hybrid phenomenon and does not involve sex-chromosome meiotic drive ( data not shown ) . We thus conclude that there is no evidence for unmasked cryptic sex-ratio distortion among our fertile introgression lines . We cannot , of course , entirely exclude the possibility that past bouts of conflict have occurred in the D . mauritiana or D . sechellia lineages . Although no D . mauritiana autosomal introgressions released cryptic D . sechellia X-linked distorters , we were obviously unable to test most X-linked D . mauritiana introgressions for their ability to cause sex-ratio distortion in an otherwise D . sechellia background , as most of these introgressions were completely sterile . A second possible explanation of the large X effect involves dosage compensation [5] . In particular , dosage compensation in the germline might be easily disrupted in hybrids . The X chromosome of Drosophila males is hyper-transcribed to equalize gene dose between males and females . In the soma , this process is under the control of the male-specific lethal ( MSL ) complex of proteins [84] . Divergence of the MSL machinery between species could thus cause a breakdown in dosage compensation in interspecific hybrids . Because this would affect nearly all X-linked loci , the X chromosome would have a disproportionately high density of hybrid incompatibilities . Although previous studies provide evidence that disruption of the MSL-mediated dosage compensation pathway does not cause inviability in D . melanogaster–D . simulans hybrids [85] , recent work suggests that dosage compensation does occur in the germline by an MSL-independent mechanism [86] . A breakdown of dosage compensation specifically in the germline could , therefore , potentially produce a large X effect for hybrid male sterility . Finally , X inactivation—the condensation of the X chromosome during early spermatogenesis—could be disrupted in hybrid males . It has been suggested that spermatogenesis may be an inherently sensitive developmental process , rendering hybrid males particularly prone to sterility [8 , 30] . Genetic and cytogenetic evidence within species supports this idea: spermatogenesis appears sensitive to genetic perturbations , particularly with respect to the X chromosome [87] . In D . melanogaster , for instance , translocations from the autosomes to the X almost always result in dominant male sterility . Sterility in these cases is thought to result from improper X inactivation in primary spermatocytes [87] . It seems possible , then , that foreign genetic material that is recognized as “non-X” by the X inactivation machinery could disrupt X inactivation , causing hybrid male sterility . If introgressions on the X from one species are not recognized as X-linked material by the inactivation machinery of the other species , X-linked introgressions could result in hybrid male sterility , giving rise to a large X effect . The present data do not allow us to distinguish among the three potential evolutionary causes of the large X effect presented here . Resolution of the ultimate cause must , therefore , await further molecular studies .
Construction of the D . mauritiana–D . sechellia introgression lines was performed using D . sechellia w ( kindly provided by J . A . Coyne , University of Chicago ) and the D . mauritiana P-element insertion stocks described in True et al . [21 , 74] . Each of these stocks contains a single P[lac-w+] insertion in a D . mauritiana w background . The P[w+] insert acts as a semi-dominant visible marker; flies heterozygous for the insert on a w background show orange eye color , whereas flies homozygous for the insert show red eye color . The inserts are randomly distributed over all four chromosomes at 87 known cytological positions [21] . Figure 1 shows our introgression procedure , which closely followed that of True et al . [21] . We began with the 84 ( of the original 103 ) P[w+] inserts that are still in existence . Briefly , fertile F1 hybrid females were generated by crossing D . sechellia w females to D . mauritiana males homozygous for the P[w+] markers . We established 2–4 independent replicate sublines for each P[w+] insert by crossing F1 hybrid females to D . sechellia w males . Each subline was then backcrossed independently for 15 generations by crossing hybrid females heterozygous for the P[w+] insert with D . sechellia w males . This crossing scheme produced hybrid introgression lines that have a mostly D . sechellia genetic background , but that carry a small chromosome region from D . mauritiana marked by the P[w+] insert . To make introgressions homozygous , five heterozygous ( hemizygous ) male progeny were selected from the G15 backcross and mated individually to ten D . sechellia w virgin females ( cross H1; Figure 1B and 1C ) ; five white-eyed male siblings were also mated individually to ten D . sechellia w virgin females as controls . ( These white-eyed males share the same genetic background as their P[w+] siblings , including any unmarked D . mauritiana material that is not linked to the P[w+] insert . So long as there is no unmarked D . mauritiana material , these white-eyed males should be fertile . Indeed , these control males were always fertile . ) If at least one H1 hybrid male produced offspring , progeny from a single fertile male were selected for cross H2 . For autosomal introgressions ( Figure 1B ) , if cross H2 proved fertile , homozygous progeny were scored for viability and fertility ( crosses H3 ) . If homozygous viable and fertile H2 progeny of both sexes were produced , they were crossed to establish a homozygous hybrid introgression line . For X-linked introgressions ( Figure 1C ) , heterozygous H2 females were crossed with D . sechellia w males to produce heterozygous females and hemizygous males . These flies were used in cross H3 . If the resulting homozygous female progeny were viable and fertile , they were crossed with their hemizygous brothers to establish a homozygous hybrid introgression line ( cross H4 ) . For most X-linked introgressions , however , all five H1 hybrid males failed to produce offspring . Therefore , hybrid male fertility was also tested in mass matings by crossing ten hemizygous hybrid males with ten D . sechellia w virgin females ( H5 ) . Viability and fertility were scored after autosomal ( Figure 1B ) and X-linked ( Figure 1C ) introgressions were made homozygous . Each introgression subline was classified into one of four categories: lethal , female-sterile , male-sterile , or fertile . Hybrid lethality was scored as the absence of red-eyed progeny of both sexes from cross H2 for autosomal introgressions , or the absence of red-eyed males from backcross G15 for X-linked introgressions ( Figure 1 ) . Homozygous hybrid male and hybrid female fertility were measured as the ability to produce offspring in mass matings with D . sechellia w flies . Crosses that did not produce progeny were scored as sterile . Because of the large number of crosses performed in a short period of time , we were unable to simultaneously score sperm motility , as is commonly done . For some inserts , it was impossible to distinguish homozygous individuals from heterozygous individuals . For these cases , male and female progeny from cross H2 ( autosomes ) and cross H3 ( X ) were mated individually with D . sechellia w females and males , respectively , to test their viability and fertility . Homozygous and heterozygous individuals could then be distinguished by progeny testing , as segregation of the P[w+] marker will only occur among the progeny of heterozygotes . The D . mauritiana P[w+] inserts were localized originally to the resolution of cytological bands [21 , 74] . Recently , however , flanking regions from 95 inserts have been sequenced and were kindly provided by Y . Tao ( Emory University ) and L . Araripe ( Harvard University ) . These data provide precise genomic locations for each P[w+] insert . At the time of our analyses , the genome sequence for D . sechellia was not available . Because both D . mauritiana and D . sechellia are closely related to ( and homosequential with ) D . simulans ( 1%–2% sequence divergence ) , we used the D . simulans genome assembly to obtain genomic coordinates for the sequenced inserts . We used species-specific microsatellite repeat length differences as markers to roughly determine introgression sizes . We identified potential markers approximately 1 Mb , 500 kb , and 100 kb to the left and to the right of each P[w+] insert using Tandem Repeats Finder [88] ( Figure S1 ) . We then designed primers flanking microsatellites from D . simulans genomic sequence . We used a standard protocol to amplify markers , and PCR fragments were separated on 8% polyacrylamide gels to identify length differences . Although we were unable to find suitable markers for every region , in several cases the genotype at those locations could be inferred from the genotypes at adjacent markers . When scoring markers , we assumed that unobserved double recombination events are sufficiently rare as not to occur between adjacent markers ( i . e . , no double crossovers within ∼500-kb windows ) . When we could not score or reliably infer genotype , species identity was treated as unknown . We were able to score a total of 485 microsatellite markers from 55 P[w+] insert regions . Because we know the genomic coordinates of the P[w+] inserts and the microsatellite markers , we can obtain good estimates of introgression size for each subline . We calculated a minimum and a maximum size of D . mauritiana material within a 2-Mb region around each insert ( Figure S1 ) . With respect to the insert's location , minimum introgression size was calculated as the distance between the farthest markers to the left and the right sides of the insert that showed a D . mauritiana allele . Maximum introgression size was calculated as the distance between the nearest markers to the left and the right sides of the insert that showed a D . sechellia allele . Twenty-two sublines showed the D . mauritiana allele at all six markers . For these cases , minimum introgression size was calculated as the distance between the left- and right-most distal markers ( L3 and R3 , respectively ) , and the maximum introgression size was calculated by adding two base pairs to this distance . Likewise , five sublines showed the D . sechellia allele at all six markers . In this case , maximum size was calculated as the distance between the left- and right-most proximal markers ( L1 and R1 , respectively ) , and minimum size was calculated as two base pairs . Because our introgression size estimates have a maximum size limit of roughly 2 Mb , our genotype data might underestimate true introgression size . The distribution of recombination events within this 2-Mb window , however , shows that 81% ( 54/67 ) of the complete sublines experienced at least one recombination event and 48% experienced two recombination events . Thus , our data appear to capture a reasonably accurate sample of introgression size . We present the results from our statistical tests using the maximum size estimates for simplicity . Our conclusions do not depend on the scoring procedure because the results using minimum estimates of introgression size remain similar . The distributions of introgression sizes are non-normal . We therefore tested for differences in means ( e . g . , X versus autosomes or fertile versus sterile ) using unpaired t-tests with null distributions generated by 1 , 000 randomizations of the data . Means are reported ± one standard error .
|
The evolution of reproductive isolation is a fundamental step in the origin of species . One kind of reproductive isolation , the sterility and inviability of species hybrids , is characterized by two of the strongest rules in evolutionary biology . The first is Haldane's rule: for species crosses in which just one hybrid sex is sterile or inviable , it tends to be the sex defined by having a pair of dissimilar sex chromosomes ( e . g . , the “XY” of males in humans ) . The second rule is the large X effect: the X chromosome has a disproportionately large effect on hybrid fitness . We dissected the genetic causes of these two rules of speciation by replacing many small chromosomal segments of the fruit fly Drosophila sechellia with those of a closely related species , D . mauritiana . Together , these segments cover 70% of the genome . We found that virtually all segments causing hybrid sterility or inviability act recessively and that hybrid male sterility is by far the most common type of hybrid incompatibility , confirming two leading theories about the causes of Haldane's rule . We also found that X-linked segments are more likely to cause hybrid male sterility than similarly sized autosomal segments . These results show that the large X effect is caused by a higher density of hybrid incompatibilities on the X chromosome .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Materials",
"and",
"Methods"
] |
[
"arthropods",
"eukaryotes",
"evolutionary",
"biology",
"drosophila",
"animals",
"insects"
] |
2007
|
High-Resolution Genome-Wide Dissection of the Two Rules of Speciation in Drosophila
|
A fundamental question in evolutionary genetics concerns the extent to which adaptive phenotypic convergence is attributable to convergent or parallel changes at the molecular sequence level . Here we report a comparative analysis of hemoglobin ( Hb ) function in eight phylogenetically replicated pairs of high- and low-altitude waterfowl taxa to test for convergence in the oxygenation properties of Hb , and to assess the extent to which convergence in biochemical phenotype is attributable to repeated amino acid replacements . Functional experiments on native Hb variants and protein engineering experiments based on site-directed mutagenesis revealed the phenotypic effects of specific amino acid replacements that were responsible for convergent increases in Hb-O2 affinity in multiple high-altitude taxa . In six of the eight taxon pairs , high-altitude taxa evolved derived increases in Hb-O2 affinity that were caused by a combination of unique replacements , parallel replacements ( involving identical-by-state variants with independent mutational origins in different lineages ) , and collateral replacements ( involving shared , identical-by-descent variants derived via introgressive hybridization ) . In genome scans of nucleotide differentiation involving high- and low-altitude populations of three separate species , function-altering amino acid polymorphisms in the globin genes emerged as highly significant outliers , providing independent evidence for adaptive divergence in Hb function . The experimental results demonstrate that convergent changes in protein function can occur through multiple historical paths , and can involve multiple possible mutations . Most cases of convergence in Hb function did not involve parallel substitutions and most parallel substitutions did not affect Hb-O2 affinity , indicating that the repeatability of phenotypic evolution does not require parallelism at the molecular level .
When multiple species evolve similar changes in phenotype in response to a shared environmental challenge , it suggests that the convergently evolved character state is adaptive under the changed conditions and that it evolved under the influence of directional selection . A key question in evolutionary genetics concerns the extent to which such cases of phenotypic convergence are caused by convergent or parallel substitutions in the underlying genes . This question has important implications for understanding the inherent repeatability of evolution at the molecular level [1–9] . In principle , the convergent evolution of a given phenotype may be attributable to ( i ) unique substitutions , ( ii ) parallel substitutions ( where identical-by-state alleles with independent mutational origins fix independently in different lineages ) , or ( iii ) collateral substitutions ( where shared , identical-by-descent alleles fix independently in different lineages ) [8] . In the last case , allele-sharing between species may be due to the retention of ancestral polymorphism or a history of introgressive hybridization—either way , the function-altering alleles that contribute to phenotypic convergence do not have independent mutational origins . One especially powerful means of assessing the pervasiveness of repeated evolution at the sequence level is to exploit natural experiments where phylogenetically replicated changes in protein function have evolved in multiple taxa as an adaptive response to a shared environmental challenge . For example , there are good reasons to expect that vertebrate species living at very high altitudes will have convergently evolved hemoglobins ( Hbs ) with increased O2-binding affinities [10 , 11] . Under severe hypoxia , an increased blood-O2 affinity can help ensure tissue O2 supply by safeguarding arterial O2 saturation while simultaneously maintaining the pressure gradient that drives O2 diffusion from the peripheral capillaries to the cells of respiring tissues [12–18] . Evolutionary adjustments in blood-O2 affinity often stem directly from structural changes in the tetrameric ( α2β2 ) Hb protein . Genetically based changes in the oxygenation properties of Hb can be brought about by amino acid mutations that increase intrinsic Hb-O2 affinity and/or mutations that suppress the sensitivity of Hb to the inhibitory effects of allosteric co-factors in the red blood cell [19–22] ( S1 Fig ) . Derived increases in Hb-O2 affinity have been documented in some high-altitude birds and mammals [23–30] , but other comparative studies have not revealed consistent trends [31–34] . Additional comparisons between conspecific populations and closely related species are needed to assess the validity of empirical generalizations about the relationship between Hb-O2 affinity and native elevation in vertebrates . Previous surveys of sequence variation in the globin genes of Andean waterfowl documented repeated amino acid substitutions in the major Hb isoforms of multiple high-altitude taxa [35 , 36] , but the functional effects of the substitutions were not assessed so it was not known whether the repeated changes contributed to convergent changes in the oxygenation properties of Hb . Here we report a comparative analysis of Hb function in eight phylogenetically replicated pairs of high- and low-altitude waterfowl taxa to test for convergent changes in biochemical phenotype , and to assess the extent to which convergent changes in phenotype are attributable to repeated amino acid substitutions . We measured the functional properties of native Hb variants in each population and species , and we used protein engineering experiments based on site-directed mutagenesis to measure the functional effects of repeated substitutions that were implicated in convergent increases in Hb-O2 affinity in high-altitude taxa . In six of the eight taxon pairs , the high-altitude taxa evolved derived increases in Hb-O2 affinity that were caused by a combination of unique , parallel , and collateral amino acid replacements . In comparisons involving high- and low-altitude populations of three different species , function-altering amino acid polymorphisms emerged as highly significant outliers in genome scans of nucleotide differentiation , with derived , affinity-enhancing mutations present at high frequency in the high-altitude populations . In combination with results of the functional experiments , the population genomic analyses provide an independent line of evidence that the observed changes in Hb function are attributable to positive directional selection .
To characterize the red cell Hb isoform composition of each species , we analyzed blood samples from individual specimens using a combination of isoelectric focusing ( IEF ) and tandem mass spectrometry ( MS/MS ) . Consistent with data from other anseriform birds [38 , 39] , the waterfowl species that we examined expressed two distinct isoforms , HbA ( pI = 8 . 0–8 . 2 ) and HbD ( pI = 7 . 0–7 . 2 ) with the major HbA isoform comprising ~70–80% of total Hb ( S1 Table ) . The major HbA isoform incorporates α-chain products of the αA-globin gene and the minor HbD isoform incorporates products of the tandemly linked αD-globin gene; both isoforms incorporate β-chain products of the same βA-globin gene [38 , 39] . Since avian HbD has a consistently higher O2-affinity than HbA in all avian taxa examined to date [28 , 30 , 32 , 39] , upregulating HbD expression could be expected to provide an efficient means of increasing blood-O2 affinity in response to environmental hypoxia . However , it appears that high-altitude Andean waterfowl do not avail themselves of this option , as we observed no difference in relative isoform abundance between pairs of high- and low-altitude sister taxa ( Wilcoxon signed-rank test , W = 12 , N = 7 pairwise comparisons , P>0 . 05; S1 Table ) . MS/MS analysis confirmed that subunits of the two adult Hb isoforms represent products of the adult-expressed αA- , αD- , and βA-globin genes; products of the embryonic α- and β-type globin genes were not detected . By combining αD-globin sequences with previously published αA- and βA-globin sequences for the same individual specimens , we identified all amino acid differences that distinguish the HbA and HbD isoforms of each pair of high- and low-altitude taxa ( Fig 1 ) . Full alignments of αA- , αD- , and βA-globin amino acid sequences are shown in S2 Fig , and the direction of changes in character state at all substituted sites are shown in S3–S5 Figs . Comparisons of the South American species revealed repeated amino acid replacements at five sites that distinguish the HbA isoforms of high- and low-altitude sister taxa , including repeated replacements at one site in the αA-globin gene ( α77Ala→Thr in Andean goose , torrent duck , Puna teal , and speckled teal ) and four sites in the βA-globin gene ( β13Gly→Ser in ruddy ducks and speckled teal , β94Asp→Glu in crested duck and Puna teal , and both β116Ala→Ser and β133Leu→Met in yellow-billed pintail and speckled teal ) ( Fig 1 ) . The derived pair of βA-globin amino acid variants ‘116Ser-133Met’ that are shared between sympatric high-altitude populations of yellow-billed pintails and speckled teal are clearly identical-by-descent ( S6 Fig ) . Independent evidence for hybridization between the two species [40 , 41] suggests that the ‘116Ser-133Met’ βA-globin allele in high-altitude yellow-billed pintails was derived via introgression from high-altitude speckled teals . The same is true for a shared β13 ( Gly/Ser ) polymorphism , although the derived Ser variant is present at low-frequency in yellow-billed pintails . The repeated amino acid changes at βA-globin sites 13 , 116 and 133 therefore represent collateral replacements , rather than true parallel replacements , as they do not have independent mutational origins in each species . Three of the eight pairs of high- and low-altitude taxa had structurally distinct HbD isoforms due to 1–2 amino acid substitutions in the αD-globin gene ( Fig 1 ) . Repeated substitutions at αD96 occurred in Orinoco goose ( Val→Ala ) and silver teal ( Ala→Val ) , but the direction of the change in character-state was different in each case ( S5 Fig ) . In both interspecific comparisons ( Andean goose vs . Orinoco goose , and Puna teal vs . silver teal ) , αD96Ala is associated with a higher HbD O2-affinity . However , the individual effects of amino acid replacements at αD96 could not be isolated in either comparison because of potentially confounding replacements in the β-chain ( β86Ala→Ser in Andean goose , and β94Asp→Glu in Puna teal ) ( Fig 1 ) . We measured the O2-binding properties of purified HbA and HbD variants from each taxon and we estimated P50 ( the PO2 at which Hb is half-saturated with O2 ) as an index of Hb-O2 affinity . We focus primarily on measures of Hb-O2 affinity in the presence of Cl- ions and IHP ( P50 ( KCl+IHP ) ) , as this is the experimental treatment that is most relevant to in vivo conditions in avian red blood cells . The experiments revealed that O2-affinities of HbD were consistently higher ( P50 values were lower; S2 Table ) than those of HbA , consistent with data from other birds [28 , 30 , 32 , 39] . Comparisons between high- and low-altitude sister taxa revealed appreciable differences in the O2-affinity of the major HbA isoform in six of eight cases , and in each of these six cases the HbA of the high-altitude taxon exhibited the higher O2-affinity ( i . e . , lower P50 ) ( Fig 2A; S2 Table ) . The only two taxon pairs that did not exhibit appreciable differences in Hb-O2 affinity were those involving conspecific populations of ruddy ducks and torrent ducks ( Fig 2A; S2 Table ) . In contrast to the altitudinal trend for HbA , O2-affinities of the minor HbD isoform were not consistently higher in high-altitude taxa ( Fig 2B ) . However , there were three taxon pairs in which O2-affinities of HbA and HbD were both markedly higher in the high-altitude taxa than in the corresponding low-altitude taxa ( crested duck , Puna teal , and speckled teal ) , a pattern that implicates causative mutations in the β-chain subunit , which is shared by both isoforms . Comparisons involving purified Hb variants from birds with known genotypes provide a means of identifying the specific amino acid mutations that are responsible for evolved changes in Hb-O2 affinity . Below we describe the functional effects of unique and repeated replacements , and we report model-based inferences about the structural mechanisms responsible for the observed changes in Hb-O2 affinity . The fact that high-altitude taxa exhibited higher Hb-O2 affinities than their lowland sister taxa in six of eight pairwise comparisons is an intriguing trend and is suggestive of adaptive convergence , but the overall pattern does not permit conclusive inferences about the adaptive significance of observed changes in Hb function in any particular high-altitude population or species . In principle , genome-wide analyses of nucleotide differentiation between individual pairs of high- and low-altitude populations can provide an independent means of assessing whether altitudinal differences in globin allele frequencies may be attributable to a history of spatially varying selection . Accordingly , we used restriction-site associated DNA sequencing ( RAD-Seq ) to survey genome-wide patterns of nucleotide differentiation between high- and low-altitude populations of three separate species: cinnamon teal , yellow-billed pintail , and speckled teal . In each of these three pairwise population comparisons , function-altering amino acid polymorphisms in the αA- and/or βA-globin genes emerged as highly significant outliers in the genome-wide distribution of site-specific FST values ( Fig 7 ) . Indirect inferences about the adaptive significance of these polymorphisms are corroborated by results of the functional experiments , which demonstrated that the derived variants at these sites contributed to increases in Hb-O2 affinity in high-altitude populations of all three species ( αA9 in cinnamon teal and the two-site ‘116–133’ βA-globin haplotypes shared by yellow-billed pintail and speckled teal ) . Since the βA-globin allele of high-altitude yellow-billed pintail was derived via introgressive hybridization with high-altitude speckled teal , the combined results of our functional experiments and population genomic analyses provide strong evidence for positive selection on introgressed allelic variants . This finding contributes to a growing body of evidence that introgressive hybridization can provide an important source of adaptive genetic variation in animal populations [58–60] . Convergent increases in Hb-O2 affinity in high-altitude waterfowl taxa were caused by a combination of unique amino acid replacements ( as in the case of cinnamon teal , where the causative mutation was not shared with other highland taxa ) , parallel replacements ( as in the case of high-altitude crested ducks and Puna teal that shared independently derived β94Asp→Glu mutations ) , and collateral replacements ( as in the case of yellow-billed pintail and speckled teal that shared identical-by-descent β-globin alleles due to a history of introgressive hybridization ) . Andean goose appears to represent another case where the evolution of a derived increase in Hb-O2 affinity is attributable to one or more unique substitutions , although additional experiments will be required to pinpoint the causative change ( s ) . These results demonstrate that convergent changes in protein function can occur through multiple historical paths involving multiple possible mutations . Among the Andean waterfowl taxa that we examined , we identified only a single case where a convergent increase in Hb-O2 affinity was attributable to a true parallel amino acid substitution ( β94Asp→Glu in high-altitude crested ducks and Puna teal ) . The limited number of function-altering parallel substitutions in the Hbs of Andean waterfowl stands in contrast to patterns of functional evolution in vertebrate opsin proteins , where convergent changes in the wavelengths of maximum absorbance ( spectral tuning ) are very often attributable to parallel amino acid substitutions [61 , 62] . In vertebrate opsins , the more pervasive patterns of parallelism may reflect the fact that genetically based changes in spectral tuning can only be achieved via specific mutational replacements at a limited number of key residues in the active site [63] . Our findings are more consistent with results of experimental evolution studies in microbes and yeast where replicated changes in fitness involved little to no parallelism at the underlying sequence level [64 , 65] . Our comparative survey also identified numerous parallel substitutions that had no effect on the inherent oxygenation properties of Hb , although we cannot rule out the possibility that the derived variants contributed to changes in other structural or functional properties . Our results for waterfowl Hbs provide two important lessons about repeated evolution at the molecular level: ( i ) most cases of convergence in protein function did not involve true parallel substitutions ( indicating that similar phenotypic outcomes can be produced by multiple possible mutations ) , and ( ii ) most parallel substitutions produced no change in Hb-O2 affinity ( convergent or otherwise ) . These findings demonstrate that parallel substitutions cannot be interpreted as prima facie evidence for adaptive evolution [66 , 67] , and that the functional significance ( and , hence , adaptive significance ) of specific substitutions needs to be experimentally tested in order to support conclusions about the molecular basis of phenotypic evolution .
Blood and tissue samples were obtained from Andean waterfowl at high- and low-altitude localities as described previously [35] . Samples from Orinoco geese , Abyssinian blue-winged geese , and Hartlaub’s ducks were obtained from Sylvan Heights Waterfowl Park ( Scotland Neck , North Carolina ) . Animals were handled in accordance with protocols approved by the Institutional Animal Care and Use Committee of the University of Alaska ( certification numbers 02-01-152985 and 05-05-152985 ) . We characterized Hb isoform composition in the mature erythrocytes of 106 wild-caught birds ( median sample size = 14 individuals per species ) ( S1 Table ) . Native Hb components were separated by means of IEF using precast Phast gels ( pH 3–9 ) ( GE Healthcare; 17-0543-01 ) . IEF gel bands were excised and digested with trypsin , and MS/MS was used to identify the resultant peptides , as described previously [26 , 28 , 32 , 68] . Database searches of the resultant MS/MS spectra were performed using Mascot ( Matrix Science , v1 . 9 . 0 , London , UK ) ; peptide mass fingerprints were queried against a custom database of avian globin sequences , including the full complement of embryonic and adult α- and β-type globin genes that have been annotated in avian genome assemblies [38 , 69–73] . We identified all significant protein hits that matched more than one peptide with P<0 . 05 . After separating the HbA and HbD isoforms by native gel IEF , the relative abundance of the two isoforms was quantified densitometrically using Image J [74] . The αA- and βA-globin genes were amplified and sequenced according to protocols described previously [35 , 36] . For all specimens used as subjects in the experimental analyses of Hb function , we extracted RNA from whole blood using the RNeasy kit ( Qiagen , Valencia , CA ) , and we amplified full-length cDNAs of the αD-globin gene using a OneStep RT-PCR kit ( Qiagen , Valencia , CA ) . We designed paralog-specific primers using 5’ and 3’ UTR sequences , as described by Opazo et al . [38] . We cloned reverse transcription ( RT ) -PCR products into pCR4-TOPO vector using the TOPO TA Cloning Kit ( Invitrogen , Carlsbad , CA ) , and we sequenced at least five clones per sample in order to recover both alleles . This enabled us to determine full diploid genotypes for αD-globin in each specimen . The sequences were analyzed using Geneious Pro ver . 5 . 4 . 3 . All new sequences were deposited in GenBank under accessions numbers KT988975-KT988992 and KU160516-KU160529 . For each amino acid difference between pairs of high- and low-altitude sister taxa , we identified ancestral and derived states by comparison with orthologous sites in a large number of other waterfowl species ( n = 117 sequences for αA-globin , 96 for βA-globin , and 57 for αD-globin ) . Alignments of variable sites in the αA- , βA- , and αD-globin genes are shown in S3 , S4 and S5 Figs , respectively . For each divergent site in each pair of sister taxa , unordered parsimony ( using the trace character function in Mesquite [75] ) yielded unambiguous inferences of character polarity . One notable case of homoplasy in the βA-globin gene involved sites 116 and 133 in high-altitude yellow-billed pintails and speckled teal ( S4 Fig ) , two species that are known to hybridize in nature [40 , 41] . To assess whether identical two-site ‘β116Ser-β133Met’ haplotypes from the two species were identical-by-descent , we reconstructed haplotype networks of βA-globin coding sequence using the median-joining algorithm [76] , as implemented in the program Network 4 . 6 ( Fluxus Technology , Suffolk , UK ) . We conducted the analysis on a sample of 257 βA-globin sequences ( 116 from yellow-billed pintails and 141 from speckled teal ) obtained from sympatric high- and low-altitude populations of both species . Sixty individuals representing three species of Andean ducks ( speckled teal , cinnamon teal , and yellow-billed pintail ) were selected for genome-wide surveys of nucleotide variation using single-digest RAD-Seq [77] . For each species , ten male specimens were selected from high-altitude ( ≥3 , 211 m above sea level ) , and ten were selected from low-altitude ( ≤ 914 m ) . Total genomic DNA was extracted from muscle tissue using a DNeasy Tissue Kit ( Qiagen , Valencia , California , USA ) and normalized using a Qubit Fluorometer ( Invitrogen , Grand Island , New York , USA ) . DNA samples were submitted to Floragenex ( Eugene , Oregon , USA ) for single-digest RAD-Seq using SbfI , which recognizes an 8-nucleotide ( CCTGCAGG ) restriction site . Digested DNAs were ligated to barcodes and sequencing adaptors and then sequenced on the Illumina HiSeq 2000 with single-end 100 bp chemistry . Following Illumina sequencing , sequences were demultiplexed and trimmed to yield RAD sequences of 90 bp . Data analysis and bioinformatics pipelines were provided by Floragenex [77–79] . The Floragenex RAD unitag assembler and BSP pipelines v . 2 . 0 were used to create a RAD-Seq ‘unitag’ assembly and Bowtie alignments of SAMtools pileup sequences to the reference assembly . Genotypes at each nucleotide site were determined using the VCF popgen v . 4 . 0 pipeline to generate a customized VCF 4 . 1 ( variant call format ) database with parameters set as follows: minimum AF for genotyping = 0 . 075 , minimum Phred score = 15 , minimum depth of sequencing coverage = 10x , and allowing missing genotypes from up to 10% of individuals at each site . To filter out base calls that were not useful due to low quality scores or insufficient coverage , genotypes at each nucleotide site were inferred using the Bayesian maximum likelihood algorithm described by Hohenlohe et al . [79] . This algorithm calculates the likelihood of each possible genotype at each site using a multinomial sampling distribution , which gives the probability of observing a set of read counts ( n1 , n2 , n3 , n4 ) for a particular genotype , where ni is the read count for each of the four possible nucleotides at each site , excluding ambiguous reads with low quality scores . The genotyping algorithm incorporates the site-specific sequencing error rate , and assigns the most likely diploid genotype to each site using a likelihood ratio test and significance level of α = 0 . 05 . A total of 372 million sequence reads were obtained with an average depth of 7 . 6 ( ±2 . 4 SD ) million reads per sample for yellow-billed pintail and speckled teal and 3 . 3 ( ±1 . 4 SD ) million reads per sample for cinnamon teal , corresponding to an average of 140 , 671 ( ±27 , 856 ) RAD loci . After filtering and genotyping , the RAD-Seq survey yielded 49 , 670 SNPs associated with 18 , 998 distinct loci in yellow-billed pintail , 47 , 731 SNPs associated with 19 , 433 distinct loci in speckled teal , and 18 , 145 SNPs associated with 9 , 300 distinct loci in cinnamon teal , respectively . The mean depth of coverage was 36 . 8 ( ±10 . 0 SD ) reads per site with an average per site quality score of 166 . 2 ( ±31 . 3 SD ) for yellow-billed pintail and speckled teal , and 39 . 8 ( ±24 . 4 SD ) reads per site with an average per site quality score of 177 . 6 ( ±26 . 4 SD ) for cinnamon teal . Illumina reads were submitted to the European Nucleotide Archive and can be accessed under the short read archive ( SRA ) accession number PRJEB11624 . Sequencing coverage and quality scores were summarized using the software VCFtools v . 0 . 1 . 11 [80] . Custom perl scripts were first used to filter triploid or tetraploid sites and convert the Floragenex-generated VCF file to a biallelic , VCF v4 . 0 compatible format . We then calculated Weir and Cockerham’s [81] estimator of FST for each SNP in comparisons between high- and low-altitude population samples . We purified HbA and HbD variants from hemolysates of 1–4 specimens per species , all of which had known αA- , αD- , and βA-globin genotypes . In the case of ruddy ducks and yellow-billed pintails , previous population surveys of sequence polymorphism in the αA- and βA-globin genes had revealed multiple amino acid haplotypes segregating within high- and/or low-altitude populations [35 , 36] . In each case we purified HbA and HbD variants from individuals that were homozygous for each of the alternative allelic variants . Hemolysates of each individual specimen were dialyzed overnight against 20 mM Tris buffer ( pH 8 . 4 ) . The two tetrameric HbA and HbD isoforms were then separated using a HiTrap Q-HP column ( GE Healthcare; 1 ml 17-1153-01 ) and equilibrated with 20 mM Tris buffer ( pH 8 . 4 ) . HbD was eluted against a linear gradient of 0–200 mM NaCl . The samples were desalted by means of dialysis against 10 mM HEPES buffer ( pH 7 . 4 ) at 4°C , and were then concentrated by using a 30 kDa centrifuge filter ( Amicon , EMD Millipore ) . We measured O2-equilibria of purified Hb solutions under standard conditions ( 37°C , pH 7 . 4 , 0 . 3 mM heme ) using a modified diffusion chamber where absorption at 436 nm was monitored during stepwise changes in equilibration gas mixtures generated by precision Wösthoff gas-mixing pumps [28 , 32 , 39 , 56 , 82 , 83] . In order to characterize intrinsic Hb-O2 affinities and mechanisms of allosteric regulatory control , we measured O2-equilibria in the presence of Cl- ions ( 0 . 1M KCl ) , in the presence of IHP ( IHP/Hb tetramer ratio = 2 . 0 ) , in the simultaneous presence of both effectors , and in the absence of both effectors ( stripped ) . Free Cl- concentrations were measured with a model 926S Mark II chloride analyzer ( Sherwood Scientific Ltd , Cambridge , UK ) . We estimated values of P50 and n50 ( Hill’s cooperativity coefficient at half-saturation ) by fitting the Hill equation Y = PO2n/ ( P50n + PO2n ) to the experimental O2 saturation data by means of nonlinear regression ( Y = fractional O2 saturation; n , cooperativity coefficient ) . The model-fitting was based on 5–8 equilibration steps between 30% and 70% oxygenation . The αA- and βA-globin sequences of yellow-billed pintail were synthesized by Eurofins MWG Operon ( Huntsville , AL , USA ) after optimizing the nucleotide sequences in accordance with E . coli codon preferences . The synthesized αA-βA globin gene cassette was cloned into a custom pGM vector system along with the methionine aminopeptidase ( MAP ) gene , as described by Natarajan et al . [27 , 84] . We engineered each of the β-chain codon substitutions using the QuikChange II XL Site-Directed Mutagenesis kit from Stratagene ( LaJolla , CA , USA ) . Each engineered codon change was verified by DNA sequencing . Recombinant Hb expression was carried out in the JM109 ( DE3 ) E . coli strain as described in Natarajan et al . [27 , 84] . To ensure the post-translational cleaving of N-terminal methionines from the nascent globin chains , we co-transformed a plasmid ( pCO-MAP ) containing an additional copy of the MAP gene . Both pGM and pCO-MAP plasmids were cotransformed and subject to dual selection in an LB agar plate containing ampicillin and kanamycin . The expression of each rHb mutant was carried out in 1 . 5 L of TB medium . Bacterial cells were grown in 37°C in an orbital shaker at 200 rpm until absorbance values reached 0 . 6–0 . 8 at 600 nm . The bacterial cultures were induced by 0 . 2 mM IPTG and were then supplemented with hemin ( 50 μg/ml ) and glucose ( 20 g/L ) . The bacterial culture conditions and the protocol for preparing cell lysates were described previously [27–29 , 32 , 84] . The bacterial cells were resuspended in lysis buffer ( 50 mM Tris , 1 mM EDTA , 0 . 5 mM DTT , pH 7 . 6 ) with lysozyme ( 1 mg/g wet cells ) and were incubated in the ice bath for 30 min . Following sonication of the cells , 0 . 5–1 . 0% polyethylenimine solution was added , and the crude lysate was then centrifuged at 15000 g for 45 min at 4°C . The rHbs were purified by two-step ion-exchange chromatography . Using high-performance liquid chromatography , the samples were passed through a prepacked anion-exchange column ( Q-Sepharose ) followed by passage through a cation-exchange column ( SP-Sepharose ) . The clarified supernatant was subjected to overnight dialysis in CAPS buffer ( 20 mM CAPS with 0 . 5mM EDTA , pH 9 . 7 ) at 4°C . The samples were passed through the Q-column and the rHb solutions were eluted against a linear gradient of 0–1 . 0 M NaCl . The eluted samples were desalted by overnight dialysis with 20 mM HEPES pH 7 . 4 ( 4°C ) . Dialyzed samples were then passed through the SP-Sepharose column ( HiTrap SPHP , 1 mL , 17-1151-01; GE Healthcare ) equilibrated with 20 mM HEPES ( pH 7 . 4 ) . The rHb samples were eluted with a linear gradient of 20 mM HEPES ( pH 9 . 2 ) . Samples were concentrated and desalted by overnight dialysis against 10 mM HEPES buffer ( pH 7 . 4 ) and were stored at -80°C prior to the measurement of O2-equilibrium curves . The purified rHb samples were analyzed by means of sodium dodecyl sulphate ( SDS ) -polyacrylamide gel electrophoresis . After preparing rHb samples as oxyHb , deoxyHb , and carbonmonoxy derivatives , we measured absorbance at 450–600 nm to confirm that the absorbance maxima match those of the native HbA samples . Results of isoelectric focusing analyses indicated that each of the purified rHb mutants was present as a tetrameric assembly , and this was further confirmed by cooperativity coefficients ( n50 ) >1 . 00 in the O2-equilibrium experiments . In vitro measurements of O2-binding properties were conducted in the same manner for rHbs and native Hb samples . Homology-based structural modeling was performed with Modeller 9 . 15 [85] using human Hbs in different ligation states ( PDB , 2hhb and 1hho ) as templates . Models were evaluated on the SWISS-MODEL server [86] . All models had QMEAN values between 0 . 71 and 0 . 78 . Structural mining was performed using PISA [87] , PyMol ( Schrödinger , New York , NY ) , and SPACE [88] .
|
The convergent evolution of similar traits in different species could be due to repeated changes at the genetic level or different changes that produce the same phenotypic effect . To investigate the extent to which convergence in phenotype is caused by repeated mutations , we investigated the molecular basis of convergent changes in the oxygenation properties of hemoglobin ( Hb ) in eight pairs of high- and low-altitude waterfowl taxa from the Andes . The results revealed that convergent increases in Hb-O2 affinity in highland taxa involved a combination of unique and repeated amino acid replacements . However , convergent changes in Hb function generally did not involve parallel substitutions , indicating that repeatability in the evolution of protein function does not require repeatability at the sequence level .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
Convergent Evolution of Hemoglobin Function in High-Altitude Andean Waterfowl Involves Limited Parallelism at the Molecular Sequence Level
|
The goal of human genome re-sequencing is obtaining an accurate assembly of an individual's genome . Recently , there has been great excitement in the development of many technologies for this ( e . g . medium and short read sequencing from companies such as 454 and SOLiD , and high-density oligo-arrays from Affymetrix and NimbelGen ) , with even more expected to appear . The costs and sensitivities of these technologies differ considerably from each other . As an important goal of personal genomics is to reduce the cost of re-sequencing to an affordable point , it is worthwhile to consider optimally integrating technologies . Here , we build a simulation toolbox that will help us optimally combine different technologies for genome re-sequencing , especially in reconstructing large structural variants ( SVs ) . SV reconstruction is considered the most challenging step in human genome re-sequencing . ( It is sometimes even harder than de novo assembly of small genomes because of the duplications and repetitive sequences in the human genome . ) To this end , we formulate canonical problems that are representative of issues in reconstruction and are of small enough scale to be computationally tractable and simulatable . Using semi-realistic simulations , we show how we can combine different technologies to optimally solve the assembly at low cost . With mapability maps , our simulations efficiently handle the inhomogeneous repeat-containing structure of the human genome and the computational complexity of practical assembly algorithms . They quantitatively show how combining different read lengths is more cost-effective than using one length , how an optimal mixed sequencing strategy for reconstructing large novel SVs usually also gives accurate detection of SNPs/indels , how paired-end reads can improve reconstruction efficiency , and how adding in arrays is more efficient than just sequencing for disentangling some complex SVs . Our strategy should facilitate the sequencing of human genomes at maximum accuracy and low cost .
The human genome is comprised of approximately 6 billion nucleotides on two pairs of 23 chromosomes . Variations between individuals are comprised of ∼6 million single nucleotide polymorphisms ( SNPs ) and ∼1000 relatively large structural variants ( SVs ) of ∼3 kb or larger and many more smaller SVs are responsible for the phenotypic variation among individuals [1] , [2] . Most of these large SVs are due to genomic rearrangements ( e . g . duplication and deletion ) , and a few others contain novel sequences that are not present in the reference genome [3] . The goal of personal genomics is to determine all these genetic differences between individuals and to understand how these contribute to phenotypic differences in individuals . Making personal genomics almost a reality over the past decade , the development of high throughput sequencing technologies has enabled the sequencing of individual genomes [3] , [4] . In 2007 , Levy et al . reported the sequencing of an individual's genome based on Sanger [5] whole-genome shotgun sequencing , followed by de novo assembly strategies . Wheeler et al . in 2008 presented another individual's genome sequence constructed from 454 sequencing reads [6] and comparative genome assembly methods . In the mean time , other new sequencing technologies such as Solexa/Illumina sequencing [7] have become available for individual genome sequencing with corresponding , specially-designed sequence assembly algorithm designed [8]–[12] . These projects and algorithms , however , mostly relied on a single sequencing technology to perform individual re-sequencing and thus did not take full advantage of all the existing experimental technologies . Table 1 gives a summary of the characteristics of several technologies in comparative individual genome sequencing . At one extreme , performing long Sanger sequencing with a very deep coverage will lead to excellent results at high cost . In another , performing only the inexpensive and short Illumina sequencing may generate good and cost-efficient results in SNP detection , but will not be able to either unambiguously locate some of the SVs in repetitive genomic regions or fully reconstruct many of the large SVs . Moreover , array technologies such as the SNP array [1] and the CGH array at different resolutions [13]–[16] can also be utilized to identify the SVs: the SNP arrays can detect SNPs directly , and the CGH array is able to detect kilobase- ( kb ) to megabase- ( mb ) sized copy number variants ( CNV ) [17] , which can be integrated into the sequencing-based SV analysis . It is thus advantageous to consider optimally combining all these experimental techniques into the individual genome re-sequencing framework and to design experiment protocols and computational algorithms accordingly . Due to the existence of reference genome assemblies [18] , [19] and the high similarity between an individual's genome and the reference [3] , the identification of small SVs is relatively straightforward in comparative re-sequencing with the analysis of single split-reads covering small SVs . Meanwhile , although there exist algorithms to detect large SVs with paired-end reads [2] , the complete reconstruction of a large SV requires the integration of reads spanning a wide region , often involving misleading reads from other locations of the genome . If there were no repeats or duplications in the human genome , the reconstruction of such large SVs would be trivially accomplished by the de novo assembly with a high coverage of inexpensive short reads around these regions . With the existence of repeats and duplications in the human genome , however , a set of longer reads will be required to accurately locate some of these SVs in repetitive regions , and a hybrid re-sequencing strategy with both comparative and de novo approaches will be necessary to identify genomic rearrangement events such as deletions and translocations , and also to reconstruct large novel insertions in individuals . Such steps are thus much harder than the others , and will be the main focus of this paper . Here we present a toolbox and some representative case studies on how to optimally combine the different experimental technologies in the individual genome re-sequencing project , especially in reconstructing large SVs , so as to achieve accurate and economical sequencing . An “optimal” experimental design should be an intelligent combination of the long , medium , and short sequencing technologies and also some array technologies such as CGH . Some of the previous genome sequencing projects [20] , [21] have already incorporated such hybrid approaches using both long and medium reads , although the general problem of optimal experimental design has not yet been systematically studied . While it is obvious that combining technologies is advantageous , we want to quantitatively show the potential savings based on different integration strategies . Also , since the technologies are constantly developing , it will be useful to have a general and flexible approach to predict the outcome of integrating different technologies , including the new ones coming in the future . In the following sections , we will first briefly describe a schematic comparative genome re-sequencing framework , focusing on the intrinsically most challenging steps of reconstructing large SVs , and then use a set of semi-realistic simulations of these representative steps to optimize the integrated experimental design . Since full simulations are computationally intractable for such steps in the large parameter space of combinations of different technologies , the simulations are carried out in a framework that can combine the real genomic data with analytical approximations of the sequencing and assembly process . Also , this simulation framework is capable of incorporating new technologies as well as adjusting the parameters for existing ones , and can provide informative guidelines to optimal re-sequencing strategies as the characteristics and cost-structures of such technologies evolve , when combining them becomes a more important concern . The simulation framework is downloadable as a general toolbox to guide optimal re-sequencing as technology constantly advances .
The hybrid genome assembly strategy incorporates both comparative [22] and de novo methods . On one hand , most of the assembly can be done against the reference , and it will be unnecessary to perform a computationally intensive whole genome de novo assembly . Comparative approaches will be capable of identifying small SVs and large rearrangement events . On the other hand , de novo assembly will sometimes still be useful in reconstructing regions with large and novel SVs . Fig . 1 shows the schematic steps of SV reconstruction in the context of the genome sequencing/assembly process . The data from different sequencing/array experiments can be processed in the following way: As shown in Fig . 1A and 1B , with errors corrected [23] and short reads combined into “unipaths” [10] , all the reads ( long/medium/short ) from the individual's genome can be mapped back to the reference genome . In Fig . 1C , the SNPs can then be identified immediately based on the reads with single best matches , and the boundaries of deletions or small insertions will be detected by such reads as well ( allowing gaps in alignment ) . Meanwhile , haplotype islands can also be extracted based on the paired-end information [3] , [24] , [25] and the prior knowledge of the population haplotype patterns revealed by previous work [26] . Further analysis of the single/paired-end reads are required to reconstruct the large SVs ( Fig . 1D and 1E ) , which are by nature more complicated than identifying small SVs . First of all , locations of such SV events need to be detected by analyzing the split-reads ( shown in Fig . 2A and 2B ) that cover their boundaries . Second , two distinct types of SVs need to be handled separately: de novo assembly is required to reconstruct large novel insertions , and comparative algorithms should be utilized to identify genomic rearrangement events ( e . g . segmental duplication/deletion ) . The homozygosity/heterozygosity of such SVs can be determined based on the existence of the reads that map back to the corresponding reference sequences . Fig . 2A–C show the overall process of de novo assembly for large novel insertions . While the reconstruction of such regions mostly depends on the spanning-reads from the new inserted sequence , misleading-reads from elsewhere in the genome can often hinder the full reconstruction process . These reads usually comes from the highly represented regions in the genome , which also exist in the insertion . In such cases , reads longer than such regions and appropriate assembly strategies are needed to ensure the unambiguous and correct assembly output . Paired-end reads with an appropriate gap size can also help the unambiguous mapping of the reads inside novel insertions [2] . Fig . 2D illustrates the comparative identification of rearrangements from the reference sequence . CGH array data can be integrated into the reconstruction process of such SVs . For long rearrangements detected by sequencing data , the CGH data can be utilized in both validation and correction of large segmental duplications/deletions . What is more , incorporating the CGH data can also lower the coverage depth requirement of sequencing experiments , since the inner ( i . e . not close to SV boundaries ) regions of segmental duplications/deletions not covered by sequencing reads can still be identified by CNV results . An example is shown in Fig . 2D: Although the sequence reads can detect the SV event in region A , B and C , they may not be sufficient to distinguish deletions from translocations when the sequencing coverage is relatively low . The copy numbers of the genomic regions inferred from CGH array data can be integrated into the rearrangement analysis , and provide additional evidence of the SV types . It is important for us to define a reasonable performance metric so that the re-sequencing approach can be designed in such a way that its outcome will be optimized according to that metric . For large SVs , the metric can be defined based on the alignment result of the actual variant sequence and the inferred variant sequence . For a large SV due to genomic rearrangements ( e . g . deletion , duplication ) , it is natural to define its recovery rate as either 1 ( detected ) or 0 ( missed ) . For a large novel insertion , on the other hand , we may want to take into account cases where the insertion is detected but its sequence content is not reconstructed with full accuracy . Hence , we define the recovery rate of such a large novel insertion as follows based on its reconstruction percentage:in which SV is the actual insertion ( in simulations , it is already known; in reality , it will need to be identified in a validation step ) , SVinf is the insertion sequence inferred by the genome re-sequencing approach , mismatch returns the number of mismatches of two aligned sequences , wflanking returns a sequence with its flanking sequences on both ends , and size returns the size of a sequence . The purpose of introducing flanking sequences is to take into account the accuracy of the predicted location of the SV . Based on the schematic assembly strategy and the performance measure defined in the previous sections , we can simulate the sequence assembly process in order to obtain an optimal set of parameters for the design of the sequencing experiments ( e . g . the amount of long ( Sanger ) , medium ( 454 ) and short ( Illumina ) reads , the amount of single and paired-end reads ) and the array experiments ( e . g . the incorporation of CGH arrays ) to achieve the desired performance with a relatively low cost in the individual genome re-sequencing project . Here we present the results of a set of simulation case studies on reconstructing large SVs , which are in general much more challenging problems compared to the detection of small SVs . In order to fully reconstruct a long novel insertion , for instance , one needs to not only detect the insertion boundaries based on the split-reads , but also assemble the insertion sequence from the spanning- and misleading-reads . For the identification of genomic rearrangements such as deletion/translocations , one may also want to incorporate array data to increase the confidence level of such analysis . The simulations described in this section are based on large ( ∼10 kb , ∼5 Kb and ∼2 Kb ) novel insertions and deletions discovered by Levy et al . [3] , and they perform semi-realistic whole genome assembly representative using the sequence characteristics of both the NCBI reference genome [18] and the target HuRef genome [3] . The sequencing/array technologies considered in these simulations are long , medium and short sequencing methods and CGH arrays , as shown in Table 1 . Paired-end reads are also included in these simulations . One major challenge in implementing these simulations is to design them in a computationally realistic way . Brute-force full simulations of whole-genome assembly in this case would be unrealistic: thousands of possible combinations of different technologies will need to be tested , and for each of these combinations hundreds of genome assembly simulations need to be carried out to obtain the statistical distributions of their performance . Since a full simulation of one round of whole-genome assembly will probably take hundreds of CPU hours to finish , the full simulation to explore the full space of technology combinations will then require hundreds of millions ( ∼108 ) of CPU hours , equivalent to ∼10 years with 1000 CPUs . We designed the simulations using analytical approximations of the whole-genome assembly process in order for them to be both time and space efficient , and the gain in efficiency is summarized in Table 2 and will be described in details later in the Materials and Methods section . We have also made this simulation framework publicly available as a toolbox that can incorporate technology advancements as well as other SV regions .
The simulation results in the previous section are based on three sequencing technologies and an idealized array technology , and assume a specific parameterization of their characteristics and costs . Thus , the particular optimal solutions found may not be immediately applicable to a real individual genome re-sequencing project . However , these results illustrate quantitatively how we can design and run simulations to obtain guidelines for optimal experimental design in such projects . Since our simulation approach is based on the general concept of mapability map and comparative SV reconstruction instead of on a specific organism , it can also be adapted to the comparative sequencing of a non-human genome with regard to a closely related reference . In such a study , we can first construct an artificial target genome based on estimations of its divergence from the reference , and then compute the mapability maps of those representative SVs as input to the simulation framework to find the optimal combination of technologies . Obviously , the closer the two genomes are , the more informative the simulation result would be . In cases where it is hard to estimate the divergence of the target genome from the reference , a two-step approach can be conducted: First , combined sequencing experiments will be carried out using an optimal configuration obtained from the simulation based on the “best guess” , such as another closely related genome . Second , by using the target genome constructed in the previous step , a new set of simulations can be executed and their results can guide a second round of combined sequencing which can provide a finer re-sequencing outcome when combined with the previous sequencing data . Meanwhile , our simulation framework specifically focuses on the effects of misleading reads in the SV reconstruction process , and it will be the most helpful in cases where the target and reference genome both have complex repetitive/duplicative sequence characteristics which will introduce such reads . In this paper , we propose to optimally incorporate different experimental technologies in the design of an individual genome-sequencing project , especially for the full reconstruction of large SVs , to achieve accurate output with relatively low costs . We first describe a hybrid genome re-sequencing strategy for detecting SVs in the target genome , and then propose how we can design the optimal combination of experiments for reconstructing large SVs based on the results of semi-realistic simulations with different single and paired-end reads . We also present several examples of such simulations , focusing on the reconstruction of large novel insertions and confirmation of large deletions based on CNV analysis , which are the most challenging steps in individual re-sequencing . The simulations for actual sequencing experimental design can integrate more technologies with different characteristics , and also test the sequencing/assembly performance at different SV levels . By doing so , a set of experiments based on various technologies can be integrated to best achieve the ultimate goal of an individual genome re-sequencing project: accurately detecting all the nucleotide and structural variants in the individual's genome in a cost-efficient way . Such information will ultimately prove beneficial in understanding the genetic basis of phenotypic differences in humans .
The NCBI assembly v36 [18] and the HuRef assembly [3] were used as reference and target genomes , respectively . Three sequencing technologies , long ( Sanger ) , medium ( 454 ) , and short ( Illumina ) sequencing , were considered with the characteristics summarized in Table 1 . We also assumed that the per-base sequencing error rate increases linearly from the start to the end of a read similar to ReadSim [28] , and assigned error types ( insertion , deletion or substitution ) randomly according to the characteristics of the sequencing technique used [6] , [7] , [28] . The novel SVs used in the novel insertion reconstruction simulation are ∼10 Kb , ∼5 Kb and ∼2 Kb insertion sequences in the HuRef genome [3] with variant IDs 1104685256488 , 1104685222085 and 1104685613186 , respectively . The deletion used in the CNV analysis simulation is a ∼18 Kb sequence in the HuRef genome with variant ID 1104685125828 . Since we would be testing thousands of possible combinations of the long , medium and short sequencing technologies , it would be unrealistic ( both time and space consuming ) to generate for each combination all the reads from the whole target genome and then apply any existing assembler to these reads . We decided to semi-realistically simulate the assembly process of large novel insertions to achieve relatively accurate estimates in an affordable amount of time . Several difficulties need to be addressed by such a simulation: 1 ) One of the most time-consuming step in a real assembler is the read overlap-layout step . 2 ) The whole-genome sequencing experiment introduces large numbers of misleading reads that are partially similar to the reads from the targeted genomic region , which would require an huge storage space in a real assembly process . In this simulation , we assume that the boundaries of a large deletion event have already been identified by sequence reads , and we are simulating the process of determining whether this is a deletion or translocation event , based on the short reads alone or on the idealized CGH data . The reads are generated in a similar fashion as described in the previous section , without considering sequencing errors for simplicity . The idealized CGH signal of a corresponding region r is defined as Gaussian variable with mean M ( r , TargetG ) , and noise/standard deviation = 0 . 05 , 0 . 1 , 0 . 2 . For each dataset , the log-ratio of the posterior probability of the deletion event is computed to represent the confidence level provided by each dataset for determining that deletion . These confidence levels are computed according the following formulas:where sub ( s , a , b ) returns the sub-sequence of s from a to b-1 ( 1-based index ) , l is the length of the short read , SV stands for the deleted region , coveragereads is the sequencing coverage , obs ( r ) is the number of observed reads that are the same as r , sig ( r ) is the normalized CGH-array signal of probe r , PDF{D , v} is the probability density/mass function of the distribution D at value v , and RefG/TargetG refers to the reference/target genome .
|
In recent years , the development of high throughput sequencing and array technologies has enabled the accurate re-sequencing of individual genomes , especially in identifying and reconstructing the variants in an individual's genome compared to a “reference” . The costs and sensitivities of these technologies differ considerably from each other , and even more technologies are expected to appear in the near future . To both reduce the total cost of re-sequencing to an affordable point and be adaptive to these constantly evolving bio-technologies , we propose to build a computationally efficient simulation framework that can help us optimize the combination of different technologies to perform low cost comparative genome re-sequencing , especially in reconstructing large structural variants , which is considered in many respects the most challenging step in genome re-sequencing . Our simulation results quantitatively show how much improvement one can gain in reconstructing large structural variants by integrating different technologies in optimal ways . We envision that in the future , more experimental technologies will be incorporated into this simulation framework and its results can provide informative guidelines for the actual experimental design to achieve optimal genome re-sequencing output at low costs .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biotechnology",
"molecular",
"biology/bioinformatics",
"computer",
"science",
"computational",
"biology/genomics"
] |
2009
|
Integrating Sequencing Technologies in Personal Genomics: Optimal Low Cost Reconstruction of Structural Variants
|
Kinetoplastid parasites—trypanosomes and leishmanias—infect millions of humans and cause economically devastating diseases of livestock , and the few existing drugs have serious deficiencies . Benzoxaborole-based compounds are very promising potential novel anti-trypanosomal therapies , with candidates already in human and animal clinical trials . We investigated the mechanism of action of several benzoxaboroles , including AN7973 , an early candidate for veterinary trypanosomosis . In all kinetoplastids , transcription is polycistronic . Individual mRNA 5'-ends are created by trans splicing of a short leader sequence , with coupled polyadenylation of the preceding mRNA . Treatment of Trypanosoma brucei with AN7973 inhibited trans splicing within 1h , as judged by loss of the Y-structure splicing intermediate , reduced levels of mRNA , and accumulation of peri-nuclear granules . Methylation of the spliced leader precursor RNA was not affected , but more prolonged AN7973 treatment caused an increase in S-adenosyl methionine and methylated lysine . Together , the results indicate that mRNA processing is a primary target of AN7973 . Polyadenylation is required for kinetoplastid trans splicing , and the EC50 for AN7973 in T . brucei was increased three-fold by over-expression of the T . brucei cleavage and polyadenylation factor CPSF3 , identifying CPSF3 as a potential molecular target . Molecular modeling results suggested that inhibition of CPSF3 by AN7973 is feasible . Our results thus chemically validate mRNA processing as a viable drug target in trypanosomes . Several other benzoxaboroles showed metabolomic and splicing effects that were similar to those of AN7973 , identifying splicing inhibition as a common mode of action and suggesting that it might be linked to subsequent changes in methylated metabolites . Granule formation , splicing inhibition and resistance after CPSF3 expression did not , however , always correlate and prolonged selection of trypanosomes in AN7973 resulted in only 1 . 5-fold resistance . It is therefore possible that the modes of action of oxaboroles that target trypanosome mRNA processing might extend beyond CPSF3 inhibition .
Kinetoplastid protists cause severe human diseases affecting millions of people . Trypanosoma cruzi causes Chagas disease in South America , and various Leishmania species cause a spectrum of diseases throughout the tropics . Salivarian trypanosomes , the subject of this study , cause sleeping sickness in humans and economically important diseases in cattle , horses and camels [1–3] . Approximately 70 million people , living in sub- Saharan Africa , are estimated to be at risk of contracting human African trypanosomosis , which is caused by Trypanosoma brucei subspecies [4 , 5] . As a result of sustained international activities to control the disease [4–6] , less than 3000 cases were reported in 2016 ( http://www . who . int/trypanosomiasis_african/en/ ) . Trypanosomosis in cattle , caused by infection with Trypanosoma congolense , Trypanosoma vivax and , to a lesser extent , T . brucei , is in contrast a major problem , with wide-reaching effects on human well-being: cattle are used not only as a source of milk and meat , but also for traction . Elimination of cattle trypanosomosis could create economic benefits estimated at nearly 2 . 5 billion US$ per year [2] . Within Africa , trypanosomosis is transmitted by tsetse flies , but outside Africa , variants of T . brucei are transmitted venereally or by biting flies , and T . vivax can also be transmitted non cyclically by non-tsetse biting flies with massive economic losses affecting draught and milk animals from Argentina to the Philippines [7] . Control of cattle trypanosomosis currently relies on reducing the tsetse population by means of traps and insecticidal dips , together with treatment as required . The most popular treatment is with the diamidine diminazene aceturate ( Berenil ) , the alternative being the DNA-intercalating molecule isometamidium [3] . Suramin is also sometimes used [3] . Development of new animal therapeutics is constrained by the need for cure after a single intramuscular injection [3] . In the last ten years , benzoxaboroles have generated considerable excitement for antimicrobial and other applications . Benzoxaboroles have a range of known effects . For example , they are able to bind cis-diols , such as those found in sugars , yielding stable spiro complexes [8] . This activity is the basis of the mode of action of the antifungal drug Tavaborole ( AN2690 ) [9] , which binds to the editing site of leucyl tRNA synthetase [10] . Other oxaboroles inhibit bacterial leucyl tRNA synthetases by the same mechanism [11–13] . Other benzoxaborole classes interact with ATP-binding pockets , but with different modes of binding . Crisaborole , approved for the treatment of atopic dermatitis [14 , 15] , selectively inhibits phosphodiesterase PDE4 , with the oxaborale oxygen atoms coordinating the zinc and magnesium ions within the active site [16] . The aminomethylphenoxy benzoxaboroles inhibit Rho-activated protein kinases through hydrogen bond interactions with the hinge region of the protein and the aminomethyl group interacting with the magnesium/ATP-interacting aspartic acid [17] . The first trypanocidal oxaboroles to be described were the oxaborole 6-carboxamides [18 , 19] . Acoziborole ( trade name of SCYX-7158 or AN5568 , 1 ) [20] is orally available and can cross the blood brain barrier [18]; it is now in phase IIb/III clinical trials for sleeping sickness ( see https://www . dndi . org/diseases-projects/portfolio/ ) . AN7973 ( Fig 1 ) , the main subject of this paper , is efficacious against T . congolense and T . brucei and was considered as a candidate for treatment of cattle trypanosomosis , but was later replaced by AN11736 ( Fig 1 ) , which can achieve single-dose cure of both T . congolense and T . vivax infection in cattle [21] . At present , not enough is known about structure-activity relationships in benzoxaboroles to enable predictions to be made about their modes of action . Jones et al . [22] obtained T . brucei lines with 4-5-fold resistance to AN2965 ( Fig 1 , named oxaborole-1 in their paper ) . They observed numerous single nucleotide polymorphisms ( SNPs ) , genome rearrangements and amplifications in the resistant lines . Affinity purification with a benzoxaborole column yielded 14 proteins that bound specifically , but since none of these was affected in the resistant mutants , there was no clear indication as to which might be relevant to benzoxaborole action [22] . Aminomethyl phenoxyl benzoxaboroles such as AN3056 are sequentially activated to an active carboxylic acid form by serum and intracellular enzymes [23] . Various oxaboroles are being considered for treatment of apicomplexan parasites; and AN13762 ( Fig 1 ) is in development [24 , 25] . Plasmodium and Toxoplasma that were resistant to AN3661 ( Fig 1 ) had mutations in the cleavage and polyadenylation factor CPSF3 [26 , 27] , which is implicated in 3' cleavage of mRNA precursors prior to polyadenylation [28] . The two zinc ions of CPSF3 are thought to interact with phosphate [28] . In silico molecular docking of AN3661 suggested that the boron atom occupies the position of the cleavage site phosphate of the mRNA substrate , with one hydroxyl group interacting with a zinc atom in the catalytic site—similar to binding of other benzoxaboroles to the bimetal centers of beta-lactamase and phosphodiesterase-4 [26 , 27] . Introduction of resistance mutations—which were all in or near the active site—into susceptible parasites resulted in compound resistance . These results , combined with the loss of transcripts for three trophozolite-expressed genes in treated parasites , suggest that AN3661 inhibits mRNA polyadenylation through its interaction with CPSF3 [26 , 27] . In trypanosomes , transcription of mRNAs is polycistronic . Individual mRNA 5'-ends are created co-transcriptionally by trans splicing of a 39nt leader sequence ( SL ) [29–31] . Trans splicing is spatially and mechanistically coupled to polyadenylation of the preceding mRNA . Polyadenylation sites are dictated by the positions of trans-splicing sites [32–35] , RNAi-mediated depletion of polyadenylation factors inhibits trans splicing [36 , 37] , and disruption of splicing stops polyadenylation [32–35] . The spliced leader precursor RNA ( SLRNA ) is 139 nt long and is synthesised by RNA polymerase II [38] from approximately 200 tandemly repeated genes [39] . Unlike protein-coding genes , each SLRNA gene has its own promoter [40–42] . The SLRNA cap and the following four nucleotides are methylated [43–46] . Inhibition of SLRNA methylation using S-adenosyl-L-homocysteine or Sinefungin prevents splicing [47 , 48] . This results in loss of mRNA from the cells by the normal mechanisms of mRNA turnover [49] . It has recently been shown that an accumulation of numerous metabolites , including S-adenosylmethionine , occurs in trypanosomes treated with Sinefungin , and a similar profile was identified in trypanosomes treated with acoziborole [50] . We here describe studies to discover molecular targets of several anti-trypanosomal benzoxaboroles . Our work concentrated on AN7973 ( Fig 1 ) , which is orally active and was the DNDi back-up for SCYX-7158/AN5568 for the treatment of human African trypanosomosis . We undertook a comprehensive analysis , including resistance generation and characterisation of morphology , metabolomes , macromolecular biosynthesis and molecular modeling .
AN7973 ( Fig 1 ) was selected as a candidate veterinary drug from a range of 7-carboxamido-benzoxaboroles structurally similar to AN5568 . The selection of AN7973 was based on its in vitro potency against T . congolense ( S1 Table , sheet 1 ) and its ability to cure T . congolense-infected mice with a single 10mg/kg i . p . dose ( S1 Table , sheet 3 ) . T . congolense-infected goats were also cured when AN7973 was administered as a single bolus dose injection of 10 mg/kg , but for T . vivax-infected goats , two intramuscular injections of 10 mg/kg were required ( S1 Table , sheet 3 ) . Testing in cattle was done using T . vivax and T . congolense isolates that were resistant to maximum dosages of diminazene ( 7 mg/kg ) and isometamidium ( 1 mg/kg ) . A single 10 mg/kg intramuscular injection of AN7973 cured 3/3 cattle of T . congolense infection , but two 10 mg/kg injections of AN7973 cured only one out of two T . vivax infections and a single injection failed to cure 3 animals ( S1 Table , sheet 3 ) . This meant that AN7973 would be inappropriate for field use [3] . The reduced efficacy of AN7973 against T . vivax might have been a consequence of weaker intrinsic potency: the ex vivo EC50 against T . vivax was 215 nM , as against an in vitro EC50 of 84 nM for T . congolense ( S1 Table , sheet 1 ) . The latter value is similar to the results for T . brucei ( 20–80 nM , S1 Table ) . The lack of single-dose cattle efficacy at 10mg/kg i . m . against T . vivax precluded development of AN7973 as a commercially viable treatment against cattle trypanosomosis , but it could still have potential for diseases caused by other salivarian trypanosomes . One of the most direct strategies to identify the targets of antimicrobial drugs is selection of resistant mutants . We therefore attempted to select parasites resistant to AN7973 using an over-expression library [51] . The level of resistance obtained was very modest—less than 2-fold ( S2 Table , sheet 1 ) and in the two lines that we obtained , the over-expression plasmids did not contain an in-frame coding sequence . We therefore sequenced their genomic DNA . As previously observed for trypanosomes that were mildly resistant to AN2965 [22] we found numerous single nucleotide polymorphisms , amplifications and deletions ( summarized in S2 Table , details in S3 and S4 Tables ) . The large number of genes affected made it impossible to pinpoint any particular pathway as being relevant to resistance . Examination of parasite morphology can reveal defects in DNA synthesis , cell cycle regulation , cell motility and protein trafficking , each of which can ultimately cause parasite death . For example , Jones et al . saw accumulation of parasites in the G2 phase of the cell cycle after AN2965 treatment [22] . Treatment of trypanosomes with AN7973 with 10x EC50 for 7h caused growth arrest ( Fig 2A ) , but no obvious changes in parasite morphology or motility and no significant changes in the proportions of cells in different stages of the cell cycle ( Fig 2B ) . These results suggest that AN7973 does not interfere directly with DNA synthesis , cell motility , or protein trafficking . Amino-acyl tRNA synthetase inhibition is a known mode of action of several benzoxaboroles , and inhibition of tRNA synthetase should result in cessation of protein synthesis . We therefore measured this in AN7973-treated trypanosomes by pulse labelling with [35S]-methionine followed by denaturing polyacrylamide gel electrophoresis and autoradiography . Inhibition of protein synthesis ( Fig 2C ) was clear , but some prominent protein bands were affected more than others ( compare bands labeled a and b ) . The kinetics of the inhibition , combined with apparent selectivity for particular proteins , suggested that protein synthesis was not a primary target of AN7973 , but might be inhibited in a secondary fashion . Loss of protein synthesis could be caused by loss of mRNA . For example , if mRNA synthesis were inhibited , the pattern in Fig 2C would be explained if the mRNA encoding protein "b" were less stable than that encoding protein "a" . Loss of mRNA could be caused by inhibition of either RNA transcription or processing . To investigate this possible mechanism , we incubated cells with AN7973 , prepared RNA , and hybridised Northern blots with a probe that detects the spliced leader SL . This probe detects all processed mRNAs , as well as the ~139nt precursor , called SLRNA , that donates the SL . Incubation with AN7973 for 9h had little or no effect on the total amount of rRNA , as judged by methylene blue staining ( Fig 3A ) but caused progressive loss of spliced mRNA ( Fig 3B ) . In contrast , the levels of the SLRNA remained roughly constant ( Fig 3B ) . We therefore suspected that AN7973 was inhibiting mRNA processing . To test this we re-hybridised the blot with a probe that detects beta-tubulin . The tubulin genes are arranged in alternating alpha-beta tandem repeats , and splicing inhibition through heat shock [52] or Sinefungin treatment [47] leads to accumulation of partially processed beta-alpha dimers and multimers . Partially processed tubulin mRNAs were indeed detected within an hour of AN7973 application ( Fig 3C ) . This result showed that after AN7973 addition , transcription of protein-coding genes continued but mRNA processing was impaired . At later time points , the tubulin mRNAs disappeared , suggesting that the failure in processing was complete . ( Most measurements suggest that the tubulin mRNAs have half-lives of about half an hour [53] . ) As noted above , after 1h AN7973 treatment , the level of SLRNA was not much changed . This is somewhat counter-intuitive , since one might expect processing inhibition to lead to a build-up of SLRNA . However , RNAi that targets splicing or polyadenylation factors rarely causes SLRNA increases of more than 2-fold [37 , 54 , 55] . The reason is that SLRNA synthesis is balanced by degradation . Thus when Actinomycin D ( Act D ) is used to inhibit transcription , SLRNA disappears within 30 min even though no new precursors are available for splicing [56] . Fig 3D compares the effects of AN7973 with those of Act D treatment . As expected , 30 min Act D resulted in SLRNA loss whereas AN7973 had no effect . Moreover , if Act D was added after a one-hour AN7973 treatment , SLRNA again disappeared within 30 min—whereas SLRNA was still visible after 9h in the presence of AN7973 . We concluded that treatment of the cells with AN7973 for only one hour inhibited mRNA processing but did not prevent synthesis of SLRNA . Inhibition of procyclic trypanosome mRNA processing by RNAi targeting polyadenylation factors causes growth arrest within 3 days , followed by cell death [37 , 54] . These delays in observing effects are explained by the fact that the targeted proteins remain detectable ( see e . g . [55] ) . Processing inhibition can therefore easily explain killing of trypanosomes by AN7973 . The only small molecule currently known to inhibit kinetoplastid mRNA processing in vivo is the S-adenosyl methionine analogue Sinefungin , which is a general methylation inhibitor [57] . Chemical inhibition of spliced leader methylation prevents trans splicing in a permeabilised cell system [48] , and Sinefungin does the same in vivo [47] . Since we already knew that AN5568 caused increases in methylated intermediates [50] , we wondered whether splicing inhibition by AN7973 was also caused by inhibition of spliced leader methylation . We used primer extension to assess the amount and methylation status of SLRNA . At the same time , we assayed the level of the 2'-5' branched "Y-structure" splicing product ( Fig 4A ) . cDNA synthesis was primed by a 5'-labelled oligonucleotide that is complementary to a region towards the 3' end of the SLRNA ( black arrows in Fig 4B ) . The products from full-length SLRNA form a small ladder , because 5' methylation partially blocks reverse transcriptase , while the Y-structure trans splicing intermediate gives a product of 87nt ( Fig 4B ) . These products are seen in Fig 4C , lane 1 . As expected , incubation with the methylation inhibitor Sinefungin for 30 min abolished the multiple bands caused by cap methylation , with cDNA synthesis extending cleanly to the 5' end of the SLRNA ( Fig 4C , lane 4 , arrowhead ) ; at the same time , the signal from the Y structure was decreased . The concentration of Sinefungin used , 2 μg/mL , is 4000 times the EC50 . This is the standard concentration in these types of experiments and was chosen in order to prevent cap methylation within about 10 min [58] . Importantly , AN7973 ( 10x EC50 ) had no effect on the pattern of bands from the SLRNA , showing that cap methylation was not affected ( Fig 4C ) . In contrast , Y-structure formation was clearly decreased ( Fig 4C , lanes 2 and 3 ) . The combined results so far thus showed that AN7973 acts specifically on splicing , not on production of mature splicing-competent SLRNA . Quantitation of four independent experiments showed that after 30 min , the effect of AN7973 was already significant ( p<0 . 0005 , S3A Fig ) . Remarkably , considering the differences in the doses used , a 2-h incubation with AN7973 inhibited Y-structure formation to the same extent as a 30-min incubation with Sinefungin ( Fig 4D ) . Although the effect of AN7973 on splicing was quite rapid , we nevertheless had to consider the possibility that it was secondary to inhibition of some other process . Apart from Sinefungin , the only treatment that was previously shown to cause accumulation of tubulin precursors within 30 min was severe heat shock [52 , 59] , but the mechanism is unknown and there are numerous other effects including general translation arrest [59] and transcription inhibition [60] . To find out whether trans splicing inhibition was a general feature of parasites stressed through addition of trypanocidal or trypanostatic drugs , we measured Y structure abundance in cells treated with DFMO , diminazene aceturate , pentamidine or suramin , all at 10x EC50 . No significant inhibition of Y-structure formation was observed ( Fig 4E–4H , S3 Fig ) . This result demonstrated conclusively that splicing inhibition is not a general side-effect of growth inhibition . The effect of AN7973 on splicing therefore indicates a specific mechanism . Another possible cause of splicing inhibition might be severe metabolic disruption . The fact that the parasites retained normal motility for several hours indicated that there were no immediate effects upon ATP generation , and we already knew that both mRNA transcription and SLRNA modification were unaffected . Analysis of the metabolome after 5h AN7973 treatment revealed , however , that the main effects were—as with AN5568—related to methylation , with substantial increases in S-adenosylmethionine ( SAM ) methyl thioadenosine ( MTA ) , methyl lysine ( ML ) , dimethyl lysine ( 2ML ) , trimethyl lysine ( 3ML ) and acetyl lysine ( AL ) ( Fig 5; S5 Table ) . SAM is a methyl donor and it has been shown that the methyl groups on the methylated lysines are derived from methionine [50] . It therefore seemed worthwhile to find out whether other anti-trypanosomal benzoxaboroles had similar effects . We started with a panel of 30 compounds . These were tested in vitro against T . brucei and T . congolense ( S2 Fig , S1 Table ) and metabolic effects were also assessed ( S6 Table ) . We then selected various compounds based on potency , the time taken to see an effect ( S3 Fig ) , and the effects on methylated amino acids , SAM and MTA ( S6 Table ) . Nearly all of the chosen compounds had EC50s in the low nM range ( Fig 6 ) . The notable exception was AN3661 ( Fig 1 ) , the anti-malarial lead compound that targets CPSF3 [26 , 27]: its EC50 for bloodstream-form trypanosomes was at least 20 times higher than that of AN7973 . At a concentration of 5x the EC50 determined at 48h , most of the compounds acted within 6-8h ( S3 Fig , S1 Table ) . The veterinary drug candidate AN11736 kills the parasites more slowly because it requires processing for full activity ( Giordani et al . , manuscript in preparation ) . Results for treatment of T . brucei with the selected compounds are shown in Fig 5 . There were no discernable structure-activity relationships for EC50 , time to kill , or the SAM/MTA effect , but most of the oxaboroles that induced a SAM/MTA pattern were faster killing . To investigate the possible link between splicing inhibition and the methylated metabolites , we measured the effects of the chosen compounds on splicing . We also included AN2965 ( Fig 1 ) , the compound whose mode of action had been investigated previously [22] , and the antimalarial candidate AN13762 ( Fig 1 , compound 46 in [25] ) . Initial triplicate measurements ( Fig 7A , experiment 1 ) gave reproducible results for most of the compounds; repeat assays for three that had given ambiguous results confirmed that they gave substantially less splicing inhibition than AN7973 ( Fig 7 , experiment 2 and S4A Fig ) . The clear processing inhibitors included the antimalarial candidate AN13762 ( compound 46 in [25] ) ; the EC50 of this compound for trypanosomes in the 3-day assay was 4 times higher than that for P . falciparum . Overall , there was a remarkably good correlation between splicing inhibition and the average increase in SAM and MTA ( Fig 7B ) ; only two compounds did not conform to the overall pattern . A caveat is that AN11736 is slow acting , and was used at low doses given its extreme potency , so it is possible that splicing inhibition by AN11736 might be seen upon more prolonged incubation . Why could splicing inhibition and S-adenosyl methionine-related metabolites be linked ? We could think of two explanations . First , transcription and processing of the SLRNA genes absorbs substantial cellular resources , since the cell needs to make at least 10 , 000 SLRNAs per hour [53] . It was possible that after 6-8h ( the time of most metabolome measurements ) , there might have been some feedback inhibition of transcription that led to a decreased methylation requirement . Second , loss of mRNA production leads to loss of unstable mRNAs . If the protein encoded by an unstable mRNA has a high turnover rate , then that protein will disappear; and if that protein is a metabolic enzyme , its substrates will accumulate . This too could have led to metabolic changes as we noted . To address the second hypothesis , we directly measured the effect of mRNA synthesis inhibition on the metabolome , using Actinomycin D ( 10 μg/mL ) for 6h . Interestingly , there were again large increases in the amounts of methylated amino acids and significant ( but smaller ) increases in SAM and MTA as well ( Fig 5 ) . Comparison with the well-characterised effects of AN5568 revealed a clear correlation ( Fig 7C , S7 Table ) . These results suggest that the increases in methylated and acetylated lysine—and perhaps also in SAM and MTA—that were seen after treatment with benzoxaboroles could indeed be an indirect consequence of loss of mRNA . As an alternative way to find out whether the effect of AN7973 on mRNA processing was direct or indirect , we measured trans splicing in permeabilised procyclic-form trypanosomes [48] . ( Equivalent assays are not established for bloodstream forms . ) As for most benzoxaboroles that have been tested ( Fig 6 , S1 Table ) the EC50 of AN7973 for procyclic forms was 5–10 times higher than that for bloodstream forms . Treatment with 10x EC50 clearly inhibited splicing in procyclic forms , with 70% loss of Y structure after 2h ( S4B and S4C Fig ) . To test splicing in vitro , procyclic trypanosomes were permeablised with lysolecithin , pre-incubated with AN7973 or DMSO , then transcription was allowed to proceed for 10 min in the presence of [alpha-32P]-UTP [61] . RNA was separated on 7% polyacrylamide-urea gels and visualized by autoradiography . Under these conditions , the SL intron is visible as a ~100nt species , which disappears if incubation is continued for a further 20 min [61]; it is also not made if cap methylation is inhibited by S-adenosyl homocysteine [48] . We do not know the intracellular concentration of AN7973 , and in the in vitro transcription reaction the density of permeabilised parasites is 1000 times higher than in the experiments with cultures . For the in vitro assays we therefore chose to use a concentration of 100 μM ( 500x the EC50 ) , which gives a AN7973:parasite ratio that is equivalent to the ratio at 5x EC50 . This treatment reproducibly prevented formation of the Y structure without preventing transcription of SLRNA or smaller RNAs ( Fig 8 ) . An additional band ( arrowhead ) , which appeared in the presence of AN7973 and was slightly shorter than SLRNA , might be a 3' degradation product that was previously described [48] . AN7973 also reproducibly inhibited labeling of RNAs longer than 500nt ( indicated by a question mark ) . These RNAs are thought to include mRNAs and rRNA [62] , and the inhibition could either be a consequence of trans splicing inhibition or another effect of AN7973 . In future it would be interesting to repeat these studies at a variety of concentrations and with other benzoxaboroles . When trypanosomes are stressed by heat shock or starvation , their mRNAs accumulate in RNA-protein particles called stress granules containing a helicase , DHH1 [63] . Generally , these are throughout the cytoplasm , but after treatments that inhibit splicing , they transiently cluster around the nuclear envelope [64] . To find out whether AN7973 had the same effect we followed localization of YFP-tagged DHH1 in treated cells ( Fig 9; the YFP signal is coloured magenta ) . A few granules were seen in untreated cells ( Fig 9A ) but after 30 min Sinefungin treatment , the nuclear periphery granule pattern was clear ( Fig 9B ) . AN7973 also induced formation of perinuclear granules in a subset of cells ( Fig 9C ) . However , when the remaining compounds were tested ( S5 Fig , S6 Fig , S7 Fig ) we observed no correlation between peri-nuclear granule formation and Y structure inhibition ( Fig 9D ) . This may partly be explained by the transient nature of the perinuclear localization , but the clear perinuclear granule formation after AN11736 treatment also suggests that the pattern can be caused by stresses other than splicing inhibition . We concluded that without careful time-course studies , this assay could not be used to distinguish between specific splicing effects and more general stress responses . To try to identify possible targets implicated in mRNA processing , we re-examined the genomes of our partially-resistant lines ( S2–S4 Tables ) . Since loss of polyadenylation stops splicing [36 , 37] , we looked for changes in genes associated with both processes . No mutations of the U snRNAs were found . A missense mutation was found in the Tb927 . 10 . 9660 open reading frame encoding a putative CRN/SYF3; this protein co-purified with the PRP19 complex , but did not co-sediment with the complex on a sucrose gradient [65]; its function is thus unclear . The Sm complex forms the core of spliceosomal snRNPs , and our two cell lines that had been selected in AN7973 both had 1–2 extra copies of the genes encoding four out of the seven components: Sm-B , Sm-E , Sm-F and Sm-G . The other three genes were , however , not amplified . After selection of proteins on an oxaborole affinity column , Jones et al . found enrichment of RBSR1 ( Tb927 . 9 . 6870 ) , a protein with an SR-domain that is potentially involved in splicing , and of the U2 splicing auxiliary factor ( U2AF35 , Tb927 . 10 . 3200 ) [22] . Our resistant lines had no changes in RBSR1 or U2AF35 . In P . falciparum and T . gondii , mutations in CPSF3 gave resistance to AN3661 [26 , 27] . Moreover , in one of their AN2965 resistant lines , Jones et al . [22] observed a two-fold amplification of the gene encoding CPSF3 ( Tb927 . 4 . 1340; also designated CPSF73 ) ; and AN2965 was a strong inhibitor of splicing ( Fig 7A ) . We therefore compared the T . brucei sequence with those of Plasmodium and humans , concentrating on the residues that were mutated in AN3661-resistant Apicomplexa . H36 and Y408 of the P . falciparum sequence were mutated in AN3661-resistant lines but are retained as H and Y the human and trypanosome sequences ( S8 Fig ) . The remaining important Plasmodium residues , however , are already different in trypanosomes . Y252 ( mutated to C for AN3661 resistance ) is N in T . brucei; T406 ( mutated to I ) is A; T409 ( mutated to A ) is C; and D470 ( mutated to N ) is already N in T . brucei ( S8 Fig ) . From these changes alone one could predict that trypanosomes would be quite resistant to AN3661—as is indeed the case ( S1 Fig ) . Although AN3661 did inhibit trypanosome mRNA processing and give a SAM/MTA effect , the concentrations used were 10–20 times higher than for AN7973 ( S1 Table ) . To test the role of CPSF3 we first attempted to make the equivalent of the Y408S mutation in trypanosomes by homologous gene replacement . Interestingly , although the transfections yielded several transgenic clones , none had the mutation . This , together with the fact that neither we nor Jones et al . [22] found the mutation after resistance selection , suggests that in the context of the trypanosome sequence , the Y408S equivalent ( Y383S in the T . brucei sequence , S8 Fig ) results in an unacceptable decrease in CPSF3 activity . If so , the result suggests either that the mutant CPSF3 has dominant-negative effects , or that CPSF3 is present in limiting amounts such that mutation of one gene copy results in haplo-insufficiency . We next assessed how the differences between T . brucei and P . falciparum CPSF3 sequences would affect protein conformational dynamics . To do this we investigated a homology model of CPSF3 by elastic network normal mode analysis ( S9A and S9B Fig ) , and found a relative rotational breathing motion of the domains ( S9C Fig ) . Interestingly , in the model , many residues associated with AN3661 resistance , such as Y408 and N252 ( P . falciparum numbering ) were found to line the cleft between the breathing domains . Relative domain rotation might thus be able to enhance the accessibility of the binding site . Thus , mutations associated with AN3661 resistance might affect the conformational dynamics of the enzyme , the accessibility of the active site and the stability of the interdomain contacts . As an alternative to mutation , we inducibly over-expressed RBSR1 , U2AF35 and CPSF3 in bloodstream forms , as C-terminally myc-tagged versions . Expression of RBSR1-myc and U2AF35-myc ( S10A Fig ) did not affect the EC50 of AN7973 . In contrast , expression of CPSF3-myc caused at least 3-fold increases in the EC50s of four tested benzoxaboroles , including AN7973 ( Fig 10 , S10B Fig , S8 Table ) . Statistically significant increases of 2-fold or more were also seen for several other benzoxaboroles ( Fig 10 , S10B Fig , S8 Table ) and there was a moderate correlation with Y-structure inhibition ( Fig 10B ) . This suggests that the effective intracellular concentration of the benzoxaboroles is reduced through binding to excess CPSF3 . The modest level of resistance may be explained by the fact that CPSF3 normally functions as part of a complex: we do not know the extent to which CPSF3 can be accumulated independently , and its conformation might be influenced by protein-protein interactions . Over-expression of CPSF3-myc had almost no effect on the EC50 of AN3661 . Because of its low potency , AN3661 was always tested at relatively high ( micromolar ) concentrations , so this result is difficult to interpret . However poor binding of AN3661 to CPSF3 is expected , since the trypanosome protein has residues equivalent to those found in the AN3661 resistant version of the Plasmodium falciparum protein ( S8 Fig ) . To investigate whether AN7973 could interact with CPSF3 , we performed induced fit docking to our comparative model of T . brucei CPSF3 ( TbCPSF3 ) . For AN3661 , it was proposed that boron occupies the position of an mRNA substrate phosphate group and that the ring oxygen , as well as hydroxylations , of the benzoxaborole can interact with the two active site zinc ions [27] . We applied constraints accordingly when docking AN7973 in the TbCPSF3 active site . In our model , in addition to metal-oxygen interactions and hydrogen bonding contacts with D67 , which stabilize the benzoxaborole binding mode , the pyrazole ring of AN7973 forms a hydrogen bond with the backbone amide nitrogen of A406 ( S11 Fig ) . ( All residue numbers mentioned correspond to those of the P . falciparum sequence , PfCPSF3 , S8 Fig ) The elongated tail of AN7973 lodges in a subpocket that is largely lined by hydrophobic residues such as F267 , P284 , F286 , L294 , A406 and F443 . Notably , the same interaction pattern was observed for both the tetrahedral form of the compound with a formal negative charge on the boron atom and the trigonal planar neutral variant ( S11A and S11B Fig , respectively ) . The benzoxaborole core of the planar variant however has a rotated orientation compared to the tetrahedral form , which leads to a displacement of the loop containing Y408 . The results of the docking therefore suggested that inhibition of TbCPSF3 by AN7973 is feasible .
Benzoxaboroles are important drug candidates for both human and ruminant African trypanosomosis . Our results show that AN7973 inhibits mRNA processing in trypanosomes . Expression of additional CPSF3 increased the EC50 of AN7973 , suggesting that AN7973 can bind to CPSF3 . These results suggest that mRNA processing is an important target of AN7973 , which might operate through CPSF3 inhibition . AN7973 also caused metabolite changes indicative of disturbed methylation , similar to those observed for acoziborole . When this work was started , AN7973 was under consideration as a candidate for treatment of cattle trypanosomosis , but it was later found to be less effective against T . vivax . In the available T . vivax genome sequence there is a gap in the CPSF3 open reading frame . This gap spans N231 in the T . brucei sequence , which is Y in Apicomplexa and humans and mutated to C in resistant Apicomplexa ( S8 Fig ) . All other CPSF3 residues that are implicated in AN3661 resistance in Apicomplexa are conserved between T . vivax , T . congolense and T . brucei ( S8 Fig ) . Residues predicted to be in contact with AN7973 in our structural model of TbCPSF3 were also largely conserved in the T . vivax sequence . Only in the loop 378–385 ( or 403–410 in the P . falciparum numbering ) are two valine residues replaced by isoleucine , which might slightly restrain the space accessible to the compound in the active site ( S8 Fig ) . Our results suggested a mechanistic link between splicing inhibition and accumulation of specific methylated metabolites ( Fig 7B ) . One hypothesis we had was that the effects on methylation intermediates could be caused by decreased requirements for cap methylation . However we found no evidence , either in vivo or in vitro , that SLRNA transcription , capping or methylation were affected by AN7973 . Another hypothesis was that loss of mRNA production will lead to a selective reduction in the activities of enzymes that have relatively high mRNA and protein turnover rates , and that these enzymes are required for various aspects of methylation . This was partially supported by accumulation of methylated amino acids after Actinomycin D treatment . Strangely , our results revealed no clear structure-function relationships for benzoxaborole effects on trypanosome viability , metabolites , or splicing . The diversity in structures of compounds that inhibited trypanosome mRNA processing , combined with molecular modelling , suggests to us that this might be an intrinsic property of the pharmacophore . Nevertheless , some trends with respect to molecules with the same scaffolds can be noted . For example the two compounds AN15368 and AN11736 are both L-valinate amide benzoxaboroles [21] and both showed very little Y structure inhibition ( Fig 1 , Fig 7 ) . These two compounds also showed little change in EC50 with overexpression of CPSF3 ( Fig 10 ) suggesting that the L-valinate amide benzoxaborole series may not act through inhibition of mRNA processing . AN7973 and AN4169 , both of the carboxamide scaffold , also showed similar decreases in Y structure . An important caveat is that the uptakes and metabolisms of the various compounds are likely to differ . For example , after 2h incubation , AN3056 and the veterinary drug candidate AN11736 had less effect on processing than did AN7973: but their action might be delayed , since both are subject to activation within the parasites . So far , selection of trypanosomes resistant to benzoxaboroles has met very limited success . Substantial resistance was obtained only for compounds which require intracellular metabolic activation , and the mutations responsible were in the activating enzymes . For benzoxaboroles that are probably not metabolised , such as AN7973 and AN2965 , only very limited resistance could be obtained . A priori , one would expect that if these compounds have a single active site target , selection for resistance should be relatively straightforward . It is however possible that the necessary mutations are incompatible with function . Inhibition of mRNA processing was the earliest effect that was seen after AN7973 treatment and is therefore likely to make a vital contribution to parasite killing . In addition , both docking studies and results from over-expression identify CPSF3 as a likely target . This raises the possibility that similar modes of action might be seen in oxaboroles under development against other Kinetoplastids , including Leishmania and Trypanosoma cruzi .
All in vivo mouse experiments were carried out in accordance with the strict regulations set out by the Swiss Federal Veterinary Office , under the ethical approval of the Canton of Basel City , under license number #2813 . The experimental protocols used for goat studies were approved by the ethics committee for animal experimentation by the Veterinary Faculty of the University of Las Palmas in Gran Canaria , Spain on July 21 , 2012 with the reference number 240/030/0121-36/2012 . The studies were conducted under the strict guidelines set out by the FELASA for the correct implementation of animal care and experimentation . Tests on cattle were done in accordance with the principles of veterinary good clinical practice ( http://www . vichsec . org/guidelines/pharmaceuticals/pharma-efficacy/good-clinical-practice . html ) . The ethical and animal welfare approval number was 00 l-2013/CE-CIRDES . Testing of T . congolense in vitro ( 72 hrs ) and T . vivax ex vivo ( 48 hrs ) assays were done as described in [66] . Testing of T . congolense and T . vivax in vivo in mice was done as described in [21] . The proof of concept efficacy studies in goats using AN7973 were conducted as previously described in [67] , but established and modified for T . congolense and T . vivax models of infection . The trials took place from January to May 2013 within the Veterinary Faculty of the University of Las Palmas and the Agricultural farm ( Granja Agricola ) of the Canarian Island Government in Arucas , Gran Canaria , Spain . In total , 45 female Canarian goats , weighing between 12–35 kg and no less than four months old , were purchased from a local dairy farmer and transported to the study site . The goats were placed in fly-proofed pens and allowed to acclimatise for two weeks , before being randomly selected and divided into test groups of four . Goats were experimentally infected intravenously from two highly parasitaemic donor goats , with 106 and 105 parasites per goat for T . congolense and T . vivax , respectively . AN7973 was administered intramuscularly accordingly , as either two injections of 10 mg/kg or as a single bolus dose of 10 mg/kg , on days 7 and 8 post-infection . Thereafter , the parasitaemia was monitored in the goats for up to 100 days post-treatment , after which any aparasitaemic and surviving goats were considered cured . Relapsed goats were removed immediately from the trial and humanely euthanized with an intravenous injection of sodium phenobarbital . The efficacy of AN7973 against T . congolense and T . vivax in cattle was tested as described in [68] . The studies were conducted in fly-proof facilities and included negative ( saline ) controls; and the staff were blinded with regard to allocation of animals to treatment groups . Assessments were made for 100 days post treatment unless animals relapsed sooner . Bloodstream-form T . brucei brucei 427 Lister strain were cultured in HMI-9 plus 10% foetal calf serum or CMM [69] plus 20% foetal calf serum at 37°C , 5% CO2 . PCF T . b . brucei were cultured in SDM79 plus 10% foetal calf serum at 28°C . Compounds were dissolved at 20 mM in DMSO and aliquoted to avoid excessive freeze-thaw cycles . EC50s were measured in two different ways . To obtain the EC50s in S1 Fig , compounds were serially diluted over 24 doubling dilutions in 100 μL culture medium in 96 well opaque plates from a starting concentration of 100 μM . Bloodstream form trypanosomes were added at a final density of 2x104/mL ( 100 μL ) and incubated for 48 hours , while procyclic forms were added at a final density of 2x105/mL and incubated for 72 hours . After incubation , 20 μL of 0 . 49 mM Resazurin sodium salt in PBS was added to each well and plates were incubated for a further 24 hours . Plates were read on a BMG FLUOstar OPTIMA microplate reader ( BMG Labtech GmbH , Germany ) with λexcitation = 544 nm and λemission = 590 nm . For the results in S1 Table , sheet 1 , column E , bloodstream form trypanosomes were diluted to 4000/mL in the presence of compounds ( diluted in water from DMSO ) and incubated for 72h . 3-4h before the end of the incubation , Resazurin ( Sigma ) was added ( final concentration 44 μM ) . Resazurin fluorescence was measured to assess the number of surviving viable cells [70 , 71] . Each assay was performed with 3 technical and 3 biological replicates . For the time to kill assay ( S1 Fig , S2 Fig ) bloodstream-form T . b . brucei were cultured in 24 well plates in triplicate . Cultures were seeded at 5 x 105/mL and compound was added at 5xEC50 . Cells in each well were counted at 2 , 4 , 6 , 8 and 24 hours using a haemocytometer . For the all assays except the metabolomes shown in Fig 1 and S6 Table , compounds were used at 10x the 72-h EC50 . To allow for variations between drug aliquots , EC50s were measured prior to every experimental series . The concentrations of compounds used in different experiments are listed in S1 Table and S6 Table . For protein analysis , 2-3x106 cells were collected for each sample , resuspended in Laemmli buffer heated and subjected to SDS-PAGE gel electrophoresis . All assays of macromolecular biosynthesis and RNA processing were done at densities of less than 2 x 106/mL . Pulse-labeling was done as described in [72] . Total RNA was extracted from roughly 5x107 cells using peqGold TriFast ( peqLab ) following the manufacturer's instructions . The RNA was separated on formaldehyde gels and then blotted on Nytran membranes ( GE Healthcare ) . Following crosslinking and methylene blue staining ( SERVA ) , the northern blots were hybridized with the appropriate probes . For mRNA detection , the membranes were incubated with [α-32P]dCTP radioactively labelled DNA probes ( Prime-IT RmT Random Primer Labelling Kit , Stratagene ) overnight at 65°C . For spliced leader detection , a 39mer oligonucleotide complementary to the spliced leader was labelled with [γ-32P]ATP using T4 polynucleotide kinase ( NEB ) and incubated with the membrane overnight at 42°C . After washing the blot , it was exposed to autoradiography films and detection was performed with FLA-7000 ( GE Healthcare ) . The images were processed with ImageJ . Compound treatments were all done at cell densities of about 0 . 9x106 cells/mL . For each condition , 8-10x107 cells were used . Primer extension was done approximately as described in [47]; primers were: ACCCCACCTTCCAGATTC for SLRNA ( KW01 or CZ6364 ) and TGGTTATTTCTCATTTAAGAGG ( CZ6491 ) for U3 snRNA . Both primers and the ladder were radioactively 5'-end-labelled with [γ-32P]ATP . For extension , 10 μg of RNA was incubated for 5’ at 65° with 2 μL of dNTPs ( 10 mM ) and roughly 200 000 counts per minute ( cpm ) of the corresponding primer . Afterwards , RNasin ( Promega ) , SuperScript III Reverse Transcriptase ( Thermo Fischer ) , DTT and buffer were added according to the manufacturers instructions . The mixture was incubated 60’ at 50°C and then inactivated 15’ at 70°C . The samples were run in 35 cm long 6% polyacrylamide gels , dried , and analysed by phosphorimaging . The images were analysed using Fuji / ImageJ . In vitro transcription in permeabilised cells was done following the published procedure [48 , 61] with minor modifications [73] . Briefly , cells ( 2 . 5 x 108/reaction ) were permeabilised with lysolecithin for 1 min on ice , washed , then resuspended in 60 μL transcription buffer . 1 μL of either AN7973 or DMSO alone were added , and the reaction pre-incubated at 28°C for 2 min . After addition of 100 μL transcription cocktail containing 1 μL of either AN7973 or DMSO , the reaction was allowed to proceed for 10 min at 28°C . The permeabilised cells were pelleted ( 45 sec ) and resuspended in 1 mL of TriFast . After the first phase separation , the aqueous fraction was re-extracted with phenol to remove residual protein; this was necessary to obtain good separation during gel electrophoresis . The final RNA volume was 20 μL . 10 μL of each reaction were separated on a 7% polyadcrylamide/urea gel and the products were detected by phosphorimaging with low molecular weight DNA markers ( New England Biolabs ) . For genomic DNA sequencing , libraries were prepared at the Cell Networks Deep Sequencing Core Facility ( University of Heidelberg ) and subjected to paired-end MiSeq ( Illumina ) at EMBL . The quality of sequencing was evaluated with FastQC and the reads were trimmed using Trimmomatic . The output was aligned to the T . b . brucei TREU927 genome ( version 9 . 0 ) using bowtie2 . Results are available in ArrayExpress under accession number E-MTAB-6307 . The Picard option AddOrReplaceReadGroups was used to create a valid . bam file to then be piped into GATK for obtaining , as output , . vcf files containing SNP and indel information . SnpSift filtered the features of interest , excluding for example synonymous mutations and intergenic regions . Identified variations from all cell lines were pooled to look for mutations found in all strains compared to the wild type , and in addition , reads from each strain were processed separately to find all the mutated genes . The lists were then compared . The genes taken into consideration were identified with annotation and categories from the Clayton lab in-house annotation list . Based on these annotations , the genes were filtered in Excel , where highly repetitive genes were excluded from the analysis . At first , variant surface glycoproteins ( VSG ) , expression site-associated genes ( ESAG ) , receptor-type adenylate cyclase GRESAG , UDP-Gal or UDP-GlcNAc-dependent glycosyltransferases and pseudogenes were removed . In the case of non-homozygous mutations , the genes were further selected filtering out other repetitive genes such as leucine-rich repeat proteins ( LRRP ) , various invariant surface glycoproteins ( ISG ) , nucleoside transporters ( TbNT ) and retrotransposon hot spot ( RHS ) proteins . The list of gene IDs was compared with their translation level based on ribosome profiles [74 , 75] and genes with values below 10 ( non-translated ) were excluded from the analysis . For all assays at 5x EC50 , bloodstream-form T . b . brucei were inoculated into medium at 1 x 106/mL ( for a six or eight hour incubation ) or 2 x 105/mL ( for a 24 hour incubation ) and compounds were added . Cells were incubated with compounds under normal growth conditions . At the desired time point , 1 x 108 cells were taken and cooled rapidly in a dry-ice ethanol bath to 4°C . Samples were centrifuged at 1250 g twice to remove all medium before 200 μL chloroform:methanol:water ( 1:2:1 ) were added . Extracts were shaken for one hour at max speed before cell debris was removed by centrifugation at 16 , 000 g . Metabolite extracts were stored at -80°C under argon gas . Mass spectrometry–Metabolite samples were defrosted and run on a pHILIC column coupled to an Orbitrap mass spectrometer as previously described [76] . In batch 1 an Orbitrap Exactive ( Thermo Scientific ) was used with settings including mass range: 70–1400 , a lock mass of 74 . 0964 , capillary: 40V , Tube lens: 70V , Skimmer: 20V , Gate lens: 6 . 75V and C-trap RF: 700V . In batch 2 an Orbitrap QExactive ( Thermo Scientific ) was used with settings including mass range: 70–1050 , lock masses of 74 . 0964 and 88 . 0757 , S-lens: 25V , Skimmer: 15V , Gate lens: 5 . 88V and C-trap RF: 700V . For fragmentation analysis an MS2 isolation window of 4 m/z , an intensity threshold of 3 . 3e5 and dynamic exclusion of 10 seconds were used . Metabolites were putatively annotated using IDEOM software [77] before verification of annotations using mass , retention time , isotope distribution and fragmentation pattern . Xcalibur ( Thermo ) was used to explore the raw data , MzCloud ( mzcloud . org ) was used to match fragments to database spectra . Metabolite analysis was done using four biological replicates per condition and cell line . Relative metabolite levels were based on raw peak height relative to the average raw peak height of untreated cells . Lists of metabolites were mapped on metabolic pathways using Pathos ( http://motif . gla . ac . uk/Pathos/ ) . Intersections between samples were found using bioVenn ( http://www . cmbi . ru . nl/cdd/biovenn/ ) . ( http://www . cmbi . ru . nl/cdd/biovenn/ ) . Metabolite identities are consistent with standards from the Metabolomics Standards Initiative and evidence for each identity is shown in S6 Table . For over-expression of C-terminally myc-tagged proteins , the open reading frames encoding proteins of interest were amplified by PCR from genomic DNA , and cloned into pRPa-6xmyc [78] . After transfection , cells were selected and expression of myc-tagged protein was induced overnight with 100 ng/mL tetracycline . The plasmid for creation of cells expressing YFP-DHH1 [59] was a kind gift from Susanne Krämer ( University of Würzburg ) . It was transfected into bloodstream-form trypanosomes and two stable cell lines expressing the protein were selected . Trypanosomes ( maximum density less than 1 million/mL in 10 mL ) were treated for 30 min with 2 μg/mL Sinefungin , or for 1-2h with compounds at 10x EC50 . After collection ( 5 min 1000g ) and washing in PBS cells were resuspended in 20 μL PBS . 500 μL of 4% paraformaldehyde solution in PBS was added , cells were incubated without shaking for 18 min , washed 3x with PBS , then distributed onto poly-lysine coated chamber glass slides ( all cells divided in two chambers ) and left at 4°C over night . The PBS was then removed and cells were permeabilised using 0 . 2% Triton X-100 ( w/v ) in PBS , with shaking at room temperature for 20 min . After a further 3 washes , slides were incubate at room temperature with 200 ng/mL DAPI in PBS ( 15 min shaking ) , washed twice more , air-dried , embedded and covered for microscopy . Slides were viewed and the images captured with Olympus IX81 microscope . The 100x oil objective was used . Digital imaging was done with ORCA-R2 digital CCD camera C10600 ( Hamamatsu ) and using the xcellence rt software . Bright field images were taken using differential interference contrast ( DIC ) , Exposure time 30 ms , Lamp 4 . 0 . Fluorescent images were made using DAPI and YFP filters . They were taken as Z-stacks , 30–40 images in a 8 μm thick layer , Exposure time 40 ms , Light intensity 100% , and afterwards they were deconvoluted ( Numerical aperture 1 . 45 , Wiener filter , Sub-Volume Overlap 20 , Spherical Aberration Detection Accurate ) . The fluorescence images were processed using ImageJ . For YFP and DAPI channels , layers containing signals ere selected , then the projection of maximum intensity of the deconvoluted stack was used . The images were saved in an 8-bit range then overlayed with DAPI in cyan and YFP as magenta . The colour balance was then adjusted with variable maxima for DAPI/cyan , but a set maximum of 80 for YFP/magenta . After initial assessment after AN7973 and Sinefungin treatment , three independent experimental series , each including a negative control , were processed , and the images were read blinded . Cells were classified as having nuclear periphery granules if there were at least four strong granules on top of , or within one granule diameter of , the nucleus . Cells were classified as "strong" if nearly all granules were around the nucleus , and "possible" if there were several granules in other positions as well . For granule counts , only structures with at least four adjacent pixels at maximum intensity were considered .
|
Trypanosomes and leishmanias infect millions of humans and cause economically devastating diseases of livestock; the few existing drugs have serious deficiencies . Trypanosomosis of cattle caused by Trypanosoma congolense and Trypanosoma vivax , is a serious problem in Africa because cattle are used not only for food but also for traction , and new drugs are needed . A single injection of the benzoxaborole compound AN7973 cured T . congolense infection in cattle and goats . Although slightly lower effectiveness against T . vivax precludes development of AN7973 as a commercially viable treatment against cattle trypanosomosis , it could still have potential for diseases caused by other trypanosomes . We used a large range of methods to find out how AN7973 kills trypanosomes , and compared it with other benzoxaboroles . AN7973 and some of the other compounds had effects on parasite metabolism that resembled those previously seen for a benzoxaborole that is being tested for human sleeping sickness . The most rapid effect of AN7973 , however , was on processing of trypanosome mRNA . As a consequence , amounts of mRNA decreased and synthesis of proteins stopped . We conclude that AN7973 and some other benzoxaboroles kill trypanosomes by stopping gene expression .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"medicine",
"and",
"health",
"sciences",
"messenger",
"rna",
"parasitic",
"protozoans",
"parasitology",
"trypanosoma",
"brucei",
"protozoans",
"methylation",
"pharmacology",
"drug",
"metabolism",
"veterinary",
"science",
"veterinary",
"parasitology",
"gene",
"expression",
"chemistry",
"pharmacokinetics",
"biochemistry",
"rna",
"trypanosoma",
"eukaryota",
"polyadenylation",
"nucleic",
"acids",
"genetics",
"biology",
"and",
"life",
"sciences",
"chemical",
"reactions",
"physical",
"sciences",
"trypanosoma",
"brucei",
"gambiense",
"organisms"
] |
2018
|
The trypanocidal benzoxaborole AN7973 inhibits trypanosome mRNA processing
|
Synaptic activity can boost neuroprotection through a mechanism that requires synapse-to-nucleus communication and calcium signals in the cell nucleus . Here we show that in hippocampal neurons nuclear calcium is one of the most potent signals in neuronal gene expression . The induction or repression of 185 neuronal activity-regulated genes is dependent upon nuclear calcium signaling . The nuclear calcium-regulated gene pool contains a genomic program that mediates synaptic activity-induced , acquired neuroprotection . The core set of neuroprotective genes consists of 9 principal components , termed Activity-regulated Inhibitor of Death ( AID ) genes , and includes Atf3 , Btg2 , GADD45β , GADD45γ , Inhibin β-A , Interferon activated gene 202B , Npas4 , Nr4a1 , and Serpinb2 , which strongly promote survival of cultured hippocampal neurons . Several AID genes provide neuroprotection through a common process that renders mitochondria more resistant to cellular stress and toxic insults . Stereotaxic delivery of AID gene-expressing recombinant adeno-associated viruses to the hippocampus confers protection in vivo against seizure-induced brain damage . Thus , treatments that enhance nuclear calcium signaling or supplement AID genes represent novel therapies to combat neurodegenerative conditions and neuronal cell loss caused by synaptic dysfunction , which may be accompanied by a deregulation of calcium signal initiation and/or propagation to the cell nucleus .
Physiological levels of synaptic activity are required for neurons to survive [1] . Activity-dependent neuroprotection is induced by calcium entry through synaptic NMDA receptors and requires that calcium transients invade the cell nucleus [2]–[6] . Procedures that interfere with electrical activity and compromise NMDA receptor function or nuclear calcium signaling can have deleterious effects on the health of neurons both in vitro and in vivo . For example , blockade of NMDA receptors in vivo following intraperitoneal injections of the NMDA receptor antagonist MK-801 into seven day-old rats triggers , within 24 hours , a wave of apoptotic neurodegeneration in many brain regions , including the parietal and frontal cortex , the thalamus and the hippocampus [7] . Likewise , the selective blockade of nuclear calcium signaling prevents cultured hippocampal neurons from building up anti-apoptotic activity upon synaptic NMDA receptor stimulation [2] , [3] , [6] . Conversely , enhancing neuronal firing and synaptic NMDA receptor activity is neuroprotective: networks of cultured hippocampal neurons that have experienced periods of action potential bursting causing calcium entry through synaptic NMDA receptors are more resistant to cell death-inducing conditions [2]–[4] . Moreover , stimulating synaptic activity in vivo by exposing rats to enriched environments reduces spontaneous apoptotic cell death in the hippocampus and protects against neurotoxic injuries [8] . Neuronal activity and NMDA receptor-induced calcium signaling pathways can suppress apoptosis and promote survival through two mechanistically distinct processes . One process is independent of on-going gene transcription and involves the phosphatidylinositide 3′-OH kinase ( PI3K ) -AKT signaling pathway which promotes survival while neurons are being electrically stimulated [3] . However , the principal pathway conferring long-lasting neuroprotection requires the generation of calcium transients in the cell nucleus [2]–[6] , [9] . The aim of this study was to investigate how nuclear calcium promotes neuroprotection . Using tools to selectively block nuclear calcium signaling in hippocampal neurons in conjunction with microarray technologies and bioinformatics , we uncovered a genomic survival program that is induced by calcium transients in the cell nucleus . The core components of this program , referred to as Activity-regulated Inhibitor of Death ( AID ) genes , can provide neurons with a broad-spectrum neuroprotective shield against cell death .
To identify genes regulated by nuclear calcium signaling in hippocampal neurons , we carried out comparative whole-genome transcriptional profiling . Hippocampal neurons were infected with a recombinant adeno-associated virus ( rAAV ) expressing either the calmodulin ( CaM ) binding-peptide , CaMBP4 ( rAAV-CaMBP4 [6] ) or β-galactosidase ( rAAV-LacZ ) as a control . CaMBP4 is a nuclear protein that contains 4 repeats of the M13 calmodulin binding peptide derived from the rabbit skeletal muscle myosin light chain kinase; it binds to and inactivates the nuclear calcium/CaM complex [10] . Inhibition of nuclear calcium signaling with CaMBP4 in hippocampal neurons blocks synaptic activity-evoked CREB-mediated transcription and prevents the induction of a genomic neuroprotective program by neuronal activity [3] , [6] . Hippocampal neurons were stimulated by exposing the network to the GABAA receptor antagonist , bicuculline . GABAergic interneurons , which represent about 11% of the neuron population , impose a tonic inhibition onto the network [11] . Removal of GABAAergic inhibition with bicuculline leads to action potential ( AP ) bursting , which stimulates calcium entry though synaptic NMDA receptors , generates robust cytoplasmic and nuclear calcium transients , induces CREB-dependent transcription , and strongly promotes neuronal survival [2]–[4] , [6] , [11] , [12] . RNA isolated from these hippocampal neurons was used for microarray analyses on Affymetrix GeneChips . The Affymetrix microarray data were analyzed by a two-step process; details of the data analysis are described in Text S1 . First , we determined all genes induced or repressed by AP bursting ( which gives rise to robust nuclear calcium signals ) in control-infected ( rAAV-LacZ ) hippocampal neurons . A threshold of 2 . 0 fold was chosen , which , given that microarray data are compressed and generally underestimates the fold differences in gene expression [6] , [13] , filters out genes that are likely to undergo signal-induced changes in their expression that are in the range of at least 2 . 5 to 3 fold . This analysis revealed 302 genes that were induced and 129 genes that were repressed in rAAV-LacZ infected hippocampal neurons 4 hours after the induction of AP bursting . A color-coded map provides an overview of these 431 AP bursting-regulated genes ( Figure 1A ) . A comparison of the genes identified in this study using rAAV-LacZ infected hippocampal neurons with the pool of activity-regulated genes described in a previous study [6] revealed a high degree of overlap . However , due to the higher threshold applied in this analysis ( 2 fold vs . 1 . 5 fold change used in our previous study [6] ) , the current analysis filtered out fewer genes . In a second analysis step , we compared the expression of genes regulated by AP bursting in hippocampal neurons infected with rAAV-LacZ and in hippocampal neurons infected with rAAV-CaMBP4 to block the nuclear calcium/CaM complex . The regulation of a gene was considered dependent on nuclear calcium signaling if based on the microarray data its induction or repression by AP bursting is reduced by at least 40% in rAAV-CaMBP4 infected neurons compared to rAAV-LacZ infected neurons . We found that 183 genes ( plus Btg2 and Bcl6; see below ) fulfill these criteria ( Figure 1B ) ; a list of those genes including their fold changes following AP bursting and percent inhibition by CaMBP4 is given in Table 1 . We are likely to underestimate the total number of genes regulated by nuclear calcium signaling for several reasons . First , our screen did not identify genes controlled by the downstream regulatory element antagonist modulator ( DREAM ) [14] . DREAM is a transcriptional repressor that can directly bind calcium through its EF hands . In its calcium-bound form , DREAM is released from the DNA allowing transcription to be activated in a nuclear calcium-dependent but calmodulin-independent manner [14] . Second , our analysis was restricted to one time point ( i . e . 4 hours after induction of AP bursting ) and therefore it is possible that we have missed genes that have peak expression levels at time points significantly earlier or later than 4 hours . Although those genes may also show changes in expression levels at 4 hours after AP bursting , if this induction was less than two fold they were not scored as induced in our study . Third , a possible regulation by nuclear calcium could also have been missed because for genes weakly induced by AP bursting , the accuracy of the microarray data–based assessment of nuclear calcium regulation decreases and the necessary statistical criteria may not be met . Indeed , these conditions apply to two neuronal survival-promoting genes , Btg2 and Bcl6 [6] . Btg2 is induced with fast kinetics after AP bursting; the induction peaks at 2 hours [about 8 fold induction based on microarray data and 17 fold induction based on quantitative reverse transcriptase ( QRT ) -PCR analysis [6]] , whereas at the 4 hours time point the change in expression based on microarray data relative to unstimulated control is only about 2 fold [6] . In the case of Bcl6 , the fold changes observed at 4 hours after the on-set of AP bursting are about 1 . 5 fold based on microarray data analysis and about 2 . 5 fold based on QRT-PCR analysis [6] . Both , Btg2 and Bcl6 , are regulated by nuclear calcium signaling [6] and have therefore been included in the list of nuclear calcium-regulated genes ( Table 1 ) . The identified nuclear calcium-regulated gene pool comprises 43% of all activity-regulated genes . It contains a large variety of gene products with different catalytic and binding activities ( Figure 1C ) . Because nuclear calcium signals evoked by AP bursting strongly promote neuronal survival [2]–[4] , [6] , we next aimed at identifying putative neuroprotective genes present in the nuclear calcium-regulated gene pool . Using a Gene Ontology ( GO ) analysis with the GO term ‘Apoptosis’ and a literature search we were able to identify 20 nuclear calcium-regulated genes that have been implicated in cell death/survival processes in non-neuronal or neuronal cells ( Table 1 ) . Our attention was drawn in particular to 8 nuclear calcium-regulated genes that , based on the microarray data , showed a very robust induction ( more than 10 fold changes ) of expression following neuronal activity . This includes 6 genes with known or putative functions in the cell nucleus ( Atf3 , GADD45β , GADD45γ , Interferon activated gene 202b ( Ifi202b ) , Npas4 , and Nr4a1 ) and 2 genes encoding secreted proteins ( Inhibin β-A and Serpinb2 ) . Atf3 ( activating transcription factor 3 ) is a member of the ATF/cAMP-responsive element-binding protein ( CREB ) family of transcription factors [15] that has been implicated in survival processes in neuronal and non-neuronal cells [16]–[18] . The family of growth arrest and DNA damage-inducible 45 ( GADD45 ) genes comprises three members ( GADD45α , GADD45β and GADD45γ ) that are expressed in response to stress stimuli and DNA damage . GADD45 genes have been implicated in DNA excision-repair processes [19] , [20] but may also contribute to gene transcription via a process that involves DNA demethylation [21] . Ifi202b belongs to the interferon-activated p202 gene family that plays a role in cell survival and the regulation of caspase activation [22] , [23] . Npas4 ( also known as NxF ) is a member of the basic helix-loop-helix/Per-Arnt-Sim ( bHLH-PAS ) homology protein family [24]; it functions as a transcriptional regulator with possible roles in cell survival and differentiation [25] , [26] . Nr4a1 ( also known as nur77 or NGFIB ) is an orphan nuclear hormone receptor with possible pro-apoptotic and anti-apoptotic functions [27]–[29] . Serpinb2 [also known as plasminogen activator inhibitor type-2 ( PAI-2 ) ] is a serine proteinase inhibitor that can influence cell proliferation , differentiation and cell death [30] , [31] . Inhibin β-A is a member of the transforming growth factor ( TGF ) -β superfamily [32] that can protect human SH-SY5Y neuroblastoma cells from chemical-induced death and may mediate neuroprotective actions of basic FGF [33] , [34] . We considered these 8 genes plus the previously identified pro-survival gene Btg2 ( which is robustly induced by neuronal activity in a nuclear calcium-dependent manner [6] ) as the core components of the putative neuroprotective gene program and refer to this set of genes hereafter as Activity-regulated Inhibitor of Death ( AID ) genes ( Table 1; AID genes are boxed ) . Using QRT-PCR , we confirmed the regulation of each AID gene by AP bursting and nuclear calcium signaling; the regulation of Btg2 by nuclear calcium signaling has been established previously [6] . The expression of GADD45β , GADD4γ , and Nr4a1 increased about 20 to 35 fold after AP bursting , which were the weakest inductions among this group of genes ( Figure 2 ) . Significantly higher fold changes relative to unstimulated control in either uninfected or rAAV-LacZ infected hippocampal neurons were observed for Atf3 ( 63±4 fold , uninfected; 49±4 fold , rAAV-LacZ ) , Npas4 ( 203±17 fold , uninfected; 186±9 fold , rAAV-LacZ ) , Ifi202b ( 288±19 fold , uninfected; 246±30 fold , rAAV-LacZ ) , and Inhibin β-A ( 432±41 fold , uninfected; 388±30 fold , rAAV-LacZ ) ( Figure 2 ) . For Serpinb2 , QRT-PCR revealed a very dramatic , 1839±45 fold increase in expression following AP bursting in uninfected hippocampal neurons and 1781±48 fold increase in rAAV-LacZ infected hippocampal neurons ( Figure 2 ) . This is - to the best of our knowledge - the highest fold change ever observed for a signal–regulated gene . For all AID genes , we confirmed the requirement for nuclear calcium signaling for their induction by AP bursting ( Figure 2 ) . Most dramatic inhibitions of well over 80 percent were observed for Inhibin β-A ( 93±3% inhibition ) , Npas4 ( 92±3% inhibition ) , Ifi202b ( 87±4% inhibition ) , and Serpinb2 ( 85±3% inhibition ) , and Atf3 ( 80±5% ) ( Figure 2 ) . We also included prostaglandin-endoperoxide synthase 2 ( Ptgs2; also known as Cox2 ) in the QRT-PCR analysis . Expression of Ptgs2 is robustly induced by neuronal activity ( 80±24 fold , uninfected; 78±18 fold , rAAV-LacZ ) , and this induction was inhibited by CaMBP4 by 68±20% ( Figure 2 ) . Because there is no available evidence for a role of Ptgs2 in promoting survival of neuronal or non-neuronal cells , this gene served as one of the negative controls in the in vivo survival experiments ( see below ) . Given the importance of CREB and its co-activator CREB binding protein ( CBP ) in mediating transcriptional activation by synaptic activity and nuclear calcium signaling [12] , [35] , [36] and the critical role of CREB in neuronal survival [3] , [37] , [38] , we carried out data base searches to determine whether the nuclear calcium-regulated genes , in particular AID genes , contain putative CREB binding sites . Information retrieved from two databases ( the CREB ‘regulon’ ( http://saco . ohsu . edu/ ) [39] and the CREB target gene database ( http://natural . salk . edu/creb/ ) [40] indicated that a large fraction of the nuclear calcium regulated gene pool ( 56% ) and all AID genes except Npas4 contain one or several CREs or CRE-like sequences , suggesting that they could be CREB target genes ( Table S1 ) . In addition to CREB , other CBP-recruiting transcription factors may contribute to the regulation of AID genes by nuclear calcium signaling . We next investigated the role of the nuclear calcium/calmodulin-dependent protein kinase IV ( CaMKIV ) in the regulation of AID genes . CaMKIV is one important mediator of nuclear calcium/CREB-regulated transcription [36] , [41]–[46] . To inhibit CaMKIV activity , we infected hippocampal neurons with a rAAV containing an expression cassette for a kinase-inactive form of CaMKIV ( CaMKIVK75E ) that functions as a negative interfering mutant of CaMKIV [36] , [42] , [43]; immunoblot analysis of expression of CaMKIVK75E in hippocampal neurons infected with rAAV-CaMKIVK75E is shown in Figure 3A . We found that in hippocampal neurons infected with rAAV-CaMKIVK75E the induction by AP bursting of all AID genes and induction of Ptgs2 was inhibited ( Figure 2 ) . For 5 AID genes ( i . e . GADD45β , GADD45γ , Serpinb2 , Inhibin β-A , and Ifi202b ) and for Ptgs2 , the percent inhibition by rAAV-CaMKIVK75E was very similar to the inhibition obtained by the blockade of nuclear calcium signaling using CaMBP4 ( Figure 2 ) . CaMKIVK75E was slightly less potent than CaMBP4 in inhibiting induction of Atf3 , Npas4 , and Nr4a1 ( Figure 2 ) , suggesting that targets of nuclear calcium other than CaMKIV may contribute to regulation of these genes by neuronal activity . These results indicate that nuclear calcium-CaMKIV is an important regulatory module of AID genes . We next investigated the role of AID genes in neuronal survival . We first carried out gain-of-function experiments in which we used rAAV-mediated gene delivery to over-express Flag-tagged AID proteins in cultured hippocampal neurons . Expression of the proteins was assessed with immunoblots using antibodies to the Flag-tag ( Figure 3A ) . Infection rates were determined immunocytochemically and ranged from 80 to 95 percent of the viable neurons ( data not shown ) . We used two types of assays to assess apoptotic cell death: growth factor withdrawal and treatment of cultured hippocampal neurons with a low concentration of staurosporine [2] , [6] , a classical inducer of apoptotic cell death . We found that compared to control ( i . e . non-infected neurons or neurons infected with rAAV-LacZ ) , cell death induced by either growth factor withdrawal or staurosporine treatment was inhibited in neurons infected with rAAV carrying Flag-tagged AID proteins ( Figure 3B ) . Inhibition of apoptosis ranged from about 30 to 95% for growth factor withdrawal-induced apoptosis and from about 40 to 80% for apoptosis induced by staurosporine ( Figure 3B ) . Over-expression of Atf3 , Nr4a1 , GADD45β , and GADD45γ yielded the most potent inhibition for growth factor withdrawal-induced apoptosis , whereas expression of Npas4 was most efficient in protecting against staurosporine-induced apoptosis ( Figure 3B ) . These results indicate that AID proteins can confer robust neuroprotection to cultured hippocampal neurons . We next investigated whether AID genes contribute to activity-dependent survival induced by AP bursting and activation of synaptic NMDA receptors [2]–[4] , [6] . For this analysis , we selected three genes ( Atf3 , GADD45β , and GADD45γ ) that protected efficiently against growth factor withdrawal-induced apoptosis and Npas4 that protected efficiently against staurosporine induced apoptosis ( see Figure 3B ) . RNA interference ( RNAi ) was used to inhibit expression of these genes following synaptic activity . DNA sequences encoding short hairpin RNAs ( shRNAs ) designed to appropriate target regions were inserted downstream of the U6 promoter of a rAAV vector that also harbors an expression cassette for humanized Renilla reniformis green fluorescent protein ( hrGFP ) [6] ( for details see Text S1 ) . To control for non-specific effects of infections with rAAVs carrying an expression cassette for shRNAs , a rAAV was used that contains a universal control shRNA ( rAAV-Control-RNAi ) , which has no significant sequence similarity to the mouse , rat , or human genome . For all rAAVs carrying an expression cassette for shRNAs , infection rates of 80 to 95 percent of the neuron population were obtained ( data not shown ) . QRT-PCR analysis revealed that RNAi was effective in eliminating induction of Atf3 , GADD45β , and GADD45γ , and Npas4 by AP bursting in hippocampal neurons . The inhibition of expression was in the range of 85% for all 4 genes; rAAV-Control-RNAi had no significant effect ( Figure 4A ) . Given the neurotropism of rAAVs used in the study [47] , the results also indicate that the induction of these genes occurs in hippocampal neurons and not in glial cells . To assess activity-dependent survival , apoptotic cells were counted after treatment with staurosporine or withdrawal of growth factors with and without previous periods of neuronal activity ( Figure 4B and 4C ) [2] , [3] , [6] . In the staurosporine assays , we detected about 10 to 15% of apoptotic cells in the control condition , which increased to about 60% after treatment ( Figure 4B ) . In the growth factor withdrawal assays , basal cell death was slightly higher ( about 20 to 25% ) and also increased to about 60% ( Figure 4C ) . Upon subjecting the neurons to a period of 12–16 hours of synaptic activity ( induced by bicuculline treatment in the presence of 4-amino pyridine , which increases the burst frequency [2] ) prior to growth factor withdrawal or staurosporine exposure , far fewer cells underwent apoptosis ( Figure 4B and 4C ) . As shown previously , this activity-dependent survival is triggered by calcium entry into the neurons through synaptic NMDA receptors and involves nuclear calcium signaling [2]–[4] , [6] . In uninfected neurons and neurons infected with rAAV-Control-RNAi , we observed the typical stimulus-induced increase in apoptotic cells; following the period of synaptic activity prior to staurosporine exposure or growth factor withdrawal , the stimulus-induced cell death was reduced ( Figure 4B and 4C ) . In contrast , in neurons infected with rAAV-Atf3-RNAi , rAAV-GADD45β-RNAi , rAAV-GADD45γ-RNAi , and rAAV-Npas4-RNAi the basal level of cell death was slightly elevated and activity-dependent survival was severely compromised ( Figure 4B and 4C ) . These results indicate that the AID genes Atf3 , GADD45β , GADD45γ and Npas4 are important for neuronal survival and represent key components of the synaptic NMDA receptor-induced genomic neuroprotective program . Virtually all cell death processes involve the deregulation of mitochondrial functions . One important early event in excitotoxic cell death is the collapse of the mitochondria membrane potential and the shift in the mitochondrial membrane permeability , known as mitochondrial permeability transition ( MPT ) [48]–[50] . Using Rhodamine 123 ( Rh123 ) imaging techniques to monitor the mitochondrial membrane potential , we have recently shown that one mechanism through which down-regulation of the tumor suppressor gene , p53 , or increasing expression of Btg2 can enhance the survival of hippocampal neurons involves inhibition of the NMDA-induced break-down of the mitochondrial membrane potentials [51] . We therefore investigated whether the identified AID genes can also act through a process that guards mitochondria against toxic insults . Rh123 imaging of uninfected hippocampal neurons and hippocampal neurons infected with rAAV-LacZ revealed the typical increase in Rh123 fluorescence after application of NMDA ( 30 µM ) , which is indicative of the break—down of mitochondrial membrane potential [2] , [51] , [52] . In contrast , in hippocampal neurons that had been infected with rAAV to express the AID genes Npas4 , Inhibin β-A , Ifi202b , and Nr4a1 , or to express the previously identified neuroprotective gene , Bcl6 [6] , the NMDA-induced loss of mitochondrial membrane potential occurred with slower kinetics and reached significantly lower magnitudes ( Figure 5A–5C ) . A quantitative analysis of the imaging data revealed that expression of Npas4 and Bcl6 had the largest inhibitory effects on the NMDA-induced break—down of mitochondrial membrane potential ( Figure 5B and 5C ) . The inhibitions by Inhibin β-A , Ifi202b , and Nr4a1 were smaller but also significant , whereas no significant reductions were observed for Atf3 , GADD45β , GADD45γ , and Serpinb2 under the conditions used ( Figure 5B and 5C ) . These results suggest that one converging point common to several AID genes is the mitochondria , which are rendered more resistant against death signal-induced dysfunction . We next analyzed the neuroprotective activity of AID genes in vivo . Stereotaxic injection was used to deliver rAAVs carrying expression cassettes for AID genes or appropriate negative controls ( i . e . rAAV-LacZ and rAAV-Empty ) to the dorsal hippocampus of male Sprague-Dawley rats weighing 230 to 250 g . We also included Ptgs2 in our in vivo analysis as an additional negative control . Expression of Ptgs2 is robustly induced by neuronal activity in a nuclear calcium/CaMKIV dependent manner ( Figure 2 ) . However , there is no available evidence for a role of Ptgs2 in promoting survival of neuronal or non-neuronal cells and we therefore expected that expression of Ptgs2 in vivo would not provide neuroprotection . Two weeks after viral delivery , the rats were injected intra-peritoneally with kainic acid ( KA ) , which induces seizures leading to cell death in the hippocampus [53] . The animals were sacrificed three days after KA injection . The brains were removed , cut into slices , and stained with the histofluorescent label , Fluoro-Jade C , which serves as a very reliable marker for degenerating neurons [54] . The slices were immunostained with antibodies to the neuronal marker , NeuN , and antibodies to the Flag tag to detect the over-expressed proteins . We found widespread KA-induced cell death in the CA1 region of the hippocampus of animals injected with rAAV-Empty , rAAV-LacZ , and rAAV-Ptgs2 , as well as in the CA1 region of the non-injected side of the hippocampus ( Figure 6 ) . In contrast , expression of AID genes in the CA1 area of the injected side of the hippocampus protected against KA-induced cell death ( Figure 7 ) . Quantification of the Fluoro-Jade C signals revealed inhibitions of KA-induced cell death of 85±7% ( Atf3 ) , 87±5% ( Btg2 ) , 92±6% ( GADD45β ) , 93±2 ( GADD45γ ) , 96±1% ( Ifi202b ) , 70±8% ( Inhibin β-A ) , 92±3% ( Npas4 ) , 90±7% ( Nr4a1 ) , and 96±1% ( Serpinb2 ) ; over-expression of Ptgs2 or LacZ did not inhibit KA-induced cell death ( Table 2 ) . Expression of Btg2 , Ifi202b , and Serpinb2 consistently led to neuroprotection also on the contralateral ( i . e . non-injected ) side ( Figure 7 , Table 2 ) . This could be due to secretion of the neuroprotective protein , which may be the case for the serine proteinase inhibitor , Serpinb2 . It is also conceivable that expression of neuroprotective proteins in the processes of infected neurons , which project to the contralateral hippocampus , can promote survival of target neuron; this may be the case for flag-tagged Ifi202b which is readily detectable in the hippocampus of the non-injected , contralateral side ( see Figure 7 ) .
Activity-dependent , long-lasting neuroprotection as well as other adaptive responses in the nervous system require the dialogue between the synapse and the nucleus . There is growing evidence to suggest that calcium signals propagating from their site of activation at the plasma membrane towards the cell soma and the nucleus are key mediators of synapse-to-nucleus communication . Nuclear calcium , most likely acting via nuclear calcium/calmodulin dependent protein kinases , controls CREB/CBP-dependent transcription [12] , [35] , [36] , [43]–[46] and is thought to regulate genomic programs critical for neuronal survival , synaptic plasticity , memory formation , and emotional behavior [3] , [6] , [55]–[59] . In this study , we identified the nuclear calcium-regulated gene pool in hippocampal neurons . The large number of 185 genes induced or repressed by nuclear calcium signaling is not unexpected . A genome-wide analysis revealed that CREB – the principal target of nuclear calcium signaling - can occupy between 4000 and 6000 promoter sites in the rat or human genome , although the transcription of only a subset of those genes is signal-induced in a given cell type perhaps due to preferential recruitment of CBP [39] , [40] , [60] . Indeed , 56% of nuclear calcium-regulated genes and all AID genes except Npas4 are known or putative CREB targets ( Table S1 ) , underscoring the importance of the nuclear calcium-CREB axis in neuronal survival . The AID genes characterized in this study fall into two functional categories: regulators of gene transcription and secreted proteins . Although they may act in concert to collectively provide full neuroprotection , over-expression of individual AID genes is sufficient to promote survival . Moreover , RNAi-based loss-of-function experiments indicate that the selective reduction of individual AID genes can compromise the activity-induced build-up of a neuroprotective shield . These results could be explained by a possible convergence of AID genes on one or a small number of targets that execute protection . Given that 7 out of 9 AID genes are putative regulators of gene expression , the existence of common target genes that are part of a ‘second wave’ transcriptional response is conceivable . Under physiological condition , a signal-regulated , coordinate induction of transiently expressed genes may be required for the regulation of common targets; thus interference with one or a small number of AID genes would disturb the system . However , high-level , constitutive expression of individual AID genes may be sufficient to activate or inactivate down-stream regulators of survival . The regulation of a putative common target could involve direct trans-activation by AID gene products through binding to the target genes' promoter elements , although other modes of regulation ( such as control of mRNA or protein stability or activity-regulating post-translational modifications ) are conceivable and may involve secondary responses triggered by AID genes . Additional components of such a regulatory network may include other survival-promoting transcriptional regulators such as C/EBPβ [61]–[64] , or the secreted AID genes Serpinb2 and Inhibin β-A , or Bdnf [65] ( see Table 1 ) ; these genes may contribute through transcription-dependent and transcription-independent processes to the funneling of information flow and the reduction of complexity to few molecules implementing neuroprotection . Mitochondria , which are vital for supplying the energy required to maintain life , may be the end-point of neuroprotective processes . In this study we show that expression of AID genes renders the mitochondria of hippocampal neurons more resistant to harmful conditions . Thus , neuroprotection may ultimately guard mitochondria against stress and toxic insults to prevent mitochondrial dysfunction . The finding that activation of synaptic NMDA receptors and calcium signaling to the cell nucleus builds up a strong and lasting neuroprotective shield may change our view of neurodegenerative disorders and cell death associated with aging . Proper functioning of the endogenous neuroprotective machinery requires a sequence of events that can be disturbed at the level of synaptic transmission , synaptic NMDA receptor activation , the generation of calcium signal and their propagation to the cell nucleus , and the regulation of gene transcription . Malfunctioning of calcium signaling towards and within the cell nucleus may lead to neurodegeneration and neuronal cell death . In Alzheimer's disease cell death may be caused by compromised endogenous neuroprotection due to impaired synaptic transmission and synapse loss caused by A-β or changes in calcium homeostasis or calcium signaling in neurons expressing mutant presenilin-1 [66]–[71] . Consistent with this concept is the observation that compared to an age-matched healthy control group , individuals with Alzheimer's disease have reduced levels of the activated ( i . e . phosphorylated ) form of CREB [72]; calcium signaling to the cell nucleus is the key inducer of CREB phosphorylation on its activator site serine 133 [12] . Similarly , in aged neurons , calcium signaling may be altered at the level of calcium signal generation and/or calcium signal propagation [73]–[75] . This could explain the reduced levels of serine 133-phosphorylated CREB in the hippocampus of aged , learning-impaired rats [76]–[78] , which could lead to compromised endogenous neuroprotection , progressive cell loss and cognitive decline . The development of strategies to boost the endogenous neuroprotective machinery may lead to effective therapies of neurodegenerative condition . Both in disease and aging , health and functionality of neurons may be preserved by expressing AID genes or by restoring or enhancing key signals , in particular nuclear calcium .
Hippocampal neurons from newborn C57/Black mice were cultured in Neurobasal media ( Invitrogen , Carlsbad , CA , USA ) containing 1% rat serum , B27 ( Invitrogen , Carlsbad , CA , USA ) , and penicillin and streptomycin ( Sigma ) . The procedure used to isolate and culture hippocampal neurons has been described [79] , [80] . The hippocampal cultures used for this study typically contained about 10 to 15% glial cells and therefore a fraction of the RNA isolated from the cultures was derived from glial cells . Stimulations were done after a culturing period of 9 to 12 days during which hippocampal neurons develop a rich network of processes , express functional NMDA-type and AMPA/Kainate-type glutamate receptors , and form synaptic contacts [12] , [81] . Action potential bursting was induced by treatment with the GABAA receptor antagonist bicuculline ( Sigma ) ( 50 µM ) as described previously [2] , [11] , [12] . In the survival experiments , neurons were treated for 16 hours with bicuculline in the presence of 250 µM 4-amino pyridine ( 4-AP; Calbiochem ) [2] . 4-AP increases the frequency of the bicuculline-induced action potentials bursts , thereby enhancing nuclear calcium , CREB-mediated transcription , and activity-induced neuroprotection [2] , [11] , [12] . DNA microarray analysis was done using Affymetrix GeneChip Mouse Genome 430 2 . 0 Arrays . See Text S1 for details . The vectors used to construct and package rAAVs have been described previously [6] . The rAAV cassette for mRNA expression contains a CMV/chicken β actin hybrid promoter . The following rAAVs were generated and confirmed by DNA sequencing: rAAV-Atf3 , rAAV-CaMKIVK75E , rAAV-GADD45β , rAAV-GADD45γ , rAAV-Ifi202b , rAAV-Inhba , rAAV-LacZ , rAAV-Npas4 , rAAV-Nr4a1 , rAAV-Ptgs2 , and rAAV-Serpinb2 . rAAV-Btg2 and rAAV-CaMBP4 have been described [6] . All rAAV-expressed proteins except hrGFP carry a Flag tag . For shRNA expression , a rAAV vector was used that contains the U6 promoter for shRNA expression and a CMV/chicken β actin hybrid promoter driving hrGFP expression [6] . For details on the construction of rAAVs expressing shRNAs see Text S1 . Hippocampal neurons were infected with rAAVs at 4 days in vitro ( DIV ) . Infection efficiencies were routinely determined immunocytochemically at 9 DIV or 10 DIV using antibodies to the Flag tag or to hrGFP or by analyzing the fluorescence of hrGFP; they ranged from 80 to 95 percent of the viable neurons [6] . To determine the mRNA expression levels of Atf3 , GADD45β , GADD45γ , Ifi202b , Inhba , Npas4 , Nr4a1 , Ptgs2 , Serpinb2 , Gusb , and glyceraldehyde-3-phosphate dehydrogenase ( Gapdh ) , QRT-PCR was performed using real-time TaqMan technology with a sequence detection system model 7300 Real Time PCR System ( Applied Biosystems , Foster City , California , USA ) . For further details , see Text S1 . As described previously [2] , [3] , two types of assays were used to investigate apoptotic cell death and the protection from cell death afforded by a period of action potential bursting . At 10 DIV , activity-dependent survival was induced by treatment of the neurons for 16 hours with bicuculline ( 50 µM ) and 4-AP ( 250 µM ) . All electrical activity of the network was subsequently stopped using tetrodotoxin ( TTX; TOCRIS Bioscience ) ( 1 µM ) followed by keeping the cells either in regular medium ( containing growth and trophic factors ) with or without staurosporine ( Calbiochem ) ( 10 nM ) , or in medium lacking growth and trophic factors , all in the presence of TTX ( 1 µM ) . The principal growth and trophic factors in the regular , serum-free hippocampal medium [termed transfection medium ( TM ) [76]] are insulin , transferrin , and selenium . Staurosporine-induced and growth factor withdrawal-induced apoptosis was assessed after 36 hours and 72 hours , respectively , by determining the percentage of hippocampal neurons with shrunken cell body and large round chromatin clumps characteristic of apoptotic death [2] , [3] . In the growth factor withdrawal assays , basal cell death is slightly higher due to the differences in the time that the neurons are kept in serum free , TTX-containing media . At least 20 visual fields from each coverslip ( corresponding to 1500–2000 cells per coverslip ) were counted with Hoechst 33258 ( Serva ) and the percentage of dead cells was determined . TUNEL assays ( Roche , Mannheim , Germany ) were done according the instructions provided by the manufacturer and were used to validate the analysis of cell death using the Hoechst 33258 stain . Photomicrographs of examples of healthy and apoptotic hippocampal neurons stained with TUNEL and with Hoechst 33258 are shown in Figure S1 . All cell death analyses were done without knowledge of the treatment history of the cultures . All results are given as means±SEM; statistical significance was determined by ANOVA . Imaging of mitochondrial membrane potential was done using Rhodamine 123 ( Rh123; Molecular Probes , Eugene , OR ) as described [2] , [51] , [52] . Imaging and data analysis were performed without knowing the experimental conditions . Quantitative measurements are given as means±SEM from n≥4 experiments , with at least 100 cells analyzed each . Statistical significance was determined by ANOVA . rAAVs were delivered by stereotaxic injection into the dorsal hippocampus of male Sprague-Dawley rats weighing 230–250 g . Rats were randomly grouped and anaesthetized with ketamine . A total volume of 3 µl containing 3×108 genomic virus particles were injected unilaterally over a period of 30 min at the following coordinates relative to Bregma: anteroposterior , −3 . 8 mm; mediolateral , 2 . 8 mm; dorsoventral , −2 . 8 to −3 . 8 mm from the skull surface . Procedures were done in accordance with German guidelines for the care and use of laboratory animals and to the respective European Community Council Directive 86/609/EEC . Two weeks after rAAV delivery , rats were injected with kainate ( Sigma; 10 mg/kg i . p . ) or vehicle ( phosphate-buffered saline , PBS ) , and monitored for at least 4 hours to categorize the severity of epileptic seizures according to following criteria: level 1 , immobility; level 2 , forelimb and/or tail extension , rigid posture; level 3 , repetitive movements , head bobbing; level 4 , rearing and falling; level 5 , continuous rearing and falling; level 6 , severe tonic-clonic seizure [82] . Only animals that exhibited at least level 4 or 5 of epileptic seizure behavior were analyzed further . Three days after seizure induction , animals were deeply anesthetized with an overdose of Nembutal ( 300 mg/kg ) , pre-perfused transcardially with PBS , and perfused with 200 ml of neutral phosphate buffered 10% formalin ( Sigma ) . Brains were removed and post-fixed overnight in the same fixative solution . For cryoprotection , brains were incubated in 30% sucrose in PBS for 2 days . Brains were rapidly frozen on dry ice . Frozen sections ( 40 µm thick ) were collected in PBS . Three consecutive sections separated by a 240 µm distance were used for immunostaining and Fluoro-Jade C staining ( Histo-Chem , Inc . , Jefferson , Arkansas , USA ) , which selectively stains degenerating neuronal cell bodies and processes , regardless of the mechanism of cell death . Transgene expression was detected with anti-Flag antibodies ( 1∶2500 , M2 mouse monoclonal; Sigma ) . Neuronal cell loss was assessed with NeuN immunostaining ( 1∶500 , mouse monoclonal; Chemicon ) . Immunostaining was done using standard procedures . Fluoro-Jade C staining was done as described previously [54] . Images of Fluoro-Jade C staining were taken and the signals in the CA1 area of the hippocampus were quantified . The quantification was performed without knowing the experimental conditions . Images ( 10× objective , 1600*1200 pixels ) of Fluoro-Jade C stained sections were taken at central CA1 region infected with rAAVs . Sections were collected every 240 µm by cryostat sectioning at the level of 3 . 0∼5 . 0 mm posterior to Bregma . Five sections from each brain hemisphere were chosen for image analysis . Fluoro-Jade C signals from the images were quantified with NIH ImageJ software ( National Institute of Health , Bethesda , MD , USA ) . Background intensity was measured from the CA1 area lacking positively-stained neuronal cell bodies . Threshold level was set as means of the background +3 SD . The CA1 pyramidal cell layer was encircled manually and particle analysis was performed . Particles were defined as 30 pixels to infinity , roundness 0∼1 . 0 . Total pixel of Fluoro-Jade C positive particles from each section was obtained to calculate the mean cell death area for each hemisphere . All pixel values were normalized to the average kainate-induced Fluoro-Jade C signal from the control ( i . e . non-injected ) hemispheres of kainic acid treated animals; the signal obtained from the non-injected hemispheres of the animals injected with rAAV-Btg2 , rAAV-Ifi202b , and rAAV-Serpinb2 was not included into the calculation of the average kainate-induced Fluoro-Jade C signal because for rAAV-Btg2 , rAAV-Ifi202b , and rAAV-Serpinb2 we observed neuroprotection on the contralateral , non-injected hemispheres . All animal experiments were done in accordance with the international ethical guidelines for the care and use of laboratory animals and were approved by the local animal care committee of the Regierungspräsidium Karlsruhe . We minimized the number of animals used and their suffering .
|
The dialogue between the synapse and the nucleus plays an important role in the physiology of neurons because it links brief changes in the membrane potential to the transcriptional regulation of genes critical for neuronal survival and long-term memory . The propagation of activity-induced calcium signals to the cell nucleus represents a major route for synapse-to-nucleus communication . Here we identified nuclear calcium-regulated genes that are responsible for a neuroprotective shield that neurons build up upon synaptic activity . We found that among the 185 genes controlled by nuclear calcium signaling , a set of 9 genes had strong survival promoting activity both in cell culture and in an animal model of neurodegeneration . The mechanism through which several genes prevent cell death involves the strengthening of mitochondria against cellular stress and toxic insults . The discovery of an activity-induced neuroprotective gene program suggest that impairments of synaptic activity and synapse-to-nucleus signaling , for example due to expression of Alzheimer's disease protein or in aging , may comprise the cells' own neuroprotective system eventually leading to cell death . Thus , malfunctioning of nuclear calcium signaling could be a key etiological factor common to many neuropathological conditions , providing a simple and unifying concept to explain disease- and aging-related cell loss .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"neuroscience/neuronal",
"signaling",
"mechanisms",
"neuroscience/neurobiology",
"of",
"disease",
"and",
"regeneration",
"cell",
"biology/gene",
"expression",
"cell",
"biology/neuronal",
"signaling",
"mechanisms"
] |
2009
|
Nuclear Calcium Signaling Controls Expression of a Large Gene Pool: Identification of a Gene Program for Acquired Neuroprotection Induced by Synaptic Activity
|
There is a growing recognition for the importance of proteins with large intrinsically disordered ( ID ) segments in cell signaling and regulation . ID segments in these proteins often harbor regions that mediate molecular recognition . Coupled folding and binding of the recognition regions has been proposed to confer high specificity to interactions involving ID segments . However , researchers recently questioned the origin of the interaction specificity of ID proteins because of the overrepresentation of hydrophobic residues in their interaction interfaces . Here , we focused on the role of polar and charged residues in interactions mediated by ID segments . Making use of the extended nature of most ID segments when in complex with globular proteins , we first identified large numbers of complexes between globular proteins and ID segments by using radius-of-gyration-based selection criteria . Consistent with previous studies , we found the interfaces of these complexes to be enriched in hydrophobic residues , and that these residues contribute significantly to the stability of the interaction interface . However , our analyses also show that polar interactions play a larger role in these complexes than in structured protein complexes . Computational alanine scanning and salt-bridge analysis indicate that interfaces in ID complexes are highly complementary with respect to electrostatics , more so than interfaces of globular proteins . Follow-up calculations of the electrostatic contributions to the free energy of binding uncovered significantly stronger Coulombic interactions in complexes harbouring ID segments than in structured protein complexes . However , they are counter-balanced by even higher polar-desolvation penalties . We propose that polar interactions are a key contributing factor to the observed high specificity of ID segment-mediated interactions .
In cells , communication is established principally by protein-protein interactions [1] . It is clear that proteins have to interact in a specific manner in order for messages/signals to be transmitted correctly . Therefore , significant efforts have been made to understand the driving mechanisms of protein-protein interactions [2]–[7] . The picture that has emerged from these studies illustrates the removal of non-polar residues from the aqueous environment as a major thermodynamic driving force for protein binding [8] , [9] . Consistently , interaction surfaces have been shown to be enriched in hydrophobic residues , especially in the most buried regions of interfaces [10]–[12] . In contrast , specificity in interactions is believed to rely on shape complementarity , hydrogen bonding , and salt-bridge formation [13] , [14] . In this context , the role of electrostatics in protein-protein interactions has been studied extensively [6] , . It has been shown that salt bridges in protein interfaces can contribute favorably to protein stability and the free energy of binding through Coulombic interactions , but that this effect is often counterbalanced by very unfavorable desolvation [16]–[21] . Hence , the electrostatic component of the free energy of binding often destabilizes the protein complex . Despite that , salt bridges are still important for binding because of their contribution to interaction specificity [21] . This contribution is explained by the large energetic penalty for burying but not compensating for charged residues . Some of the mechanisms and principals of protein-protein interactions derived from previous studies are likely to be challenged for interactions that involve intrinsically disordered ( ID ) segments of proteins [22]–[25] . One obvious reason is that ID segments lack a unique three-dimensional structure when free in solution and are likely to fluctuate between different conformations that lack any secondary structure or visit them only transiently [26] , [27] . A few recent studies analyzed the interfaces of ID segments that are in complex with folded proteins [28]–[30] . In contrast to typical ID regions , which are generally enriched in charged residues and depleted in large hydrophobics , it was revealed that ID segments involved in protein binding tend to be enriched in hydrophobic residues [28]–[30] . Given the dominance of hydrophobic residues in ID segments that are part of interfaces , researchers have proposed that interactions mediated by ID segments may be less specific than interactions between folded proteins [29] . However , this idea seems to be at odds with results from various studies . For instance , intrinsic disorder has been shown to be important in specific protein-DNA interactions [31] . Specificity in interactions mediated by ID regions is often explained by the mechanism of coupled folding and binding , which gives ID binding regions the ability to mold into a precise fit for a given binding surface [32] , [33] . However , there is a hint of paradoxical nature in this argument because the ability to mold implies the flexibility to fit a wide selection of binding surfaces promiscuously . Logic dictates that the sequence of the ID segment should encode determinants that constrain their promiscuity in binding . A reasonable source of specificity is the polar properties of ID segments . Here , we devised a new structure-based computational method to identify ID segments that are in complex with other macromolecules . Making use of the extended nature of most ID segments when in complex with globular proteins , we identified complexes between globular proteins and ID segments ( ID complexes henceforward ) from the Protein Data Bank ( PDB ) by using radius-of-gyration-based selection criteria . The method was first benchmarked on 52 complexes where one partner is experimentally proven to be ID and then applied to a large non-redundant PDB dataset to identify new ID complexes . Consistent with previous studies , we find the interfaces of ID complexes to be enriched in hydrophobic residues , and that these residues contribute significantly to the stability of the interaction interfaces . However , our results show that polar interactions play a larger role in ID complexes than in structured protein complexes . Computational alanine scanning and salt-bridge analysis indicate that interfaces in ID complexes are highly complementary with respect to electrostatics , more so than interfaces of globular proteins . Follow-up calculations of the electrostatic contributions to the free energy of binding with DelPhi uncovered high desolvation penalties in ID complexes . However , these penalties are often nearly compensated by favourable Coulombic interactions that are significantly stronger than those in structured-structured protein complexes . In the light of the magnitude of the electrostatic energy terms that we estimated for ID complexes , we suggest that strong electrostatic interactions are a key component of the highly complementary interactions between ID segments and their partners that translates to high specificity .
It has been shown that ID segments bury more solvent accessible surface area relative to their length than do structured proteins when they interact with other macromolecules [28] . In addition , several examples have been reported in which ID segments wrap around binding partners ( e . g . p27Kip1 and p21Cip1 [34] , [35] , Figure 1b ) . This motivated us to hypothesize that ID segments could be identified based on their geometry when they are bound to structured proteins . Specifically , the radius of gyration ( Rg ) of a protein is a measure of its size and will reveal the extendedness of the protein chain when divided by chain length ( N ) . We tested this hypothesis on a set of 52 long ID segments found in the literature for which the structure when in complex with a partner protein ( mainly a globular one ) is known . As a negative set , we selected 762 complexes from the 3D Complex database ( Figure 1c ) , which is a database of proteins classified based on sequence , structure and topology . This should give a negative set that is enriched in structured proteins with different folds . We will refer to this dataset as the 3D complexes . Consistent with our hypothesis and previous observations [30] , we found that ID segments in complex with structured proteins tend to have larger Rg values for a given protein length than do structured proteins ( Figure 1a ) . In order to see whether Rg/N can be used as an effective classifier of these structures , we used receiver operating characteristic ( ROC ) curves . The ROC curve constructed from calculating the Rg/N of the positive ID segment and the negative structured protein datasets has an area under the curve ( AUC ) of 0 . 986 ( Figure S1a , Table S1 ) . However , Rg/N may not provide the best classifying results because Rg does not scale linearly with N . Flory has shown that for heteropolymers , Rg and N can be related by a simple scaling law [36]:where Ro is a constant correlated with persistent length of the polymer and ν is a scaling factor that varies with different solvents due to resulting changes in the compaction of the proteins . Denatured proteins , which are in a collapsed state in poor solvent , have a scaling factor ν of 0 . 33 [36] . Figure S2 shows that the 3D complex structured proteins have , as expected [36] , a low scaling factor ν of 0 . 35 . On the other hand , the ν of the ID segments is 0 . 5 due to their expanded conformations . In order to determine whether a scaling factor ν different from 1 would improve the discrimination of ID segments from structured proteins , we constructed ROC curves at different scaling factors . We determined an optimal cutoff value at each scaling factor by selecting the threshold with the highest Matthew's correlation coefficient ( MCC ) [37] , [38] . Out of all the scaling factors that we tested , ν of 1 results in a classifier that has the highest AUC ( Table S1 ) . At the chosen cutoff values , ν of 1 also gives the highest MCC with a low false discovery rate and high sensitivity . We cannot deny the possibility that there are other values of ν with greater performance . Nevertheless , our calculations clearly demonstrate that Rg/N is a very effective classifier for distinguishing extended ID segments from structured proteins when in complex with partner molecules . Despite the excellent performance , we were interested in determining which proteins were falsely identified as ID . A close inspection of Figure 1a revealed that most of the false positives are short protein chains . Rg/N as a classifier appears to have difficulty discerning short ID segments from small , folded structures . Folded proteins shorter than 100 residues are often stabilized by disulfide bonds [39] , which allow them to adopt relatively expanded conformations . Indeed , proteins with disulfide bonds are enriched among the false positives that we identified with the Rg/N classifier . Therefore , we removed proteins that are rich in disulfide bonds from the negative set . In addition , coiled coils often form long stretches of helical structure ( Figure 1d ) . Although some coiled coils are known to be intrinsically disordered [40] , this is not the case for all coiled coils . Consequently , coiled coils were also removed from the negative set . As a result of the exclusion of disulfide-rich complexes and coiled coils , the AUC rises slightly to 0 . 99 ( Figure S1b , Table S2 ) and the MCC at an Rg/N of 0 . 26 Å ( Figure S1c ) is maximized to a value of 0 . 87±0 . 04 . The false discovery rate and sensitivity at this threshold are 0 . 11 and 0 . 87 respectively . Next , we applied the Rg/N classifier with a 0 . 26 Å cutoff on a non-redundant set of 6379 PDB files ( see methods ) , which provided 330 potential ID complexes . Table 1 presents the list of datasets that we created and studied in this article . A motivation for using a measure of geometry to select interacting ID segments of proteins is to have a method that does not rely on sequence information , which avoids biases in the analyses presented below . Sequence-based disorder predictions can instead be used to validate the selected structures . First , we predicted the disorder for the selected polypeptide chains with known coordinates ( Figure 2 ) . The selected chains have significantly higher predicted disorder content than a control set of non-redundant structures from the PDB ( x˜ID = 34% , x˜nrPDB = 6% , p value = 7 . 8×10−56; Wilcoxon test , x˜ is the median ) . Moreover , the selected polypeptide chains are also predicted to be significantly more disordered when compared to chains that are not expanded and have Rg/N<0 . 26 Å ( x˜Rg/N<0 . 26 Å = 8% , p value = 8 . 1×10−42; Wilcoxon test ) . Nevertheless , there are still sequences in the identified dataset that have low intrinsic disorder predicted based on their primary structure . The lower than expected predicted disorder may be explainable by the enrichment of ID interaction regions in the selected sequences . Interaction-prone regions within ID segments are known to be highly hydrophobic and have residue compositions more similar to the buried regions of structured proteins than to the rest of the ID segments [29] . To test this reasoning , we extended the sequence of the identified ID segments by 30 residues at their N and C-termini ( i . e . by residues lacking coordinates in PDB files ) and repeated the analysis . As expected , the distribution of percentage disorder increases significantly ( Figure S3 , x˜ID = 34% , x˜extended ID = 41% , p value = 8 . 1×10−3; Wilcoxon test ) , which suggests that the flanking regions of these ID segments are often more disordered than the interacting regions . As a considerable number of ID segments in our set do not have both flanking regions , i . e . , they are at the termini , or the extended sequences were not readily found , we expect the difference to be greater still if we included more flanking regions . The overall amino acid composition of the ID segments , which includes only residues with coordinates in the PDB files , is in agreement with the disorder prediction . The composition of the identified protein segments shown in Figure 3 is enriched in disorder-promoting residues compared to 3D complex proteins [41] , especially charged residues such as R and K ( R p value = 4 . 2×10−9 , K p value = 1 . 4×10−4; Wilcoxon Test ) . The order-promoting residues , which are generally the hydrophobic amino acids , are depleted in our dataset ( W p value = 2 . 6×10−13 , F p value = 3 . 7×10−8 , I p value = 4 . 5×10−3 , L p value = 7 . 7×10−2 , V p value = 9 . 1×10−29 , Y p value = 1 . 9×10−11; Wilcoxon Test ) . When we again extended the number of residues analyzed by 30 on each end of the selected polypeptide chains , there was an increase in some of the disorder promoting residues ( Figure 3 ) . In order to put the residue composition of the ID segments that we identified into perspective , Figure S4 also shows the relative residue composition for the set of 52 ID segments from our literature search that we used to evaluate the classifier as well as 1150 ID protein segments taken from the DisProt database , which is a database of proteins with experimental evidence for intrinsic disorder [42] . Although the exact magnitude of the enrichment or depletion of specific amino acids differs , order-promoting hydrophobic residues are depleted and disorder-promoting charged and polar residues are enriched in all ID protein datasets when compared to 3D complex proteins . In summary , the presented results show that our method is able to select for ID proteins , and respectively , for ID protein segments that are in complex with other macromolecules . Another known property of proteins with ID segments is their involvement in signalling and regulation [43] . By classifying gene ontology terms into nine major groups and assigning the ID segments into these groups , we observed significant enrichment of proteins involved in cytoskeleton , signalling , and transport ( Figure S5a ) . We also observed that our ID dataset is not biased towards a single group of proteins ( Figure S5b ) . The main focus of our study is on the interaction between ID segments and their structured binding targets , so we analyzed the interface residue composition . Interface residues are categorized into the rim and core as defined by Levy [10] . The residues in the interface core are the most buried residues upon protein binding and are generally at the central region of the interface . The residues on the outer edges of the interface that remain partially exposed to solvent are part of the interface rim . For complexes between globular proteins , it has been shown that the interface rim defined in this way has a residue composition similar to the protein surface [10] . In contrast , the core has a distinctive residue composition that is an intermediate between the surface and interior of folded proteins . Figure 4 shows that both ID and 3D complexes are significantly enriched in hydrophobic residues in the core interface regions when compared to the rim . The extent of enrichment in hydrophobic residues is especially notable in the core region of the ID segment ( e . g . ID core vs . ID rim: W p value = 1 . 0×10−4 , F p value = 3 . 4×10−13 , I p value = 2 . 3×10−14 , Y p value = 0 . 16 , V p value = 1 . 4×10−6 , L p value = 1 . 2×10−19 , A p value = 1 . 8×10−6; Wilcoxon test ) . Importantly , the core residues of ID segments are more often hydrophobic than the core residues in 3D complexes . Compared to the 3D interface cores , the ID segment cores have lower percentage composition of charged and polar residues ( ID core vs . 3D core: K p value = 7 . 5×10−3 , E p value = 2 . 6×10−7 , D p value = 7 . 4×10−7; R p value = 6 . 1×10−3 , H p value = 7 . 9×10−8 , N p value = 5 . 5×10−3 , Q p value = 9 . 3×10−5; Wilcoxon test ) . Instead , there are greater proportions of hydrophobic residues such as F , I , L and A ( ID core vs . 3D core: F p value = 5 . 8×10−2 , I p value = 9 . 0×10−11 , L p value = 5 . 7×10−11 , A p value = 1 . 0×10−4; Wilcoxon test ) . These findings advocate that hydrophobic residues play a critical role in ID complex interfaces , which is in agreement with previous studies by Vacic et al . and Mészáros et al . [29] , [30] . However , comparison of Figure 4a and 4b clearly reveals that the rim contains significantly more charged and polar residues than the core and this is the case not only for 3D complexes , as shown previously [10] , but also for ID complexes . Compared to the core , K , E , D , N , R and H are significantly enriched in the rim of ID segments ( ID core vs . ID rim: K p value = 4 . 7×10−26 , E p value = 8 . 2×10−24 , D p value = 3 . 4×10−14 , N p value = 8 . 6×10−7 , R p value = 3 . 8×10−11 , H p value = 2 . 2×10−4; Wilcoxon test ) . These distinctions are evidence for the good efficacy of Levy's interface definitions . In contrast to the interface core , we find less significant differences in the distribution of residues in the rim when comparing ID and 3D complexes . Accounting for the total size of the interface core and rim regions provides another perspective for understanding the differences between ID and 3D complexes ( Table S3 ) . One very distinctive feature is the large number of residues in the rim of binding partners of ID segments ( BPs ) . On average , the rim of BPs includes 12% more residues than the rim in classical protein complexes ( p value = 6 . 6×10−8; Wilcoxon test ) while the number of residues on the rim of the ID segment is closer to that of 3D complex proteins ( p value = 0 . 71; Wilcoxon test ) . As a result , the average number of most charged residue types in the rim is significantly greater in the rim of BPs compared to the rim of ID segments ( E p value = 1 . 1×10−8 , D p value = 1 . 4×10−7 , H p value = 8 . 8×10−3; Wilcoxon test ) as well as the rim of 3D complexes ( E p value = 8 . 6×10−6 , R p value = 1 . 7×10−2 , D p value = 3 . 7×10−3; Wilcoxon test ) ( Figure S6 ) . The interface core of ID segments , on the other hand , contains 17% and 22% fewer residues compared to the BPs' and 3D complex proteins' interface core , respectively ( ID vs . ID partner p value = 4 . 8×10−6 , ID vs . 3D p value = 2 . 2×10−3; Wilcoxon test ) . In summary , the ID complex interface appears to consist of a small but very hydrophobic core on the ID segment's side and a large and polar rim on the BP's side . To gain further insights into the interactions in ID and 3D complexes , we calculated the number of salt bridges and hydrogen bonds ( Table 2 ) . An average of 48% more salt bridges are found in ID complexes compared to complexes with only structured members ( p value = 5 . 0×10−3; Wilcoxon test ) . This difference is still present after normalization by the average interface area , though not statistically significant . There are also slightly more hydrogen bonds , by 13% on average , in ID complex interfaces ( p value = 0 . 082; Wilcoxon test ) . The numbers of hydrogen bonds in ID complexes is comparable when we normalize by interface areas . Overall , these analyses confirm that the core of ID interfaces are highly enriched in hydrophobic residues , but they also reveal that charged residues are abundant in the rim , especially on the BP , and are involved in salt bridge or hydrogen bond formation . To analyze the contributions of individual residues to the stability of the interaction interface , we carried out computational alanine scans . It is well understood that , in general , only a small number of interface residues , called “hot spots” , are making the essential interactions . We defined hot spot residues as those which have a ΔΔGbind of >1 . 5 kcal/mol when mutated into alanine . In order to avoid artefacts due to low-resolution data , we performed the ALA-scan on subsets of the ID and 3D complexes that contain only high-resolution crystal structures ( resolution <2 . 5 Å; see methods ) . We also analyzed our results through comparisons with the 3D complex proteins to minimize any bias in the calculations . Moreover , we analyzed the effect of alanine mutations for residues in the ID segments and their BP separately . Analysis of our high-resolution datasets reveals a greater percentage of interface residues qualifying as hot spots in ID complexes than in 3D complex proteins at 27% and 20% , respectively ( p value = 1 . 3×10−7; Wilcoxon test ) . The percentage of interface residues qualifying as hot spots averages at 40% and 21% ( p value = 4 . 4×10−21; Wilcoxon test ) for the ID segment and their BPs respectively . Next , we dissected the contribution of hydrophobic and charged residues to the interaction energy . Figure 5a shows the distribution of change in binding free energies ( ΔΔGbind ) for the mutation of hydrophobic residues ( V , L , I , F , M , Y , W ) to alanine in ID complexes and 3D complexes . ALA mutations of hydrophobic residues in ID segments are generally more destabilizing than ALA mutations of hydrophobic residues in 3D complexes ( x˜ID = 2 . 0 kcal/mol , x˜3D = 0 . 97 kcal/mol , p value = 1 . 8×10−59; Wilcoxon test ) . This difference may be expected given the enrichment for large hydrophobic residues with large surface area in the interface core of ID segments . Another possible explanation may be that the packing of the side chains in the interface is better compared to the rigid binding in structured proteins . A previous study has shown a greater number of atoms in contact per residue in ID complex interfaces [30] , which could give rise to greater interaction energy for non-polar residues . Regardless of the reason , the fact that our results show higher ΔΔGbind for the hydrophobic residues in ID segments once again confirms that hydrophobic interactions are a key driving force for ID segment binding . A more complex picture emerges when analyzing the effect of mutations of charged residues . Figures 5b and 5c reveal the change in binding free energies upon alanine mutations of charged residues ( E , D , R , K , H ) and only those charged residues that are involved in salt bridges , respectively . As expected , the comparison of the ALA-scan results for these two groups shows that the ion-pairing residues tend to have much higher ΔΔGbind ( x˜charged = 0 . 43 kcal/mol , x˜salt bridge = 1 . 32 kcal/mol ) . The results of the ALA-scan for all charged residues ( Figure 5b ) demonstrate that mutations of charged residues on the ID segment side of the interface are , on average , significantly more destabilizing than ALA mutations of charged residues in 3D complexes ( x˜ID = 0 . 93 kcal/mol , x˜3D = 0 . 35 kcal/mol , p value = 1 . 9×10−28; Wilcoxon test ) . Moreover , mutations of these charged interface residues in the ID segment result in significantly higher ΔΔGbind when compared to mutations of charged residues on the BP's side ( x˜ID = 0 . 93 kcal/mol , x˜BP = 0 . 39 kcal/mol , p value = 4 . 3×10−18; Wilcoxon Test ) . This finding suggests that a greater proportion of the charged residues on the ID segment's interface are making specific interactions than the charged residues on the partner protein . Indeed , the disparity in ΔΔGbind for charged residues in the ID segments or their BP is smaller , though still significantly different ( x˜ID = 1 . 7 kcal/mol , x˜BP = 1 . 4 kcal/mol , p value 6 . 6×10−3; Wilcoxon Test ) , when we analyzed only the salt-bridge-forming residues ( Figure 5c ) . Importantly , the ALA mutations of salt-bridging residues in ID segments are significantly more destabilizing than ALA mutations of salt-bridging residues in 3D complexes ( x˜ID = 1 . 7 kcal/mol , x˜3D = 1 . 1 kcal/mol , p value = 9 . 0×10−7; Wilcoxon Test ) . As pointed out above , there is a relationship between change in solvent accessible surface area and ΔΔGbind , and this is explored in Figure S7 . The correlation between ΔΔGbind and change in accessible surface area is , as expected , strong for hydrophobic residues ( R2ID = 0 . 67 , R2BP = 0 . 62 ) but extremely weak for charged residues ( R2ID = 0 . 25 , R2BP = 0 . 16 ) and those that form salt bridges ( R2ID = 0 . 13 , Figure S7a ) . The correlation is worst for the bridging residues on the BPs' side of interfaces ( R2BP = 0 . 036 ) because these residues often have small changes in solvent accessible surface area upon binding as part of the partially solvent-exposed regions . Hence , normalizing ΔΔGbind by the change in solvent accessible surface area has a pronounced effect for hydrophobic residues ( Figure S7b versus Figure 5a ) , but less so for charged residues ( Figure S7c versus Figure 5b ) . Indeed , the normalized ΔΔGbind of charged residues in ID segments are still significantly larger than those in 3D complexes ( x˜ID = 0 . 019 kcal/molÅ2 , x˜3D = 0 . 013 kcal/molÅ2 , p value = 7 . 8×10−9; Wilcoxon Test ) ( Figure S7c ) . It is clear that the results of single mutation experiments involving charged residues do not reveal whether their polar interactions are stabilizing or destabilizing . The results would only reflect the effect of the removal of this particular charged residue on the binding affinity [20] , [21] . Moreover , charged residues on the proteins can appear to have weak complementarity across the complex interface while significant complementarity can still be found in their electrostatic potential [15] . Hence , ALA-scan results cannot be used to compare the importance of electrostatic interactions for the stability of ID and 3D complexes . Consequently , we used DelPhi for continuum electrostatic calculations in order to get more accurate insights into the contribution of electrostatics to the binding of ID segments to partner proteins . To estimate the electrostatic free energies of binding , we used the same approach that has previously been used to study the binding of folded proteins [20] . This approach assumes that the structure of both binding partners do not change upon complexation . It has to be stressed that this is a considerably big assumption to make , particularly in the case of ID segments that often fold upon binding ( see also below ) . However , Sheinerman and Honig previously outlined how calculations based on this assumption can be used to compare the electrostatic contributions to binding between different protein complexes [20] . Consistently , we did not directly compare the results with experimental data but compared the electrostatic contributions to the interface stability of ID and 3D complexes . The distribution of the polar-desolvation energies of binding is shown in Figure 6a . As expected , the electrostatic desolvation energy of binding is higher for ID complexes than for 3D complexes ( x˜ID = 368 . 11 kcal/mol , x˜3D = 68 . 59 kcal/mol , p value = 2 . 2×10−10; Wilcoxon test ) . The high electrostatic desolvation energies reflect the cost of removing the polar surfaces of the larger ID complex interfaces from the high dielectric environment of the polar solvent and burying them within the low dielectric protein environment . It is interesting to note that , while most complexes have very unfavorable desolvation energies of binding , they are favorable for 32% of 3D complexes . Favorable polar solvation free energies of binding are less intuitive but were observed previously [44] . Unfavorable polar desolvation may be compensated by inter-chain electrostatic interactions , such as salt bridges . Indeed , the contribution of the Coulombic energies of binding in ID complexes is generally high enough to overcome most of the high polar-desolvation energies . Figure 6b shows that Coulombic contributions to binding are also greater in ID complexes than in conventional complexes ( x˜ID = −346 . 92 kcal/mol , x˜3D = −42 . 86 kcal/mol , p value 1 . 3×10−10; Wilcoxon test ) . Taken together , the sum of solvation free energy and the Coulombic energy result in a total contribution of electrostatics to binding that is , on average , not favorable and very similar in ID complexes and 3D complexes ( x˜ID = 24 . 64 kcal/mol , x˜3D = 31 . 02 kcal/mol , p value = 0 . 13; Wilcoxon test; Figure 6c ) . These findings are also confirmed in a subset of the ID complexes with literature support that we used to train the classifier ( Figure S8 ) . Furthermore , these trends hold true even upon normalization of the binding energy components by the interface areas ( Figure S9 ) .
In contrast to protein folding , interactions between proteins are more strongly driven by polar interactions . As pointed out by Sheinerman and colleagues [6] , protein folding is largely driven by the burial of large hydrophobic areas , which is necessary to offset the entropic cost of folding . In the binding of folded monomers , entropic penalties are much smaller , which reduces the requirements for compensatory energy gains upon binding , for instance , through the burial of large hydrophobic areas . The binding mechanism of ID segments can be considered as an intermediate between protein folding and the interaction between folded proteins in terms of entropic costs . This line of reasoning explains why ID complex stability must be afforded by the burial of extensive hydrophobic surfaces . Therefore , the interface size and packing of hydrophobic residues are key factors in determining complex affinity . The enrichment of hydrophobic residues others [28] , [29] and we found in the interface of ID complexes is a testimony of this mechanism . However , protein binding of large ID segments in extended conformations involves the formation of extensive interface-rim regions that are only partially accessible to solvent . These large , polar interface areas can cause high desolvation costs that have to be offset by complementary electrostatic interactions . In this way , the ID segments can only bind to specific binding partner ( s ) with sufficient electrostatic complementarity . This may be the key for interacting with multiple partners while being specific to every one of them . Hence , packing and size of the interface as well as its electrostatic complementarity are partners in the fine-tuning of the affinity and selectivity of ID segment interactions [22] , [33] , [57] .
The structured protein dataset was derived from the 3D Complex database [58] . 3D Complex is a database of protein complexes classified by their known three-dimensional structure in a hierarchical way . We selected dimers out of the Quaternary Structure Families of complexes and downloaded the structure coordinates from the Protein Data Bank ( PDB ) [59] . The Quaternary Structure Families grouping has little to no relation to the sequence identity of the proteins . Therefore , we applied a sequence alignment and clustering procedure on the dataset to remove redundancies . We used EMBOSS Needle [60] , a pairwise alignment tool , with the scoring matrix EBLOSUM62 and default gap penalties . Needle outputs sequence percentage similarity and identity for each chain pair . We defined redundant sequences using a sequence identity threshold defined by Rost [61]:where n is the number of percentage points above the default curve and L is the protein length . Using a threshold determined at n = 3 returned 782 non-redundant structures . With protein chains that are shorter than 20 residues removed , a final dataset of 762 protein chains was created . The structures for the test set of ID complexes , i . e . protein complexes with one interaction partner that has been identified as intrinsically disordered in experiments , were selected through literature search . We identified 40 ID complexes by searching primary literature . We further extended the dataset by including complexes identified by Mészáros and coworkers , which resulted in a total of 74 complexes [30] , [62] . Sequence alignment and clustering reduced the final test set of ID complexes to 52 . The radius of gyration ( Rg ) was calculated using the Perl script rgyr . pl from the MMTSB tool set [63] . It is defined as:where the position of the atom i and the center of mass are and respectively . The mass of atom i is and the total mass is M . Rg/N was calculated for the α-carbon coordinates of the proteins . The performance of Rg/N for separating the structures of the ID complex test set from those of the 3D complex set was first evaluated using Receiver Operating Characteristic ( ROC ) curves . Hence , the ID complex test set and 3D complex dataset were used as the positive and negative datasets , respectively . The area under the ROC curve ( AUC ) was used for evaluating the ROC curves . AUC is correlated to the accuracy of the classifier and has the advantage of being insensitive to class skew [64] . However , the ROC curve does not readily identify the optimal Rg/N threshold . A Matthew's correlation coefficient ( MCC ) curve was then used to identify the optimal threshold of Rg/N [38] . The MCC is defined as:where TP is the true positive , TN is the true negative , FP is the false positive , and FN is the false negative . The Rg/N of 0 . 26 Å was selected as the optimal threshold for identifying ID complexes . To get an averaged ROC curve and confidence levels for the MCC curve , we randomly sampled half of the ID and 3D complex datasets for 1000 repetitions . The threshold averaged ROC curves and the averaged MCC curve were calculated by using the ROCR package [65] . In order to identify new ID complexes with the Rg/N classifier , a list of non-redundant PDBs was taken from NCBI nr table ( ftp:/ftp . ncbi . nlm . nih . gov/mmdb/nrtable/ ) [66] . The nrlevel 0 was chosen , which means sequences were grouped using BLAST p values of 10−7 as a cutoff . We applied the classifier only on protein chains that have more than 20 residues with coordinates present and are interacting with at least one other partner in the PDB file . At least one of the interacting chains also has to have more than 70 residues . Protein structures from the 3D complex set that overlap in their Rg/N with structures from the ID complex test set ( Figure 1a ) are often coiled coils or disulfide-rich domains . Therefore coiled coils and disulfide-rich domains were removed from the non-redundant protein dataset . We used Socket [67] to identify coiled coils . Socket is a program that identifies coiled coils by recognizing the ‘knobs-into-holes’ packing patterns formed by interlocking α-helices in PDB file structures . Disulfide-rich protein domains are small domains whose structures are stabilized by disulfide bonds . These stable protein domains tend to be shorter than 100 residues [39] . Consequently , all protein chains shorter than 100 residues with two or more disulfide bonds were removed . Finally , transmembrane proteins were also removed because they also have high Rg/N . We used the HMMTOP server for the prediction of transmembrane helices [68] . Disopred2 [45] was used to predict disorder from protein sequence . Intrinsic disorder content for protein segments of interest ( e . g . , segments with coordinates ) was extracted from predictions for the entire protein . Whenever possible , the complete native protein sequence from the Uniprot database was used for the prediction . Otherwise , the full protein chain sequence from the PDB file was used . Proteins were mapped to gene ontology terms using the Gene Ontology ( GO ) database [69] . The gene association file for Protein Data Bank Ids was used . For unannotated PDB Ids , the Uniprot Id taken from the PDB file or the sequence information of each chain was used to find the appropriate gene ontology terms . The gene ontology terms were classified into nine groups based on each term's position on the gene ontology graph . Statistical enrichment of each group in the ID set relative to non-redundant PDB set was calculated based on hypergeometric distributions [70] . The interface of a protein complex was defined as the region with a change in solvent accessible surface area ( SASA ) upon binding . Similarly , we defined all residues with a change in SASA upon binding to be interface residues . For ID complexes and 3D complexes , the SASA was calculated for the PDB structure of the whole complex and the binding partners in isolation . The change in SASA upon protein binding is defined as:SASA was calculated with Areaimol [71] , [72] using a probe radius of 1 . 4 Å . In addition to standard van der Waals radii used by Areaimol , the radii of zinc , calcium , and sodium ions were taken from CHARMM22 parameters [73] . Importantly , when calculating the size of the interfaces , we take the average between the two sides of the protein complex . Interface residues were further divided into core and rim based on definitions by Levy [10] . The relative SASA ( rSASA ) was calculated by normalizing the residues' SASA by their SASA in a Gly-X-Gly peptide . Residues with rSASA greater than 0 . 25 in the complexed state were assigned to the rim . Residues that have rSASA less than 0 . 25 in the complexed state were assigned to the core if rSASA is greater than 0 . 25 in the uncomplexed state . The high-resolution datasets of both 3D complex dimers and ID complexes consist of X-ray structures that are higher in resolution than 2 . 5 Å . We also excluded structures that have heterogens because they cannot be readily modeled using CHARMM and other programs . We included some of the ID complex structures containing heterogens with resolutions higher than 2 . 0 Å to maximize the size of the high-resolution ID set . Coordinate ions such as calcium and zinc were included in CHARMM structures . Other small molecules that are not part of the native structure or do not interact with the ID complex interface were ignored ( e . g . glycerol from crystallization process ) . The high-resolution datasets are used for all the procedures described below . We defined salt bridges as any donor nitrogen and acceptor oxygen side chain atom pairs that are closer than 4 . 0 Å apart . Terminal charged groups were also included in the analysis . But despite the prevalence of terminal groups on relatively short ID segments , the terminal groups did not contribute significantly . Distances between each charged interface residue defined through SASA calculations and all other charged residues were calculated . We used the program HBPlus to calculate hydrogen bonds from the high-resolution protein complex structures [74] . We used the default criteria and all hydrogen bonds formed across protein complex interfaces were tabulated . The analysis did not include aromatic hydrogen bonds . FoldX version 3 . 0 beta3 was used to for the ALA-scan . Calculations were carried at 298K , pH 7 , and ionic strength of 0 . 05M [75] . The FoldX repair procedure was used on each complex before introducing alanine mutations . ALA-scan using FoldX outputs a change in free energy of folding ( ΔGmut ) for each residue . The change in binding energy upon mutation ( ΔΔGbind ) for each of the interface residue was calculated by subtracting the ΔGmut of the isolated protein subunit from that of the complex . Interface residues with a ΔΔGbind>1 . 5 kcal/mol were classified as interaction hot spots . Alanine , glycine , and proline residues were not analyzed . Polar solvation free energy and total electrostatic free energy of binding were calculated using DelPhi with a procedure similar to that of Sheinerman and Honig [20] . As discussed in their article , this is a simplified approximation that assumes the components of the complex do not undergo significant conformational changes upon binding . They described the free energy of binding ( ΔGbind ) as a sum of the free energy due to changes in conformation during binding ( ΔGstrain ) and free energy of rigid binding ( ΔGrigid ) . ΔGstrain consists of enthalpic and entropic changes upon binding and is always positive in value , which is unfavorable for binding [6] . Similar to their study , the electrostatic calculations performed in this study also consisted only of ΔGrigid . We used the CHARMM param22 charges and radii for the calculations [73] . The scale is set to 2 grids/Å with a grid size of 401 . The interior-dielectric and exterior-dielectric constants are 2 and 80 , respectively . In accordance to the method used by Sheinerman and Honig , we used an interior dielectric constant of 2 . We also tested the calculation with an interior dielectric constant of 4 . We saw the same pattern in the free energies of binding among the ID complexes and the 3D complexes , but the magnitudes of the binding free energies were much smaller ( not shown ) . For each complex , electrostatic calculations were done for the complex and both binding partners in isolation . All three structures were placed in the same grid , which was centered on the geometrical center of the protein complex . The probe radius for determining the solvent accessible surface area was 1 . 4 Å . Each calculation consisted of 2000 iterations of the linear Poisson-Boltzmann Equation ( PBE ) . Generally , 1000 iterations were more than enough for our systems to converge . Similar to the alanine scanning calculations , the electrostatic free energy of binding was calculated as the change in electrostatic free energy between the complex and the two subunits in the unbound state . Wang and Kollman [76] showed that the electrostatic contribution to the free energy of binding of proteins in water with no salt could be calculated as follows:where , and are the corrected reaction field energies of the whole complex ( a:b ) , protein a in isolation , and protein b in isolation calculated by DelPhi with interior and exterior dielectric constants of 2 and 80 , respectively . The subtraction of the reaction field energy of the protein subunits from the complex results in the change in free energy of solvation from binding . is the Coulombic interaction energy , which is calculated as:where and are the Coulombic energy of the complex , subunit a , and subunit b respectively . The Coulombic and solvation energies were calculated in zero salt concentration . The ionic contributions were calculated by subtracting the total grid energy calculated in the 0 . 1M salt condition by the total grid energy in the zero-salt condition . In the 0 . 1M salt condition , an ion exclusion ( Stern ) layer of 2 . 0 Å surrounds the protein where the ion concentration is zero . In salt solution , the electrostatic contributions to the free energy of binding is equal to:where is the salt contribution to the electrostatic component of the binding free energy . The continuum electrostatic calculations were done on structures minimized in the bound state with CHARMM [77] . We used Param22 with topology and parameter files from the CHARMM-GUI [73] , [78] and the FACTS implicit solvation model [79] . First , missing atoms were added with the build command . Subsequently , the structures were minimized . All backbone atoms were fixed at their crystal coordinates before applying 30 steps of steepest descent minimization at 0 . 05 kcal/mol gradient tolerance . Afterwards , all heavy atoms with coordinates in the original PDB file were constrained with force of 50 kcal/mol/Å2 before the whole structure was minimized with 5000 conjugate steps at 0 . 002 kcal/mol gradient tolerance .
|
Protein-protein interactions are essential to communication and signal integration in cells . For these processes to be precise , interactions between proteins have to be specific and well coordinated . In order to understand the specificity in protein interactions , researches have focused on interfaces between two or more folded proteins . It has been shown that specificity in interactions between folded proteins relies on shape complementarity , hydrogen bonding , and salt-bridge formation . However , many proteins lack a unique folded structure; the so-called intrinsically disordered proteins . These proteins fluctuate between different conformations in isolation but often adopt a single structure when interacting with partner proteins . As many intrinsically disordered proteins are involved in signaling and regulation , their interactions have to be highly specific . The finding that the interaction interfaces of intrinsically disordered proteins are enriched in hydrophobic residues has led to questions regarding the specificity of interactions mediated by this group of proteins . Here , we show that polar and charged residues play a larger role in interfaces that involve intrinsically disordered proteins compared to interfaces that involve only folded proteins . Our results suggest that polar interactions are key contributors to the specificity of interactions that involve intrinsically disordered proteins .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"macromolecular",
"complex",
"analysis",
"biochemistry",
"protein",
"chemistry",
"protein",
"interactions",
"proteins",
"biology",
"computational",
"biology",
"biophysics",
"macromolecular",
"structure",
"analysis"
] |
2013
|
On the Importance of Polar Interactions for Complexes Containing Intrinsically Disordered Proteins
|
Type 2 diabetes ( T2D ) is more prevalent in African Americans than in Europeans . However , little is known about the genetic risk in African Americans despite the recent identification of more than 70 T2D loci primarily by genome-wide association studies ( GWAS ) in individuals of European ancestry . In order to investigate the genetic architecture of T2D in African Americans , the MEta-analysis of type 2 DIabetes in African Americans ( MEDIA ) Consortium examined 17 GWAS on T2D comprising 8 , 284 cases and 15 , 543 controls in African Americans in stage 1 analysis . Single nucleotide polymorphisms ( SNPs ) association analysis was conducted in each study under the additive model after adjustment for age , sex , study site , and principal components . Meta-analysis of approximately 2 . 6 million genotyped and imputed SNPs in all studies was conducted using an inverse variance-weighted fixed effect model . Replications were performed to follow up 21 loci in up to 6 , 061 cases and 5 , 483 controls in African Americans , and 8 , 130 cases and 38 , 987 controls of European ancestry . We identified three known loci ( TCF7L2 , HMGA2 and KCNQ1 ) and two novel loci ( HLA-B and INS-IGF2 ) at genome-wide significance ( 4 . 15×10−94<P<5×10−8 , odds ratio ( OR ) = 1 . 09 to 1 . 36 ) . Fine-mapping revealed that 88 of 158 previously identified T2D or glucose homeostasis loci demonstrated nominal to highly significant association ( 2 . 2×10−23 < locus-wide P<0 . 05 ) . These novel and previously identified loci yielded a sibling relative risk of 1 . 19 , explaining 17 . 5% of the phenotypic variance of T2D on the liability scale in African Americans . Overall , this study identified two novel susceptibility loci for T2D in African Americans . A substantial number of previously reported loci are transferable to African Americans after accounting for linkage disequilibrium , enabling fine mapping of causal variants in trans-ethnic meta-analysis studies .
The prevalence of type 2 diabetes ( T2D ) among adults in the USA is currently 11 . 3% , with substantially higher prevalence in African Americans ( 18 . 7% ) than in European Americans ( 10 . 2% ) [1] . To date , genome-wide association studies ( GWAS ) have identified >70 susceptibility loci for T2D [2]–[8] . While it is known that T2D is heritable in African Americans [9] , it is unclear how much heritability is explained by the known genetic associations discovered primarily from European ancestry populations and whether there are risk loci specific to African Americans . Given that individuals of African ancestry tend to harbor more genetic diversity than individuals of other ancestries [10] , we hypothesized that large-scale association analyses in African Americans could shed light on the genetic architecture of T2D and the risk attributable to cosmopolitan vs . population-specific variants .
We conducted a meta-analysis of 17 African American GWAS on T2D comprising 8 , 284 cases and 15 , 543 controls ( Tables S1 and S2 ) . Missing genotypes in individual studies were imputed to one of the HapMap reference panels ( Phase II release 21–24 CEU+YRI , Phase II release 22 all populations , Phase II+III release 27 CEU+YRI , Phase II+III release 27 CEU+YRI+ASW or Phase II+III release 27 all populations ) using MACH , IMPUTE2 or BEAGLE ( Table S3 ) . Genomic control corrections [11] were applied to each study ( λ = 1 . 01–1 . 08 ) and after meta-analysis ( λ = 1 . 06 ) due to modest inflated association results ( Table S3 ) [12] . Association results for ∼2 . 6M SNPs were subsequently examined . From stage 1 meta-analysis , 49 SNPs moderately associated with T2D ( P<1×10−5 ) and two candidate SNPs near the p value threshold ( rs231356 at KCNQ1 , P = 2 . 84×10−5 and rs2244020 at HLA-B , P = 1 . 02×10−5 ) totaling 51 SNPs in 21 loci were followed up for replication . rs231356 is 14 kb downstream of the reported T2D index SNP , rs231362 , in Europeans [3] . Moderate associations have also been observed across the HLA region in Europeans [3] . The stage 2 replication included in silico and de novo replication in up to 11 , 544 African American T2D cases and controls , as well as in silico replication in 47 , 117 individuals of European ancestry from DIAGRAMv2 [3] ( Table S4 ) . Meta-analyses were performed to combine results from African Americans ( stage 1+2a , n≤35 , 371 , Table S4 ) and both African Americans and Europeans ( stage 1+2a+2b , n≤82 , 488 , Table S4 ) . Five independent loci reached genome-wide significance ( P<5×10−8 ) . Stage 1 meta-analysis identified the established TCF7L2 locus . Stage 1+2a meta-analysis identified the established KCNQ1 and HMGA2 loci . Stage 1+2a+2b meta-analysis identified a second signal at KCNQ1 and a novel HLA-B locus . Secondary analysis including body mass index ( BMI ) adjustment in stage 1+2a meta-analysis identified the second novel locus at INS-IGF2 ( Table 1 and Figure 1 ) . None of the most strongly associated SNPs at these loci demonstrated significant heterogeneity of effect sizes among studies within each stage , between African Americans in stages 1 and 2a , or between African Americans in stage 1+2a and Europeans in stage 2b after Bonferroni correction of multiple comparisons ( Phet>0 . 001 ) ( Figure S1 ) . At the TCF7L2 locus , the most strongly associated SNP in stage 1+2a African Americans samples was rs7903146 ( OR = 1 . 33 , P = 4 . 78×10−44 , Table 1 and Figure 2 ) . rs7903146 is also the index SNP ( most significantly associated with T2D in prior studies ) in Europeans ( OR = 1 . 40 , P = 2 . 21×10−51 ) [3] , South Asians ( OR = 1 . 25 , P = 3 . 4×10−19 ) [4] and East Asians ( OR = 1 . 48 , P = 2 . 44×10−15 ) [13] . Two association signals were observed at KCNQ1 ( Table 1 and Figure 2 ) . The first association signal was represented by rs2283228 located at the 3′ end of KCNQ1 ( stage 1+2a OR = 1 . 20 , P = 9 . 90×10−11; stage 1+2a+2b OR = 1 . 19 , P = 4 . 87×10−13 ) . Using data from individuals of African ancestry in Southwest USA ( ASW ) from the 1000 Genomes Project ( 1KGP ) [14] , rs2283228 mapped to the same linkage disequilibrium ( LD ) -based interval as index SNPs from other populations ( rs2283228 [15] and rs2237892 [16]–[17] in Japanese , rs2237892 in Hispanics [18] , rs163182 [19] and rs2237895 [20] in Han Chinese ) . The second association signal was represented by rs231356 ( r2 = 0 with rs2283228 in both ASW and CEU ) ( stage 1+2a OR = 1 . 11 , P = 1 . 94×10−5; stage 1+2a+2b OR = 1 . 09 , P = 3 . 93×10−8 ) , located 144 kb upstream of the first signal . rs231356 is located at the same LD interval as the index SNPs rs231362 in Europeans [3] and rs231359 in Chinese [20] . At the HMGA2 locus , the most strongly associated SNP was rs343092 ( stage 1+2a OR = 1 . 16 , P = 8 . 79×10−9; stage 1+2a+2b OR = 1 . 14 , P = 2 . 75×10−12; Table 1 and Figure 2 ) . rs343092 is located 76 kb downstream and at the same LD interval as of the index SNP rs1531343 reported in Europeans [3] . Two novel T2D loci were identified . The effect sizes of rs2244020 located near HLA-B were similar in African Americans and Europeans ( OR = 1 . 11 vs . 1 . 07 , Phet = 0 . 26; stage 1+2a+2b P = 6 . 57×10−9 ) ( Table 1 and Figure 2 ) . HLA-B encodes the class I major histocompatibility complex involved in antigen presentation in immune responses . The most strongly associated SNP near INS-IGF2 was rs3842770 in African Americans ( OR = 1 . 14 , P = 2 . 78×10−8 , stage 1+2a BMI adjusted , Table 1 and Figure 2 ) but the risk A allele was absent in the CEU population . Insulin plays a key role in glucose homeostasis . Mutations at INS lead to neonatal diabetes , type 1 diabetes , and hyperinsulinemia [21] . Insulin-like growth factor 2 ( IGF2 ) is involved in growth and development . IGF2 overexpression in transgenic mice leads to islet hyperplasia [22] and IGF2 deficiency in the Goto–Kakizaki rat leads to beta cell mass anomaly [23] . We investigated index SNPs from 158 independent loci associated with T2D and/or glucose homeostasis from prior genome-wide and candidate gene studies in individuals of European , East Asian , South Asian , or African American ancestry ( Table S5 ) . Among the 104 T2D-associated index SNPs , 19 were associated with T2D in stage 1 African American samples ( P<0 . 05 ) . Most of the 17 T2D-associated SNPs that showed consistent direction of effects had similar effect sizes between this study and prior reports , despite that rs10440833 at CDKAL1 had substantially stronger effect size in Europeans ( OR = 1 . 25 ) than in African Americans ( OR = 1 . 06 , Phet = 5 . 86×10−6 ) . Additionally , 3 out of 54 trait-increasing alleles from glucose homeostasis-associated index SNPs were associated with increased T2D risk in African Americans ( P<0 . 05 ) . We also performed a locus-wide analysis to test for associations of all SNPs within the LD region at r2≥0 . 3 with the previously reported index SNPs and results were corrected for the effective number of SNPs [24] . Since the causal variant ( s ) at each locus may be different or reside on different haplotypes across populations with different LD structures , this approach allows the identification of the most strongly associated SNPs in African Americans that may or may not be in LD with the index SNPs reported in other populations . A total of 55 T2D- and 29 glucose-associated loci were associated with T2D in African Americans ( Plocus<0 . 05 , corrected for LD in ASW for SNPs within a locus; Table S6 ) . We compared the genetic architecture between the previously reported index SNPs and our fine-mapped SNPs for these 84 loci . The respective average risk allele frequencies were 0 . 51 and 0 . 46 , and the distributions or pairwise differences of risk allele frequencies were not significantly different ( P = 0 . 255 , Wilcoxon rank sum test; and P = 0 . 295 , Wilcoxon signed-rank test , respectively , Figure S2 ) . In contrast , the average odds ratios for the risk alleles were higher for the fine-mapped SNPs as compared to the index SNPs ( 1 . 14 vs . 1 . 05 ) . The distributions and pairwise differences of risk allele odds ratios were significantly different ( P = 1 . 18×10−19 and 5 . 55×10−14 , respectively , Figure S2 ) . Thus , the locus-wide analysis identified variants with larger effect sizes and similar allele frequencies . We leveraged differences in LD between African Americans and Europeans to fine-map and re-annotate several established loci . The association signal spanning ∼100 kb at INTS8 in African Americans overlapped the ∼200 kb TP53INP1 T2D locus in Europeans [3] . The most strongly associated SNP in MEDIA tended to have larger effect size in African Americans than in Europeans ( rs17359493 , OR = 1 . 13 vs . 1 . 06 , P = 1 . 39×10−7 vs . 3 . 20×10−2 , respectively , Phet = 0 . 06 ) ( Table S4 ) . However , rs17359493 at intron 10 of INTS8 was only in weak LD with the reported index SNP rs896854 in Europeans ( r2 = 0 . 21 in CEU , 0 . 10 in ASW ) . Neither the reported index SNP rs896854 nor its proxies from the CEU data demonstrated significant association to T2D in African Americans ( Table S6 and Figure S3a , b ) , suggesting that rs17359493 may be an independent novel signal . INTS8 encodes a subunit of the integrator complex which is involved in the cleavage of small nuclear RNAs . At KCNQ1 , the most strongly associated SNP rs231356 was in weak LD with the index SNP rs231362 reported in Europeans [3] ( r2 = 0 . 24 in CEU and 0 . 17 in ASW ) . Given rs231362 was modestly associated with T2D in African American ( P = 0 . 04 ) and was in weak LD ( r2 = 0 . 21 to 0 . 46 in CEU ) with other associated SNPs in this region ( Table S6 and Figure S3c , d ) , the results suggest a refinement of the localization of causal variant ( s ) to variants in strong LD with rs231356 . At HMGA2 , the most strongly associated SNP rs343092 was in moderate LD with the index SNP rs1531343 ( r2 = 0 . 60 in CEU and 0 . 32 in ASW ) . Despite rs1531343 and its proxies in high LD were not associated with T2D in African Americans ( P>0 . 05 ) , several SNPs in moderate LD , including rs343092 , showed nominal to strong associations ( Table S6 and Figure S3e , f ) . Trans-ethic fine mapping will be particularly useful to dissect the causal variant ( s ) at this locus . We investigated the influence of obesity by comparing the stage 1 meta-analysis results with or without adjustment for BMI at the 51 most significantly associated SNPs from the GWAS for follow up ( Tables S4 and S7 ) and 158 established T2D or glucose homeostasis index SNPs ( Table S5 ) . Association results were highly similar with and without BMI adjustment ( correlation coefficients were 0 . 99 for both effect sizes and −logP values ) . Of particular note , FTO is suggested to influence T2D primarily through modulation of adiposity in Europeans [3] , [25] , but evidence is contradictory across multiple ethnic groups [26]–[28] . The index SNP rs11642841 was not significantly associated with T2D in African Americans without and with BMI adjustment ( P = 0 . 06 and 0 . 23 , respectively ) ( Table S5 ) . The frequency of the risk A allele was 0 . 13 in this study . It had 100% power to detect association at the reported OR of 1 . 13 at type 1 error rate of 0 . 05 , suggesting that FTO is unlikely a key T2D susceptibility gene in African Americans . Among the six genome-wide significant loci ( Table 1 ) , we found no coding variants in the most significantly associated SNPs or their proxies . These SNPs demonstrated only weak associations with expression quantitative trait loci ( eQTLs ) ( P>0 . 001 , Table S8 ) . Examination of the ENCODE data [29] revealed that several SNPs at TCF7L2 , KCNQ1 , and HMGA2 were located at protein binding sites or were predicted to alter motif affinity for transcription factors implicated in energy homeostasis ( Table S9 ) . The most strongly associated SNP rs7903146 in TCF7L2 is predicted to alter the binding affinity for a POU3F2 regulatory motif [30] . POU3F2 is a neural transcription factor that enhances the activation of genes regulated by corticotropin-releasing hormone which stimulates adrenocorticotropic hormone ( ACTH ) . ACTH is synthesized from pre-pro-opiomelanocortin ( pre-POMC ) which regulates energy homeostasis . For the 3′ signal at KCNQ1 , several tag SNPs are predicted to alter the binding affinity for regulatory motifs , including SREBP , CTCF and HNF4A . SREBP is a transcription factor involved in sterol biosynthesis . CTCF regulates the expression of IGF2 [31] . HNF4A is a master regulator of hepatocyte and islet transcription . The tag SNP rs2257883 at HMGA2 is predicted to alter the binding affinity of MEF2 , which regulates GLUT4 transcription in insulin responsive tissues [32] .
We have performed the largest genetic association analysis to date for T2D in African Americans . Our data support the hypothesis that risk for T2D is partly attributable to a large number of common variants with small effects [7] . We identified HLA-B and INS-IGF2 as novel T2D loci , the latter specific to African Americans . We found evidence supporting association for 88 previously identified T2D and glucose homeostasis loci . Taken together , these 90 loci yielded a sibling relative risk of 1 . 19 . The phenotypic variance measured on the liability scale is substantially larger in African Americans than in European Americans ( 17 . 5% vs . 5 . 7% ) [7] due to larger effect sizes upon fine-mapping as well as higher disease prevalence in African Americans . The two novel T2D loci , HLA-B and INS-IGF2 , have been implicated in type 1 diabetes ( T1D ) risk in Europeans [33]–[35] . One limitation of our study is the lack of autoantibody measurement . However , our results are unlikely to be confounded by the presence of misclassified patients . Among diabetic youth aged <20 years , T2D characterized by insulin resistance without autoimmunity is more prevalent in African Americans ( 40 . 1% ) than in European Americans ( 6 . 2% ) , while African Americans less often present with autoimmunity and insulin deficiency resembling T1D compared to European Americans ( 32 . 5% vs . 62 . 9% , respectively ) [36] . Autoimmunity is also uncommon in African American diabetic adults [37] . Furthermore , associations for T1D are stronger at HLA class II ( HLA-DRB1 , -DQA1 , and -DQB1 ) than HLA class I regions in Europeans [33]–[34] , [38]–[41] ( http://www . t1dbase . org ) . In African Americans , T1D individuals showed both shared and unique risk and protective HLA class II haplotypes as compared to European T1D individuals [42]–[43] . More importantly , these individuals also showed substantially stronger associations at HLA class II ( P<1×10−25 ) than class I regions ( P<1×10−5 ) [42] , which is in contradiction with our finding of stronger associations at HLA class I than class II regions in T2D individuals ( HLA-B , Figure S4 ) . The observed HLA-B association may be due to LD with nearby causal gene ( s ) since there is long range LD in this region . Recently , rs3130501 near POU5F1 and TCF19 was reported for association with T2D in a trans-ancestry meta-analysis [8] . rs3130501 was located 211 kb upstream of rs2244020 and mapped to the same LD interval . However , the two SNPs were not correlated in both CEU ( D′ = 0 . 57 , r2 = 0 . 05 ) and ASW ( D′ = 0 . 68 , r2 = 0 . 16 ) from 1KGP nor strongly associated with T2D in the stage 1 meta-analysis ( P = 0 . 04 ) . Other potential non-HLA candidate genes may include TNFA which regulates immune and inflammatory response . It has been hypothesized that activated innate and adaptive immune cells stimulate release of cytokines such as TNFα and IL-1β , which promote both systemic insulin resistance and β-cell damage [44] . On the other hand , evidence has implicated T1D loci HLA-DQ/DR , GLIS3 and INS in the susceptibility of latent autoimmune diabetes in adults ( LADA ) and/or T2D [7] , [34] , [45]–[46] , while T2D loci such as PPARG and TCF7L2 was associated with T1D [47] and LADA [46] , [48] , respectively . More comprehensive studies are needed to understand the shared and distinct genetic risks in different forms of diabetes which will facilitate diagnosis and personalized treatment . Our results have several implications regarding the genetic architecture of T2D . First , fine-mapping suggests that currently known loci explain more of the risk than previously estimated . Second , the loci conferring the largest risk for T2D appear to act through regulatory rather than protein-coding changes . Third , many , but not all , of the previously identified T2D loci are shared across ancestries . The differential LD structure of African-ancestry populations at shared loci provides an opportunity for fine mapping in trans-ethnic meta-analysis . Fourth , the ∼2 . 6M MEDIA SNPs achieved only 43 . 3% coverage of the 1KGP ASW common SNPs , suggesting that risk loci that are specific to African-ancestry individuals are difficult to discover with the genotyping arrays being used . Large-scale sequencing studies , such as those focusing on whole genomes , exomes , and targeted resequencing for associated non-coding regions , will be necessary to further delineate the causal variants for T2D risk in African Americans .
Stage 1 discovery samples included 17 T2D GWAS studies ( ARIC , CARDIA , CFS , CHS , FamHS , GeneSTAR , GENOA , HANDLS , Health ABC , HUFS , JHS , MESA , MESA Family , SIGNET-REGARDS , WFSM , FIND , and WHI ) with up to 23 , 827 African American subjects ( 8 , 284 cases and 15 , 543 controls ) . Stage 2 replication samples included up to 11 , 544 African American subjects ( 6 , 061 cases and 5 , 483 controls ) , using in silico replication of GWAS data from eMERGE and IPM Biobank and de novo genotyping in IRAS , IRASFS , SCCS , and WFSM . In general , T2D cases were defined as having at least one of the following: fasting plasma glucose ≥126 mg/dl , 2 hour glucose during oral glucose tolerance test ( OGTT ) ≥200 mg/dl , random glucose ≥200 mg/dl , oral hypoglycemic agent or insulin treatment , or physician-diagnosed diabetes . All cases were diagnosed at ≥25 years ( or age at study ≥25 years if age at diagnosis was not available ) . For cohort studies , individuals who met the criteria at any of the visits were defined as cases . Controls with normal glucose tolerance ( NGT ) were defined by satisfying all the following criteria: fasting plasma glucose <100 mg/dl , 2 hour OGTT<140 mg/dl ( if available ) , no treatment of diabetes , and age ≥25 years . For cohort studies , individuals who met the criteria at all visits were defined as controls . All study participants provided written informed consent , except for eMERGE that use an opt out program , and approval was obtained from the institutional review board ( IRB ) from the respective local institutions . Detailed descriptions of the participating studies are provided in Text S1 . For stage 1 and 2 GWAS studies , genotyping was performed with Affymetrix or Illumina genome-wide SNP arrays . Imputation of missing genotypes was performed using MACH [49] , IMPUTE2 [50] or BEAGLE [51] using HapMap reference haplotypes . For each study , samples reflecting duplicates , low call rate , gender mismatch , or population outliers were excluded . In general , SNPs were excluded by the following criteria: call rate <0 . 95 , minor allele frequency ( MAF ) <0 . 01 , minor allele count <10 , Hardy-Weinberg P-value <1×10−4 , or imputation quality score <0 . 5 ( Table S3 ) . For de novo replication studies , genotyping was performed using the Sequenom MassArray platform ( Sequenom; San Diego , CA ) . Sample and SNP quality controls were performed as with GWAS data . Single SNP association was performed for each study by regressing T2D case/control status on genotypes . To account for uncertainty of genotype calls during imputation , genotype probabilities or dosage were used for association tests in imputed SNPs . The association tests assumed an additive genetic model and adjusted for age , sex , study centers , and principal components . Principal components were included to control for confounding effects of admixture proportion and population structure . Secondary analysis with additional adjustment for BMI was performed for SNPs with P<1×10−5 in stage 1 meta-analysis and index SNPs previously reported to be associated with T2D or glucose homeostasis traits . BMI adjustment allows increasing power to detect T2D loci independent of BMI effect and diminish associations at T2D loci with effects modulated through BMI . Logistic regression was used for samples of unrelated individuals . Generalized estimating equations [52] or SOLAR [53] were used for samples of related individuals . Association results with extreme values ( absolute beta coefficient or standard error >10 ) , primarily due to low cell counts resulting from small sample sizes and/or low minor allele frequencies , were excluded ( Table S3 ) . In stage 1 , association results were combined by a fixed effect model with inverse variance weighted method using the METAL software [12] . Genomic control correction [11] was applied to each study before meta-analysis , and to the overall results after meta-analysis . Results from SNPs genotyped in <10 , 000 samples and those with allele frequency difference >0 . 3 among studies were excluded . A total of 2 , 579 , 389 SNPs were analyzed in the meta-analysis ( Table S3 ) . In stage 2a , association results from African American replication studies were also combined using a fixed effect inverse variance weighted method . To assess the overall effects in African Americans ( stage 1+2a ) and both African Americans and Europeans ( stage 1+2a+2b ) , association results from studies in the respective stages were combined using a fixed effect inverse variance weighted method . Genome-wide significance is declared at P<5×10−8 from the meta-analysis result of all stages , which has better power than the replication-based strategy [54] . Among the 51 SNPs carried forward for replication , heterogeneity of effect sizes across studies within each stage was assessed using Cochran's Q statistic implemented in METAL . Meta-analysis results from stages 1 and 2a , stage 1+2a and 2b were used to assess heterogeneity of effect sizes between discovery and replication stages in African Americans , and between African Americans and Europeans , respectively . For SNPs with significant heterogeneous effect size after multiple comparison corrections ( Phet<0 . 001 ) , meta-analysis results including studies of all stages assessed by the random effect model implemented in GWAMA [55] were reported . Heterogeneous associations may partly due to differences in ascertainment scheme across studies . For index SNPs reported in prior studies , assessment of heterogeneity using Cochran's Q statistic between prior studies and this study were also reported . Index SNPs associated with T2D or glucose homeostasis traits from prior GWAS and candidate gene studies were examined for association with T2D in African Americans ( Table S5 ) . For the index SNP association tests , a per-SNP P value <0 . 05 was defined as significant . In the locus-wide analysis , the boundaries of a locus were defined by the most distant markers ( within ±500 kb ) using the 1KGP CEU data with r2≥0 . 3 with the index SNP . All MEDIA SNPs within these bounds were examined for association analysis . All pairwise LD values within each locus were estimated using the 1KGP CEU and ASW data . To estimate the effective number of SNPs at a locus , we retrieved genotypes from the 1KGP ASW data for markers present in MEDIA , estimated the sample covariance matrix from those genotypes , and spectrally decomposed the covariance matrix [24] . The effective number of SNPs was estimated using the relationship , in which λk is the kth eigenvalue of the K×K covariance matrix for the K SNPs in the locus [24] . The per-locus significance level was defined as 0 . 05/effective number of SNPs ( Table S6 ) . By accounting for all SNPs within the bounds of LD , the per-locus significance level is corrected to account for markers in LD with the index SNP as well as markers not in LD with the index SNP , thereby potentially allowing for discovery of new associations at markers not tagged by the index SNP . For each independent locus , we estimated the sibling relative risk using the most strongly associated SNP within that locus . Let pi and ψi be the risk allele frequency and the corresponding odds ratio at the ith SNP , respectively . Assuming the additive genetic model and independence between SNPs , the contribution to the sibling relative risk λs for a set of N SNPs is given by [56] . Let K be the disease prevalence . The liability-scale variance explained by the set of N SNPs is given by , in which , , and , with representing the standard normal quantile function and z representing the standard normal density at T [57] . The coverage of MEDIA SNPs to the human genome was estimated using HaploView [58] via pairwise tagging at the r2 = 0 . 8 threshold . We used all SNPs with minor allele frequencies ≥1% in both MEDIA and the 1KGP ASW sequence data . Coverage was estimated using non-overlapping bins of 1 , 000 SNPs . Study power was calculated using the genetic power calculator [59] . For SNPs with MAF≥0 . 3 , our study had >80% power to detect odds ratios for T2D at OR≥1 . 06 and ≥1 . 13 at P<0 . 05 and P<5×10−8 , respectively , in stage 1 samples under an additive model . The observed odds ratios among our stage 1 most significantly associated SNPs with P<1×10−5 ranged from 1 . 11 to 1 . 56 ( Table S4 ) . Given our African American sample size in stage 1+2a , our study had >80% power to detect OR≥1 . 1 at P<5×10−8 at MAF≥0 . 3 , thus provided good power to detect genome-wide significance among the most significantly associated SNPs using all African American samples . For T2D SNPs reported from the literature , power was also calculated from the reported effect size using the risk allele frequency from this study for stage 1 samples at P<0 . 05 and P<5×10−8 , respectively ( Table S5 ) . The MuTHER resource ( www . muther . ac . uk ) includes lymphoblastoid cell lines ( LCLs ) , skin , and adipose tissue derived simultaneously from a subset of well-phenotyped healthy female twins from the TwinsUK adult registry [60] . Whole-genome expression profiling of the samples , each with either two or three technical replicates , was performed using the Illumina Human HT-12 V3 BeadChips ( Illumina Inc . ) according to the protocol supplied by the manufacturer . Log2-transformed expression signals were normalized separately per tissue as follows: quantile normalization was performed across technical replicates of each individual followed by quantile normalization across all individuals . Genotyping was performed with a combination of Illumina arrays ( HumanHap300 , HumanHap610Q , 1M-Duo , and 1 . 2MDuo 1M ) . Untyped HapMap2 SNPs were imputed using the IMPUTE2 software package . In total , 776 adipose and 777 LCL samples had both expression profiles and imputed genotypes . Association between all SNPs ( MAF>5% , IMPUTE info >0 . 8 ) within a gene or within 1 Mb of the gene transcription start or end site and normalized expression values were performed with the GenABEL/ProbABEL packages [61]–[62] using the polygenic linear model incorporating a kinship matrix in GenABEL followed by the ProbABEL mmscore score test with imputed genotypes . Age and experimental batch were included as cofactors . Genotype and gene expression in LCL in HapMap samples were also available [63] . Association of genotypes and gene expression of transcripts within 1 MB of tested SNPs were analyzed separately for CEU and YRI populations . The variance components model implemented in SOLAR was used for association analysis which accounts for correlation among related individuals [53] . In this study , we examined the association of the most significantly associated SNPs from the six genome-wide significant loci and their proxies ( r2≥0 . 8 in ASW ) within 1 Mb of the associated SNPs with cis-expression quantitative trait loci ( eQTLs ) in peripheral blood leukocytes ( LCL ) and adipose tissue ( Table S8 ) . We examined putative function of non-coding genome-wide significant SNPs and their proxies within 1 Mb ( r2≥0 . 8 in 1KGP ASW ) using HaploReg [30] and RegulomeDB [64] . These databases interrogated multiple chromatin features from the Encyclopedia of DNA Elements ( ENCODE ) project [29] . High priority was given to variants annotated as protein-binding via ChIP-seq , and motif-changing via position weight matrices , with the respective transcription factors implicated in diabetes pathogenesis and related biological processes .
|
Despite the higher prevalence of type 2 diabetes ( T2D ) in African Americans than in Europeans , recent genome-wide association studies ( GWAS ) were examined primarily in individuals of European ancestry . In this study , we performed meta-analysis of 17 GWAS in 8 , 284 cases and 15 , 543 controls to explore the genetic architecture of T2D in African Americans . Following replication in additional 6 , 061 cases and 5 , 483 controls in African Americans , and 8 , 130 cases and 38 , 987 controls of European ancestry , we identified two novel and three previous reported T2D loci reaching genome-wide significance . We also examined 158 loci previously reported to be associated with T2D or regulating glucose homeostasis . While 56% of these loci were shared between African Americans and the other populations , the strongest associations in African Americans are often found in nearby single nucleotide polymorphisms ( SNPs ) instead of the original SNPs reported in other populations due to differential genetic architecture across populations . Our results highlight the importance of performing genetic studies in non-European populations to fine map the causal genetic variants .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"diabetes",
"mellitus",
"medicine",
"and",
"health",
"sciences",
"diabetic",
"endocrinology",
"epidemiology",
"endocrinology",
"genetics",
"biology",
"and",
"life",
"sciences",
"genetics",
"of",
"disease",
"type",
"2",
"diabetes",
"metabolic",
"disorders",
"genetic",
"epidemiology"
] |
2014
|
Meta-Analysis of Genome-Wide Association Studies in African Americans Provides Insights into the Genetic Architecture of Type 2 Diabetes
|
Fibrinogen is an essential part of the blood coagulation cascade and a major component of the extracellular matrix in mammals . The interface between fibrinogen and bacterial pathogens is an important determinant of the outcome of infection . Here , we demonstrate that a canine host-restricted skin pathogen , Staphylococcus pseudintermedius , produces a cell wall-associated protein ( SpsL ) that has evolved the capacity for high strength binding to canine fibrinogen , with reduced binding to fibrinogen of other mammalian species including humans . Binding occurs via the surface-expressed N2N3 subdomains , of the SpsL A-domain , to multiple sites in the fibrinogen α-chain C-domain by a mechanism analogous to the classical dock , lock , and latch binding model . Host-specific binding is dependent on a tandem repeat region of the fibrinogen α-chain , a region highly divergent between mammals . Of note , we discovered that the tandem repeat region is also polymorphic in different canine breeds suggesting a potential influence on canine host susceptibility to S . pseudintermedius infection . Importantly , the strong host-specific fibrinogen-binding interaction of SpsL to canine fibrinogen is essential for bacterial aggregation and biofilm formation , and promotes resistance to neutrophil phagocytosis , suggesting a key role for the interaction during pathogenesis . Taken together , we have dissected a bacterial surface protein-ligand interaction resulting from the co-evolution of host and pathogen that promotes host-specific innate immune evasion and may contribute to its host-restricted ecology .
Many bacteria evolve strict mutualistic relationships with their host species with limited capacity to colonize and cause disease in other hosts . In contrast , other bacteria have the ability to expand into new host-species leading to the emergence of new pathogenic clones . Our understanding of the bacterial and host factors that underpin pathogen-host ecology is very limited . However , bacterial surface proteins are central mediators of host colonization , and tissue tropism and , as such , are likely to play a critical role in determining host ecology [1 , 2] . For example , the choline-binding protein A , of the major human pathogen Streptococcus pneumoniae , binds to polymeric immunoglobulin receptor , secretory component , secretory IgA , and factor H of complement from humans but not from other animal species tested [1] . In addition , the human host-restricted Streptococccus pyogenes expresses surface-anchored M protein that binds exclusively to human CD46 mediating binding and invasion of epithelial cells . Adaptive diversification of bacterial surface proteins can also have a major impact on tissue tropism and disease manifestation . For example , uropathogenic Escherichia coli virulence has arisen due to mutations in the fimbrial adhesin FimH , promoting high affinity binding to the urinary epithelium [3] . Similarly , a single non-synonymous mutation in a fibronectin-binding autolysin of Staphylococcus saprophyticus , associated with a selective sweep , has been linked to the pathogenesis of urinary tract infection in humans [4] . Additionally , single amino acid substitutions in the fibronectin-binding protein A ( FnBPA ) of Staphylococcus aureus , are associated with cardiac device infections and bacteremia in humans due to increased binding affinity for fibronectin [5–7] . Fibrinogen is a highly abundant protein in blood and is required for blood coagulation , thrombosis and host immune defense [8] . This large glycoprotein is composed of three chains , termed the α- , β- , and γ-chains , that form a dimer of trimers [8] . During coagulation , thrombin cleaves the fibrinogen α- and β-chains allowing fibrin formation , with the γ-chain binding directly to platelets to produce the blood clot [8] . Bacterial pathogens have evolved many mechanisms to bind to host fibrinogen to disrupt blood coagulation as well as promote host cell adherence , immune evasion and abscess formation [9 , 10] . The importance of this interaction is highlighted by the large number of fibrinogen-binding proteins of bacteria that have been identified , with S . aureus encoding at least 9 fibrinogen-binding proteins [9–12] . It is unclear if each of these proteins confer an exclusive function via distinct fibrinogen-binding sites , or if convergent evolution is driving a high redundancy for fibrinogen-binding . In S . aureus there are fibrinogen-binding proteins that exhibit host-specificity and those that exhibit a broader host tropism . In the case of clumping factor B ( ClfB ) a host-restrictive fibrinogen-binding phenotype is observed due to the interaction with a sequence unique to the human fibrinogen α-chain [13] . Conversely , clumping factor A ( ClfA ) interacts with fibrinogen from multiple hosts , such as human , canine , and murine , due to an interaction with the fibrinogen γ-chain [14] . A single residue substitution of Q407A in the ovine fibrinogen γ-chain is sufficient to eliminate binding of ClfA to ovine fibrinogen [14] . As FnBPA adheres to the same region in the fibrinogen γ-chain , it is assumed that it exhibits the same host phenotype but this has not been directly investigated [15] . Staphylococcus pseudintermedius naturally colonizes the nares and perineum of healthy dogs and is a major cause of canine skin infections , particularly in canine breeds that are genetically pre-disposed to atopic dermatitis including boxers , German shepherds , golden retrievers , Dalmatians , Labrador retrievers , French bulldogs , West Highland white terriers , Jack Russell terriers , and shar-peis [16–19] . Although S . pseudintermedius occasionally causes zoonotic infections of humans via dog bite wounds , it is highly host-restricted and there is limited evidence for colonization or transmission among non-canine host species such as humans and cats [17] . Of note , S . pseudintermedius demonstrates a host-specific preference for corneocytes collected from healthy dogs in comparison to human healthy volunteers [20] , suggesting the existence of bacterial surface factors that promote a canine host-tropism . However , the bacterial factors underpinning the canine host-restricted ecology of S . pseudintermedius are unknown . Previously , we identified a complement of 18 genes encoding cell wall-associated proteins of S . pseudintermedius strain ED99 and discovered 2 fibrinogen-binding proteins , SpsD and SpsL , which exhibited host-dependent variation in fibrinogen-binding after heterologous expression by Lactococcus lactis [21] . While SpsD was encoded by closely-related species of staphylococci associated with non-canine host-species , SpsL was specific for S . pseudintermedius [21] and was shown to be required for abscess formation in a murine skin infection model [22] . Previous sequence and structural analysis of SpsL identified a similar domain architecture to the fibronectin-binding proteins of S . aureus with an amino acid identity of ~27% [21] . In S . pseudintermedius strain ED99 , SpsL is a protein of 930 amino acids that contains the typical signatures of a cell wall-associated protein with an N-terminal signal peptide and C-terminal LPKTG anchor motif [21] . The N-terminal of SpsL consists of an A-domain with N1 , N2 , and N3-subdomains that would be predicted to mediate fibrinogen-binding [21] . The C-terminal contains 7 tandem repeats that share 91–100% pairwise identity and 24% protein identity to the fibronectin-binding repeats of FnBPA [21] . These C-terminal repeats of SpsL have been demonstrated to confer fibronectin-binding that can mediate internalization of S . pseudintermedius into canine epithelial cells [23] . Here we dissect the interaction of SpsL with fibrinogen and demonstrate its functional consequences . Using atomic force microscopy ( AFM ) we quantify the enhanced binding force of SpsL for canine fibrinogen , which is dependent on a tandem repeat region in the fibrinogen α-chain that is genetically diverse in mammalian species . Further , we demonstrate that the strong host-specific interaction with canine fibrinogen is required for SpsL-mediated aggregation and biofilm formation , and promotes neutrophil opsonophagocytosis suggesting a key role for SpsL-fibrinogen binding in the pathogenesis and canine host ecology of S . pseudintermedius .
Previously , heterologous expression of SpsL in L . lactis revealed a host-dependent binding to fibrinogen [21] . In order to investigate this preliminary observation , we examined the capacity of wild-type S . pseudintermedius strain ED99 to bind to immobilized fibrinogen from different host species . At mid-exponential growth phase , the highest binding was observed to canine and human fibrinogen , with very limited binding to bovine and ovine fibrinogen ( Fig 1A ) . A S . pseudintermedius ED99 mutant deficient in expression of a fibrinogen-binding protein SpsD ( ED99ΔspsD ) , cultured to early-exponential growth phase , demonstrated binding to fibrinogen that was equivocal to wild-type ED99 for canine , ovine and human fibrinogen but reduced for bovine fibrinogen ( p<0 . 001 ) ( Fig 1B ) . In contrast , a mutant deficient in SpsL ( ED99ΔspsL ) , cultured to mid-exponential growth phase , exhibited highly reduced binding to fibrinogen from all host species ( p<0 . 001 ) ( Fig 1C ) , with complete loss of fibrinogen-binding by a mutant deficient in both SpsL and SpsD ( ED99ΔspsLΔspsD ) , at both early-exponential ( Fig 1B ) and mid-exponential growth phases ( Fig 1C ) . Re-introduction of the deleted spsL gene restored fibrinogen-binding as did complementation of ED99ΔspsLΔspsD with a plasmid ( pALC2073::spsL ) encoding SpsL ( Fig 1C and 1D ) . In summary , these data indicate that S . pseudintermedius ED99 has host-specific interactions with fibrinogen that are primarily mediated by SpsL . However , these adherence assays do not allow quantification of the strength of the binding interaction between SpsL and fibrinogen . In order to compare the molecular forces driving the binding of SpsL to canine and human fibrinogen , we used atomic force microscopy ( AFM ) [24 , 25] . Firstly , for single-cell force spectroscopy ( SCFS , Fig 2A ) , single bacteria were attached onto AFM cantilevers , and force-distance curves were collected between the cell probes and fibrinogen-coated surfaces ( Fig 2A ) . The adhesion forces obtained for three representative cells interacting with either human or canine fibrinogen are presented ( Fig 2A ) . While there was substantial variation between cells , the binding probability was always higher for canine rather than for human fibrinogen ( 85% vs 56%; means ± 12 and 28 , from a total of n = 1 , 139 and 1 , 176 curves ) . Also , binding forces were stronger for canine fibrinogen ( 355 ± 354 pN from n = 228 adhesive curves; 2 , 077 ± 1 , 157 pN ( n = 388 ) , and 1 , 024 ± 427 pN ( n = 352 ) , for cell #1 , cell #2 and cell #3 , respectively ) , than for human fibrinogen ( 149 ± 84 pN ( n = 85 ) , 744±467 pN ( n = 362 ) , and 541±266 pN ( n = 216 ) ) . Next , we used single-molecule force spectroscopy ( SMFS ) with fibrinogen-coated AFM tips to quantify the strength of single bonds ( Fig 2B and 2C ) . Canine fibrinogen ( Fig 2B ) always showed very large forces ( 1 , 237 ± 754 pN from n = 258 adhesive curves; 1 , 554 ± 828 pN ( n = 308 ) , and 2 , 630 ± 1 , 393 pN ( n = 137 ) , for cell #1 , cell #2 and cell #3 , respectively ) . Of note , these high forces are in the range of values reported previously for the high-affinity “dock , lock and latch” binding of SdrG to fibrinogen [26] . For human fibrinogen , these strong forces were also observed but much less frequently ( Fig 2C ) . Taken together these data are consistent with a high affinity dock , lock and latch-based mechanism for the binding of SpsL to canine fibrinogen . To investigate the region of SpsL required for fibrinogen-binding , an array of recombinant truncates of SpsL were generated and purified from E . coli as described in the Supplemental Methods section ( S1A Fig ) . From structural and functional studies of related staphylococcal surface proteins , we predicted that the N2N3 subdomains of SpsL would be sufficient for fibrinogen-binding [27–29] . However , in ELISA-like assays , none of the purified recombinant truncates of SpsL exhibited binding to fibrinogen , with all peptides tested demonstrating binding equivocal to the negative control ( fibronectin-binding domain of SpsD ) ( S1B–S1D Fig ) . In contrast , fibronectin-binding could be detected in the SpsL recombinant protein construct containing a single fibronectin-binding repeat ( S1E Fig ) [23] . Similarly , full-length or truncated SpsL A-domain fragments expressed and purified from S . pseudintermedius ED99ΔspsLΔspsD supernatant did not adhere to canine fibrinogen , in contrast to a positive control of recombinant SpsD N2N3 purified from E . coli ( S1F Fig ) . However , heterologous overexpression of SpsL on the surface of a fibrinogen-binding deficient S . aureus strain ( SH1000ΔclfAΔclfBΔfnbAΔfnbB ) [30] promoted high levels of adherence to canine fibrinogen ( S1G Fig ) . Taken together , these data suggest that SpsL requires bacterial cell surface attachment to mediate fibrinogen-binding . Accordingly , subsequent experiments employed SpsL constructs expressed on the surface of S . pseudintermedius ED99 . SpsL fragments representing the A-domain and N2N3 subdomains were expressed on the surface of the S . pseudintermedius fibrinogen-binding deficient mutant ED99ΔspsLΔspsD ( Fig 3A ) . As reported for other bacterial cell wall-associated proteins , we considered that the C-terminal repeat region may be required to project the fibrinogen-binding domain from the cell surface , and that a small region of the N1 subdomain may be required for secretion and cell surface expression [31 , 32] . To address this issue , chimeric proteins were generated that replace the SpsL fibronectin-binding repeats with the ClfA SD repeats ( that do not exhibit any known ligand-binding activity ) [33] and that contain a 21 amino acid region of the N1 subdomain ( N121; 181 VSKEENTQVMQSPQDVEQHVG 201 ) ( Fig 3A ) . Analysis of the binding of these constructs to immobilized fibrinogen and expression analysis by Western blot indicated the requirement for the N121 peptide for cell surface expression , with the N2N3+SD and N2N3 constructs not expressed on the cell surface ( Fig 3B and 3C ) . Importantly , both the chimeric A-domain ( A+SD ) and N2N3-subdomain ( N121+N2N3+SD ) proteins exhibited binding to canine fibrinogen that was equivocal to the full length SpsL protein ( Fig 3B ) . The binding of the chimeric N121+N2N3+SD protein to fibrinogen from multiple host species indicate that the SpsL N2N3 subdomains are sufficient for host-specific fibrinogen-binding , with the 21 amino acids of the N1 subdomain required for cell surface expression , ( Fig 3D ) suggesting that SpsL mediates ligand-binding in a manner analogous to the dock , lock and latch-binding mechanism described for other staphylococcal cell wall-associated proteins [34] . To investigate this further , we modelled the structure of SpsL N2N3 subdomains , based on the crystal structure of ClfA ( pdb 1N67 ) [35] . The structural model predicted classical DE-variant IgG folds made up of β-sheets typical of staphylococcal fibrinogen-binding proteins ( S2A Fig ) . From this model we identified a putative latch region , 502 NSASGSG 508 , required for the dock , lock and latch binding mechanism ( S2A Fig ) . Deletion of this putative latch region in a surface-expressed SpsL construct had no effect on surface expression or fibronectin-binding but abrogated adherence to both canine and human fibrinogen ( p<0 . 001 ) ( Fig 3E and S2B Fig ) . In addition to the AFM data , these results , suggest that the SpsL N2N3-subdomains expressed on the bacterial surface mediate fibrinogen-binding via a mechanism analogous to the dock , lock , and latch binding model . Staphylococcal proteins have evolved the ability to bind fibrinogen through multiple distinct interactions with different regions of host fibrinogen [13 , 14 , 36] . Previously it has been identified that S . pseudintermedius strain 326 is capable of binding to the fibrinogen α-chain with binding to the β- , and γ-chains not investigated [37] . To identify the binding site of SpsL , recombinant versions of the α- , β- , and γ-chains of human fibrinogen were expressed in and purified from E . coli and employed in bacterial binding assays . Both S . pseudintermedius ED99 and ED99ΔspsD demonstrated specific adherence to the human α-chain , but not to the β- or γ-chains revealing the α-chain as the receptor for SpsL binding ( Fig 4A ) . To further refine the location of the SpsL binding site , 6 overlapping fragments of the canine fibrinogen α-chain were synthesized ( NCBI reference sequence: XP_532697 . 2 ) , purified from E . coli , and analyzed for adherence to ED99ΔspsLΔspsD expressing full length SpsL ( Fig 4B and 4C ) . SpsL demonstrated binding to two of the overlapping fragments ( 250–450 and 400–600 amino acids ) that span the α-connector region of fibrinogen containing unordered tandem repeats ( residues P283-S419 ) ( Fig 4B ) . Purification of equivalent fragments derived from the human α-chain revealed equivocal binding to the 400–600 fragment but reduced binding to the human 250–450 fragment ( Fig 4D ) . These data indicate that the canine fibrinogen α-chain contains strong ( 250–450 ) and weaker ( 400–600 ) SpsL binding sites while human fibrinogen contains just the weaker binding site ( 400–600 ) . To confirm that the canine tandem repeat region is responsible for the host-specific interaction of SpsL with fibrinogen , we generated chimeric full-length proteins where the tandem repeat region from human and canine fibrinogen α-chains , respectively ( P283-G421 ) , were exchanged . The addition of the canine α-chain tandem repeats provides stronger binding to the human fibrinogen α-chain and in contrast the addition of the human α-chain tandem repeats provides weaker binding to the canine fibrinogen α-chain ( Fig 4E ) . As a control , we examined the binding of ClfB expressed on the surface of SH1000ΔclfAΔclfBΔfnbAΔfnbB . ClfB is a S . aureus fibrinogen-binding surface protein that binds specifically to repeat 5 of the human α-chain tandem repeats [13] . The host-specificity of ClfB was confirmed with specific binding observed to the human 250–450 fragment ( S3A Fig ) , and the canine chimeric protein containing the human tandem repeat sequence ( S3B Fig ) but not to the canine 250–450 fragment or the canine alpha chain . These data demonstrate that the tandem repeat region of the fibrinogen α-chain is responsible for the host-specific interaction of SpsL . The canine tandem repeat region of the fibrinogen α-chain contains 7 repeats of 18 amino acids and a partial repeat of 11 amino acids ( S3C Fig ) [38] . The generation of recombinant fragments spanning the tandem repeats of the fibrinogen α-chain ( S3D Fig ) , revealed that SpsL is capable of binding to multiple regions in the canine tandem repeat region ( S3E Fig ) . In addition , deletion of the whole tandem repeat region , in the canine α-chain , confirmed the presence of a weaker binding site ( Fig 4F ) , which we localized to the region adjacent to the tandem repeats ( S423-E474 ) in both canine and human fibrinogen ( S3F and S3G Fig ) . Overall , these data demonstrate that SpsL mediates binding to multiple locations in the fibrinogen α-chain , and that the strong canine-specific interaction is dependent on the unique tandem repeat sequence present in the canine fibrinogen α-chain . The number of tandem repeats in the bovine fibrinogen α-chain have been reported to differ between cattle breeds [39] . To investigate if this is also the case for dogs we investigated publicly available canine sequences of the fibrinogen α-chain but , due to the repetitive nature of the tandem repeat region , the paired-end short sequence reads were not sufficient to support assembly and robust analysis . To overcome this , we isolated genomic DNA from 11 different canine breeds and PCR-amplified DNA specific for the P283-E474 region of the fibrinogen α-chain , followed by DNA sequencing . Sequence analysis , in comparison to NCBI reference sequence: XP_532697 . 2 , revealed that the region of weaker binding ( S423-E474 ) is conserved among the canine breeds examined ( S4A Fig ) . In contrast , the French bulldog and Labrador retriever exhibited heterozygous alleles that contain an additional repeat unit in the stronger binding site ( Fig 4G ) . This heterozygous allele , common to both breeds , contains a duplication of repeat 4 and amino acid substitutions that result in the replacement of repeats 6 and 7 with repeat 8 ( XP_532697 . 2:p . [S347_S348insTRPGSTGPGSAGTWS;S373N;L394P] ) ( Fig 4G ) . The unique French bulldog allele contains substitutions that convert repeat 5 to repeat 4 , and repeat 7 to repeat 8 ( XP_532697 . 2:p . [S347_T351del;G352R;T361A;L394P] ) , with a unique bulldog allele replacing repeat 6 with repeat 8 ( XP_532697 . 2:p . S373N ) ( Fig 4G ) . Overall , these analyses demonstrate that the canine-specific binding site of SpsL in the tandem repeat region of the canine fibrinogen α-chain has undergone genetic diversification during the evolution of different breeds of dog . The evolution of a strong canine-specific fibrinogen-binding interaction for SpsL suggests an important role in canine host-pathogen interactions . Accordingly , we investigated the impact of the interaction on phenotypes relevant to pathogenesis . Firstly , we considered if the interaction could promote inhibition of opsonophagocytosis as reported previously for fibrinogen-binding proteins of S . aureus ClfA and Efb [40 , 41] . Accordingly , FITC-labelled bacteria were opsonized with bovine , canine , human , or ovine fibrinogen or bovine fibronectin and analyzed for phagocytosis by human neutrophils . As expected , full length SpsL , but not the chimeric A-domain protein ( A+SD ) , inhibited phagocytosis in the presence of fibronectin ( p<0 . 001 ) ( Fig 5A ) . However , the ability of SpsL to inhibit neutrophil phagocytosis in the presence of fibrinogen was demonstrated to be host-specific with opsonophagocytosis inhibited in the presence of canine and human fibrinogen ( p<0 . 001 ) but not in the presence of bovine or ovine fibrinogen ( Fig 5A ) . We next examined the role of SpsL-canine fibrinogen-binding on S . pseudintermedius aggregation and biofilm formation . The aggregation of S . aureus has been demonstrated to be important for the development of bloodstream infections [42 , 43] and catheter-related infections [44] . In particular , it has been reported that fibrinogen-dependent S . aureus aggregation can stimulate the activation of virulence through a quorum-sensing dependent mechanism [45] . In order to examine the potential role of the canine-specific fibrinogen-binding in S . pseudintermedius aggregation , we attempted to block binding of S . pseudintermedius to the canine fibrinogen α-chain by including soluble fibrinogen in a bacterial adherence assay . Instead of blocking adherence , we found that soluble canine fibrinogen α-chain , but not human fibrinogen α-chain , supported the formation of surface bound aggregates ( Fig 5B ) . Deletion of the weaker binding site ( S423-E474 ) in the canine fibrinogen α-chain , had no effect on bacterial aggregation ( Fig 5B ) . However , deletion of the stronger binding site ( the tandem repeat region ) , resulted in complete abrogation of bacterial aggregation ( Fig 5B ) . This demonstrates that SpsL promotes surface bound bacterial aggregation in a host-restricted manner . To further investigate the impact of fibrinogen on the aggregation of S . pseudintermedius we performed static biofilm assays in the presence or absence of fibrinogen from different host species ( Fig 5C ) . Coating with canine fibrinogen supported enhanced biofilm formation among bacterial cells expressing SpsL than wells coated with either human or bovine fibrinogen demonstrating that the strong interaction of SpsL with canine fibrinogen promotes the initial attachment stage of biofilm formation ( Fig 5C ) . Overall , these data demonstrate that the high strength interaction of SpsL with canine fibrinogen promotes bacterial aggregation and biofilm formation . In order to investigate if other staphylococcal fibrinogen-binding proteins exhibit similar host-specificity , we generated constructs expressing chimeric SpsL proteins that contain the fibrinogen-binding N2N3 subdomains of ClfB or FnBPA but maintain the SpsL promoter , signal peptide , fibronectin-binding repeats , and cell wall anchor ( Fig 6A ) . The generation of chimeric proteins was favored over the expression of native ClfB or FnBPA proteins to limit variation in cell surface expression . The N2N3 subdomains of these proteins were selected because of the host-restriction of ClfB to human fibrinogen α-chain and the similar domain architecture of SpsL and FnBPA . As expected from our previous analysis , SpsL showed similar binding to both canine and human fibrinogen ( Fig 6B ) . However , the high strength interaction of SpsL with canine fibrinogen is essential for bacterial aggregation ( Fig 6C ) and biofilm formation ( Fig 6H ) . SpsL-ClfBN2N3 demonstrated a similar binding curve to SpsL but exhibited specific binding to human fibrinogen ( Fig 6D ) as previously predicted [13] . This human fibrinogen-binding was not sufficient to mediate bacterial aggregation ( Fig 6E ) or biofilm formation ( Fig 6H ) demonstrating that not all staphylococcal fibrinogen-binding proteins are capable of mediating these infection-related phenotypes . In contrast , SpsL-FnBPAN2N3 exhibited a binding pattern predicted from an interaction with the fibrinogen γ-chain with equivocal binding to bovine , canine , and human fibrinogen and reduced ovine fibrinogen-binding ( Fig 6F ) . SpsL-FnBPAN2N3 is capable of mediating bacterial aggregation ( Fig 6G ) and biofilm formation ( Fig 6H ) in the presence of fibrinogen from all hosts tested suggesting that FnBPA does not have a host-restricted tropism . From this comparative analysis we can conclude that SpsL is unique in promoting bacterial aggregation and biofilm formation in a manner that corresponds to the host-restricted ecology of S . pseudintermedius .
The factors underpinning bacterial host-tropism are not well understood but often involve surface proteins mediating interactions with host cells and the extracellular matrix [1] . The genus Staphylococcus includes species such as S . aureus which have a multi-host tropism with the capacity to switch between different host species . In contrast , some species such as S . pseudintermedius are highly host-restricted and although S . pseudintermedius can occasionally cause zoonotic infections of humans ( typically through dog bite wounds ) , the capacity to spread in human populations has not been reported . The bacterial factors underpinning the host-restricted ecology of S . pseudintermedius are unknown . Previously , we demonstrated that SpsL contributed to abscess formation in a murine model of subcutaneous infection indicating that it is a virulence factor during the pathogenesis of skin infection [22] . The poor binding of SpsL to murine fibrinogen suggests that this effect is not mediated by the interaction of SpsL with murine fibrinogen [22] . Here we demonstrate that SpsL mediates high strength binding to canine fibrinogen in a host-specific manner and that this host-adaptation conferred the ability to mediate bacterial aggregation and biofilm formation . The role of SpsL-fibrinogen binding in canine pathogenesis cannot be formally tested in vivo by experimental infections of dogs in the UK due to ethical constraints . However , our ex vivo binding and cellular infection data reveal multiple pathogenic traits that depend on the host-specific interaction of SpsL and canine fibrinogen , suggesting a key role in the host ecology of S . pseudintermedius . Cell surface proteins of S . aureus have been reported to contribute to tissue or disease tropism in humans . For example , the fibrinogen- and loricrin-binding protein , ClfB , exhibits greater adherence to skin corneocytes taken from atopic dermatitis patients with low levels of natural moisturizing factor suggesting a role in niche adaptation [46 , 47] . ClfB interacts with the human tandem repeat region of the fibrinogen α-chain [13] but unlike SpsL , ClfB binds to a single site; namely repeat unit 5 , and exclusively binds to human fibrinogen [13] . In addition to ClfB , the bone sialoprotein-binding protein ( Bbp ) and the extracellular fibrinogen-binding protein ( Efb ) also bind to the fibrinogen α-chain via distinct RGD-integrin-binding sites inhibiting thrombin-induced coagulation and platelet aggregation , respectively [36 , 48] . In contrast , SpsL interacts with multiple sites in the canine fibrinogen α-chain; namely within the tandem repeats and their flanking regions , ( Fig 4 ) . Similarly , the serine-rich repeat glycoproteins of Streptococcus agalactiae , Srr1 and Srr2 , bind to repeat units 6 , 7 , and 8 of the tandem repeat region of the human fibrinogen α-chain via a variation of the dock , lock , and latch binding mechanism , with Srr2 displaying a stronger binding affinity than Srr1 [49] . The enhanced binding affinity of Srr2 was linked with increased adherence to endothelial cells , which may be important for Group B Streptococcus-associated meningitis [49] . The ability of the Srr and SpsL proteins to adhere to more than one site in the tandem repeat region of the fibrinogen α-chain may have evolved as a mechanism for overcoming extant genetic diversity in this region between individuals within a host species as observed in the current study for SpsL ( Fig 4G ) [39 , 50] . We were unable to detect binding of soluble SpsL proteins to canine fibrinogen by ELISA suggesting that immobilization and surface presentation is essential for SpsL functionality , even when full length SpsL is expressed as a recombinant protein ( S1 Fig ) . To address this , we utilized AFM , demonstrating that bacterial surface-associated SpsL binds to fibrinogen via extremely strong binding forces ( around 2000 pN ) that are in the range of the strength measured for the dock , lock and latch interaction between fibrinogen and the structurally-related SdrG and ClfA [26 , 51] . Dock , lock and latch forces have been shown to originate from hydrogen bonds between the ligand peptide backbone and the adhesin [52 , 53] , and are activated by mechanical tension , as observed with catch bonds [54] . Of note , ClfB has much greater affinity for loricrin when expressed on the bacterial cell surface rather than as a recombinant protein with the C-terminal stalk enhancing binding affinity [25] . A similar mechanism may be required for SpsL adherence to fibrinogen with the C-terminal repeat domain enhancing the ligand-binding affinity of the N2N3 subdomains . It is increasingly being recognized that analysis of protein-protein interactions on the bacterial cell surface is more physiologically relevant than testing the interaction of recombinant polypeptides [25 , 55] . Our data reveal that the high strength canine-specific binding of SpsL facilitates several virulence phenotypes not previously reported for S . pseudintermedius including surface-bound bacterial aggregation . When S . aureus forms fibrinogen-dependent aggregates , agr-mediated quorum sensing is activated leading to the up-regulation of virulence gene expression [45] . Consequently , the inhibition of S . aureus aggregation in vivo has been linked with decreases in mortality from sepsis and protection from lethal lung injury [43 , 56] . We also discovered that SpsL facilitates fibrinogen-dependent biofilm formation , a phenotype not previously reported for S . pseudintermedius . Such fibrinogen-dependent biofilms are observed in S . aureus strains isolated from skin infections [57] , a phenomenon implicated in indwelling medical device infections [58] . In this regard , inhibition of fibrin formation reduced the development of S . aureus biofilms in a murine catheter infection model [58] , and molecules targeting SpsL could be beneficial in preventing canine indwelling device infections caused by S . pseudintermedius . Finally , we have demonstrated that SpsL binding to soluble fibrinogen inhibits neutrophil phagocytosis , suggesting a role for SpsL in innate immune evasion . Taken together , we have dissected the host-dependent binding of a bacterial surface protein and demonstrated its importance for multiple pathogenic traits , providing new insights into the host-specific ecology of a major bacterial pathogen .
Chicken immunization was performed using unembryonated hen’s eggs at the Scottish national blood transfusion service ( Pentland Science Park , Midlothian , UK ) . The procedures performed were carried out under the authority of the UK Home Office Project License PPL 60/4165 and Animals ( Scientific Procedures ) Act 1986 regulations . Human venous blood was taken from consenting adult healthy volunteers in accordance with a human subject protocol approved by the national research ethics service ( NRES ) committee London City and East under the research ethics committee reference 13/LO/1537 . Passive volunteer recruitment was conducted at the Roslin Institute ( University of Edinburgh ) . Written consent was taken from each volunteer before blood collection and after an outline of the risks was provided . All blood collection samples were anonymized . The bacterial strains and plasmids used in this study are listed in S1 Table . S . pseudintermedius and S . aureus strains were routinely cultured in Brain Heart Infusion broth at 37°C with shaking and supplemented with 10 μg ml-1 chloramphenicol as required . E . coli strains were cultured in Luria broth at 37°C with shaking supplemented with 100 μg ml-1 ampicillin , 15 μg ml-1 tetracycline , or 25 μg ml-1 kanamycin as required . Fibrinogen isolated from bovine , human , and ovine plasma ( Sigma-Aldrich ) and bovine fibronectin ( EMD Millipore ) was sourced commercially . Canine fibrinogen was purified from Beagle sodium citrate whole blood ( Lampire Biological Products ) using a previously described method [59] . All fibrinogen samples were purified to remove contaminating fibronectin using Gelatin-Sepharose 4B ( GE Healthcare ) . Depletion of fibronectin was confirmed by Western blot analysis using 1 μg ml-1 rabbit anti-fibronectin IgG ( abcam ) and 0 . 2 μg ml-1 goat anti-rabbit IgG-HRP ( abcam ) . Solid phase adherence assays were performed using S . pseudintermedius and S . aureus strains expressing pALC2073 or pCU1 constructs cultured to an OD600nm of 0 . 6 and induced for protein expression with 3 μg ml-1 anhydrotetracycline for 2 h . Cells were washed and suspended in PBS to OD600nm of 1 . 0 . Wells were coated overnight at 4°C with fibrinogen from multiple hosts or recombinant α-chain fragments in a 96-well MaxiSorp plate ( Nunc ) . After blocking with 8% ( w/v ) milk-PBS , bacteria were applied to the wells for 2 h at 37°C . After washing , adherent cells were fixed with 25% ( v/v ) formaldehyde ( Sigma ) for 30 min and stained with 0 . 5% ( v/v ) crystal violet ( Sigma ) for 3 min . The cell-associated stain was solubilized with 5% acetic acid ( v/v ) and analyzed using a Synergy HT plate reader ( BioTek ) at 590 nm wavelength . For aggregation experiments , the same procedure was followed as stated above with the addition of either soluble fibrinogen or recombinant fibrinogen α-chain to the bacteria using two-fold serial dilution and then incubation for 2 h at 37°C . The primers used in this study are listed in S2 Table . Initial expression constructs of the human and canine fibrinogen α-chains were synthesized by Integrated Design Technologies ( IDT ) using the DNA sequence of a female Boxer ( NCBI reference sequence: XP_532697 . 2 ) as highlighted in S1 Table . For typical restriction-ligation cloning procedures , the region of interest was amplified ( PfuUltra II Fusion HS Polymerase—Agilent ) and blunt cloned into pSC-B using the StrataClone Blunt PCR Cloning Kit ( Agilent ) . Restriction digestion of the plasmid of interest ( pQE30 , pT7 , or pALC2073 ) and the blunt cloned PCR product was performed at 37°C for at least 2 h and purified using the Monarch Gel Extraction Kit ( NEB ) . All digested plasmids were treated with Antarctic Phosphatase ( NEB ) before overnight ligation with T4 DNA Ligase ( NEB ) at a 3:1 molar ratio of insert:plasmid . Dialysis of the 20 μl ligation reactions was performed using 0 . 025 μm filter circular discs ( Millipore ) before electroporation into the appropriate E . coli strain–DC10B [60] , DH5α ( Invitrogen ) , or XL-1 Blue ( Agilent ) . All plasmid constructs were verified by Sanger sequencing ( Edinburgh Genomics , University of Edinburgh ) before transformation into E . coli BL21 DE3 ( Invitrogen ) , or appropriate S . pseudintermedius strain . Some expression constructs were also produced using sequence ligase independent cloning ( SLIC ) as described previously [61] . Briefly , primers were designed to amplify the gene of interest as well as sequence complementation to the expression plasmid . Primers were also designed to amplify the plasmid of interest , pQE30 , pALC2073 or pCT using Platinum PCR Supermix ( Invitrogen ) or PfuUltra II Fusion HS Polymerase ( Agilent ) . All PCR products were purified using Monarch PCR & DNA Cleanup kit or Monarch Gel Extraction kit ( NEB ) . T4 DNA Polymerase ( NEB ) was used to generate DNA overhangs on both the insert and plasmid PCRs with step-wise temperature increments used to anneal the complementary DNA sequences . The heat annealed constructs were electro-transformed into E . coli DC10B [60] or DH5α ( Invitrogen ) and verified using Sanger sequencing ( Edinburgh Genomics , University of Edinburgh ) . S . pseudintermedius and S . aureus competent cells were produced using a method outlined previously [60] . Plasmids for electroporation were concentrated to 1 μg μl-1 using Pellet Paint co-precipitant ( Novagen ) and 5 μg used for the electro-transformation as previously described [23] . Recombinant hexa-Histidine-tagged proteins expressed in E . coli were cultured to OD600nm of 0 . 6 and induced using 1 mM IPTG at either 37°C for 4 h or 16°C overnight . Recombinant α-chain proteins were purified under denaturing conditions ( 8M urea , 100 mM monosodium phosphate , 10 mM Tris-HCl ) using Ni-NTA agarose ( Invitrogen ) and gravity flow columns ( Bio-Rad ) . Bacterial lysis was performed in pH 8 . 0 binding buffer at room temperature with tilting for at least 1 h . Lysates were pelleted at 16000 x g for 20 min and the supernatant filter sterilized . Lysates were tilted at room temperature with conditioned Ni-NTA agarose for 1 h . The column was washed with pH 6 . 3 wash buffer and the protein eluted with pH 4 . 5 elution buffer . After analysis by 4–20% Mini-PROTEAN TGX precast gel ( Bio-Rad ) , protein quantification was performed using a BCA assay ( Novagen ) . S . pseudintermedius cells were cultured to exponential phase ( OD600nm of 0 . 4–0 . 6 ) . Cells were washed with PBS and suspended in lysis buffer ( 50 mM Tris-HCl , 20 mM MgCl2 , pH 7 . 5 ) supplemented with 30% ( w/v ) raffinose and cOmplete protease inhibitor ( Roche ) . Cell wall proteins were solubilized by incubation with 400 μg ml-1 lysostaphin at 37°C for 20 min . Supernatant samples were collected after protoplast recovery by centrifugation at 6000 x g for 20 min . The production of cell lysate samples was generated by lysing cell pellets in PBS on the One-Shot cell disruptor ( Constant Systems ) with 2 passes at 40 Kpsi . Recombinant His-tag SpsL N2N3 protein was used as antigen for chicken immunization and antibody generation at the Scottish national blood transfusion service ( Pentland Science Park ) . The Eggspress IgY purification kit ( Gallus Immunotech ) was used to purify antibody from egg yolk . Further purification of the antibody was performed using CNBr-activated Sepharose 4B ( GE Healthcare ) . This antibody was used in Western blot analysis to detect the expression of SpsL using 1 μg ml-1 chicken anti-SpsL N2N3 IgY and 0 . 5 μg ml-1 F ( ab’ ) 2 rabbit anti-chicken IgG-HRP ( Bethyl Laboratories ) . Genomic DNA was isolated from whole canine blood using the method described previously [65] . The region of interest in the fibrinogen α-chain was amplified using Q5 Hot Start high-fidelity DNA polymerase ( NEB ) and purified using Monarch PCR & DNA Cleanup kit ( NEB ) . Purified PCR products were analyzed by Sanger sequencing ( Eurofins ) and DNAStar SeqMan Pro 14 ( Lasergene ) . Sequence alignment was performed using MegAlign ( Lasergene ) and PRALINE [66] . Biofilm assays were performed using S . pseudintermedius strains expressing pALC2073 constructs of full length SpsL or A-domain+SD . Strains were grown in TSB supplemented with 0 . 5% ( v/v ) glucose and 3% ( v/v ) NaCl . 96-well tissue culture plates were coated overnight at 4°C with 100 nM bovine , canine , human , or ovine fibrinogen with some wells left uncoated . Overnight cultures were diluted to an OD600nm of 0 . 05 and 100 μl applied to the plate and incubated at 37°C for 24 h . The plates were washed three times with PBS and the bacteria fixed with 25% ( v/v ) formaldehyde ( Sigma ) for 30 min . After washing , the plates were stained with 0 . 5% ( v/v ) crystal violet ( Sigma ) for 3 min and then solubilized with 5% acetic acid ( v/v ) . Plates were analyzed using a Synergy HT plate reader ( BioTek ) at 595 nm wavelength . 50 ml of venous blood was drawn from healthy volunteers and mixed with 6 ml of acid-citrate-dextran ( Sigma ) . Human neutrophils were isolated as outlined previously [67] and suspended to a final concentration of 2 . 5 x 106 cells ml-1 in RPMI-1640 ( Gibco ) containing 0 . 05% human serum albumin ( Sigma ) . 2 . 5 x 106 CFU of bacteria , previously labelled with FITC using a method previously described [68] , were opsonized with 50 nM of extracellular matrix protein at 37°C for 15 min and diluted to 1 ml in RPMI-1640 containing 0 . 05% human serum albumin . 2 . 5 x 105 CFU were then opsonized with 10% human serum in 2 ml 96-well v-bottomed plates ( Corning ) at 37°C for 15 min . 2 . 5 x 105 neutrophils were added to the opsonized bacteria ( MOI of 1 ) and incubated at 37°C for 15 min with shaking at 750 rpm . The samples were fixed with 1% ( v/v ) paraformaldehyde ( Fisher Scientific ) and incubated at 4°C for at least 30 min . Phagocytosis was measured in comparison to serum-only controls using the BD LSRFortessa X20 cell analyzer . Data is presented in Prism 6 ( Graphpad ) with statistical analysis performed using Minitab 16 . All data was analyzed for normality , using the Anderson-Darling test , and equal variance before choosing the method of statistical analysis . Multiple comparisons were performed were appropriate . ELISA-type binding assays and bacterial adherence assays were analyzed at one protein concentration . For data displaying statistical significance , the following symbols are used , * p≤0 . 05 , ** p≤0 . 01 , and *** p≤0 . 001 .
|
Many bacterial pathogens are specialized for a single host-species and rarely cause infections of other hosts . Our understanding of the bacterial factors underpinning host-specificity are limited . Here we demonstrate that a canine host-restricted bacterial pathogen , Staphylococcus pseudintermedius , produces a surface protein ( SpsL ) that has the ability to preferentially bind to canine fibrinogen with high strength . This host-specific interaction has evolved via binding to a tandem repeat region of the fibrinogen α-chain which is divergent among mammalian species . Importantly , we found that the strong binding interaction with canine fibrinogen promotes bacterial aggregation and biofilm formation as well as inhibiting neutrophil phagocytosis . Our findings reveal the host-adaptive evolution of a key bacterium-host interaction that promotes evasion of the host immune response .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"bacteriology",
"biofilms",
"cell",
"physiology",
"cell",
"binding",
"medicine",
"and",
"health",
"sciences",
"chemical",
"characterization",
"pathology",
"and",
"laboratory",
"medicine",
"fibrinogen",
"pathogens",
"microbiology",
"staphylococcus",
"aureus",
"glycoproteins",
"bacteria",
"bacterial",
"pathogens",
"research",
"and",
"analysis",
"methods",
"staphylococcus",
"medical",
"microbiology",
"microbial",
"pathogens",
"pathogenesis",
"repeated",
"sequences",
"binding",
"analysis",
"biochemistry",
"bacterial",
"biofilms",
"tandem",
"repeats",
"cell",
"biology",
"host-pathogen",
"interactions",
"genetics",
"biology",
"and",
"life",
"sciences",
"genomics",
"glycobiology",
"organisms"
] |
2019
|
Host-specialized fibrinogen-binding by a bacterial surface protein promotes biofilm formation and innate immune evasion
|
Computational approaches to tune the activation of intracellular signal transduction pathways both predictably and selectively will enable researchers to explore and interrogate cell biology with unprecedented precision . Techniques to control complex nonlinear systems typically involve the application of control theory to a descriptive mathematical model . For cellular processes , however , measurement assays tend to be too time consuming for real-time feedback control and models offer rough approximations of the biological reality , thus limiting their utility when considered in isolation . We overcome these problems by combining nonlinear model predictive control with a novel adaptive weighting algorithm that blends predictions from multiple models to derive a compromise open-loop control sequence . The proposed strategy uses weight maps to inform the controller of the tendency for models to differ in their ability to accurately reproduce the system dynamics under different experimental perturbations ( i . e . control inputs ) . These maps , which characterize the changing model likelihoods over the admissible control input space , are constructed using preexisting experimental data and used to produce a model-based open-loop control framework . In effect , the proposed method designs a sequence of control inputs that force the signaling dynamics along a predefined temporal response without measurement feedback while mitigating the effects of model uncertainty . We demonstrate this technique on the well-known Erk/MAPK signaling pathway in T cells . In silico assessment demonstrates that this approach successfully reduces target tracking error by 52% or better when compared with single model-based controllers and non-adaptive multiple model-based controllers . In vitro implementation of the proposed approach in Jurkat cells confirms a 63% reduction in tracking error when compared with the best of the single-model controllers . This study provides an experimentally-corroborated control methodology that utilizes the knowledge encoded within multiple mathematical models of intracellular signaling to design control inputs that effectively direct cell behavior in open-loop .
The ability to predictably manipulate intracellular signaling pathways would provide an unprecedented level of control of cellular processes and could potentially generate new approaches for therapeutic design and research tools in medicine and systems biology . Intracellular signaling networks are complex assemblies of interconnected molecular components that relay information and coordinate responses to environmental cues . For example , T lymphocytes are critical regulators of the immune response against the threats of invading pathogens and cancerous host cells . Their response to external stimuli is coordinated through several mediators including extracellular signal-regulated kinases ( Erk ) , which are particularly noteworthy as they have been implicated in a number of autoimmune diseases and cancers [1]–[4] . Phenotypic change due to extracellular perturbation is a robust property of normal cell behavior and involves considerable feedback and crosstalk and is highly nonlinear . To help resolve the uncertainty and understand the complexity inherent within these signaling pathways , many researchers have developed mathematical models of signaling processes [5]–[10] . These models can be used to inform control strategies that try to predictably manipulate the intracellular signaling response , but also give rise to a new set of challenges in systems biology and control engineering . To date , the majority of model-based control of cellular processes and systems has focused on biomass production in bioreactors [11] , [12] or were largely theoretical . Within the past decade , research has started to evaluate engineered control strategies for single and multiple cell signaling processes within experiments . Noble and Rundell [13] used closed-loop ( i . e . in silico feedback ) control to direct HL60 cell differentiation through periodic boluses of a differentiation-inducing agent determined by nonlinear model predictive control ( MPC ) . In 2012 , they revised the initial approach to improve the transient response of the differentiating cells over 20 days by using a multi-scenario adaptive model predictive control [14] . Uhlendorf et al . [15] applied open-loop ( i . e . non-feedback ) and closed-loop control to provide long-term regulation of gene expression on the single-cell and population level by manipulating the osmotic stress on cells in a microfluidic environment . The open-loop approach failed to regulate mean fluorescence to the desired set points in the laboratory . On the other hand , attempts using measurement feedback proved to be more successful at coping with modeling inaccuracies and inherent intracellular fluctuations . Opto-genetics and synthetic biology provide effective methods for control theory to interface with cellular processes at the genetic and signal transduction level . Milias-Argeitis et al . [16] attempted to control the activation of a light-responsive Phy/PIF module that altered the expression of a yellow florescent protein ( YFP ) activated through the Gal1 promoter in Saccharomyces . As in [15] , the closed-loop approach proved more successful in driving YFP intensity to the desired set points in the laboratory . Tettcher et al . [17] , [18] also used light as the control input to modulate the localization of PIF-tagged proteins to the cell membrane . This has the ability to alter intracellular signaling through coordination of the localization . Both of these studies achieve long-term regulation of cell activities through transcriptional control . Another closed-loop control study was proposed in Menolascina et al . [19] using pulse width modulation to specify the duration of pulses of galactose administered via microfluidics . Their experimental results confirm the predictive capabilities of their model for the synthetic gene network in S . Cerevisiae . All of the aforementioned studies rely upon computer-based feedback control: a computer in the loop uses system measurements to inform control decisions . While closed-loop control is widely understood to be more robust to disturbance and uncertainty , in many cases the rapid dynamics , scale and complexity of intracellular signaling events and absence of real-time measurement assays prohibit its use . In these cases , any control strategy must design the control inputs in advance without measurements to inform future steps and rely solely upon prior information gleaned from existing experimental data . However , this may introduce unwanted degradation in control performance due to discrepancies between the actual system and the prediction model ( i . e . plant-model mismatch ) . Methods to systematically and optimally combine this prior information with the predictive capacity of multiple mathematical models are needed . Because mathematical models are abstractions of biological reality , they may differ in dominant species , network structure , parameter values , and functional representation . For most signaling pathways , limited preexisting quantitative data and qualitative observations are insufficient to discriminate unambiguously between the mathematical models . When applying control theory techniques , the experimental perturbations ( i . e . control inputs ) predicted to elicit a desired behavior from a system may be different for each model . Selecting the “best” of these models is an important challenge to control theorists for systems biology applications [20] . Apgar et al . [21] applied control theory to discriminate among mechanism-based chemical kinetic models of epidermal growth factor receptor signaling . The method designs dynamic stimuli to delineate the system's response to subtle differences in the network topology . The model associated with the best controller is deemed the best representative of the original system . However , in using this model alone , we would implicitly assume that the model is also the most accurate in alternate operating regions , which may or may not be the case . Without performing the experiments there is no way to know a priori which model is best; furthermore , the best model may change depending upon the experiment planned . How to optimally combine information from these network models to design control inputs that , when applied to the cell , force the signaling dynamics along a desired path is the subject of much debate . Growing attention in systems biology has been given to control methodologies considering multiple prediction models . Multiple models , or scenarios , have been previously used to improve robustness to parametric uncertainty in closed-loop model-based control [14] , [22]–[25] . The approach proposed by Rao et al . [24] computed control inputs by weighting multiple step-response models using a Bayesian algorithm to control hemodynamic variables in hypertensive subjects . These concepts were extended for disturbance rejection in a van de Vusse reactor using Bayesian methods to produce a weighted-average linear prediction model [25] . The recursive weighting system was effective at eliminating the “hard switch” between controllers . Noble et al . [14] employed adaptive nonlinear MPC based on multiple data-consistent parameter characterizations to manipulate cell differentiation experimentally . Control inputs were chosen such that the average tracking error and resource efficiency were optimized , which resulted in superior controller performance over single-model MPC . However , this approach assumes all parameter scenarios are equally likely and does not consider that their accuracy in predicting actual system dynamics often varies between distinct regions of the state and control input spaces . Furthermore , the aforementioned approaches consider essentially single model structures with little to no variation in the mathematical equation structures . While methods considering model uncertainty explicitly are numerous , it is generally in the form of disturbances and process noise with very little consideration given to qualitatively distinct biological hypotheses . This is a critical flaw as these hypotheses could translate into qualitatively distinct equation structures and input/output and state/output relationships . Finally , all of these approaches employ real-time feedback that is not available for most intracellular signaling pathways because the dynamics are often too rapid for standard measurement assays . In this paper , we present a practical open-loop control framework with a novel method for employing existing experimental data to confidently combine multiple model predictions to form effective control inputs . That is , an automated control input selection process is developed for the open-loop case; this process is advised by information regarding model accuracy in regions of the input space where potential control inputs are likely to be present . Akaike weights , based on an information-theoretic metric penalizing model complexity and lack-of-fitness used for model discrimination [26] , [27] , are employed for this purpose . In the results , we successfully demonstrate the algorithm with several simulated test cases and corroborate a subset of these with in vitro experiments in Jurkat T lymphocytes . Conclusions and future work are presented in the discussion and detailed and illustrated descriptions of the algorithm and experimental protocols are provided in the materials and methods .
The signaling pathway considered herein is the T cell receptor ( TCR ) -activated extracellular signal-regulated kinase ( Erk , or MAPK ) pathway ( generalized in Figure 1 ) . Activated Erk is an important condition in lymphocyte development and activation processes because it is a highly conserved and ubiquitous mechanism for transferring extracellular signals from membrane-bound receptors to the nuclear domain for gene regulation . The stimulation of TCRs by antigenic peptides ( e . g . αCD3 , shown in green in Figure 1 ) initiates a number of molecular reactions involved in signal transduction through Erk . During ligand binding , the TCR recruits and is phosphorylated by the tyrosine kinase Lck . A second tyrosine kinase , ZAP70 , binds the tyrosine phosphorylated TCR subunits and is phosphorylated and activated by Lck . The receptor-kinase complex recruits and phosphorylates the adapter proteins LAT and Grb2 and phosphorylates and activates PLCγ leading to the formation of GTP-bound Ras . Activated Ras initiates the canonical Raf/Mek/Erk signaling cascade which results in T cells in the activation of the gene for interleukin-2 . Three mathematical models of the TCR-mediated signaling cascade are used for the basis of the prediction model bank [5]–[7] . The model proposed by Zheng [5] ( herein referred to as Model Z ) contains primarily first- and second-order mass action kinetics with 24 ODEs and 53 reaction parameters . The second model ( herein referred to as Model L ) is the deterministic version of the model proposed by Lipniacki et al . [6] , which explicitly incorporates SHP-mediated negative feedback and Erk-mediated positive feedback . Model L consists of 32 reaction parameters and 37 ODEs derived from mass action kinetics . The original version of the third model , presented by Klamt et al . [7] , uses Boolean logic to describe the main steps involved in the activation of CD4+ helper T cells . CellNetAnalyzer , a Matlab software package [7] for structural and functional analysis of signaling networks , and the Odefy toolbox [28] were used to convert the logical model to a continuous homologue ( herein referred to as Model K ) with 40 states and 147 reaction parameters . Values for these reaction parameters were taken from Table 1 in [29] . All prediction models were modified to contain control inputs that simulate the actions of sanguinarine and U0126 . In addition , model outputs ( i . e . total concentration of phosphorylated Erk ) are normalized so that the peak uncontrolled response scaled to unity in order to account for the differences in scale ( see Section 1 in Text S1 for further details ) . All programming and simulation was performed in Matlab R2011b ( 7 . 13 . 0 ) and code is available in Dataset S1 . Herein , the computational burden of repeatedly evaluating large nonlinear ODE models is mitigated by using sparse grid interpolation . Sparse grids have been used as computational cost-cutting tools for control applications in systems biology [14] , [30] by serving as surrogates for slow-evaluating models and objective functions to allow rapid screening of the design space . In traditional model predictive control ( MPC ) , also referred to as receding horizon control , the controller surveys the possible trajectories stemming from the current state and selects the control input sequence so that the predicted model outputs track the desired trajectories over a finite prediction horizon . The first control input of the selected sequence is used to update the prediction model state and the procedure is repeated as the prediction horizon slides along for the remaining time intervals . When based on a single model , the controller is at risk of degraded performance because of mismatch between the predicted and actual system behaviors . The proposed control strategy , illustrated in Figure 2 , employs multiple prediction models to mitigate the effects of model uncertainty . In our approach , first , the model bank is populated with a set of relevant models , each of which predicts the system response to possible control inputs ( in control theory terms this is referred to as the plant response ) . During the initial stage , training data and the corrected-Akaike Information Criterion ( AICc ) are used to generate weight maps for the prediction models . These weight maps give the relative probability that a given model is consistent with training data at any given point in the feasible input space . In essence , these maps inform the controller of the tendency for models to differ in their ability to accurately reproduce the system dynamics under different control inputs . We then use a model predictive control framework in which the performance metrics for the models are optimized simultaneously using a multiobjective technique . Optimal control inputs are selected from the resulting Pareto solution set by prioritizing the solutions according to the pre-computed weight maps . Model weights are automatically recalibrated using the portion of training data that most closely corresponds to the proposed control input . Optimization and input selection cycles repeat for subsequent time intervals as the prediction horizon slides along until the entire open-loop control sequence is specified and ready to be applied ( see Materials and Methods for further details ) . As discussed previously , the purpose of this manuscript is to present a computational strategy to aid in the design of experimental input regimens to elicit predictable dynamical behaviors from biological processes . Herein , we demonstrate our approach using the well-known TCR signaling pathway both through simulation and laboratory experiments . First , we explored two in silico case studies , each considering a pair of control reagents acting on different regions of the T cell signaling pathway , to demonstrate the functionality of the proposed control strategy . In both case studies , we performed a series of experiments in which one model was selected from the aforementioned set of three models to serve as the simulated system . This model was also used to design the optimal control input regimen as a controller based on this model would match the actual system exactly and represent the best possible scenario . The remaining two models were then used to form the basis of the control strategies we wished to compare: single-model control ( mismatched model ) , multiple-model control with fixed equal weights , and our proposed multiple-model control with adaptive Akaike weights . For the in vitro case study , a subset of the control input regimens derived from the simulated case studies were tested on populations of Jurkat T cells for experimental corroboration . In each of our case studies of the TCR signaling pathway , we desired the system readout ( i . e . total concentration of phosphorylated Erk ) to follow a series of predefined time course trajectories . These target trajectories were characterized by the equation , where t = 0 refers to the time at which the initial stimulation dose of αCD3 is administered . The term pss represents the desired steady-state fraction of maximal activation that is to be achieved by the end of the 30-minute experiment . The term toff represents the time in minutes , following an initial interval of maximal activation , at which the controller should begin driving the output to the desired steady-state fraction . The ten parameter pairs chosen for this study are ( toff , pss ) = { ( 8 , 0 ) , ( 15 , 0 ) , ( 22 , 0 ) , ( 8 , 0 . 25 ) , ( 15 , 0 . 25 ) , ( 22 , 0 . 25 ) , ( 8 , 0 . 5 ) , ( 15 , 0 . 5 ) , ( 22 , 0 . 5 ) , ( 30 , 1 ) } . The five possible input dosing times began at 3 min post-αCD3-stimulation and were spaced 5 min apart to accommodate both the rapid dynamics of TCR signaling and the limitations of experimental input dosing and measurement rates . The control inputs used to perturb the TCR signaling pathway were chosen based on both the need to demonstrate the efficacy of our methodology and for experimental corroboration . For our simulated and experimental case studies , we chose two commercially-available reagents known to control the dynamics of phosphorylated Erk: sanguinarine and U0126 . Sanguinarine ( Figure 1 ) is a small-molecule inhibitor of Erk dual-specificity phosphatase-1 ( MKP-1 ) and leads to elevated Erk phosphorylation [31] . U0126 ( Figure 1 ) on the other hand is a Mek inhibitor with high selectivity , which effectively inhibits activation of Erk [32] . To further evaluate controller performance using a variety of objectives , our second simulated case study replaces these reagents with two hypothetical reagents that modulate the function of phosphorylated ZAP70 . The two reagents , aZAP and iZAP ( Figure 1 ) , act to promote and inhibit the function of pZAP70 , respectively . Sanguinarine and U0126 concentrations were constrained to the intervals [0 , 50] µM and [0 , 10] µM , respectively , the upper limits of which were estimated from experimental results that indicate saturation effects at levels above the specified concentrations ( see Figure 8 in Materials and Methods ) . aZAP and iZAP were normalized to the interval [0 , 1] as they are hypothetical . For details on our in silico implementation of all control reagents , readers are referred to Section 1 in Text S1 and to the Matlab code provided in Dataset S1 . The control strategy was first tested by considering combinations of two models at a time to control an unknown system , which is simulated by the remaining third model . In the following discussion , we will use S to denote single-model control with a subscript to denote the model ( Z , L , or K ) and M to denote multiple-model control with the subscript indicating either equal weights ( eq ) or adaptive weights ( aw ) . For illustrative purposes , the case in which Model K was the simulated system and the target profile corresponded to full termination at 22 min ( i . e . ( toff , pss ) = ( 22 , 0 ) ) is described and illustrated in detail ( results for all experiments are provided in Sections 3 and 4 in Text S1 ) . Figure 3 shows the control input dosing regimens for ( A ) the matched single-model controller SK , ( B–C ) the two mismatched single-model controllers SZ and SL , ( D–E ) the multiple-model controller with fixed equal weights ( Meq ) and with adaptive Akaike weights ( Maw ) , ( F ) the Akaike weights for Maw , ( G ) the simulated system responses , and ( H ) the squared error values for all five control strategies . The ideal scenario was defined to be one in which the model used to derive the control inputs is the same as the model used to simulate the actual system response , although it is generally not feasible in the laboratory and thus considered only for the purpose of theoretical comparison . For this ideal scenario , the control input regimen necessary to track the target profile included negligible input quantities to maintain initial Erk activation , quickly followed by a large bolus of U0126 at 23 min to promote Erk dephosphorylation near the end of the experiment ( Figure 3A ) . As shown in Figure 3 , the single-model controllers specified qualitatively different control inputs profiles , each causing qualitatively different simulated system responses . To achieve the sustained pErk activation phase of the target profile , SZ required a ramp-up in the sanguinarine doses ( Figure 3B ) , which consequently caused the simulated system to systematically overshoot the target ( Figure 3G , blue ) . On the other hand , SL predicted that negligible quantities of either reagent are necessary ( Figure 3C ) , which caused the simulated system to track the target moderately well , undershooting the target only slightly ( Figure 3G , green ) . For the rapid dephosphorylation phase , SZ specified a large dose of U0126 at 23 min while SL specified relatively small doses over the final two intervals . In contrast to the activation phase , the small doses specified by SL were insufficient to track the desired rapid transient behavior; the large dose specified by SZ produced significantly better tracking . Considering the experiment as a whole , however , neither controller adequately controlled the simulated system over the entire time interval . The multiple-model controller with equal weights ( Meq ) considered the predictions from both models equally in specifying the control inputs ( Figure 3D ) , but tracked only marginally better than either of the mismatched controllers SZ or SL ( Figure 3G , cyan ) . On the other hand , the adaptive weighting strategy allowed the multiple-model controller to inherit the best characteristics of both single-model controllers ( Figure 3E ) . According to Figure 3F , the weights tended toward Model L as it was the better representative of the simulated system initially , then shifted toward Model Z as it more accurately described the rapid dephosphorylation dynamics possessed by the simulated system . As a result , Maw tracked the target significantly better with at least a 74% reduction in the squared error ( see Figure 3H ) than any of the mismatched single-model and fixed equal weight controllers as indicated in dynamics shown in Figure 3G . Controller performances for the simulated case studies involving the real commercially-available control reagents sanguinarine and U0126 and the hypothetical control reagents aZAP and iZAP are summarized in Figures 4A and 4B , respectively . Both plots show the target tracking error between the predicted system dynamics and the target trajectories for all target and system model combinations . Matched single-model scenarios ( Sm ) are cases when the model used to design the control inputs is identical to that which is used to simulate the system response while the mismatched single-model scenarios ( Smis ) use different models in both roles , the latter case tending to be the better representation of biological reality . As shown in both plots of Figure 4 , Smis had relatively large error values and tracked quite poorly . Meq were able to partially mitigate these effects by averaging out some of the inconsistencies , but still performed no better than Smis . The proposed controller strategy ( Maw ) performed significantly better than Smis and Meq by preferentially selecting predictions from the models that were known to match the desire behavior at any point in time . This effect was more pronounced in scenarios where the dynamics differ among the prediction models because the controller was better able to filter out the inadequate models , thus improving tracking performance relative to the other controllers . Notably , the only scenarios not outperformed by Maw were those in which a matched prediction model was used; however , these cases would be unrealistic in practice . We conducted a set of experiments where control input regimens were computed considering all three prediction models and implemented in vitro in Jurkat T cells according to the experimental protocol described in Materials and Methods . Since the equal weighted multiple-model controller design did not improve performance when compared with the single-model controllers for the simulated experiments , we did not include those in the more expensive in vitro study . The model weight maps were trained using preexisting experimental data ( see Figure 8 in Materials and Methods ) . Figure 5 shows for a representative experiment ( A–D ) the computed control input dosing regimens for SZ , SL , SK and Maw , and ( E ) the Akaike weights for Maw and ( F ) quantitative Western blot data . The specified target in the illustrated case required that Erk undergo rapid and sustained phosphorylation for 22 min and then rapidly return to steady state at basal levels . The quantitative Western blot data for this exemplar experiment are shown in Figure 6C . Without any external manipulation , pErk returned to its basal level slower and sooner than desired ( Figure 6C , cyan triangle ) . Of the three single-model controllers , only SZ ( Figure 5A ) correctly predicted that an initial ramp-up in sanguinarine was required to sustain pErk levels as desired . Based on the a priori information contained in the model weight maps , Maw was preferential towards the predictions made by Model Z because it most accurately reflects the cell behavior ( Figures 5D and 5E ) . However , Model Z was unable to replicate the rapid transient behavior of pErk in Jurkat cells in response to U0126 as accurately as the other models . In this case , Maw deferred to the predictions made by Model K . Because of this ability to adapt the weights based on the current conditions , Maw was able to control pErk much more tightly over the entire experiment than any of the models considered in isolation ( Figure 6C , magenta dot ) . Overall controller performances , defined as the squared error between the target output profiles and the corresponding observed plant dynamics , for the aforementioned experiment as well as the two other performed corroborating experiments are summarized in Figure 6D ( descriptions of all experiments are provided in Section 5 of Text S1 ) . Unsurprisingly , the largest deviations from the target trajectory were the uncontrolled responses . The single-model controllers improved tracking performance , but varied quite widely due primarily to the degree of how accurately the mathematical model predicted the signaling response ( extent of plant-model mismatch ) . On the other hand , the intracellular signaling dynamics were much more tightly controlled by our multiple-model controller with adaptive Akaike weights , corroborating our in silico findings . The adaptive weighting strategy reduced the squared error by at least 63% over the best performing single-model controller . This indicates that our proposed control strategy is able to successfully filter out the prediction models in situations where they are known to be inaccurate and include them otherwise .
We have presented a method that was shown to improve target-tracking performance in a biochemical system with relatively high modeling uncertainty and measurement noise . Naturally , we employed our own knowledge of the system as well as that taken from literature to ease the implementation of the control method in the laboratory setting . Even so , the proposed framework is general enough to be employed in a wide variety of engineering applications . The method is suitable for any physical system with possible dynamics that can be characterized by a set of mathematical models . The models should include feasible control inputs that are capable of manipulating the output dynamics and can be structurally unique . Furthermore , the ensemble of models should adequately recapitulate the relevant behaviors of the system . The method does not require real-time observability in its current open-loop configuration , although observations should not be ignored when available . While we have tested and corroborated the method only with nonlinear ODE models , we believe the modeling format is an application-related issue and not restricted by the method . Although our proposed framework is widely applicable , its efficacy for a given system depends on a variety of modeling and experimental constraints ( i . e . problem-specific information used to inform the controller ) , and computational constraints ( i . e . control problem dimensionality ) . From a modeling and experimental perspective , efficacy depends on how well the system is characterized by the prediction model bank , the accuracy of the model weight maps and the availability of quantitative measurement tools or assays and control reagents or actuators . First , the prediction model bank should be formulated in such a manner that all desired system behaviors are within the reachable dynamics of the ensemble of models . Any characteristic behaviors or operating points not included in the model bank would not be able to be recapitulated , regardless of the adaptive model weighting scheme . However , these behaviors or operating points may be captured by including models with alternate parameter values or equation structures . In the case studied in this manuscript , each model had a different equation structure , but represented alternate plausible mechanisms that are at least partially supported by data . Second , the accuracy of the model weight maps depends on the quantity and quality of preliminary data . Without any prior information , the models are considered equally when selecting control inputs . As experiments are performed , the gathered data can be used to train the weight maps to the actual model-system relationship . Models can then be selected according to its capacity to accurately predict the effects of the inputs on the system . An implicit assumption here is that if a model is good at predicting the effect of some input at some time , it will be good at predicting the effect of another input at another time . While this assumption may not always be true , the known dynamics , developed and refined based on considerable experimental work , provide extensive constraints on what the possible unknown dynamics could be . Finally , while our method does not require data in real time because of the open-loop formulation , accurate quantitative assays and specific control reagents are highly beneficial in the development of accurate models and model weight maps help decrease the effects of plant-model mismatch , a common failure mode in open-loop control . Even so , with the proposed adaptive weighting procedure , each model need only be partially data-consistent . We were able to demonstrate control using quantitative Western blots to measure Erk phosphorylation dynamics in T cells , despite the fact that Western blots are notoriously difficult to use and chemical reagents generally have some off-target effects . Computational tractability is another important area to consider . To get the most benefit from the adaptive model weighting strategy , our control strategy involves a multiobjective optimization problem ( of dimension nM×ny ) at every time interval . In practice , this is equivalent to solving a series of single-objective optimization problems ( of dimension nu×Hu ) , the number of which depends on the desired Pareto front resolution . In this manuscript , we greatly reduce the complexity of our control problem by considering only three models , a single controlled output , and two discretized control inputs and a control horizon of one interval ( at a time ) . However , it would not be far-fetched to have a control problem with potentially dozens of models , multiple inputs and outputs , and an extended control horizon . In such a case , it would be critical to reduce the control problem down to a computationally tractable size . This can be achieved a number of ways . For instance , dimensions corresponding to different model outputs can be scalarized using weighted aggregation . Also , dimensions corresponding to models that are redundant or dominated ( i . e . low Akaike weights throughout the input space ) should be removed . As previously mentioned , the purpose of the weight map strategy is to help estimate an optimal blend of the considered models to create the best possible compromise input solution given modeling uncertainty . The accuracy of these maps depends heavily on the quantity and quality of the preliminary data used to train the maps . Without any prior information , there would be no evidence supporting one model over another and the resulting weight map topology would be flat . Inversely , if the system dynamics are known with certainty , but requires more than one model to recapitulate them all , then the resulting weight map topology would appear digital . That is , only one model would dominate in any given experimental scenario and the transition between which model dominates would be immediate . In our presented experimental study , we considered a case in which a substantial amount of time-course data ( 432 individual measurement points in total ) was used to train the model weight maps . Due to the relatively large number of data , the AICc tends to show a strong sensitivity to any differences in model fitness values , causing the weight map topography to appear somewhat digital in nature ( Figure 7A–C ) . However , let us consider a case in which we have fewer data ( 144 individual measurement points in total ) , or alternatively , more uncertainty in our models' ability to recapitulate the observed system behaviors . With fewer data , the AICc tends to show a weaker sensitivity to differences in model fitness values , causing the weight map topography to appear smoother ( Figure 7D–F ) . Although the weight maps for these two scenarios are clearly different quantitatively , the order in which the models are prioritized ( i . e . model rankings ) are essentially identical ( Figure 7G–L ) . To illustrate the effects of these characteristics , let us consider one exemplar experiment ( Figure 7M ) . When utilizing the full dataset , the control strategy tends to heavily favor one model over the others at any given time . However , when the limited dataset is considered , the time course is covered by a non-trivial ( i . e . non-digital ) combination of the models , each representing a portion of actual system behavior . Even so , the orders in which the models are prioritized are very similar between the two cases . This means that for any given region of the input space , the choice of model providing the most reliable information is somewhat robust to the amount or quality of the training data . As a result , the predicted control regimens ( Figure 7N ) and their corresponding target tracking performance values ( Figure 7O ) also exhibit this same qualitative robustness . T lymphocytes are an integral part of the human body's natural defense against the threats of invading pathogens and cancerous host cells . The function of these specialized immune cells largely depends on their phenotypic response to external stimulating signals . Propagating extracellular signals to their target substrates takes the coordinated effort of very large networks of molecular species with complex interactions ranging across vastly different spatial domains and temporal scales . While there are numerous mediators of signal transduction in lymphocytes , Erk is particularly noteworthy as it is an evolutionarily-conserved and ubiquitous group of signaling proteins critical to T cell development , proliferation and differentiation . Studies have linked Erk-mediated regulation to the differentiation of helper T cells into certain subtypes , particularly Th1 and Th2 , and to allergies , asthma and serious immune disorders if improperly subtyped [1] , [2] . Furthermore , it has been recognized that controlling the Ras/Raf/Mek/Erk pathway maybe beneficial towards advancing effective therapies for leukemia [37] . Constitutive activation of the Erk pathway is present in a high frequency ( >50% ) in patients suffering from acute myeloid leukemia ( AML ) and is associated with a marked reduction in survival duration [3] , [4] . Conversely , blocking Erk activation has been shown to cause cell death in leukemia cell lines [3] . Treatments based on methods that work to balance these opposing forces to restore proper Erk-mediated regulation in T cells would be highly beneficial to patients suffering from such pathologies . Historically this has been the subject of experimental research [38] , [39] . This study has confirmed that when used in combination the existing mathematical models of the Erk/MAPK pathway in T cells can support the engineering of control inputs to manipulate the activation and deactivation time course in a desired manner . We have developed a practical framework for controlling uncertain nonlinear systems using multiple models to generate predictable open-loop dynamical responses . The embedded model weight maps enable the controller to estimate the likelihood of each model in any feasible control scenario based on prior training data . The adaptive weighting strategy allows the controller to purposefully select subsets of the training data so that control decisions are made considering only the most relevant information at each time interval . Our open-loop controller design pairs model predictive control with an adaptive model weighting system based in information theory to create a cohesive strategy for systematically utilizing the most relevant knowledge embedded within limited training data in a computationally tractable manner . In both simulated and laboratory experiments this multiple-model control strategy and adaptive weighting scheme successfully reduced the open-loop target tracking error by more than half relative to multiple-model control with fixed weights ( simulation only ) and single-model control .
Erk phosphorylation ( pErk ) data were collected from Jurkat T leukemia cell line ( Jurkat clone E6 . 1; ATCC ) . Cells were grown in RPMI 1640 ( Sigma ) supplemented with 7 . 5% heat-inactivated fetal bovine serum ( BioWest ) , 1 mM sodium pyruvate ( Gibco ) , 12 . 5 mM HEPES pH 7 . 4 ( Sigma ) , 12 µM sodium bicarbonate ( Sigma ) 50 µM 2-Mercaptoethanol ( Sigma ) , 50 µg/ml streptomycin and 50 units/ml penicillin in an incubator at 37°C in humidified air containing 5% carbon dioxide . Cells were harvested in log-phase growth at a density of 2×107 cells per treatment . Cells were stimulated using anti-human CD3 ( 10 µg/ml , clone: UCHT-1 , eBioscience ) as the stimulatory signal at 37°C in a water bath . Cells were treated with the Mek1/2 inhibitor U0126 ( Calbiochem ) or the MKP inhibitor sanguinarine ( Sigma ) , depending on the protocol , dissolved in DMSO at the indicated time points with the indicated concentrations . Experimental control samples were treated with the same amount of DMSO . Samples of 2×106 cells were taken at the indicated time points and lysed in 1% NP40 lysis buffer ( 1% NP40 , 25 mM Tris , pH 7 . 4 , 150 mM NaCl , 5 mM EDTA , 1 mM NaV , 10 mM NaF , 10 µg/ml each of aprotinin and leupeptin ) for 15 min on ice . Lysates were centrifuged for 5 min at 18000 g at 4°C . The supernatant was added to the same volume of 2 protein solubilizing mixture ( PSM , 25% ( w/v ) sucrose , 2 . 5% ( w/v ) sodium dodecyl sulfate , 25 mM Tris , 2 . 5 mM EDTA , 0 . 05% bromophenol blue ) and boiled for five minutes . Proteins were separated via SDS-PAGE , blotted for phospho-Erk1/2 ( Cell Signaling ) , phospho-ZAP-70 ( pY319 , Cell Signaling ) and GAPDH ( Ambion ) . IRDye 800 and 680 secondary anti-mouse and anti-rabbit antibodies ( Li-Cor ) were used for signal detection using an Odyssey infrared scanner . Images of the blots were analyzed using ImageJ to produce quantitative data for model comparison . Model predictions were scaled to compensate for the fact that the data represent relative quantities only rather than absolute concentrations . Training experiments were designed to rapidly screen the responses of T cell populations to potential control reagent combinations . Fourteen different experiments were conducted: an experimental control ( i . e . no control inputs ) , five individual doses of sanguinarine ( 0 . 5 , 2 , 5 , 20 and 50 µM ) at the 15 minute mark , five individual doses of U0126 ( 0 . 5 , 1 , 2 , 5 and 10 µM ) at the 6 minute mark , and three combined doses of 0 . 5 and 1 , 50 and 1 , and 50 and 10 µM for sanguinarine and U0126 , respectively , at the 6 minute mark . These data were collected according to Experimental Protocol . The addition of low doses of sanguinarine had negligible effects on pErk concentrations . Only the moderate to high doses tested caused elevated to sustained phosphorylation of Erk ( Figure 8A ) . On the other hand , U0126 produced an immediate reduction in the Erk phosphorylation rate even at low doses ( Figure 8B ) and tended to overpower the effects of sanguinarine when added together ( Figure 8C ) . The representation of the controller input functions in the prediction models were modified to exhibit these trends . Due to the prevalence of observation noise , experimental data were smoothed using cubic smoothing splines to filter spurious oscillations from the time courses while retaining primary trends . This was performed in Matlab using the csaps function with the smoothing parameter p set to 0 . 6 . The smoothing splines were sampled at 31 evenly spaced time points to increase the density of the time course data . The proposed control strategy is based on a set of two or more mathematical models , , of a given system with the general form: ( 1 ) where the superscript i denotes the model number . The state variables , control inputs , system parameters , measured outputs and controlled outputs are , , , and , respectively , and , and are twice continuously differentiable functions for the system dynamics , measured outputs and controlled outputs , respectively . Note that and are functions of u , t and . For our purposes we will use the notation and . Sparse grids were implemented using the Sparse Grid Interpolation Toolbox for Matlab , version 5 . 1 . 1 [40] , which is available at http://www . ians . uni-stuttgart . de/spinterp/ . Our approach for approximating model dynamics using interval-based sparse grid interpolation is similar to that of [14] . The control inputs are assumed to lie within the bounded nu-dimensional space containing all feasible control input vectors , defined by ( 2 ) where and is the jth value of the input vector . For each model , nz output variables are selectively evaluated at points in the nu-dimensional input space and nt-dimensional time domain to form a series of grids of dimension nu ( see [14] for further details ) . On each grid , weighted Lagrange basis functions are combined at the support nodes to construct an input-domain interpolant with which the value of an output at a single time point can be estimated for any point in the input space . In this interval-based approach , interpolation between grids placed at various time points ensures a continuous trajectory over the prediction horizon for any point in the input space . To prevent excessive computational expense during grid construction , limits on absolute and relative error tolerances and allowable interpolation depth were specified ( 0 . 01 , 1% , and 6 , respectively ) . For problem set-up , we estimate weight maps on Ω for the prediction models using Akaike weights , which are based on the Akaike Information Criterion ( AIC ) . AIC provides a practical measure of the tradeoff between model fitness and complexity by estimating the theoretical Kullback-Leibler ( KL ) “distance” , or loss of information , between an approximating model and full reality [26] . Assuming normally distributed errors with constant variance and small sample sizes , the corrected-AIC ( AICc ) can be estimated as: ( 3 ) where is the total number of experimental data and is the number of sampled time points for the jth measured output . The second term , where denotes the number of uncertain parameters for the ith model , is the bias-correction factor for the AIC ( the factor 2 was introduced for historical reasons ) and the third term is an additional correction factor for small sample sizes . Note that the first term includes the squared residuals between experimental data ( ) and their ith model counterpart ( ) . Under realistic conditions , are sampled at discrete input quantities and time points and the number of data can vary between outputs . To generate a continuous approximation of over Ω , a piecewise linear interpolant is constructed using the Matlab function griddatan . That is , ( 4 ) where and denote the discrete experimental input and observation time spaces , respectively , and is an operator denoting ( nu+1 ) -dimensional linear interpolation between existing points over and by means of Delaunay triangulation . In regions not explicitly measured , interpolated data ( ) are used in place of when computing AICc values . It is important to note that only relative AICc values are meaningful due to the metric being a relative rather than absolute estimate of KL distance . It follows that the relative likelihoods of the models are given by the Akaike weights , ( 5 ) where . The weight is interpreted as the strength of evidence in support , or relative probability , of the ith model being the KL best model from the set of models given the supporting data [26] . The weight maps for Model Z , Model L and Model K based on the data described in Training Experiments are provided in Figure 3 in Text S1 . The multiple-model control strategy is built around the conventional MPC framework to reduce the computational complexity of the open-loop control problem . At each time interval , the controller surveys the possible trajectories stemming from the current state and selects the control input sequence over the control horizon ( Hu ) so that the predicted outputs track the desired trajectories over the finite prediction horizon ( Hp ) . The first control input of the selected sequence is used to update the states of the prediction models and then stored as one entry in the final control sequence U* so the procedure can repeat as the prediction horizon slides along for the remaining time intervals . The objective functions used to quantify controller performance penalizes the error between the predicted outputs for the ith model and the desired trajectories and a measure of the control effort over the prediction horizon starting at time as given by , where ( 6 ) The vectors and are the predicted and target outputs over Hp , respectively , are the discrete controller inputs over Hu . The proposed formulation assumes , j = 1 , … , nu , and m = Hu , … , Hp to reflect that manipulated variables are often applied as boluses in the considered biological context . The horizons Hu and Hp where each chosen to be one to prevent controllers from being overly conservative . Q and R are diagonal weighting matrices associated with the error and control effort , respectively , which were each chosen to be identity . Objective values are converted to log-space to compress the cost surfaces to facilitate optimization . Each objective is approximated using sparse grid interpolation ( similar to section Approximating Model Dynamics with Sparse Grids ) to form an input-domain interpolant with which the value of each objective can be estimated for any point in the input space . If the approximations are sufficiently accurate , no further evaluations of the objective function or underlying state space model are required since the interpolants are generated prior to optimization during each time interval . We define the multiobjective optimization problem at the kth time interval as follows ( with the standard Pareto interpretation of minimizers ) : ( 7 ) The variable denotes the objective function for the ith model defined by ( 6 ) and the design variable Uk is constrained to Ω defining biologically relevant limits . Herein we employ the normalized normal constraint method ( NNC ) for generating the Pareto solutions for its ability to generate a well-distributed set of global Pareto points ( [41] , refer to the original manuscript for further details ) . NNC provides a geometrically intuitive approach to multiobjective optimization that is illustrated in Figure 4 in Text S1 . It first builds a plane in the normalized objective space ( called the utopia hyperplane ) through all individual ( normalized ) minima , and second , generates equally distributed points in this plane by systematically varying weights for each objective . Then for each point , the corresponding solution on the Pareto front is found by minimizing the single ( normalized ) objective with added constraints . In addition to the original constraints , the feasible space is further restricted by adding nM −1 hyperplanes through that are each normal to the nM −1 utopia plane vectors . Successive optimization runs are performed for the remaining points in . By translating the constraining normal hyperplanes between runs , we can see that the corresponding solution set along the leading edge of the objective space is generated . Since some of these points may represent non-Pareto optimal or dominated solutions , the NNC algorithm is coupled with a Pareto filter to remove such points . The optimal control sequence for the kth time interval is selected from the set of Pareto solutions by ranking them using the objective ( 8 ) where and are vectors of objective function values and Akaike weights , respectively , corresponding to the control input vector . The models weights are adapted to accommodate the most relevant experimental data to ensure the best possible open-loop performance ( Figure 9 ) . At the first time interval , u is undetermined so the initial weights are computed considering the entire training data set . After the Pareto points are specified by solving ( 7 ) , they are ranked using ( 8 ) with the initial weights ( Figure 9A ) and the input vector corresponding to the best ranked point is taken as a temporary solution u1 ( Figure 9B ) . The weights are then recalibrated at the value u1 ( Figure 9C ) and ranked again with the best ranked input vector taken as the new temporary solution u2 and so on . Updates continue until the model weights or control inputs no longer change above a prescribed threshold or the maximum allowable updates is reached . If a limit cycle is detected , ( 8 ) is recomputed by averaging the models weights in the cycle as a tie-breaker . The final input sequence is used to update the prediction models and appended to the growing open-loop control sequence U* as the prediction horizon slides along . All statistical analysis was performed using SigmaStat v3 . 5 ( Systat Software , Inc ) . Time-course data are shown as mean ± standard error at each time point . Statistical differences between groups ( p≤0 . 05 ) are determined using one-way analysis of variance ( ANOVA ) followed by the Tukey multiple comparisons test . Target tracking performance values , as measured by squared error between target profiles and controlled plants , were log-transformed where appropriate to satisfy the normality and equal variance conditions for the ANOVA and Tukey tests .
|
Most cell behavior arises as a response to external forces . Signals from the extracellular environment are passed to the cell's nucleus through a complex network of interacting proteins . Perturbing these pathways can change the strength or outcome of the signals , which could be used to treat or prevent a pathological response . While manipulating these networks can be achieved using a variety of methods , the ability to do so predictably over time would provide an unprecedented level of control over cell behavior and could lead to new therapeutic design and research tools in medicine and systems biology . Hence , we propose a practical computational framework to aid in the design of experimental perturbations to force cell signaling dynamics to follow a predefined response . Our approach represents a novel merger of model-based control and information theory to blend the predictions from multiple mathematical models into a meaningful compromise solution . We verify through simulation and experimentation that this solution produces excellent agreement between the cell readouts and several predefined trajectories , even in the presence of significant modeling uncertainty and without measurement feedback . By combining elements of information and control theory , our approach will help advance the best practices in model-based control applications for medicine .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biotechnology",
"bioengineering",
"biomedical",
"engineering",
"systems",
"biology",
"computer",
"and",
"information",
"sciences",
"network",
"analysis",
"engineering",
"and",
"technology",
"biology",
"and",
"life",
"sciences",
"signaling",
"networks",
"computational",
"biology"
] |
2014
|
Multiple Model-Informed Open-Loop Control of Uncertain Intracellular Signaling Dynamics
|
The spatial arrangement of Ca2+ channels and vesicles remains unknown for most CNS synapses , despite of the crucial importance of this geometrical parameter for the Ca2+ control of transmitter release . At a large model synapse , the calyx of Held , transmitter release is controlled by several Ca2+ channels in a "domain overlap" mode , at least in young animals . To study the geometrical constraints of Ca2+ channel placement in domain overlap control of release , we used stochastic MCell modelling , at active zones for which the position of docked vesicles was derived from electron microscopy ( EM ) . We found that random placement of Ca2+ channels was unable to produce high slope values between release and presynaptic Ca2+ entry , a hallmark of domain overlap , and yielded excessively large release probabilities . The simple assumption that Ca2+ channels can be located anywhere at active zones , except below a critical distance of ~ 30 nm away from docked vesicles ( "exclusion zone" ) , rescued high slope values and low release probabilities . Alternatively , high slope values can also be obtained by placing all Ca2+ channels into a single supercluster , which however results in significantly higher heterogeneity of release probabilities . We also show experimentally that high slope values , and the sensitivity to the slow Ca2+ chelator EGTA-AM , are maintained with developmental maturation of the calyx synapse . Taken together , domain overlap control of release represents a highly organized active zone architecture in which Ca2+ channels must obey a certain distance to docked vesicles . Furthermore , domain overlap can be employed by near-mature , fast-releasing synapses .
Transmitter release at CNS synapses happens at active zones of sub-micrometer dimensions , which harbor docked vesicles and vesicle fusion proteins , as well as presynaptic scaffold proteins and voltage-gated Ca2+ channels [1] . The number , and distance of Ca2+ channels to readily-releasable vesicles are crucial for determining release probability , because the Ca2+ signal generated by a single open Ca2+ channel drops off steeply with distance [2–5] . There is , however , only sparse morphological information on the co-localization of individual vesicles and Ca2+ channels at synapses . For this reason , Ca2+ channel—vesicle coupling distances have often been inferred from functional measurements . One approach to functionally assess the number of Ca2+ channels controlling release utilizes the high intrinsic Ca2+ cooperativity of release [6–8] . In experiments in which the number of open Ca2+ channels is varied in presynaptic voltage-clamp experiments , the slope value in plots of transmitter release vs integral Ca2+ influx in double-logarithmic coordinates ( which we will call the "Ca2+ current release cooperativity" or simply "slope value" ) can inform about the number of Ca2+ channels involved in release control [9–13] . A large Ca2+ current release cooperativity close to the intrinsic Ca2+ cooperativity indicates that Ca2+ signals from individual channels mix to produce a graded Ca2+ signal at each docked vesicle . Conversely , a slope value close to one indicates that a few , or a single Ca2+ channel control the release of a given docked vesicle [9–13] . The diffusional distance between Ca2+ channels and vesicles can also be probed by the slowly binding Ca2+ chelator EGTA , which affects Ca2+ signals only at some distance from the Ca2+ source [14–17] . Experiments with these functional approaches have shown different coupling regimes at different CNS synapses . At the calyx of Held synapse which is amenable to direct presynaptic patch-clamp experiments , release is sensitive to EGTA at low millimolar concentrations , and high slope values are observed , indicating domain overlap control of release [18 , 19] . However , Ca2+ channel—release coupling was later shown to be developmentally regulated , such that brief presynaptic Ca2+ currents become more effective in causing transmitter release [12 , 20–22] . Electron microscopy ( EM ) has shown that active zones of various types of CNS synapses are small ( ~ 0 . 1 μm2 ) , and contain a high density of docked vesicles [23–26] . The development of EM freeze fracture replica labeling techniques has enabled the visualization of ion channels and receptors in en-face views at nanometer resolution [27] . Using this method , multiple stripe—like clusters of P/Q—type Ca2+ channels ( CaV2 . 1 subunits ) have been observed at hippocampal synapses [28] . However , with freeze fracture replica labeling , the position of docked vesicles cannot be visualized , and antibody-labelled particles do not necessarily overlap with the entire population of functionally active Ca2+ channels . Therefore , despite important advances in ultrastructural methods , the functionally relevant positions of Ca2+ channels and vesicles on the nanometer scale are still uncertain . Here , we use stochastic MCell modelling of Ca2+ influx through individual Ca2+ channels at EM—reconstructed active zones , to explore how Ca2+ channels might be positioned to produce domain overlap control of release . The position of docked vesicles was fixed by previous EM reconstructions of single active zones [29] . Other model parameters , like presynaptic Ca2+ channel gating and—conductance [18 , 19 , 30–32] , Ca2+ buffering [33–35] , and the intracellular Ca2+—sensitivity of vesicle fusion [7 , 8 , 22 , 36–39] were constrained by previous measurements at the calyx of Held and at other large model synapses . We find that random placement of Ca2+ channels leads to an excessively high release probability and low slope values . In order to enable domain overlap control of release , we rather need to assume that Ca2+ channels are kept at some distance from vesicles . We also show experimentally that the Ca2+ current—release cooperativity stays high after the onset of hearing in mice , suggesting that despite a characteristic developmental tightening in the Ca2+ channel vesicle co-localization [12 , 20–22] , fast release in more mature calyx synapses continues to be controlled by several Ca2+ channels .
We wished to explore the possible placement of Ca2+ channels compatible with domain overlap control of release . For this , it is important to know the distribution , and density of docked vesicles typically observed at active zones . To determine these parameters , we analyzed a sample of n = 15 reconstructed active zones of calyx synapses of a P11 wild-type mouse [29] ( Fig 1A ) . In this sample , the average surface area of active zones was 0 . 07 ± 0 . 03 μm2 , with an average density of docked vesicles of 110 ± 40 ves / μm2 ( n = 15 ) . Across individual active zones , the number of docked vesicles and active zone surface were correlated ( r = 0 . 74; Fig 1B ) , indicating that large active zones tend to harbor more docked vesicles [20 , 23] . Visual inspection of the active zone maps with their docked vesicles ( Fig 1A ) suggests that some active zones show sub-areas where docked vesicles are sparse . We will show below that in domain overlap control scenarios , such areas might be preferentially occupied by Ca2+ channels . To investigate whether such void spaces could arise randomly , we compared experimentally observed vesicle distributions with the same parameters derived from random x-y placements of vesicles at a given active zone ( Fig 1C , 1D , 1E , 1F , 1G and 1H ) . Analyzing nearest neighbor distances , and the largest hole radii capable to fill empty spaces within active zones suggested that the observed vesicle positions were close to the random case ( Fig 1C and 1D , Fig 1E and 1F , respectively ) . Similarly , calculating the summed area extending 30 nm away from the edge of docked vesicles resulted in similar areas for randomly distributed vesicles , and for the docked vesicles found in the data set ( Fig 1G and 1H ) . Finally , we used Ripley’s K function test , a metric for detecting deviations from spatial homogeneity [40] ( see S1 Text ) . It gave values within the 99% confidence interval for all distances tested ( Fig 1I ) , indicating no significant vesicle clustering in the reconstructed active zones . Thus , there was overall no strong vesicle clustering in the reconstructed active zones . We next investigated whether random placement of Ca2+ channels in a realistic active zone with several docked vesicles can predict release control by several channels . We developed an MCell based model which incorporated individual Ca2+ channels with realistic gating and permeation properties [19 , 31 , 41] ( Fig 2A ) . The model contained several intracellular Ca2+ buffers , and each vesicle had a Ca2+ sensor for vesicle fusion according to a highly non-linear five-site ion binding mechanism [7 , 8] ( Fig 2A ) . The parameters of the Ca2+ sensor model were determined in several independent Ca2+ uncaging studies [7 , 8 , 22 , 36 , 37] , and were fixed to the values reported in one previous study [37] . Several other parameters , including the width of the presynaptic AP ( Fig 2B ) were also tightly constrained by previous biophysical experiments at the calyx of Held synapse ( see Materials and Methods , and S2 Text ) . The Ca2+ channel density was set at a value of 280 /μm2 ( ref . [28 , 41] ) . For most simulations , we used single representative active zones drawn from the sample of reconstructed active zones . We started with active zone #5 ( see Fig 1A , star symbol ) which has an area of 0 . 05 μm2 ( slightly smaller than the mean ) , and a vesicle docking density of 120 ves / μm2 , representative of the sample ( Fig 1B , vertical arrow ) . We placed n = 14 Ca2+ channels randomly in this active zone ( Fig 2C ) , corresponding to a Ca2+ channel density of 280 /μm2 . We simulated a "Ca2+ current—release cooperativity" experiment by driving release with APs of different widths ( Fig 2D ) , and plotted the resulting average vesicle release probability , pves , as a function of Ca2+ entry in double-logarithmic coordinates ( Fig 2E ) . This yielded a slope value of only 1 . 3 ± 0 . 2 ( n = 5 independent random spatial samples ) , far smaller than a value of ~ 3–3 . 5 expected for domain overlap control of release ( see below ) . In addition , the average pves in response to the standard AP was 0 . 66 ± 0 . 18 ( Fig 2F ) , much higher than experimental estimates ( ~ 0 . 1; ref . [8] ) . As a control , we placed n = 16 Ca2+ channels in a tight cluster at 100 nm from a single vesicle . This resulted in a slope value of 2 . 8 ( Fig 2G ) , close to the maximal value of 3 . 5 which can be achieved for kinetic models with 5 Ca2+ binding sites [11 , 42] . Thus , these simulations suggest that a random placement of Ca2+ channels is unable to produce high Ca2+ current—release cooperativities close to the experimentally observed values . It has been shown that a large number of Ca2+ channels ( > 10–20 ) is needed to produce high Ca2+ current—release cooperativity [11] . If such a large number of channels controls release of a given vesicle with roughly equal strength , then it seems plausible to assume that the channels must be located at some distance to vesicles . A simple implementation of such a rule in a field of randomly placed docked vesicles at relatively high density ( Fig 1 ) is to assume an exclusion zone around each docked vesicle; a zone into which Ca2+ channels cannot enter . We implemented this exclusion zone rule by assuming that Ca2+ channels can be located anywhere at the active zone , but not below a distance of 30 nm ( Fig 3A ) . When the exclusion zone model was driven with a standard AP , brief local [Ca2+]i transients with amplitudes of 10–20 μM resulted at individual docking sites ( Fig 3B ) , similar to results based on back-calculation from Ca2+ uncaging data [7 , 8] . The vesicular release probabilities pves were variable across individual docked vesicles , with a realistic average value of ~ 0 . 1 ( Fig 3C , upper panel ) . The distribution of the number of released vesicles showed that most trials ( ~ 60% ) led to release failures at an active zone ( Fig 3C ) , again in agreement with previous estimates [43] . We then simulated the Ca2+ current—release cooperativity experiment using different AP widths , and found that the exclusion zone model with a distance of 30 nm produced high values of Ca2+ current—release cooperativity ( n = 2 . 8 ± 0 . 2; n = 4 independent spatial samples; Fig 3D ) . We attribute this to the fact that multiple Ca2+ channels ( n = 6–14 ) should influence the release of a given vesicle . Thus , assuming that Ca2+ channels can be placed anywhere except at very close distances to docked vesicles ( ~ 30 nm ) predicts a high Ca2+ current—release cooperativity , a hallmark of release control by several Ca2+ channels [9–13 , 22] . We also simulated the sensitivity of release to the slow Ca2+ buffer EGTA in the exclusion zone model ( Fig 3E ) . For this , release in response to single standard APs was modelled in the presence of EGTA , added to the Ca2+ buffers present in the standard model ( see Materials and Methods ) . These simulations showed that 0 . 5 mM EGTA suppressed release by ~ half , with an overall biphasic concentration—dependence of EGTA ( Fig 3E ) , similar to what was found at the calyx of Held synapse and at cortical synapses of young rats [15 , 18] . Finally , we tested the influence of the exclusion zone width on release probability pves , and found an inverse relation of pves with exclusion zone distance , with a value for pves of 0 . 096 reached at 30 nm ( Fig 3F , horizontal line ) . We conclude that the exclusion zone model can reproduce many features of domain overlap control of release reported in previous functional studies . In a previous model , release control by several Ca2+ channels was achieved by placing all channels in a single cluster [11] , an arrangement which we will call "supercluster" . We next tested this arrangement , by first using the docked vesicle distribution of active zone #5 . The distance of the supercluster to the nearest vesicle was 30 nm ( Fig 4A ) , which yielded an acceptable average pves of 0 . 08 ( Fig 4B ) and a high Ca2+ current—release cooperativity ( 2 . 8 ± 0 . 1; n = 4 independent spatial samples; Fig 4C ) . A characteristic property of the supercluster arrangement is , however , a very large heterogeneity between pves values for individual vesicles ( range , 0 . 005 to 0 . 24; c . v . = 1 . 1; Fig 4B ) , because far—away vesicles experienced a much lower Ca2+ signal . We next compared the exclusion zone—and supercluster channel arrangements at another example active zone with larger size ( active zone # 11; see above Fig 1A , triangle symbol ) . This active zone also had a representative density of docked vesicles ( see Fig 1B , horizontal arrow ) . In this active zone , 25 Ca2+ channels were used to again yield a Ca2+ channel density of 280 /μm2 . Assuming an exclusion zone with a distance of 30 nm ( Fig 5A ) , we found a somewhat higher average pves ( 0 . 17; Fig 5B ) as compared to the smaller active zone # 5 ( see above; Fig 3; pves ~ 0 . 1 ) . This finding , which is compatible with recent experimental findings [41] , might suggest that at comparable Ca2+ channel densities , large active zones use Ca2+ more efficiently because of smaller diffusional loss of Ca2+ away from the active zone . However , further modelling studies are needed to investigate this mechanism in more detail . In the simulations of Fig 5A–5C using an exclusion zone distance of 30 nm , the individual pves values ranged from 0 . 02 to 0 . 36 with a c . v . of 0 . 7 ( Fig 5B ) , and the Ca2+ current—release cooperativity was 2 . 7 . We next modelled Ca2+—release control at this large active zone assuming that all Ca2+ channels ( n = 25 ) are located within a single supercluster . We placed the supercluster into a void space in the bottom half of the active zone , and varied its exact position until we achieved a reasonable average pves of 0 . 1 ( Fig 5D and 5E ) . The supercluster model again yielded a high slope value ( 3 . 3; Fig 5F ) . However , the heterogeneity of pves was very large , with a c . v . value of 1 . 4 . Indeed , visual inspection of this relatively large active zone shows many vesicles that are far away from the cluster ( e . g . vesicles 1 , 2 , 7 and 8; Fig 5D ) . Correspondingly , these vesicles showed extremely low pves values ( 0 , 0 . 0001 , 0 . 0129 , and 0 . 012 , respectively ) . As expected , there was an inverse relation between pves and distance of the docked vesicle from the Ca2+ channel supercluster ( Fig 5G ) . We note , therefore , that the supercluster model produces a high heterogeneity of pves , especially in large active zones . Since the two arrangements of Ca2+ channels predict distinct heterogeneities of pves , experimental data on the heterogeneity of pves within individual active zones would be necessary to verify the supercluster versus the exclusion zone arrangement . Unfortunately , the distribution of pves within individual active zones following AP stimulation has , to our knowledge , not been estimated experimentally; a previous theoretical study has highlighted the role of pves heterogeneity for synaptic depression [44] . On the other hand , studies at the calyx using prolonged presynaptic depolarizations have revealed that release occurs in distinct kinetic phases , with time constants of ~ 2 ms and ~ 20 ms called fast- and slowly releasable pool ( FRP and SRP , respectively ) [45–47] . Some studies have interpreted the different release kinetics as a result of different distances of vesicles to Ca2+ channels ( the "positional" model; [46 , 48 , 49] ) , but other studies concluded that SRP release results from intrinsically slower release [47 , 50] ( see Discussion ) . To further validate the exclusion zone model and the supercluster model , we modelled release in response to a prolonged presynaptic depolarization to 0 mV , using the large active zone with either the supercluster—or the exclusion zone arrangement of Ca2+ channels ( Fig 6 ) . To our initial surprise , both models produced fast release , without obvious SRP component ( Fig 6A , 6B , and 6C ) . Exponential fits of the average cumulative release rates yielded time constants of 1 . 52 ms and of 0 . 78 ms for the supercluster model and the exclusion zone model , respectively ( Fig 6B , red fit lines ) . In each case , a double-exponential function did not improve the fit ( not shown ) . When we plotted release from the individual sites separately , it became apparent that the spread of release delays and 20–80% release times was larger for the supercluster model ( Fig 6C and 6D; left ) than for the exclusion zone model ( Fig 6C and 6D; right ) . This is expected , since the heterogeneity of the Ca2+ channel—vesicle distances is larger in the supercluster model than in the exclusion zone arrangement ( see above; Fig 5 ) . The faster average release time constant in the exclusion zone model as compared to the supercluster model ( 0 . 8 ms versus 1 . 5 ms ) corresponds to the lower degree of pves heterogeneity in the exclusion zone arrangement . Indeed , far-away vesicles in the supercluster arrangement have release probabilities near zero during AP stimulation ( e . g . vesicles #1 and # 2; see above Fig 5D and 5G ) , and these vesicles are released with an additional delay of ~ 2 ms upon prolonged depolarization to 0 mV ( Fig 6C and 6D ) . Taken together , simulating the response to prolonged presynaptic depolarizations does not give further clues to distinguish between the two Ca2+ channel arrangements ( exclusion zone versus supercluster ) , since both models were unable to predict SRP release with time constants of ~ 20 ms , as observed experimentally in immature calyx of Held synapses [45 , 47] . We conclude that to explain SRP release , one would either need to postulate Ca2+ channel—vesicle distances longer than 400 nm . Alternatively , SRP release could be caused by intrinsically slower release kinetics as suggested by previous Ca2+ uncaging experiments [22 , 47 , 51] . We next wanted to investigate whether transmitter release in more mature calyx of Held synapses might still be controlled in a domain overlap fashion . Previous work has shown a developmental tightening in the spatial co-localization between Ca2+ channels and docked vesicles , as demonstrated by an increased efficiency of small Ca2+ charges in inducing release [20–22] , and a reduced efficiency of EGTA in suppressing release and reduced Ca2+ current—cooperativity values [12] . The question therefore arises whether release in more mature calyx synapses is converted into a single channel control mode or else , whether several Ca2+ channels still control release of a given vesicle , albeit at somewhat shorter distance . To investigate this issue , we made measurements of Ca2+ current release cooperativity at postnatal day ( P ) P8–P11 , and in P15–P16 old mice , the oldest age group amenable to paired recordings under our conditions . We varied the number of open Ca2+ channels in Ca2+ "tail" current experiments with voltage steps of different lengths , and measured the corresponding EPSC amplitudes as a proxy of release ( Fig 7A ) . Plotting the EPSC amplitudes versus the integral Ca2+ charge in double logarithmic coordinates then allowed us to measure the Ca2+ current—release cooperativity ( slope value ) , and the Ca2+ charge ( QCa ) necessary to evoke an EPSC of 2 nA . The slope values were 4 . 75 ± 0 . 60 ( n = 6 pair recordings ) for mice at P8–P11 , and 3 . 10 ± 0 . 11 ( n = 5 pair recordings ) for P15–P16 mice ( Fig 7C ) . Thus , although there was a significant decrease of the slope value with development ( p < 0 . 05 ) , the slope of ~ 3 was still significantly higher than 1 ( p < 0 . 001; one-sample two-tailed t-test ) , the value expected for the extreme case of single-channel vesicle release control [9 , 13] . In addition , we observed that the Ca2+ charge ( QCa ) necessary to evoke an interpolated EPSC of 2 nA was shifted to smaller values ( Fig 7B and 7C; p < 0 . 05 ) , similarly as reported before for the rat calyx of Held [22] . Both findings indicate a tighter coupling in the more mature mice [12 , 20 , 21] . Despite the tighter coupling , the Ca2+ current—release cooperativity was still high ( ~ 3 ) , which indicates that several Ca2+ channels control the release of a given vesicle also in more mature synapses . Finally , we wished to investigate the EGTA sensitivity of transmitter release , and its dependence on the developmental state of the calyx of Held synapse . We used the membrane-permeable EGTA analog , EGTA-AM , which allowed us to investigate synapses from adult mice ( P60–P100 ) not amenable to paired recordings . We first investigated immature synapses ( P8–P11 ) , and found a gradual suppression of EPSC amplitudes upon acute application of 200 μM EGTA-AM ( Fig 8A1 and 8A2 ) . After applications times of 15–20 minutes , EPSC amplitudes were suppressed to 16 ± 7% of their control value ( n = 2 ) . Using lower concentrations of EGTA-AM ( 100 , and 50 μM ) , the suppression of EPSCs was smaller , which demonstrates a concentration-dependent effect of EGTA-AM with an apparent half-maximal effective concentration of 87 μM ( Fig 8B and 8C ) . These results agree with earlier findings showing that low millimolar concentrations of EGTA , or 200 μM EGTA-AM suppress transmitter release at the immature rat [18 , 52] and mouse [12] calyx of Held . We then investigated adult synapses from P60–P100 mice using the same experimental approach . Acute application of 200 μM EGTA-AM again strongly suppressed EPSC amplitudes , with 20 . 1 ± 5% of the control EPSC amplitude remaining ( Fig 8B1 and 8B2 ) . This value was slightly larger than the remaining fraction of the EPSC at the young calyx synapse , but the difference did not reach statistical significance ( p = 0 . 6 ) . The dose-dependency of the steady-state block of EPSCs by EGTA-AM indicated a half-maximal effective concentration of 92 μM ( Fig 8F ) , similar to the value found at young calyx synapses . These results show that overall , ultrafast transmitter release at the calyx synapses remains sensitive to EGTA-AM , although its efficiency might be slightly smaller in mature mice , in agreement with recent findings [53] . Therefore , the measurements of Ca2+ current release cooperativity ( Fig 7 ) and of the EGTA-AM sensitivity ( Fig 8 ) suggest that despite a developmental tightening of the Ca2+ channel—release coupling ( Fig 7; [12 , 20–22] , domain overlap remains relevant for the mature calyx of Held synapse . This also suggests that while some fast-releasing synapses employ single channel release control [54–56] , other fast-releasing synapses like the calyx of Held can rely on release control by several Ca2+ channels .
The supercluster model and the exclusion zone model predicted values of Ca2+ current—release cooperativity in the range of 2 . 8–3 . 3 ( Figs 3–5 ) . The maximal value of this parameter predicted by Ca2+ sensor models with 5 Ca2+ ion binding sites [7 , 8] is ~ 3 . 5 , as shown by previous simulations with AP-like stimuli [11 , 42] . Therefore , both models preserve the high intrinsic Ca2+ cooperativity nearly completely when the number of open Ca2+ channels is varied , as expected for domain overlap control of release [9–13] . Since the Ca2+ current—release cooperativity was therefore not a distinguishing feature between the two Ca2+ channel arrangements , our further analysis concentrated on the heterogeneity of pves between vesicles . Our model allowed us to track the individual release probabilities of multiple docked vesicles , an approach which received little attention in previous modeling studies . We assumed that all docked vesicles belonged to the readily-releasable pool consistent with experimental considerations ( see discussion in [29] ) . Furthermore , we assumed that no mechanism exists which would limit release to single fusion events at a given active zone ( thus , no "lateral inhibition" mechanism as proposed in refs . [58 , 59] ) . For simplicity , we also assumed that all vesicles have the same Ca2+ sensitivity; thus , our model would be unable to describe the biphasic release response upon Ca2+ uncaging [47] . In the exclusion zone model with a distance of 30 nm which yielded a realistic average pves of ~ 0 . 1 , we observed mostly single or no release events at the small active zone , whereas multiple fusion events ( "multivesicular release" ) were rare ( Fig 3C , bottom ) . Therefore , we think that no special depression mechanism needs to be postulated to explain that under normal conditions , there is a low degree of multivesicular release [60] . We found that the two Ca2+ channel arrangements could mainly be distinguished by the degree of heterogeneity of release probability . The large heterogeneity of pves produced by the supercluster model is expected , because placing all Ca2+ channels in a small spot inevitably produces strong gradients of local [Ca2+]i across an active zone , since active zones can extend over a few hundred nanometers . Furthermore , docked vesicles likely create a hindrance for Ca2+ diffusion towards the more distant vesicles ( Figs 4A and 5D ) . Our modelling approach included such a "shading" effect of Ca2+ diffusion by docked vesicles ( see Materials and Methods ) . To further distinguish between the two Ca2+ channel arrangements , more insights into the heterogeneity of pves amongst vesicles within individual active zones must be obtained . Earlier work using optical methods has shown a substantial variability of release probabilities [61] , but this heterogeneity likely represents differences across active zones . At the calyx synapse , prolonged presynaptic depolarizations have been used to estimate release probability heterogeneity [45–47] . Some studies concluded that SRP release with time constants of ~ 20 ms is caused by longer distances of SRP vesicles to Ca2+ channels ( the "positional" model; ref . [45 , 46] ) . We modelled the response of the supercluster- and exclusion zone models to prolonged voltage-clamp depolarizations ( Fig 6 ) . As expected , the supercluster arrangement produced a larger degree of heterogeneity in the release times than the exclusion zone model . However , even in the supercluster arrangement at the large active zone , the spread in release delays was in the range of only 2–3 ms , too small to explain SRP release with time constants of ~ 20 ms [45–47] . We conclude that the separation of FRP and SRP release cannot be used to inform about the validity of the supercluster versus the exclusion zone arrangement . The inability of realistic active zone models to predict slow SRP release with time constants ~ 20 ms under positional assumptions is consistent with the earlier view that the slow release phase is caused by an intrinsic mechanism [47] . To distinguish between the supercluster and exclusion zone arrangements , it will also be necessary to obtain more ultrastructural data on the exact localization of Ca2+ channels at active zones of CNS synapses . At hippocampal synapses , freeze-fracture replica labeling has shown several loosely organized , stripe-like clusters of particles detected by an antibody against CaV2 . 1 ( P/Q-type Ca2+ channel ) ( see Figs 6–8 in ref . [28] ) . The hypothetical distributions of Ca2+ channels produced by the exclusion zone ( Fig 3A and 5A ) could correspond to such stripe-like clusters of CaV2 . 1 particles observed in the previous study [28] . A recent study obtained evidence for the localization of P/Q-type Ca2+ channels from freeze-fracture labeling at the calyx of Held synapse . Combined with functional findings on the inhibition of release by EGTA , a release model for the calyx synapse was derived [53] . The study found clusters of Ca2+ channels; a cluster was defined if particles were located within 100 nm distance . Example images show that CaV2 . 1 particles can be located over quite long distances ( ~ 100–300nm; see e . g . Fig 1C and 6H of ref . [53] ) . Modelling then suggested that docked vesicles must be localized at a distance of 20–30 nm from a somewhat arbitrary Ca2+ channel cluster of ~ 100 nm diameter , to explain the EGTA-sensitivity of release [53] . Our finding of an exclusion zone distance of 30 nm is consistent with this previous distance estimate . Nevertheless , such distance estimates from modelling are subject to parameters which are only incompletely known , including the Ca2+ channel density [41] and the kinetics , concentration and mobility of endogenous Ca2+ buffers ( see S2 Text ) , and should therefore be taken with care . Based on our computational model , and on the placement of Ca2+ channels at active zones shown so far in EM studies [28 , 53] , we find it unlikely that all Ca2+ channels should be localized in a single cluster ( "supercluster" ) within an active zone , as predicted by a previous modelling study [11] . Rather , we regard it as more likely that Ca2+ channels are placed either randomly , or else in several sub-clusters in the void spaces in-between docked vesicles , but with a certain minimal distance to docked vesicles . Ion channels are mobile in membranes as has been found for postsynaptic glutamate receptors [62]; however , the mobile properties of presynaptic Ca2+ channels are less well known [63] . In addition , docked vesicles might also undergo lateral movement at synapses . An exclusion zone of several tens of nanometer around docked vesicles might be caused by a hindered mobility of Ca2+ channels close to the docked vesicle , maybe caused by the multitude of vesicle-near proteins which make up the docking complex and the vesicle fusion machinery [64 , 65] . In addition , specific proteins likely regulate the minimal distance between vesicles and Ca2+ channels . At the neuromuscular synapse , rib-like proteins have been found which likely hold Ca2+ channels at a certain distance from vesicles [66] , and filamentous structures playing similar roles might also be present at CNS synapses [25] . One candidate for such a structure at CNS active zones is the filamentous protein Septin-5 . When this protein was genetically inactivated , an increased release probability and a tighter spatial coupling resulted [67] , which suggested that Septin-5 normally keeps Ca2+ channels at a certain distance from docked vesicles . During developmental maturation , the Ca2+ channel—vesicle coupling distance becomes tighter ( Fig 7 ) [12 , 20–22] , which probably also requires protein re-arrangements to specifically regulate the minimal Ca2+ channel—vesicle distance with a precision of ~ 10 nm or less . Scaffold proteins at the active zone which regulate Ca2+ channel vesicle coupling distance should be investigated in future studies .
Procedures of mouse breeding , handling , and the sacrification of mice before slice preparation were approved by the Veterinary Office of the Canton of Vaud , Switzerland ( authorization no . 2063 . 3 ) . The sample of the location of docked vesicles at n = 15 active zones is the same as that previously published by [29] ( n = 15 active zones from a P11 control mouse ) . In brief , serial transmission EM images were taken from 10–20 subsequent thin sections ( 50 nm ) , which were made following standard fixation and resin embedding procedures . The images were aligned , and pre- and postsynaptic membranes and vesicles were drawn in each section using the TrackEM2 program [68] of the FIJI software [69] , and the extent of the PSD was annotated . Only active zones that were completely contained in the series were maintained in the final data set . Each active zone was 3D- reconstructed by the FIJI software , and the nearest distance between vesicle membrane and active zone membrane was measured in 3D , and plotted . Vesicles at 10 nm or less showed a distinct peak in these distance distributions [29] , and were regarded as the pool of "docked" vesicles . In order to map the x-y location of docked vesicles , we unfurled the curved surface of the active zone with its attached docked vesicles , using a Matlab program . The shortest distance from the vesicle membrane to the active zone membrane was calculated in the 3D model , and the corresponding point in the flat surface was taken as the remapped vesicle position . We used the Monte Carlo simulator MCell [70] to track the fate of individual Ca2+ ions that entered through various individual Ca2+ channels . The simulator was run on a Blue Gene/Q system , and the MCell based model is available on ModelDB ( http://senselab . med . yale . edu/modeldb/ ) . In general , 2000–4000 repetitions were run to determine an average "release response" for a given simulated stimulus , which is the release rate for all individual vesicles in response to one stimulus; see e . g . Fig 3B , bottom ) . This simulation was then repeated for each AP width ( usually , n = 9 different AP widths ) , to construct a given dose-response curve of release rates versus Ca2+ influx ( e . g . Fig 2E , bottom ) . The dose-response curve was fitted with a line in double-logarithmic coordinates to obtain the "Ca2+ current—release cooperativity" ( see main text for definition ) . The highest datapoints were usually excluded from the fit range , since they deviated from the steepest slopes , as expected from the beginning saturation of the Ca2+ sensor for vesicle fusion > 10 μM [Ca2+]i . In addition , the lowermost data points at very low release probabilities were often inaccurate despite the number of individual repetitions ( 2000–4000 , see above ) ; in this case , the lowermost data points were not included in the fit range . In order to evaluate the statistical robustness of the slope estimates , the entire simulation was repeated 3–5 times , using different random seeds for the spatial arrangements of Ca2+ channels; these repetitions are referred to as "independent spatial seeds" in the main text . The simulated compartment measured 600 nm by 600 nm and had a height of 1200 nm . Ca2+ was reflected off the walls; the size of the compartment was chosen based on published nearest-neighbor distances of active zones at the immature rat calyx of Held [71] . Models contained Ca2+ channels , endogenous Ca2+ buffers and Ca2+ sensors for release . Docked vesicles were present as spheres with 45 nm diameter inaccessible to Ca2+ ions . Ca2+ channels opened and closed stochastically during the AP according to a two-state model ( Fig 2A ) . Ca2+ channels had a realistic single channel current of 0 . 12 pA at 0 mV [31 , 41] . The standard AP had a half-width of 0 . 49 ms; the peak open probability of Ca2+ channels was ~ 0 . 6 ( Fig 2D ) . Ca2+ diffusion away from individual channels was modelled stochastically with MCell . Two mobile Ca2+ buffers , ATP and Parvalbumin , and an immobile ( fixed ) buffer were present in the simulations , with parameters as given in S1 Table . The Ca2+ sensor was a kinetic model of Ca2+ binding and vesicle fusion , with parameters which reflect extensive previous Ca2+ uncaging studies at the calyx of Held synapse . The specific kinetic model was a 5-site model ( Fig 2A ) , with parameters as reported in ref . [37] . See S2 Text for further justification of the model parameters . Transverse brain slices on the level of the medial nucleus of the trapezoid body were prepared from C57Bl6 mice of three different age groups: P8–P11; P15–P16 , and P60–P100 ( "adult" ) . Mice were killed by decapitation , sometimes following a brief isoflurane anesthesia in the oldest age group , in a protocol approved by the Veterinary office of the Canton of Vaud , Switzerland . Transverse brain slices of 200 μm thickness were made with a vibratome ( VT1000 or VT1200; Leica Microsystems , Wetzlar , Germany ) , and stored in a submerged keeping chamber containing a standard bicarbonate buffered solution ( for composition , see below ) , bubbled with 95% O2 and 5% CO2 . For paired pre- and postsynaptic recordings ( Fig 6 ) , the extracellular solution contained 50 μM D-APV , 100 μM cyclothiazide ( CTZ ) , 1 μM tetrodotoxin ( TTX; all from BIOTREND , Wangen , Switzerland ) and 10 mM TEA-Cl ( Sigma Aldrich/Fluka , Buchs , Switzerland ) , to suppress unwanted current components and AMPA-receptor desensitization ( CTZ ) . Series resistances ( Rs ) were in the range of 3–10 MΩ and 5–25 MΩ for post- and presynaptic recordings , and were compensated by up to 80 and 50% for post- and presynaptic recordings respectively , using the patch-clamp amplifier ( EPC10/2; HEKA Electronik , Lambrecht/Pfalz , Germany ) . For postsynaptic recordings , the remaining Rs error was compensated offline . Brief steps to +60 mV of various lengths were applied from a holding potential of -70 mV , resulting in the recruitment of an increasing number of Ca2+ channels giving rise to Ca2+ "tail" currents during the repolarization phase ( Fig 6A ) . In fiber stimulation experiments ( Fig 8 ) , afferent fibers were stimulated by a custom-made bipolar stimulation electrode placed close to the midline of the slice , using 0 . 2 ms long pulses of 1–10 V amplitude from an isolated stimulation unit ( ISO-STIM01D-100 , NPI electronic , Tamm , Germany ) . EGTA-AM ( Life Technologies , Zug , Switzerland ) was applied by bath perfusion , using stock solutions of 200 mM ( in DMSO , Sigma ) , at the indicated final concentrations ( 50 , 100 or 200 μM ) . Stocks of EGTA-AM solutions in DMSO were kept desiccated at–20°C for a maximum of 3 months . The extracellular solution contained ( in mM ) : 125 NaCl , 2 . 5 KCl , 25 NaHCO3 , 1 . 25 NaH2PO4 , 25 glucose , 1 MgCl2 , 2 CaCl2 , 0 . 4 ascorbic acid , 3 myo-inositol and 2 Na-pyruvate , continuously bubbled with 95% O2/5% CO2 ( pH 7 . 4 ) . The intracellular ( pipette ) solution for pre- and postsynaptic recordings contained ( in mM ) : 130 Cs-gluconate , 20 TEA-Cl , 10 HEPES , 5 Na2-phosphocreatine , 4 MgATP , 0 . 3 Na2GTP ( pH 7 . 2 adjusted with CsOH ) . For pre- or postsynaptic recordings , this solution was supplemented with 0 . 1 or 5 mM EGTA , respectively . All chemicals were from Sigma Aldrich/Fluka ( Buchs , Switzerland ) unless indicated .
|
Ca2+ channels provide the rise in intracellular Ca2+ concentration necessary to initiate the membrane fusion of transmitter—filled vesicles at synapses . Because Ca2+ diffuses away from Ca2+ channels , the distance between Ca2+ channels and vesicles on the range of tens of nanometers is a crucial determinant of the vesicle fusion probability . However , there is still little experimental evidence on how Ca2+ channels and vesicles co-localize in the nanospace of a single synapse . We show by computational modelling that the channels should be located at some distance to vesicles ( ~ 30 nm ) , to allow for release control by several channels , a release mechanism found at many synapses . In realistic synapses with a high density of docked vesicles , this translates into a likely localization of Ca2+ channels at membrane sites not occupied by docked vesicles . Thus , we present a computational model of how Ca2+ channels can be localized in an active zone with several docked vesicles , to enable control of release by several Ca2+ channels .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
An Exclusion Zone for Ca2+ Channels around Docked Vesicles Explains Release Control by Multiple Channels at a CNS Synapse
|
Hematogenous dissemination is important for infection by many bacterial pathogens , but is poorly understood because of the inability to directly observe this process in living hosts at the single cell level . All disseminating pathogens must tether to the host endothelium despite significant shear forces caused by blood flow . However , the molecules that mediate tethering interactions have not been identified for any bacterial pathogen except E . coli , which tethers to host cells via a specialized pillus structure that is not found in many pathogens . Furthermore , the mechanisms underlying tethering have never been examined in living hosts . We recently engineered a fluorescent strain of Borrelia burgdorferi , the Lyme disease pathogen , and visualized its dissemination from the microvasculature of living mice using intravital microscopy . We found that dissemination was a multistage process that included tethering , dragging , stationary adhesion and extravasation . In the study described here , we used quantitative real-time intravital microscopy to investigate the mechanistic features of the vascular interaction stage of B . burgdorferi dissemination . We found that tethering and dragging interactions were mechanistically distinct from stationary adhesion , and constituted the rate-limiting initiation step of microvascular interactions . Surprisingly , initiation was mediated by host Fn and GAGs , and the Fn- and GAG-interacting B . burgdorferi protein BBK32 . Initiation was also strongly inhibited by the low molecular weight clinical heparin dalteparin . These findings indicate that the initiation of spirochete microvascular interactions is dependent on host ligands known to interact in vitro with numerous other bacterial pathogens . This conclusion raises the intriguing possibility that fibronectin and GAG interactions might be a general feature of hematogenous dissemination by other pathogens .
Hematogenous dissemination of pathogenic organisms is an important feature of disease progression . However , dissemination is poorly understood , in large part because of the difficulty in studying this process directly in living organisms under the shear stress conditions that characterize the host vasculature . One such disseminating pathogen is the spirochete Borrelia burgdorferi , a primarily extracellular bacterium causing Lyme disease , also referred to as Lyme borreliosis [1] . Pathogenic spirochetes cause a number of emerging and re-emerging diseases , including syphilis , leptospirosis , relapsing fever and Lyme disease [2]–[5] . B . burgdorferi is transmitted to the dermis of vertebrate hosts during the blood meal of Ixodes ticks , and subsequently disseminates to other tissues and organs during the hematogenous phase of infection [1] . B . burgdorferi and other spirochetes interact with endothelial cells under static conditions in vitro [6]–[8] . However , until recently , spirochete-vascular interactions have never been directly examined in the host itself , or under the fluid shear forces that are present at dissemination sites [9] . To facilitate direct study of hematogenous dissemination we recently generated a fluorescent infectious strain of B . burgdorferi , and used intravital microscopy ( IVM ) to directly visualize its interaction with and extravasation from the microvasculature of living murine hosts ( as summarized in Fig . 1A–C ) [9] . IVM is a powerful tool for studying the dissemination and transmigration of tumor and immune cells in living hosts , but it is only recently that this technique has begun to be applied to the study of host-pathogen interactions [10] , [11] . The results of our recent study indicated that B . burgdorferi dissemination from the host microvasculature in vivo is a progressive , multi-stage process consisting of several successive steps: transient and dragging interactions ( collectively referred to as short-term interactions ) , followed by stationary adhesion and extravasation . Short-term interactions constitute the majority of spirochete-endothelial associations ( 89% and 10% for transient and dragging interactions , respectively ) , take less than one second ( transient interactions ) or 3–20s ( dragging interactions ) to travel 100 µm along the vessel wall , and occur primarily on the surface of endothelial cells and not at endothelial junctions [9] . Transient interactions are characterized by a tethering-type attachment-detachment cycle of association in which part of the spirochete adheres briefly to the endothelium before being displaced by blood flow , whereas dragging spirochetes adhere along much of the length of the bacterium , and creep more slowly along the vessel wall [9] . In contrast , stationary adhesions ( 1% of interactions ) do not move along the vessel wall , occur chiefly , but not exclusively , at endothelial junctions , and entail a more intimate association with the endothelium than short-term interactions [9] . Finally , spirochete extravasation ( <0 . 12% of interactions ) also occurs primarily , but not exclusively , at endothelial junctions , and is a triphasic process consisting of a rapid , end-first initial penetration of the endothelium , followed by a prolonged period of reciprocating movement , and ending with a rapid exit phase in which the bacterium bursts out of the vessel and migrates rapidly into the surrounding tissue [9] . In vitro studies have shown that B . burgdorferi binds several host molecules that might mediate endothelial interactions in vivo , including fibronectin ( Fn ) , integrins , heparan sulfate-type glycosaminoglycans ( GAGs ) and regulators of the complement cascade [12]-[20] . A broad array of pathogens have been shown to interact with these ubiquitous host molecules in direct binding assays and tissue culture models in vitro; most of these studies have been performed in the absence of shear forces , microvascular endothelium or a functioning immune system , and so the potential contribution of such interactions to hematogenous dissemination in the living host is unknown . To date , 19 candidate adhesin genes have been identified in B . burgdorferi , two of which are known to interact with integrins ( P66 and BBB07 ) , and two of which can associate with heparan sulfate GAGs ( BBK32 and Bgp ) [16] , [17] , [19] , [21]–[25] . B . burgdorferi encodes one characterized Fn binding protein , BBK32 , and appears to express a number of others [12] , [16] , [24] . Five B . burgdorferi CRASP proteins that interact with complement cascade regulating proteins factor H , FHL-1 and FHR-1 have also been identified [20] , [26] , [27] , but their potential contributions to endothelial cell adhesion are unknown . In the work described here we used IVM to explore the mechanistic basis for B . burgdorferi interactions with the microvasculature of living mice . We found that the initiating and stationary adhesion stages of microvascular interactions were mechanistically distinct but inter-dependent events , and that BBK32 , Fn and GAGs played a substantial role in initiation events . These findings and the methodology described here provide a framework for investigating the role of Fn and GAGs in vascular interactions during hematogenous dissemination by B . burgdorferi and possibly other pathogens .
To quantitatively analyze B . burgdorferi interactions with the microvascular endothelium in vivo we employed conventional epifluorescence IVM to examine interactions in the flank skin of mice after intravenous inoculation with 4×108 spirochetes ( Fig . 1 and Videos S1 and S2 ) . Conventional epifluorescence IVM was used instead of spinning disk confocal IVM because it is more effective for imaging the rapid associations that constitute the bulk of B . burgdorferi microvascular interactions ( Fig . 1 ) . Analysis was performed in post-capillary venules , where interactions could be most accurately quantified . For all experiments reported in this manuscript , data describing the numbers of recorded vessels and mice for each experimental condition , as well as the average time after spirochete injection at which recordings were made are provided in the Figure Legends . As we have recently reported , during the experimental observation period no signs of endothelial or leukocyte activation were detected [9] . In addition , leukocyte adhesion in dermal postcapillary venules is an indicator of local activation , and can be measured by using the dye rhodamine 6G to fluorescently label all circulating leukocytes and then counting the number of adherent leukocytes in a 100 µm length of vessel [28] . The presence of infectious B . burgdorferi in the vasculature for as long as 70 minutes after injection of spirochetes did not significantly alter leukocyte adhesion from the baseline levels observed in the absence of B . burgdorferi ( 1 . 56−/+0 . 29 vs 1 . 75−/+1 . 15 adhered leukocytes/100 µm , respectively; P = 0 . 797; N = 50 vessels from 5 mice ) . The observed number of leukocyte adhesions was normal for the dermal microvasculature of mice [28] . As shown in Fig . 2 , the ability to interact with the microvascular endothelium was specific to infectious spirochetes . When mice were injected with non-infectious B . burgdorferi exhibiting the same fluorescence intensity as infectious spirochetes ( Fig . 2A and B ) , transient interactions were reduced by 94% ( Fig . 2C ) . Furthermore , non-infectious B . burgdorferi did not form many dragging interactions and no detectable stationary adhesions ( Fig . 2B and C ) . Non-infectious spirochetes were never observed escaping the microvasculature . These observations indicated that early-stage interaction events were essential for sustained association and vascular escape . These observations also demonstrated that microvascular interactions were dependent on B . burgdorferi proteins expressed only in the infectious strain . Many bacterially-encoded proteins interact with host cells via GAGs [29] , and a number of previous in vitro studies performed under static conditions have found that B . burgdorferi can bind to GAGs and that exogenously applied GAGs can competitively inhibit interaction of B . burgdorferi with cell monolayers [13] , [15] , [30] , [31] . Therefore , we investigated the potential role of endothelial host cell GAGs in spirochete microvascular association in vivo . Interaction rates were first examined in the presence and absence of a therapeutic low molecular weight heparin compound , dalteparin ( Fragmin , average molecular weight 5 kDa ) . Dalteparin was used at a concentration previously shown to block leukocyte rolling in vivo [32] . Dalteparin ( 200 µl of a 25 I . U . /µl solution ) was injected via the femoral vein 15 minutes before intravenous inoculation with infectious spirochetes ( see Materials and Methods ) . As shown in Fig . 3A , dalteparin treatment did not cause any significant change in transient interactions between fluorescent , infectious B . burgdorferi and the vascular endothelium . However , dragging interactions were significantly reduced by 72% ( Fig . 3B ) while the number of stationary adhesions were also reduced by dalteparin treatment to a similar extent ( Fig . 3C , 76% of controls ) . Similar experiments were performed with dextran sulphate ( Fig . 3D–F ) , a 500 kDa high molecular weight GAG analogue , which interacts in vitro with infectious but not non-infectious B . burgdorferi [30] . In these experiments , dextran sulfate was incubated with infectious B . burgdorferi for 30 minutes , followed by extensive washing , prior to B . burgdorferi administration to the animal , as this compound was toxic when injected directly into the mouse bloodstream . The concentration of dextran sulfate used in preincubations ( 20 µg/ml ) was the dose that has been previously shown in vitro to maximally inhibit B . burgdorferi interaction with endothelial cells without altering spirochete morphology or motility [30] , [31] . Preincubation of infectious B . burgdorferi with dextran sulfate caused a slight ( 30% ) reduction in the number of transient interactions , but dragging interactions were reduced by 80% ( Fig . 3D and E ) . A similar reduction in the number of stationary adhesions was also observed ( Fig . 3F ) . The results from the dalteparin and dextran sulphate experiments indicated that host GAGs play an important role in dragging interactions between B . burgdorferi and the microvascular endothelium in vivo , and that competition with a high molecular weight GAG analogue ( dextran sulphate ) also inhibited transient interactions . The similar levels of inhibition of dragging interactions and stationary adhesions caused by treatment with GAGs suggested that reductions in stationary adhesion were the result of inhibition of dragging . This in turn implied that additional host and spirochete molecules might contribute to stationary adhesion . However , the results of these experiments alone did not rule out the possibility that GAGs played a role in stationary adhesion . Many bacterial adhesins can interact with GAGs , either directly , or indirectly via their association with host molecules such as fibronectin , regulators of the complement cascade and components of the coagulation system; furthermore , GAGs can act as bridging molecules that facilitate interactions between pathogen adhesins and host receptors [29] . In an effort to identify spirochete adhesins mediating GAG-dependent microvascular interactions in vivo , we PCR-amplified and sequenced all candidate B . burgdorferi adhesin genes identified to date [16] , [17] , [19] , [21]–[25] , using genomic DNA extracted from our fluorescent infectious and non-infectious strains ( Table S1 ) . This approach indicated that the genes encoding BBK32 , VlsE , OspF , ErpL and ErpK were absent or mutated in the non-infectious strain ( Table S1 ) . It is possible that VlsE , OspF , ErpL and ErpK could mediate GAG-dependent host interactions directly or through recruitment of host molecules such as complement cascade regulators; however , interaction of these proteins with GAGs has not been directly demonstrated . In contrast , BBK32 has recently been shown to bind to host GAGs and to rescue the ability of non-infectious B . burgdorferi to interact with endothelial cells in vitro [25] . It was , therefore , of interest to investigate a possible role for BBK32 in B . burgdorferi interactions with the microvasculature in the living mouse . The bbk32 coding sequence , under the control of the ospC promoter [25] , was cloned into the GFP expression construct and the resulting plasmid was used to transform the parental non-infectious B . burgdorferi strain . Both parental and complemented strains had the same endogenous plasmid content ( data not shown ) . Expression of BBK32 in the complemented strain was lower than the expression observed in the infectious strain ( 8 . 0−/+1 . 8% ) , but even this reduced expression was sufficient to restore transient and dragging interactions to the level observed with the infectious strain ( Fig . 4A and B ) . However , stationary adhesion rates in the bbk32 complementation strain did not reach the same levels as in the infectious strain ( Fig . 4C ) , implying either a greater dependence upon BBK32 or a dependence upon additional spirochete factors that were missing in the complemented non-infectious strain . Attempts to genetically disrupt the bbk32 locus in the infectious strain were successful , but did not result in usable constructs due to loss of endogenous plasmids ( lp28-1 and others ) in all recovered strains . PCR amplification and sequencing of all candidate B . burgdorferi adhesin genes identified to date also indicated that other candidate adhesin genes ( bbf32 , bbk2 . 10 , bbO39 and bbm38 , encoding VlsE , OspF , ErpL and ErpK , respectively ) were absent or mutated in the non-infectious parental strain ( Table S1 ) . Hence , these genes were not essential for transient and dragging interactions with the microvasculature of murine skin in vivo , since expression of BBK32 alone in this strain was sufficient to restore transient and dragging interactions . However , the possibility still exists that some of these genes play a role in stationary adhesion . Examination of the sequence , expression and localization of two other major adhesins , P66 and Bgp , which have been shown to associate with integrins and GAGs respectively under static conditions in vitro , indicated that these proteins were expressed and localized similarly in non-infectious and infectious strains , and were not mutated ( Fig . 4D; Table S1 ) . Therefore , neither P66 nor Bgp expression nor localization was sufficient for transient or dragging interactions in the absence of BBK32 expression . Although the genomic sequence of the P66- and Bgp-encoding genes was identical in both infectious and non-infectious strains , it remains possible that secondary mutations elsewhere in the genome of non-infectious B . burgdorferi could have negatively affected transient and dragging interactions . Because BBK32 binds Fn in addition to GAGs , we also investigated a possible role for Fn in the adhesion of B . burgdorferi to the endothelium in vivo . Rabbit serum , which contains fibronectin , is an important component of the BSK-II medium used to propagate B . burgdorferi; therefore , we investigated whether antibodies to rabbit plasma Fn could disrupt B . burgdorferi microvascular interactions in vivo . Anti-Fn IgGs did not alter spirochete morphology or motility in vitro , implying that they were not toxic to B . burgdorferi . The tethering , dragging and stationary interactions/min for infectious B . burgdorferi treated with αFn IgGs were compared to the interaction rates of untreated spirochetes , and of spirochetes treated with nonspecific goat IgGs ( Fig . 5 ) . Preincubation of infectious spirochetes for 20 minutes with the IgG fraction of goat antiserum to rabbit plasma Fn , together with intravenous injection of this IgG fraction into the blood stream of mice , reduced transient and dragging microvascular interactions by 92% and 99% , respectively ( Fig . 5A and B ) . When the same treatment regimen was performed using nonspecific goat IgGs , no effect on interaction rates was observed , indicating that the reduction in interactions following treatment with αFn IgGs was specific . Although stationary adhesions were essentially abolished by the αFn treatment ( Fig . 5C ) , the reduction in transient and dragging interaction rates was so great that we could not determine if stationary adhesion rates were specifically affected by treatment with anti-Fn IgGs . Interestingly , interaction rates returned to normal levels 15–20 minutes after injection of the spirochetes and antibody ( data not shown ) , suggesting that antibody-blocked rabbit Fn bound to spirochetes might have been replaced by mouse Fn in vivo , thus restoring microvascular interactions . The long population doubling time of B . burgdorferi ( 6–8h ) precludes the possibility that restored interaction rates were caused by spirochete replication . The dramatic reduction in transient and dragging interactions resulting from Fn antibody treatment suggested that B . burgdorferi exploits host Fn for these interactions with the host microvasculature in vivo . Although experiments performed with anti-Fn IgGs suggested that Fn played a major role in the initiation of microvascular interactions , it was possible that IgG-dependent inhibition was partly a result of factors such as steric hindrance of interactions by bulky IgGs . Therefore , we also investigated the Fn dependence of interactions using Fn peptides . Fn is a structurally and functionally complex molecule ( reviewed in [33] ) . Briefly , the N-terminal Type I Fn repeats and gelatin-binding region interact with Fn-binding proteins from B . burgdorferi , Staphylococci and Streptococci in vitro [16] , [34] , [35] . The central cell-binding domain contains multiple integrin-binding sites , including the canonical RGD sequence , which binds to most integrins that have been implicated in B . burgdorferi-host cell interactions to date [17] , [19] , [33] . Finally , the Fn C-terminus contains a high affinity heparin-binding domain that also interacts with host cell GAGs [33] . To investigate endothelial cell molecules associating with spirochete-bound Fn , we used peptides derived from the C-terminal heparin-binding domain and the integrin-interacting cell-binding domain in an attempt to block B . burgdorferi-microvascular interactions in vivo ( Fig . 6 ) . The heparin domain peptide ( FN-C/H II: KNNQKSEPLIGRKKT ) inhibits Fn-mediated cell adhesion and heparan sulfate binding [36] , and the GRGDS cell-binding domain peptide is a well-studied competitive antagonist of integrin binding [37] that also inhibits B . burgdorferi interactions with integrins αIIbβ3 , αvβ3 and α5β1 in vitro [15] . Peptides were injected via the femoral vein immediately before inoculation with infectious spirochetes , at concentrations ( ∼50 µg/ml of circulating blood ) that disrupt leukocyte adhesion and recruitment in vivo [38] . Microvascular interactions in dermal postcapillary venules were recorded for no longer than 20 minutes after injection of peptide as the effect of peptide treatment on interaction rates was diminished at later time points , presumably because linear peptides are rapidly cleared from the mouse circulation [39] . Intravenous injection of 100 µg of the heparin-binding domain peptide reduced transient interaction rates by 52% , and impaired both dragging interactions and stationary adhesion levels by 84% , confirming the role of GAGs in early stages of microvascular interaction ( Fig . 6A–C ) . In contrast , competition with the RGD peptide did not significantly inhibit any class of interaction ( Fig . 6D–F ) , even though the estimated final concentration of this peptide in the mouse circulation ( 100 µM ) was twice as high as the dose known to reduce in vitro B . burgdorferi-integrin interactions by at least 75% in vitro [15] . Administration of twice as much RGD peptide ( ∼200 µM final concentration ) did not inhibit interactions , nor did intravenous administration of anti-CD41 and CD49e antibodies that respectively target RGD-dependent B . burgdorferi-interacting integrins containing αIIb or α5 chains ( platelet glycoprotein αIIbβ3 and the α5β1 Fn receptor; data not shown ) . The effect of treatment with antibodies to αvβ3 was not examined; however , since treatment with RGD peptide in vitro has been shown to strongly inhibit B . burgdorferi binding to this integrin as well as glycoprotein αIIbβ3 and integrin α5β1 , it seems unlikely that integrin αvβ3 mediated the early stages of microvascular interactions . The conclusion that RGD-dependent integrins are not required for microvascular recruitment is consistent with the localization of known B . burgdorferi-associating RGD-dependent integrins , which are found at sites of endothelial attachment to extracellular matrix , and not in the lumen [15] . Collectively , these results implied that Fn-dependent transient and dragging interactions in vivo were mediated by host GAGs and not by RGD-dependent integrin interactions .
We found that BBK32 and its host ligands Fn and GAGs played major roles in transient and dragging interactions . Although we cannot rule out the possibility that these molecules also contribute to stationary adhesion , the results of the bbk32 complementation experiments indicate that additional spirochete molecules are likely required for stationary adhesion . This implies that stationary adhesion is a mechanistically distinct step in B . burgdorferi dissemination . This conclusion is supported by our previous observations that: 1 ) stationary adhesions form primarily at endothelial junctions , whereas short-term interactions occur chiefly on endothelial cells themselves; and 2 ) stationary adhesions associate more intimately with the endothelium than short-term interactions and appear to traverse the surface of the endothelium when these cells are labeled with PECAM-1 ( Fig . 7 ) [9] . These observations imply that B . burgdorferi dissemination shares functional similarities with the sequence of events that constitute the leukocyte recruitment cascade [46] , [47] , as well as the events associated with dissemination of circulating tumor cells [48] . Leukocyte recruitment is initiated by selectin-mediated tethering and rolling interactions that permit firm adhesion , which is mediated by integrins . The initiation phase of leukocyte recruitment is a rate-limiting step , as it is essential for all subsequent events in the recruitment cascade . Similarly , we propose that transient and dragging interactions mediated by GAGs and Fn together constitute the corresponding initiation phase of B . burgdorferi dissemination , while other host and spirochete molecules become essential at the stationary adhesion phase . Though our data indicated that transient and dragging associations were mediated by the same host and spirochete molecules , the observation that the low molecular weight heparin dalteparin inhibited only dragging interactions was surprising . The reason for this is currently unknown , but may result from differences in total charge , chain length and chemical composition of the carbohydrate moieties . This study identified a central role for the B . burgdorferi protein BBK32 , host GAGs and Fn in the initiation of microvascular interactions . This observation was unexpected , since previous studies have shown that genetic disruption of bbk32 attenuates but does not abolish infectivity [49] , [50] . However , bbk32 disruption mutants still bind Fn [49] , [50] , implying that other functionally redundant Fn-binding proteins in B . burgdorferi might also mediate the initiation of dissemination . The simplest interpretation of our data is that initiation is mediated by BBK32 interactions with GAGs , either independently or via a fibronectin bridge . It is possible that initiation might also be mediated by RGD-independent integrins such as α3β1 , which interacts with Fn , GAGs and the B . burgdorferi protein BBB07 [18] , [19]; however , this integrin is expressed at endothelial junctions [51] implying that it is more likely to mediate stationary adhesion or extravasation than initiation interactions . Furthermore , the activation of adhesive properties by endothelial integrins generally requires endothelial activation [46] , which is not detected in the short time frame of our experiments [9] . Taken together , these data make it unlikely that integrins play a role in the initiation of vascular adhesion . All molecules to date implicated in tethering under shear force conditions ( selectins , von Willebrand factor the E . coli FimH adhesin ) interact with sugar-containing ligands [52] , [53] , suggesting that Fn-dependent or -independent interactions between BBK32 and GAGs might promote B . burgdorferi tethering by a similar mechanism . The affinity of BBK32 for GAGs is unknown , but in the absence of shear forces BBK32 associates with high specificity and probable high affinity to the Fn N-terminus via a tandem β-zipper mechanism shared with Fn-binding proteins of Staphylococcus aureus and Streptococcus pyogenes [34] , [35] , [54] , [55] . In the absence of shear forces , the affinity of plasma Fn for heparin ( Kd = 0 . 1–1 . 0 µM ) is within the affinity range of P- and E-selectins for their ligands ( Kd = 1 . 5 µM and 109 µm , respectively ) [56] , [57] , suggesting that BBK32- , GAG- and Fn-dependent initiation interactions may be mechanistically feasible . Although under shear stress conditions Fn does not bind to the leukocyte Fn receptor VLA-4 , which mediates tethering to endothelial VCAM-1 under flow [58] , previous reports indicate that both platelets and Mycobacterium tuberculosis can bind to immobilized Fn in vitro under shear stress conditions that mimic those found in postcapillary venules [59] , [60]; interestingly , platelet-Fn interactions are almost completely blocked by treatment with unfractionated or high molecular weight heparin [60] . This suggests the possibility that Fn-dependent tethering interactions entail cooperative GAG binding , a conclusion that is consistent with our observation that expression of the Fn- and GAG-binding BBK32 protein was sufficient to restore initiation interactions to wild-type levels . Another possibility is that BBK32-induced conformational changes in Fn might facilitate Fn- and GAG-dependent tethering interactions . This hypothesis stems from recent data from the Höök laboratory indicating that BBK32 binding to Fn induces the formation of superfibronectin ( S . Prabhakaran and M . Höök , personal communication ) , a high molecular weight Fn complex that substantially enhances adhesion of cells to Fn by integrin-dependent and independent mechanisms [61] . Further analysis of the precise mechanisms underlying BBK32- , Fn- and GAG-dependent dissemination under shear force conditions will be required . The results of this study emphasize the importance of directly investigating host-pathogen interactions in a native context where major regulators of interaction such as fluid shear stress are present . The methodology and observations presented here provide the first direct insight into the role of host GAGs , Fn and a B . burgdorferi protein that binds both of these host components , in host microvascular interactions in situ . These results may have broad-reaching implications for our understanding of processes underlying the dissemination of a variety of other bacterial pathogens that interact with Fn and GAGs .
Plasmid pTM170 was constructed by PCR amplification of the PospC-driven bbk32 cassette from pBBK32 [25] with flanking KpnI and FspI sites , using primers B1093 ( 5′-GGTACCTTAATTTTAGCATATTTGGCTTTG-3′ ) and B1094 ( 5′-GGCCTGCGCATTAGTACCAAACGCCATTCTTG-3′ ) . The PCR product was cloned into the GeneJet plasmid , using the Gene Jet blunt cloning kit ( MBI Fermentas ) to generate pTM169 . The KpnI/FspI-digested pTM169 insert was cloned into the KpnI/FspI sites of the GFP-encoding plasmid pTM61 [9] to yield pTM170 . B . burgdorferi strains used in this study were GCB705 ( non-infectious strain B31-A transformed with pTM61 ) [9] , [62] , GCB726 ( infectious B . burgdorferi strain B31 5A4 NP1 transformed with pTM61 ) [9] , [63] and GCB769 ( non-infectious B31-A transformed with pTM170 ) . The plasmid content of these strains is noted in Table S1 . All strains were grown in BSK-II medium prepared in-house [64] . Electrocompetent B . burgdorferi strains were prepared as described [9] , [65] . Liquid plating transformations were performed with 50 µg pTM61 or pTM170 in the presence of 100 µg/ml gentamycin as described [66] , [67] . Gentamycin-resistant B . burgdorferi clones were screened for: 1 ) the presence of aacC1 sequences by colony screening PCR performed with primers B348 and B349 as described [68]; 2 ) GFP expression by conventional epifluorescence microscopy , and 3 ) BBK32 expression , as detected by immunoblotting ( for bbk32 complementation strains ) . The presence of plasmids in non-integrated form in fluorescent strains was confirmed by agarose gel electrophoresis of total genomic DNA prepared on a small scale as described [69] . PCR screening for native plasmid content was performed as described [68] , [70] . Gene sequences were amplified from genomic DNA preparations of GCB705 and GCB726 . PCR was performed with Phusion DNA polymerase ( NEB , Pickering , Ontario , Canada ) , according to manufacturer's instructions . Sequencing was performed by the University of Calgary DNA Services . All primers used for PCR amplification and sequencing are provided in Table S1 . The expression and outer membrane localization of adhesins P66 and Bgp were analyzed as previously described [16] , [25] . Briefly , for each strain , two pellets containing 5×107 spirochetes were washed twice with PBS+2% BSA . PBS containing 5 mM MgCl2 was added without dislodging pellets . Proteinase K was added to one pellet to a concentration of 4 mg/ml . After 30 min incubation at room temperature , reactions were stopped with 150 µg phenylmethylsulfonyl fluoride , spirochetes were pelleted and washed twice with PBS+0 . 2% BSA , and pellets were lysed using SDS-PAGE loading dye . Proteins were resolved by electrophoresis on 12% SDS-PAGE gels , transferred to nitrocellulose membranes , followed by immunoblotting with antibodies to P66 , Bgp or BBK32 , as previously described [17] , [22] . These conditions have been described in detail previously [9] . Quantification of spirochete interactions was performed as recently described [9] . All animal studies were carried out in accordance with the guidelines of the University of Calgary Animal Research Centre . Leukocyte recruitment studies were carried out as previously described [28] . Briefly , animals were injected with 50 µl of 0 . 05% ( i . v . ) rhodamine 6G ( Sigma-Aldrich ) . Fluorescence was visualized by epi-illumination using 510 and 560 filters . Leukocytes were considered adherent to the venular endothelium if they remained stationary for 30 s or longer . Experiments were performed in mice that had been intravenously inoculated with 4×108 infectious B . burgdorferi grown for 48h in 1% mouse blood , as previously described , and also with mice that were not inoculated with spirochetes . Leukocyte adhesions were counted in the dermal postcapillary venules of infected and non-infected mice from 5 minutes after injection of spirochetes and/or rhodamine until at least 1 hour from injection , in order to monitor leukocyte recruitment during the time frame that is used for all experiments reported in this study . GCB726 spirochetes prepared as described above were resuspended to 2×109/ml in PBS . The IgG fraction of polyclonal goat anti-rabbit plasma Fn serum or non-specific goat IgGs ( Cappel/MP Biomedicals , Solon , OH ) were added to 1 mg/ml final . After mixing for 30 min at room temperature , spirochetes were directly injected into the mouse bloodstream , along with 2 mg of Fn antiserum IgGs or non-specific IgGs . GCB726 spirochetes were prepared and injected as described above , together with 100 µg of GRGDS or FN-C/H II peptide ( Sigma Canada , Oakville , ON; catalogue numbers G4391 and F7049 , respectively ) , injected via the femoral vein . Peptides injected at this amount are known to disrupt leukocyte adhesion and recruitment in vivo [38] . Two hundred µl of a 25 I . U . /µl solution of dalteparin ( Fragmin: Pfizer Canada , Kirkland , PQ ) were injected via the femoral vein 15 minutes before intravenous inoculation with infectious spirochetes . This concentration has previously been shown to inhibit leukocyte rolling in vivo [32] . Dextran sulfate-treated spirochetes were incubated with 20 µg/ml dextran sulfate ( 500 kDa; Fisher Scientific Canada , Ottawa , ON ) in a final volume of 100 ml PBS for 30 min at RT°C , followed by 2 100ml washes with PBS . Spirochetes were resuspended to 2×109/ml in PBS , and injected as previously described [9] . The concentration of dextran sulfate used in these preincubations ( 20 µg/ml ) is the dose that maximally inhibits B . burgdorferi interaction with endothelial cells in vitro , and does not affect spirochete morphology or motility [30] , [31] . One hundred µg anti-CD41 monoclonal Ab ( Clone MwReg30; Becton Dickinson , San Diego , CA ) , or 20 µg CD49e monoclonal Ab ( clone 5H10-27; Pharmingen , Oxford , UK ) were intravenously administered immediately prior to injection of spirochetes . These quantities of anti-CD41 and CD49e antibodies are those that respectively protect against Plasmodium berghei infection in vivo [71] , and which inhibit neutrophil migration in vivo [72] . For quantitative analysis , average and standard error values for different variables were calculated and plotted graphically for all vessels from all mice using GraphPad Prism 4 . 03 ( GraphPad Software , Inc . , San Diego , CA ) . Statistical significance was calculated in GraphPad Prism using a two-tailed non-parametric Mann Whitney t-test with a 95% confidence interval .
|
Many bacterial pathogens can cause systemic illness by disseminating through the blood to distant target sites . However , hematogenous dissemination is still poorly understood , in part because of an inability to directly observe this process in living hosts in real time and at the level of individual pathogens . We recently engineered a fluorescent strain of Borrelia burgdorferi , the Lyme disease pathogen , and visualized its dissemination from the microvasculature of living mice using intravital microscopy . We found that dissemination was a multistage process that included tethering , dragging , stationary adhesion and extravasation . In the study described here , we used quantitative real-time intravital microscopy to investigate the mechanistic features of the vascular interaction stage of B . burgdorferi dissemination in living hosts . We found that tethering and dragging interactions ( collectively referred to as initiation interactions ) were mechanistically distinct from stationary adhesion . Initiation of microvascular interactions required the B . burgdorferi protein BBK32 , and host ligands fibronectin and glycosaminoglycans . Initiation interactions were also strongly inhibited by the low molecular weight clinical heparin dalteparin . Since numerous bacterial pathogens can interact with fibronectin and glycosaminoglycans in vitro , these observations raise the intriguing possibility that fibronectin and glycosaminoglycan recruitment might be a feature of hematogenous dissemination by other pathogens .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"infectious",
"diseases/bacterial",
"infections",
"microbiology/cellular",
"microbiology",
"and",
"pathogenesis",
"microbiology/medical",
"microbiology"
] |
2008
|
Molecular Mechanisms Involved in Vascular Interactions of the Lyme Disease Pathogen in a Living Host
|
There has been much effort to prioritize genomic variants with respect to their impact on “function” . However , function is often not precisely defined: sometimes it is the disease association of a variant; on other occasions , it reflects a molecular effect on transcription or epigenetics . Here , we coupled multiple genomic predictors to build GRAM , a GeneRAlized Model , to predict a well-defined experimental target: the expression-modulating effect of a non-coding variant on its associated gene , in a transferable , cell-specific manner . Firstly , we performed feature engineering: using LASSO , a regularized linear model , we found transcription factor ( TF ) binding most predictive , especially for TFs that are hubs in the regulatory network; in contrast , evolutionary conservation , a popular feature in many other variant-impact predictors , has almost no contribution . Moreover , TF binding inferred from in vitro SELEX is as effective as that from in vivo ChIP-Seq . Second , we implemented GRAM integrating only SELEX features and expression profiles; thus , the program combines a universal regulatory score with an easily obtainable modifier reflecting the particular cell type . We benchmarked GRAM on large-scale MPRA datasets , achieving AUROC scores of 0 . 72 in GM12878 and 0 . 66 in a multi-cell line dataset . We then evaluated the performance of GRAM on targeted regions using luciferase assays in the MCF7 and K562 cell lines . We noted that changing the insertion position of the construct relative to the reporter gene gave very different results , highlighting the importance of carefully defining the exact prediction target of the model . Finally , we illustrated the utility of GRAM in fine-mapping causal variants and developed a practical software pipeline to carry this out . In particular , we demonstrated in specific examples how the pipeline could pinpoint variants that directly modulate gene expression within a larger linkage-disequilibrium block associated with a phenotype of interest ( e . g . , for an eQTL ) .
Advances in next-generation sequencing ( NGS ) technologies have enabled high-throughput whole genome and exome sequencing [1] , which have led to the identification and characterization of many disease-associated mutations [2] and the vast majority of common single nucleotide variants ( SNVs ) in the human population [3 , 4] . Genome-wide association studies ( GWAS ) have found that these variants mostly lie outside of protein-coding regions [5] , emphasizing the functional importance of non-coding regulatory elements in the human genome . These advances have also led to an urgent need to develop high-throughput methods to sift through this deluge of sequencing data to quickly determine the functional relevance of each non-coding variant [6] . Evidence suggests that only a fraction of non-coding variants are functional , and the majority of functional variants show only modest effects [7] . Studies like GWAS [8] and expression quantitative trait eQTL [9] have evaluated the association of variants with traits of interest from a statistical perspective . In traditional GWAS and eQTL analyses , an association locus may host the tag-SNPs and a number of linked variants that may potentially account for the molecular mechanism underlying the association [10] . However , it remains difficult to distinguish those that are truly causal [11–13] . Thus , downstream analysis requires fine-mapping to identify the true causal variants by integrating the external genetic and epigenetic information [12 , 14] . As association studies give little information about the mechanism of a variant’s effects , it would be helpful to directly test the molecular effects of a large numbers of variants using highly quantitative assays . Luciferase reporter assays are a common method to measure the regulatory effects of functional elements [15] . Researchers can compare the difference of luciferase expression with and without a mutation to estimate the experimental molecular effect of non-coding variants lying in a functional element . By using high-throughput microarray and NGS technology , the massively parallel reporter assay ( MPRA ) has extended the scales to the genome-wide level [16–21] . Recently , Tewhey and colleagues demonstrated the capability of MPRA to identify the causal variants that directly modulated gene expression [22 , 23] . This study identified 842 expression-modulating variants ( emVARs ) showing significantly differential expression modulation effects and provided a high-quality data source for computational modeling [22 , 23] . There is an increasing need for computational methods to effectively predict the molecular effects of variants and improve our understanding of the underlying biology of these effects . Several approaches have been developed to address the problem of variant prioritization from different perspectives . Based on the target of predictions , these methods roughly fall into three major categories: 1 ) disease-causing effect predictors ( e . g . GWAVA [24] , and GenoSkyline [25] ) , which aim to prioritize causal disease variants and distinguish them from benign ones; 2 ) fitness consequence prioritization tools ( e . g . , CADD [26] , fitCons [27] and LINSIGHT [28] ) , which attempt to identify the variants based on evolutionary fitness; 3 ) comprehensive tools ( e . g . , DeepSEA [29] , FunSeq2 [6] , FUN-LDA [30] ) which integrate multiple data sources for prediction of functional variants . Many of these computational methods are designed to predict and prioritize deleterious and disease-associated variants from a phenotypic perspective , but not to highlight specific molecular consequences of these variants ( i . e . , their effects on the activities of functional elements ) . Moreover , some of these tools are cell type-agnostic , and tools that are cell type-aware depend on cell type-specific data with somewhat limited availability , such as ChIP-Seq or epigenetic features . Thus , it would be helpful to build a generalized model that can be systematically specialized to any desired cell type with only a small amount of easily obtainable cell type-specific information ( e . g . expression data ) . In this study , we addressed the problem of molecular effect prediction of variants from a different perspective . Instead of predicting phenotypic consequences from genotypes , which is a common practice , we aimed to directly predict the expression-modulating effect of the variants from various sources of information . Our model , named GRAM ( i . e . , GeneRAlized Model ) , incorporates selected transcription factor ( TF ) binding information from in vitro SELEX assays , representing the general binding affinity of TFs on the variant’s location , and cell type-specific expression profiles , representing cellular contexts . Combining cell type-independent and -dependent features makes our model both flexible and specific . When we evaluated results from MPRA and luciferase assay experiments show our model achieved high predictive performance and could be easily transferred to other cell types and assay platforms . We also demonstrated the potential application of GRAM to the fine-mapping of pre-defined variants in linkage disequilibrium . As a supplement to many general variant effect prediction methods ( which often combine disparate features ) , our model can help to precisely define the subset of prioritized variants that directly alters gene expression . For instance , after using a more general functional impact tool such as FunSeq or VEP [31 , 32] , one could use GRAM on the prioritized variants to identify the subset that has a direct expression modulating effect ( as opposed to being prioritized for other reasons such as strong association with an organismal phenotype ) . Furthermore , one could use GRAM to fine-map the key causal variant modulating gene expression from the many variants in a linkage-disequilibrium block associated with gene expression in an eQTL study .
In this study , we first collected a dataset from Tewhey et al . [22] to estimate expression modulation differences between reference allele and mutants in the GM12878 cell line . This MPRA-generated dataset contains 3 , 222 SNVs filtered by logSkew value , which measures the log-fold change of the expression-modulating differences between reference and alternative alleles . Among them , 792 variants ( named emVARs ) had a significant expression-modulating effect compared with their respective reference allele , which indicates the molecular effect of the variant . Here , we treated emVARs and non-emVARs as positive and negative dataset , respectively , in our GRAM model . As described in Fig 1 , our GRAM model is implemented in three steps: ( i ) prediction of the universal regulatory consequences of an element with variant using the SELEX TF binding score; ( ii ) prediction of a cell type modifier score in a specific cellular context by combining TF binding score with cell type-specific TF expression profiles; and ( iii ) estimation of the expression modulating effect in a cell type-specific context by integrating outputs from the previous two steps . We first investigated the potential of evolutionary conservation and transcription binding features as predictors . Evolutionary conservation is associated with deleterious fitness consequence and is widely used in prioritization algorithms of non-coding variants , such as PhyloP [33] and PhastCons [34] scores in LINSIGHT [28] and CADD [26] , and GERP [33] score in FunSeq2 [6] . We performed comparative analyses for these three conservation features across different datasets ( S1 Fig ) . We found that the PhastCons and PhyloP patterns of emVARs and non-emVARs are different from Human Gene Mutation Database ( HGMD ) [35] variants but similar to non-HGMD variants , which are thought to be benign . GERP scores show a similar pattern but have smaller variance in emVARs and non-emVARs compared to other datasets , with slightly larger values for emVARs . As we did not find differential patterns when comparing emVARs and non-emVARs , we further discovered that the correlation between logSkew and all three conservation scores was low ( close to 0 ) by linear regression . These results suggest that the conservation scores might contribute little to the molecular effects under study that focuses on expression modulation of variants in more conserved regions with homogeneous evolutionary patterns . TF binding can link the molecular effect of non-coding variants to a cascade of a regulatory network , which is thought to be an important contributing factor to the variants’ regulatory effects [26 , 29 , 36 , 37] . Tewhey et al . found that the logSkew value positively associates with TF binding scores . To thoroughly evaluate the effect of TF binding , we tested TF binding peaks overlapping with the SNVs and TF motif break events in the Tewhey dataset . We annotated and analyzed the emVAR and non-emVAR variant sets with FunSeq2 [6] , and found that the emVAR set had more TF binding events compared with the non-emVAR set ( Fig 2A ) . In addition to TF binding enrichment , we examined the motif breaking scores for these TFs . After removing TFs with insufficient observations , the differences between the distributions of motif-break scores for alternative and reference alleles in emVARs are larger than those in the non-emVAR dataset ( Fig 2B ) . According to this analysis , the emVAR set tends to have not only more TF binding events , but also larger binding alterations compared with the non-emVAR set . Our results indicate that TF binding shows high association with the expression-modulating effects of the variants and align with recent studies on the underestimated relative importance of transcription [38 , 39] . We generated a candidate training feature set from the outputs of 515 DeepBind models for TF binding , inferred from both ChIP-Seq [40] and in vitro SELEX assays [41] , on the adjacent sequences of the variant of interest . With a comprehensive feature selection framework for selection of impactful TF binding features , we prioritized these features across models with LASSO stability selection [42] and Random Forest ( shown in Fig 3A ) . The 20 most important features ( out of 515 ) with respect to the mean importance across all methods is shown in decreasing order in Fig 3A . Both ChIP-Seq and SELEX DeepBind features showed high importance , with the top two being GM12878 ChIP-Seq features ( SP1 and BCL3 ) , which are cell line specific , followed by SELEX features starting with ETP63 . The top-ranked impactful TFs tend to have more protein-protein interactions than the bottom-ranked TFs , indicating that the importance of a TF reflects its role in the TF-TF cascade regulatory network ( Fig 3B ) . Interestingly , many SELEX features , though not cell type dependent , achieved similar predictive power as cell type-specific ChIP-Seq features . We compared the predictive performances of cell type-dependent ChIP-Seq features , cell type-independent SELEX features , and a combination of both feature sets using a LASSO regressor , support vector machine ( SVM ) regressor and Random Forest . Incorporating ChIP-Seq-derived features , though introducing more cell type-specificity , did not boost the accuracy significantly for any of the three models ( Fig 3C and S1 Table ) . As the availability of ChIP-Seq data is restricted to a few cell lines ( S2 Fig ) , we instead used SELEX features to build a more generalized model that can be easily applied to different cell types . We then used the features generated from disease-association prediction tools ( CADD [43] , FunSeq2 [32] , DeepSEA [44] , GWAVA [45] , LINSIGHT [46] , and Eigen [47] ) to predict the same molecular effect target . As shown in Fig 3C , this analysis indicated that the prediction of disease-associated variants is not equivalent to that of expression-modulating variants . Using the TF binding features from DeepBind models and the MPRA dataset from Tewhey et al . [22] , we implemented our multi-step model . In the first step , we predicted the universal regulatory activity of an element with or without a variant . The 10-fold cross validation demonstrated exemplary performance of the model with an area under the receiver operating characteristic curve ( AUROC ) of 0 . 938 and an area under the precision-recall curve ( AUPRC ) of 0 . 928 ( Fig 4A and S3 Fig ) . In the second step , we calculated a cell-type modifier score as an indicator of the experimental assay’s cell-specific nature . Briefly , we defined the prediction target using a top and bottom quantile of Vodds ( S5 Fig ) . Vodds is the standard deviation of log odds for each variant’s read count in MPRA , which reflects the confidence interval of log odds ratio of an experiment . Vodds shows cell line-specific patterns , as the patterns of the two B-Lymphocyte cell lines ( NA12878 and NA19239 ) are similar while distinct from HepG2 ( S4 Fig ) ( see Methods for details ) . This indicates that Vodds can capture the cell type-specific information . We also found that variants with higher Vodds tend to include more non-emVARs ( Chi-square test p-value: 0 . 0002 ) . Hence , the cell type modifier score defined from Vodds can be used to adjust the universal regulatory effect to a cell type-specific context . Gene expression profiles , especially TF expression profiles , are more generally available and can represent the cellular environment . We incorporated TF gene expression and TF binding scores as features to predict the cell type modifier target , and got an AUROC of 0 . 66 and 0 . 8 ( Fig 4D ) , respectively , using Random Forest with a 10-fold cross-validation ( Fig 4B and S6 Fig ) . The final step is to predict the molecular effect of a variant , i . e . whether it can significantly modulate reporter gene expression . To do this , we fed the output from the first and second step into a LASSO model , with the emVAR and non-emVAR labels as targets . We found that the AUROC of a 10-fold cross-validation for the optimal model was 0 . 724 ( Fig 4C ) and the AUPRC was 0 . 602 , both of which are higher than the state-of-the-art method ( KSM ) using the same dataset ( AUROC: 0 . 684 , AUPRC: 0 . 478 ) [48] . To achieve better generalizability , we built the model with SELEX features only . We performed step ( i ) and ( ii ) on the same GM12878 dataset and another multiple-cell-line dataset ( MCL dataset: GM12878 plus HepG2 plus K562 ) . The model with cell-independent features from the SELEX assay achieved comparable performance with an AUROC = 0 . 664 ( GM12878 only ) and 0 . 658 ( MCL dataset , Fig 4D ) . We use the model based on the multiple-cell-line dataset in our final GRAM model for a better generalization potential . We next evaluated performance of the model on different cell types and assay platforms . Rather than measuring read counts as in MPRA , some other assays , such as luciferase and GFP reporter assays , measure luminescence and fluorescence readouts instead . [49 , 50] . To evaluate how our model , trained with multiple cell line MPRA data , can be transferred to these assay platforms we tested its performance on luciferase assay results of eight potential regulatory elements with mutations from the MCF7 cell line [51] . To predict expression-modulating effects , we defined the significant changes between alternative and reference alleles by using an absolute log2 ( odds ratio ) cutoff . The average AUROC value was greater than 0 . 8 for MCF7 ( Fig 5A ) and 0 . 67 for K562 given the an absolute log2 cutoff from 0 . 5 to 0 . 8 ( Fig 5B ) . This indicates that our model performs very well on the luciferase assay and MPRA dataset from different cell lines , even though these assays use different measurements . In MPRA , the element is inserted upstream ( 5’-terminal ) of the reporter gene , but for some assays , such as STARR-Seq , the element is inserted downstream ( 3’-terminal ) . Therefore , we further tested the effect of insertion location of an element in luciferase reporters in K562 cells using 14 randomly selected elements with potential regulatory activity . As shown in Fig 5C , the 5’ terminal log odds were similar to the 3’ terminal odds for region 3 , 4 , 5 , and 13 , but showed significant differences for region 6 , 8 , 9 , 10 , and 14 . The prediction of GRAM for the 5’ terminal was much better than that for the 3’-terminal insertions; the AUROC was 0 . 25 higher for universal regulatory activity and 0 . 32 higher for the expression-modulating effect prediction , indicating different mechanisms for the two ends . Therefore , GRAM model is optimal for 5’ terminal assays . As GRAM needs only gene expression and SELEX DeepBind score to predict sample-wise variants effect , it could be a flexible tool for a variety of analysis tasks . We investigated whether we could apply our GRAM model to fine-mapping of causal variants . As was described in the Methods part , we made a user-friendly pipeline GRAMMAR that could conduct the entire analysis ( S9 Fig ) . Here we mainly focused on the task of identifying the variants that are most likely to directly modulate gene expression . For our analysis , we selected five LD blocks with known risk association with prostate cancer and high enrichment of annotated eQTL SNPs reported by Dadaev et al . [10] , resulting in a set of 561 eQTL SNPs from the five LD blocks . We extracted the genotypes and gene expression data from 102 The Cancer Genome Atlas ( TCGA ) PRAD patients and ran GRAMMAR to get the prediction score for each allele in each patient ( S4 Table ) . In general , variants with high posterior probability ( ≥0 . 5 , 130 variants ) , as a causal variant , reported by Dadaev et al . [10] , generally have higher average GRAM scores as compared to those with lowest posterior probability ( <0 . 5 , 4260 variants ) ( p-value = 0 . 0545 , S7 Fig ) . Specifically , we took a closer look at region chr6:160081543–161382029 , tagged by GWAS SNP rs9364554 and enriched with 52 eQTL SNPs for genes including ACAT2 , LOC729603 , MRPL18 , SLC22A3 and WTAP . All the FunSeq2 scores ( maximum 1 . 40 ) are below 2 , an empirical threshold for confident candidate causal SNVs . GRAMMAR , however , can pinpoint three SNV candidates with the highest average GRAM scores in this region ( Fig 6A ) . Their GRAM scores differ in different patient samples , indicating different expression modulating effects of these SNVs under different personalized cellular contexts . Moreover , all three of the highest-scored variants show strong correlations between the GRAM expression modulating score and the expression of the related target gene and two of which are significant ( p-value < 0 . 05 ) ( Fig 6B–6D ) .
There has been an increasing number of computational methods that can prioritize non-coding variants . In addition , accumulating high-throughput whole-genome sequencing data have become the primary source for identifying disease-associated variants . However , we still lack an efficient prediction model for estimation of the expression-modulating effect of variants that can be universally applied to many cell lines or samples . Previous studies tend to construct one distinct model for each cell type , or predict the cell-type specificity of a variant from often very limited experimental results ( e . g . ChIP-Seq ) in different cell types [25 , 30 , 52 , 53] , which makes the generalization to other cell types challenging . In this study , we sought to represent the impact of cellular environments on variant function from a different perspective . We developed a multi-step generalized model called GRAM that can specifically predict the cell type-specific expression-modulating effect of a non-coding variant in the context of a particular experimental assay . Our model receives both cell type-dependent and independent input data and combines them with the same set of feature weights across different contexts , Thus , our model can be applied to any cellular context as long as cell type- or sample-specific expression data are provided . In this study , we aim to precisely define the expression-modulating effect as a function of the predictive variables extracted from genomic data . In line with results from recent studies [38 , 39] , a wide array of transcription-related features demonstrated high predictive power . In contrast , three selected evolutionary features demonstrated low predictive power on used datasets . This pattern is likely due to the limited variety in evolutionary patterns in the training data and also stems from the nature of GRAM , which focuses on predicting expression-modulation effects . These effects are part of the many that are related to sequence conservation [54 , 55] . In other words , the purpose of our model is to enable precise downstream analysis of molecular effects of variants in a highly conserved region , where we would not expect conservation scores to provide more additional information . We further selected a variety of TF binding features that could be useful for predicting variant effects and used direct measurements from TF binding scores and implemented a straightforward LASSO regression to assess the importance of each feature . We found that in vitro SELEX TF features ( aka non-cell-specific features ) achieve the highest predictive performance , a result further validated by SVM and Random Forest models trained in parallel . We cannot ignore the cell type-specific context when predicting a variant’s effects . Usually , a model can achieve cell type-specificity in two different ways: 1 ) building an independent model for each cell line , or 2 ) building one unified model that can accept and handle specific input data from any cell lines or samples . Which strategy to use depends on the availability of the dataset and the demand for model transferability . Our model uses the second strategy , in which cell type-specific information is incorporated as an input feature and the model learns the same set of feature weights across multiple cell lines . For such a unified model , features like histone modification and TF ChIP-Seq would limit its transferability because these features may not be available for many other cell types or samples . Thus , we would prefer features that are more easily available , such as gene expression profiles . Here , we built the model using cell type-dependent gene expression and cell type-independent TF in vitro SELEX features; thus , the model can be more easily applied to various different samples and cell lines . SELEX features represent general binding strength of the TFs on the region of interest , and gene expression profiles can represent the specific cellular context . The three-step GRAM model predicts the expression-modulating effects of variants by integrating two intermediate predictive targets: universal regulatory activity and cell type modifier score . The universal regulatory activity reflects the general regulatory effect of an element with or without a mutation in a vector-based assay without considering cell type-specific chromatin contexts or epigenomics information . Next , we modeled the cellular environment related to gene regulation with a cell type modifier score , derived from cell type-specific TF expression levels , to adjust the universal regulatory effect in the final step of the prediction model , greatly improving the performance . GRAM performed well in validations on MPRA and luciferase assay , even across different cell types . In addition to target validations , our tool enables detailed exploration of the sensitivity of these methods and the impact of vector construct . The insertion position of the element affected the outcome of the assay , which may correspond to different types of regulatory elements . Because our model is trained on 5’-terminal insertion data , the prediction is consistent with outcomes from the same position , but not for 3’-terminal assay results . This indicates different mechanisms for two insertion positions: the assay with an element inserted upstream of a reporter gene may detect either the promoter or enhancer activity of the element . However , if the element is inserted downstream of the gene’s transcriptional start site or the 3’ terminal in the assay , the reporter readout may be specifically to the enhancer activity of the element . Large-scale experimental validation is required to further elucidate the underlying mechanisms . Our GRAM model can be further applied to fine mapping of functional SNVs . Particularly , the prediction results of GRAM could aid in the identification of variants that are most likely to directly modulate gene expression in a fine-mapping study . In addition , the impact of variants on gene regulation could vary across different cell types or individuals depending on differential transcriptional factor activity , which is represented by the expression level of TFs in our model . Based on this consideration , our model could potentially be used to evaluate the molecular effect of variants in a sample-specific manner . Given a group of patients with paired genotype and gene expression data , we could evaluate for each patient the expression-modulating effect of the variants of interest , which can be used to: 1 ) evaluate the patient-specific expression modulating effect for each variant; 2 ) identify distinct expression modulating patterns among the patient population; and 3 ) evaluate the overall variant effects by integrating results from different patients . Such knowledge could potentially contribute to our understanding of the molecular mechanism underlying disease-association of variants , and guide the characterization of patient-specific candidate variants for personalized diagnosis , prognosis and medical treatments . In summary , our GRAM model will be a useful tool for elucidating the underlying patterns of variants that modulate expression in a cell type- and tissue-specific context , and can be further applied to different samples of the same cell type or tissue . By leveraging the accumulating data generated from multiple cell lines , we can further improve for in-depth investigation in the future . We will keep abreast with the growing availability of comprehensive datasets and further expand our analyses .
We downloaded the dataset from R . Tewhey et al . ’s paper [22 , 23] . From about 79K tested elements , we only kept variants for which either reference or alternative allele elements show regulatory activity . This reduced the set to 3 , 222 SNVs in the GM12878 cell line and 1124 SNVs in the HepG2 cell line . Each SNV was extended in both directions by 74bp , for a total of 149bp . We used another dataset from Ulirsch 2016 [17] , which included 2 , 756 variants tested in the K562 cell line . The protein-protein interaction network used in our downstream analysis was constructed by merging all interaction pairs identified by BioGrid [56] , STRING [57] and InBio Map [58] . GERP features were extracted using the FunSeq2 annotation pipeline , which averages over the whole genome-scale GERP score over the elements . We downloaded phyloP [33] and Phastcons [34] scores from the UCSC genome browser data portal ( http://hgdownload-test . cse . ucsc . edu/goldenPath/hg19/ ) . We performed motif enrichment analysis using a hypergeometric test . To compare the motif break and gain scores , we removed the TFs that covered less than two variants for either emVARs or non-emVARs from the list of 40 TFs with the highest p-values in hypergeometric test . Then , we performed a Wilcoxon test on the motif break score . Motif break and motif gain scores were calculated using FunSeq2 . We also calculated the motif score using DeepBind [37] with both the SELEX and ChIP-Seq motif models . SELEX motif models were identified from in vitro systematic evolution of ligands by an exponential enrichment ( SELEX ) binding assay . ChIP-Seq models were inferred from sequences of TF binding sites from different cell lines . A total of 515 motif models were investigated ( S2 Table ) . To examine the importance of features , we compared different metrics learned from various models including LASSO stability selection [42] and Random Forest regression . The feature importance for each selection method was scaled to [0 , 1]; we took the mean of all the selection methods to represent the overall ranking . We compared our models’ mean standard error ( MSE ) with CADD , Eigen , LINSIGHT , FunSeq2 , GWAVA , and DeepSea . Features from the above tools were collected and tested using both SVR and Random Forest regression with three different input feature sets: SELEX-based features , ChIP-Seq-based features , and SELEX- and ChIP-Seq-based features combined . For other variant prioritization tools , we use their outputs as features to train the SVR and Random Forest models to predict the logSkew value . We labeled emVARs as positive and non-emVARs as negative classes following the definition of [22] , where ‘expression modulating’ means having a molecular effect that significantly increases or decreases regulatory activities . We calculated the emVAR and non-emVAR for both HepG2 , GM12878 and K562 cell lines from [17] [22] . For emVAR and non-emVAR , we further filtered using logSkew with an absolute value >0 . 5849 ( skew > 1 . 5 ) . In total , we used 3 , 222 data records , including 799 positives and 2 , 423 negatives . We built a three-step GRAM model ( Table 1 ) . Step 1 predicts the universal element regulatory activity U for both reference and alternative alleles . The ground-truth of regulatory activity is determined from results of experimental assay platforms , like a luciferase assay or MPRA . In these assays , an element inserted into a plasmid , either with or without a mutation , is characterized with regulatory activity if the fold change between the vector with the inserted element and the control is larger than a statistically significant cutoff . Specifically , the predictive target is defined as follows: for the MPRA study , where expression level of the reporter gene is directly measured , a statistical test based on DESeq2 was used to indicate whether the expression change is significant; for the luciferase assay , we regarded a testing element that has a fold change of fluorescence level greater than 1 . 5 or 2 compared to control ( like eGFP ) as a regulatory element . The predictive variable is the TF binding score from reference to alternative allele , which is estimated by DeepBind . A Random Forest classifier was then trained to predict the universal regulatory activity . The predicted log odds of probability between the reference and alternative allele was calculated as log2 ( U ( imut ) 1−U ( imut ) /U ( iwt ) 1−U ( iwt ) ) . Step 2 predicts the gene expression and TF binding cell type modifier scores . The cell type modifier score is defined according to the cell specificity of the experimental assay . For each variant , an MPRA experiment is performed on both the reference and alternative alleles , each paired with a null-control , resulting in a 2x2 categorical table of read counts in the MPRA experiments . The standard deviation of log ( odds ) of the categorical table ( n1 , n2 , n3 , n4 for the average reads count , Table 2 ) is calculated as 1n1+1n2+1n3+1n4 . For three different cell lines , GM12878 , GM19239 , and HepG2 , we constructed a vector of Vodds values for all the variants that are tested . By comparing principal component loading of the Vodds from three cell lines , we found that the two GM cell lines are closer to each other relative to HepG2 ( S3 Fig ) , which indicates that Vodds could reflect cell type information . We then further compared two groups of variants above the top quartile and below the bottom quartile of Vodds in GM12878 , and found that there were more non-emVAR variants in the top quantile group , which indicates that Vodds are also associated with the molecular effects of the variants . Based on these observations , we used the top and bottom quartile variants as positive and negative training sets , respectively , to predict the cell type modifier target . The TF expression profiles were used as input features for the prediction of the cell type modifier class . For each mutation , we re-ordered the expression of TFs based on their binding scores . Given 258 TFs with a DeepBind SELEX model score S for 3 , 222 SNVs , the TF expression matrix for each variant was adjusted and re-ordered using the rank of SELEX binding scores of the TFs bound to these SNVs’ region . For each variant , this results in a vector reflecting the expression of TFs relative to their binding strengths . That is , the first value in the vector represents the adjusted expression of the most influential TF bound to this region , i . e . the one with highest rank of binding scores , and so forth , regardless of what the TFs actually are . We then used the TF binding score and re-ordered gene expression to predict the cell type modifier label . The final model predicts the molecular effect of a variant using the estimated universal odds ratio and cell type modifiers from the two previous Steps . A LASSO model was used for the prediction . The LASSO model trained with L1 regularization is more robust and tolerant to noise . To achieve optimal predictive performance , we chose the regularization parameter lambda λ that gives minimal mean cross-validated error . We hold out one-fold of same variants for all steps and perform a 10-fold cross-validation ( S8 Fig ) . We first randomly permutate all the data by rows ( variants ) , and split them into ten evenly distributed subsets T ( 1 , 2… , 10 ) . We then iteratively hold out a subset Ti ( i = 1 , 2… , or 10 ) , and make sure Ti are not used for training in any steps . We trained the model using the remaining subset T−i ( −i: excluding i ) , and predicted the results of Ti to get Ti^ . Finally , we concatenated all Ti^’s and evaluated the performance using AUROC and AUPRC . We integrated data processing pipelines and the final model into a software pipeline called GRAMMAR ( S9 Fig ) , published on ( https://github . com/gersteinlab/GRAM ) . The user provides the variant list and gene expression data of each sample . The sequences with and without the variants are then extracted from the hg19 genome and provided as input for DeepBind . The GRAM model receives the DeepBind results and gene expression data and assigns a score for each provided variant in each sample . Finally , the program outputs the sample-specific GRAM scores for each sample , along with heatmap for all variants and samples . If variants from multiple regions are provided , each region is plotted individually . The software is also made available as a fine-mapping module to the more generalized FunSeq tool ( FunSeq . gersteinlab . org ) , taking in the variants prioritized by the first tool and outputting the subset of them that have a direct expression modulating function . The work by Dadaev et al . [10] reported 75 different LD blocks characterized by a known GWAS risk association for prostate cancer . Some of the SNPs in these regions were found to be significantly co-localized with identified eQTLs , annotated as eQTL SNPs . For our analysis , we selected five regions with the largest number of eQTL SNPs , which in total contains 561 eQTL SNPs . Genotype and gene expression data for 102 TCGA PRAD patients were obtained from the TCGA data portal . These data were then provided to the GRAMMAR pipeline described above . We plotted the estimated sample-wise GRAM scores for each region , and selected variants with the highest average GRAM scores as assumed causal variants for expression modulation . As a comparison , FunSeq [6] scores for each variant were also extracted based on position and allele . To analyze the impact of these variants on gene expression , we calculated the Pearson correlation between the sample-specific GRAM scores and expression of the target genes of each eQTL variant . Each regulatory region ( both reference and alternative alleles ) was separately synthesized . Enhancer regions were designed to include 250bp upstream and 250bp downstream for each enhancer region based on the candidate SNV site . These regions were then cloned into the pGL4 . 23[luc2/minP] vector ( Promega , Cat# E841A ) . Each candidate region was placed upstream of the minP promoter to determine the effect of each putative enhancer region on luciferase expression . In total , 100ng of each candidate construct and 100ng of Nano-luc control was co- transfected into MCF-7 cells ( 5 , 000 cells per well in DMEM media containing 10% FBS and 1% Penicillin-Streptomycin antibiotic ) using the Lipofectamine 3000 reagent ( Thermo Fisher , Cat# L3000001 ) according to the manufacturer’s instructions . Cells were incubated for 48 hrs before reading the luciferase signal using the Promega Nano-Glo luciferase kit ( Promega , Cat# N1521 ) according to the manufacturer’s instructions .
|
With advances in sequencing technologies , a deluge of genomic data is available; however , only a fraction of non-coding genomic variants are functionally relevant . Sifting through this data to prioritize genomic variants with respect to function is an important but challenging task . In this study , we built GRAM , a GeneRAlized Model , to predict the expression-modulating effects of non-coding variants in a cell-specific manner . GRAM combines a universal regulatory score defined by transcription factor binding with an easily obtainable modifier defined by transcription factor binding and expression to reflect the particular cell type . We evaluated this framework on multiple cell lines with high performance and showed that it could be applied to any cell line or sample with gene expression data . We also integrated GRAM into a practical software pipeline to fine-map causal variants that directly modulate gene expression among a larger linkage-disequilibrium block associated with a phenotype of interest . GRAM complements other general variant effect prediction methods–which often combine disparate features–by helping to precisely define the subset of prioritized variants that directly alters gene expression .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Methods"
] |
[
"cell",
"binding",
"cell",
"physiology",
"genome-wide",
"association",
"studies",
"luciferase",
"assay",
"luciferase",
"chemical",
"characterization",
"enzymes",
"enzymology",
"biochemical",
"analysis",
"enzyme",
"assays",
"genome",
"analysis",
"bioassays",
"and",
"physiological",
"analysis",
"sequence",
"motif",
"analysis",
"research",
"and",
"analysis",
"methods",
"sequence",
"analysis",
"genomics",
"bioinformatics",
"proteins",
"gene",
"expression",
"oxidoreductases",
"chemistry",
"binding",
"analysis",
"biochemistry",
"cell",
"biology",
"database",
"and",
"informatics",
"methods",
"genetics",
"biology",
"and",
"life",
"sciences",
"physical",
"sciences",
"computational",
"biology",
"chemical",
"elements",
"human",
"genetics"
] |
2019
|
GRAM: A GeneRAlized Model to predict the molecular effect of a non-coding variant in a cell-type specific manner
|
The chemical-genetic profile can be defined as quantitative values of deletion strains' growth defects under exposure to chemicals . In yeast , the compendium of chemical-genetic profiles of genomewide deletion strains under many different chemicals has been used for identifying direct target proteins and a common mode-of-action of those chemicals . In the previous study , valuable biological information such as protein–protein and genetic interactions has not been fully utilized . In our study , we integrated this compendium and biological interactions into the comprehensive collection of ∼490 protein complexes of yeast for model-based prediction of a drug's target proteins and similar drugs . We assumed that those protein complexes ( PCs ) were functional units for yeast cell growth and regarded them as hidden factors and developed the PC-based Bayesian factor model that relates the chemical-genetic profile at the level of organism phenotypes to the hidden activities of PCs at the molecular level . The inferred PC activities provided the predictive power of a common mode-of-action of drugs as well as grouping of PCs with similar functions . In addition , our PC-based model allowed us to develop a new effective method to predict a drug's target pathway , by which we were able to highlight the target-protein , TOR1 , of rapamycin . Our study is the first approach to model phenotypes of systematic deletion strains in terms of protein complexes . We believe that our PC-based approach can provide an appropriate framework for combining and modeling several types of chemical-genetic profiles including interspecies . Such efforts will contribute to predicting more precisely relevant pathways including target proteins that interact directly with bioactive compounds .
The collection of yeast deletion strains has enabled systematic genomewide functional analysis [1] . In addition , strain-specific molecular barcodes allow quantitative functional profiling of pooled deletion strains by using TAG oligonucleotide microarrays [1] , [2] . One of several types of functional profiles , the chemical-genetic profile , expresses quantitative values of deletion strains' growth defects under a chemical . The compendium of chemical-genetic profiles of heterozygous and homozygous deletion strains under different chemicals has been successfully used for identifying direct target proteins of those chemicals [3] , [4] as well as exploring their common mode-of-actions [5] . By integration of synthetic lethality profiles , the chemical-genetic profiles of homozygous deletion strains were also used to discover genes and pathways targeted by specific chemicals [6] . The chemical-genetic profiles in yeast are undoubtedly a useful resource to infer drug's action mechanism in human [7] . In the previous study , however , valuable biological information such as protein-protein and genetic interactions has not been integrated with chemical-genetic profiles in a single model for drug's target pathway prediction . Also , those profiles have not been yet modeled at the molecular level in terms of biological real entity such as protein complexes . As chemical-genetic profiles of many bioactive compounds are accumulating in many species including yeast , such real molecular entity-based modeling becomes essential to the molecular level understanding of phenotypes of the eukaryotic cell exposed to different chemicals . For example , that will allow us to infer drug's mode-of-action from chemical-genetic profiles more precisely than the previous model-free approach . In order to relate hidden biological activities of some real biological entities at the molecular level to the phenotypes of deletion strains at the organism level , we assume that the growth of a deletion strain is affected by hidden activities of protein complexes in a given condition , which leads to the observable population changes of the strain . This assumption comes from the molecular rationale of protein complexes for gene-to-phenotype relationships [8] , by which similar sensitive phenotypes of deletion strains of genes comprising a protein complex were explained in various different conditions . It is plausible that protein complexes can be regarded as functional units for the phenotypes of a deletion strain . If we view the entire cell as a factory comprising various distinct machines efficiently connected for its productivity [9] , genomewide protein complexes can be regarded as a collection of “cellular machines” connected with each other for optimal growth and survival of a cell ( Figure 1 ) . Based on the assumption that each protein complex should play a proper role as well as communicate efficiently with each other for cell survival and adaptation in various treatments , we developed a Bayesian factor analysis model . Our model is similar to the network models developed in transcription regulation studies [10]–[13] ( Figure 2 ) . The basic idea of the model is that the observed growth fitness measurements of each strain are determined by combined effects of the activities of PCs in each cell in a given treatment . To implement this idea , the association relationship between deletion strains and PCs is necessary , which led us to construct the binary ( 0 or 1 ) association network between the knockout genes of strains and PCs based on their physical or genetic interactions ( Figure 2C ) . Here , we assume that the relative growth fitness of strains by a chemical are mainly affected by deleterious interactions between the knockout-gene product of a strain and PCs which are physically or genetically linked ( Figure 1 ) . By modeling chemical-genetic profiles in terms of protein complexes using physical and genetic interactions as a priori knowledge , we inferred hidden activities of a collection of PCs in each cell exposed to different chemicals . Based on those PC activities , bioactive compounds with similar mode-of-action were clustered together . It means that the binary association network in our model represents biologically meaningful relationship between knockout genes of strains and protein complexes as hidden factors . In addition , we showed that protein complexes with similar function were clustered together , which implies that the unknown functions of protein complexes can be predicted . Finally , we presented a new effective method to predict drug's target pathway using our PC-based model . For example , we were able to highlight target-protein , TOR1 , of rapamycin as well as RUB1 , UBA3 , UBC12 , and ULA1 related to protein neddylation as relevant biological pathway for cellular toxicity of camptothecin . We believe that our protein complex-based approach can provide appropriate framework for combining and modeling several types of chemical-genetic profiles including interspecies . Such efforts will contribute to predict more precisely relevant pathway including target-proteins that interact directly with bioactive compounds .
We applied our PC-based Bayesian factor model ( see the details in Method Section ) to the compendium of chemical-genetic profiles ( ∼4 , 800 haploid deletion strains , 82 chemicals ) generated in the recent study [5] . This allowed us to infer the hidden activities of 488 PCs in 82 different chemicals ( Figure 3A and Figure S3 ) . To compare our model-based approach with the previous strain-based approach for predicting common mode-of-action of drugs using the same compendium , we first performed hierarchical clustering of the inferred PC activities and strains fitness itself ( Figure 3 , see the details in Method section ) . For fair comparison , two different dendrograms of 82 drugs should be validated against the gold standard drug-drug associations in the cellular context of Yeast . However , such reliable data were not available enough to measure performance of two dendrograms quantitatively . Therefore , we marked the common clusters and different clusters on the two dendrograms ( Figure 3B and 3C ) and surveyed their literature evidences , which were categorized as follows: In strain-based clustering of the compendium [5] , deletion mutant strains with similar chemical sensitivities are clustered together as well , grouping functionally related genes . Similarly , the protein complexes with a similar function were also clustered together ( Figure 3D ) . For examples , we selected three clusters with at least two protein complexes known for their functions . Their functional annotations of Gene Ontology were summarized in Table S1 . Here , we describe more literature evidences for those clusters . As shown in the above examples , the hierarchically clustered groups of protein complexes revealed insights into similar biological processes . In other words , unknown functions of protein complexes could be inferred based on those groups closely clustered . Nonetheless , it was preliminary investigation using simple hierarchical clustering method to highlight the biological essential features of protein complex activities . If chemical-PC profiles are applied to other clustering methods such as biclustering or network reconstruction methods , such results could be different in some cases as well as give more essential features on behaviors of protein complexes perturbed by drugs . To find a drug-target pathway , we first selected significantly sensitive protein complexes to each drug ( Figure 4 ) , and then examined the biological processes of deletion genes of strains that were associated with such sensitive protein complexes in our model ( see the details in the method; Table S2 ) . This investigation of PC-constrained strains allowed us to highlight more relevant drug-target pathways than that of PC-free strains . In particular , it is clear in cases of camptothecin and rapamycin whose target-pathways are well known ( Figures 5 and 6 ) . In a seminal article in the journal Cell entitled , “The cell as a collection of protein machines: preparing the next generation of molecular biologists , ” Bruce Alberts described a cell as a factory: “Indeed , the entire cell can be viewed as a factory that contains an elaborated network of interlocking assembly lines , each of which is composed of a set of large protein machine” [9] . In recent genomewide study of yeast , two independent groups , European Molecular Biology Laboratory ( EMBL ) , Cellzome ( a spin-off company from EMBL ) , and the university of Toronto , have surveyed the first comprehensive protein complexes , called protein machines by Bruce Alberts , using tandem affinity purification ( TAP ) [8] , [44] . Furthermore , Gavin et al . [8] gives molecular rationale of protein complexes for gene-to-phenotype relationship . In our study , those protein complexes are regarded as functional units for yeast cell growth , and then protein complex based Bayesian factor analysis is performed to relate growth fitness of genomewide deletion strains to hidden activities of protein complexes in a cell . In other words , at the organism's phenotype level , the chemical-genetic profiles representing relative growth fitness of systematic deletion strains are modeled in terms of protein complexes at the molecular levels . To show that our model assumption ( Figure 1 ) is reasonable and the inferred complex activities are reliable , hierarchical clustering analysis and literature survey were performed , which showed predictive power of common mode-of-action of bioactive compounds as well as grouping of protein complexes with similar biological behavior . In addition , we performed drug's target-pathway prediction based on our model assumption ( Figure 1 ) to show how practical our model framework is . GO analysis and literature survey shows that complex-based way of drug's target-pathway prediction narrows down lots of drug-sensitive genes involved in various biological pathway to a handful of genes important in the relevant pathway perturbed by a drug . For example , we were able to highlight target-protein , TOR1 , of rapamycin as well as RUB1 , UBA3 , UBC12 , and ULA1 related to protein neddylation as relevant biological pathway for cellular toxicity of camptothecin . From the purely computational standpoint , our model and Bayesian hidden component analysis ( BHCA ) developed by Sabatti and James [11] are essentially the same , as we pointed out in Method section . We did not try to improve the BHCA , even though there may be a number of ways to improve the algorithm itself , simply because that was not our main objective . Main focus of this study is to model the chemical-genetic profiles at the molecular level , more specifically using protein complexes , and we found that BHCA is an appropriate computational framework . The reasons are as follows: first , central assumption on a hidden factor is conceptually similar . The goal of BHCA in transcriptional regulation is to infer hidden activities of transcription factors under the assumption of combinatorial regulation of a gene by a set of transcription factors . Similarly , our goal is to infer hidden activities of protein complexes under the assumption of combinatorial effect of cell growth by a set of protein complexes . Under this assumption , we were able to set up BHCA-like Bayesian factor model , as shown in Figure 2 . Key component of the model is Z matrix ( binary association matrix ) , which not only allowed us to robustly compute the equation , but also provided a window through which relevant biological information such as protein-protein interactions and genetic interactions can be integrated with experimental data . Second , the estimation of protein complex activities by Bayesian factor model is more robust and stable because estimands are obtained by tens of thousands of samplings . Deterministic factor analysis methods [10] , [12] , [13] often give rises to numerically unstable solutions especially when improper prior knowledge is used in the analysis . In our modeling , protein complexes and genetic interactions as prior information are still insufficient and inaccurate , which makes it inappropriate to use deterministic methods such as network component analysis ( NCA ) for our study [10] . To improve the hierarchical clustering , Parsons et al . also utilized a matrix decomposition method known as probabilistic sparse matrix factorization ( PSMF ) . Both PSMF and our factor model all decompose chemo-genomic profiles into some factors and their weights . These decompositions of the two methods contribute to noise-reduction of original data , and so reveal more essential features . For example , Parsons et al . discussed two cases ( “factor 6” and “factor 5” in [5] ) where PSMF method improved the results over hierarchical clustering . For our method , we have already illustrated a number of those examples in Results and Discussion Section . Moreover , if we compare the results from the strain-based clustering , PSMF , and the present method ( PC-based clustering ) all together , our PC-based clustering tend to produce the results that are more similar to those of PSMF than those of strain-based clustering ( Figure S6 ) , indicating that both decomposition-based methods represent some essential features through noise-reduction from original observation . What is different , however , is that our model uses known protein complexes as factors and fixed relationships between factors and observations , while PSMF infers both of them from observations . Therefore , our model is suitable when prior knowledge on factors and observations is available , and PSMF is suitable in case of no prior knowledge . Direct performance comparison between both methods is difficult because only four blocks in the whole “factorgram” are available from [5] . One distinctive advantage of our method over PSMF , however , is that because we use protein complexes as factors and integrate biological prior knowledge in our model , the inferred results are biologically interpretable , which allowed us to develop an effective way to predict drug's target pathway . In contrast , in PSMF biological meaning of factors is not clear . In addition , our model provides hidden activities of protein complexes from original data , which are difficult to measure in real wet experiment . In current study , we used only chemical-genetic profiles of viable yeast haploid mutants as observed data for modeling . Consequently , it has a limitation for excluding the data for ∼1 , 000 essential genes in yeast . This drawback can be overcome by combining chemical-genetic profiles of homozygous and heterozygous deletion strains . We believe that our complex-based approach can provide appropriate framework for modeling such combined fitness data , which might especially contribute to predict more precisely relevant pathway including target-proteins that interact directly with bioactive compounds .
Various types of factor models have been developed to model the transcriptional regulatory networks [10]–[13] . In factor models for transcriptional regulation , transcription factors are defined as hidden factors so that estimated effects of the factors can be biologically interpreted as the activities of transcription factors directly involved in the transcription . In a similar way , we developed a protein complex based Bayesian factor analysis ( PCBA ) for modeling chemical-genetic profiles . In essence , our model is similar to Bayesian hidden component analysis ( BHCA ) model by Sabatti et al . developed for the transcriptional regulation network analysis [11] . For convenience , we used the same notations and equations as described in the original BHCA paper . The central assumption of our model was that the observed growth fitness measurements of each strain were determined by combined effects of the activities of a collection of the protein complexes ( PCs ) in each cell under different treatments . By log-transforming strains' fitness measurements , a linear model was formulated so that eit = ΣLj = 1 aij*pjt+γit , where eit represented the relative fitness of a strain i in an experiment t; aij the association strength of the protein complex j on the strain i; pjt the relative activity for the protein complex j in the experiment t; L the number of protein complexes as hidden factors; and finally γit the measurement errors and the biological variability . We assumed that γit was an independent and identically distributed Gaussian random variable following N ( 0 , σ2 ) . If the number of strains is N and the number of experiments is M , the model can be rewritten in a matrix notation as follows: ( 1 ) Here , E denoted an N×M matrix . A was a matrix and represented unknown association strengths between strains and protein complexes . P denoted the matrix . In our model , there are the following features: ( 1 ) the “factors” , row vectors in the P matrix have a clear interpretation , as they correspond to specific protein complexes . ( 2 ) The specific value of the pjt is the primary interest for the prediction of drug's mechanism . ( 3 ) The matrix A is known to contain a large number of zeroes , corresponding to the sparseness of the network . This sparseness in our model is very important feature to solve the notorious non-identifiable problem in the factor model , by which multiple sets of different parameter values lead to the nearly identical likelihood . Therefore , the identifiability should be achieved by imposing some sort of constraints in the model . A typical way in biological analysis is to constrain the loading matrix of factors by the same size of the matrix with a specific pattern of zeros . In our factor model , the sparse network topology as such constraints was represented in Z matrix with the same size as the A matrix but with only 0 or 1 values ( Figure S4 ) . The procedure for constructing Z matrix was shown in ( A ) and ( C ) of Figure 2 . The entries , zij were assumed to be independent . If there was the association between the knockout gene of a strain i and at least one of components in a protein complex j , we set zij = 1 . Otherwise we set zij = 0 . By letting πij = Pr ( zij = 1 ) , we assigned 1 to the πij because of limited memory size and computation time . By doing so , we lost the chance to remove false positive associations . The remaining entries of Z were set to zero . The other priors on A , P and σ|Z were also defined as follows: ( 2 ) The parameters aij , pjt and were assumed mutually independent . The priors on A , P and σ were mainly used for regularization . To explore the posterior distributions of the four parameter groups Z , A , P , and , we used the collapsed Gibbs sampling based on the full conditional distributions of Z , A , P , and , which were derived from those conjugate-like priors and joint probabilities , and described in the BHCA paper . We used the compendium of chemical-genetic profiles for 82 different bioactive compounds by screening them against the yeast haploid deletion collection , of ∼5 , 000 viable strains [5] . Each value in the profiles was represented by the combined average log2 ratio ( control/experiment ) of both barcodes ( up tag and down tag ) corresponding to each strain . We excluded multidrug-resistant genes [5] , and genes of deletion strains not associated with any protein complexes . Finally , the profiles of 3 , 241 strains were used for our Bayesian factor model . We used the collection of 490 protein complexes ( PCs ) defined by the first genomewide characterization in an organism , budding yeast , using affinity purification and mass spectrometry [8] . The 488 PCs associated with at least one strains were used for our Bayesian factor model . For PC–strain associations , we used the BioGRID version 2 . 0 . 20 release containing physical and genetic interactions known in S . cerevisiae [45] . We obtained the posterior distributions of all the parameters of A and P matrices in our models using the collapsed Gibbs sampler derived in the BHCA paper [11] . The following hyperparameters were set for the Gibbs sampling: α = 0 . 7 and β = 0 . 3 for the gamma distribution as a prior of inverse , and σa = 1 , 10 , 30 , 60 , 70 , 90 , 100 and σp = 1 , 10 , 30 , 50 for a priori standard normal distributions of aij and pjt . We first ran the Gibbs sampler for 11 , 000 iterations written in statistical language R [46] to select the optimal hyper-parameters , σa and σp . The initial 1000 iterations were discarded as burn-in , and the results of one per ten iterations were recorded , as suggested by Sabatti and James [11] . To assess the overall chain behavior , we monitored the sum of squared error ( SSE ) . Given hyper parameters , σa = 30 and σp = 1 , the convergence of SSE was best ( Figure S5 ) so that we obtained multiple chains by Gibbs sampler under such hyper-parameter condition . For each chain of those multiple chains , we collected about 400 to 500 samples , one per ten iterations , after discarding initial 5 , 000 iterations as burn-in . All of samples obtained from each chain were merged so that they were used as each posterior ( or sample ) distribution of each parameter . Then , posterior means and posterior variances of each posterior distribution were used as estimands of each parameter . In this study , we focused on the analysis of posterior mean of each element in P matrix because our goal is to infer the hidden relative activities of protein complexes . Hierarchical agglomerative clustering was performed by Gene Cluster 3 . 0 [47] for ( A ) and ( D ) of Figure 3 or by hclust function of standard R package stat [46] for ( B ) and ( C ) of Figure 3 using Pearson's correlation as a distance measure , and average linkage for compounds and complete linkage for protein complexes as an agglomeration method . The figures of clustering results were generated by Java TreeView [48] or by plclust function of standard R package stat . To estimate the significance of the effect of a drug on a protein complex , we modified the error model used for haploinsufficiency-based direct drug-target identification in Yeast [4] . For this , first , the relative activities of all the protein complexes under different compounds were expressed as a matrix P with rows of 1…i…488 complexes and columns of 1…j…82 compounds , and the reference set of each complex was defined as a collection of activities of each protein complex under different 82 compounds , {Pi , j = 1‥82} . For every activities ( Pij ) of a given protein complex under a given compound , the drug effect ( eij ) on a protein complex was calculated by subtracting the mean ( X̅i ) of its reference set from Pij , and the uncertainty ( ) of the drug effect was obtained by pooling of the variance ( ) of its reference set and sample variance ( ) of Pij obtained from tens of thousands of Gibbs sampling ( Equations 3 and 4 ) . ( 3 ) ( 4 ) where N represented the number of compounds , 82 . The drug effect ( eij ) on a protein complex and its pooled variance ( ) were applied to error function ( Equation 5 ) so that significance of the effect of a given compound to a given complex was scored as a real value in the range of 0 to 1 . ( 5 ) When error function scores of complexes were less than 0 . 25 , such complexes were regarded as being significantly affected by a given compound . When a protein complex has relatively positive value of activity , it was called a “sensitive complex” . In the opposite case , it was called a “resistant complex” . In similar way , we defined “sensitive strain” and “resistant strain” , all of whose relative fitness values had greater than 0 . 5 or less than −0 . 5 . Based on known biological associations between “sensitive/resistant complexes” and “sensitive/resistant strains” , Gene Ontology ( GO ) analysis for all of 82 compounds were performed , and those results were available at http://pombe . kaist . ac . kr/CMA/ModeOfAction . pl . Each set of genes of sensitive/resistant strains associated with sensitive/resistant complexes was used for GO analysis , which was also performed against all of genes of sensitive strains to compare complex-based with strain-based GO results in terms of highlighting relevant cellular pathway targeted by a compound .
|
Finding the specific targets of chemicals and deciphering how drugs work in our body is important for the effective development of new drugs . Growth profiles of yeast genomewide deletion strains under many different chemicals have been used for identifying target proteins and a common mode-of-action of drugs . In this study , we integrated those growth profiles with biological information such as protein–protein interactions and genetic interactions to develop a new method to infer the mode-of-action of drugs . We assume that the protein complexes ( PCs ) are functional units for cell growth regulation , analogous to the transcriptional factors ( TFs ) for gene regulation . We also assume that the relative cell growth of a specific deletion mutant in the presence of a specific drug is determined by the interactions between the PCs and the deleted gene of the mutant . We then developed a computational model with which we were able to infer the hidden activities of PCs on the cell growth and showed that yeast growth phenotypes could be effectively modeled by PCs in a biologically meaningful way by demonstrating that the inferred activities of PCs contributed to predicting groups of similar drugs as well as proteins and pathways targeted by drugs .
|
[
"Abstract",
"Introduction",
"Results/Discussion",
"Method"
] |
[
"computational",
"biology/systems",
"biology",
"computational",
"biology/genomics"
] |
2008
|
Inference of Protein Complex Activities from Chemical-Genetic Profile and Its Applications: Predicting Drug-Target Pathways
|
Diverse ion channels and their dynamics endow single neurons with complex biophysical properties . These properties determine the heterogeneity of cell types that make up the brain , as constituents of neural circuits tuned to perform highly specific computations . How do biophysical properties of single neurons impact network function ? We study a set of biophysical properties that emerge in cortical neurons during the first week of development , eventually allowing these neurons to adaptively scale the gain of their response to the amplitude of the fluctuations they encounter . During the same time period , these same neurons participate in large-scale waves of spontaneously generated electrical activity . We investigate the potential role of experimentally observed changes in intrinsic neuronal properties in determining the ability of cortical networks to propagate waves of activity . We show that such changes can strongly affect the ability of multi-layered feedforward networks to represent and transmit information on multiple timescales . With properties modeled on those observed at early stages of development , neurons are relatively insensitive to rapid fluctuations and tend to fire synchronously in response to wave-like events of large amplitude . Following developmental changes in voltage-dependent conductances , these same neurons become efficient encoders of fast input fluctuations over few layers , but lose the ability to transmit slower , population-wide input variations across many layers . Depending on the neurons' intrinsic properties , noise plays different roles in modulating neuronal input-output curves , which can dramatically impact network transmission . The developmental change in intrinsic properties supports a transformation of a networks function from the propagation of network-wide information to one in which computations are scaled to local activity . This work underscores the significance of simple changes in conductance parameters in governing how neurons represent and propagate information , and suggests a role for background synaptic noise in switching the mode of information transmission .
Gain scaling refers to the ability of neurons to scale the gain of their responses when stimulated with currents of different amplitudes . A common property of neural systems , gain scaling adjusts the system's response to the size of the input relative to the input's standard deviation [1] . This form of adaptation maximizes information transmission for different input distributions [1]–[3] . Though this property is typically observed with respect to the coding of external stimuli by neural circuits [1] , [3]–[7] , Mease et al . [8] have recently shown that single neurons during early development of mouse cortex automatically adjust the dynamic range of coding to the scale of input stimuli through a modulation of the slope of their effective input-output relationship . In contrast to previous work , perfect gain scaling in the input-output relation occurs for certain values of ionic conductances and does not require any explicit adaptive processes that adjust the gain through spike-driven negative feedback , such as slow sodium inactivation [4] , [9] , [10] and slow afterhyperpolarization ( AHP ) currents [10] , [11] . However , these experiments found that gain scaling is not a static property during development . At birth , or P0 ( postnatal day 0 ) , cortical neurons show limited gain scaling; in contrast , at P8 , neurons showed pronounced gain-scaling abilities [8] . Here , we examined how the emergence of the gain-scaling property in single cortical neurons during the first week of development might affect signal transmission over multiple timescales across the cortical network . Along with the emergence of gain scaling during the first week of neural development , single neurons in the developing cortex participate in large-scale spontaneously generated activity which travels across different regions in the form of waves [12]–[14] . Pacemaker neurons located in the ventrolateral ( piriform ) cortex initiate spontaneous waves that continue to propagate dorsally across the neocortex [13] . Experimentally , much attention has been focused on synaptic interactions in initiating and propagating activity , with a particular emphasis on the role of GABAergic circuits , which are depolarizing in early development [15] , [16] . While multiple network properties play an important role in the generation of spontaneous waves , here we ask how the intrinsic computational properties of cortical neurons , in particular gain scaling , can affect the generation and propagation of spontaneous activity . Changes in intrinsic properties may play a role in wave propagation during development , and the eventual disappearance of this activity as sensory circuits become mature . A simple model for propagating activity , like that observed during spontaneous waves , is a feedforward network in which activity is carried from one population , or layer , of neurons to the next without affecting previous layers [17] . We compare the behavior of networks composed of conductance-based neurons with either immature ( nongain-scaling ) or mature ( gain-scaling ) computational properties [8] . These networks exhibit different information processing properties with respect to both fast and slow timescales of the input . We determine how rapid input fluctuations are encoded in the precise spike timing of the output by the use of linear-nonlinear models [18] , [19] , and use noise-modulated frequency-current relationships to predict the transmission of slow variations in the input [20] , [21] . We find that networks built from neuron types with different gain-scaling ability propagate information in strikingly different ways . Networks of gain-scaling ( GS ) neurons convey a large amount of fast-varying information from neuron to neuron , and transmit slow-varying information at the population level , but only across a few layers in the network; over multiple layers the slow-varying information disappears . In contrast , nongain-scaling ( NGS ) neurons are worse at processing fast-varying information at the single neuron level; however , subsequent network layers transmit slow-varying signals faithfully , reproducing wave-like behavior . We qualitatively explain these results in terms of the differences in the noise-modulated frequency-current curves of the neuron types through a mean field approach: this approach allows us to characterize how the mean firing rate of a neuronal population in a given layer depends on the firing rate of the neuronal population in the previous layer through the mean synaptic currents exchanged between the two layers . Our results suggest that the experimentally observed changes in intrinsic properties may contribute to the transition from spontaneous wave propagation in developing cortex to sensitivity to local input fluctuations in more mature networks , priming cortical networks to become capable of processing functionally relevant stimuli .
We first characterized neuronal responses of conductance-based model neurons using methods previously applied to experimentally recorded neurons driven with white noise . The neuron's gain scaling ability is defined by a rescaling of the input/output function of a linear/nonlinear ( LN ) model by the stimulus standard deviation [8] . Using a white noise input current , we extracted LN models describing the response properties of the two neuron types to rapid fluctuations , while fixing the mean ( DC ) of the input current . The LN model [18] , [19] , [23] predicts the instantaneous time-varying firing rate of a single neuron by first identifying a relevant feature of the input , and after linearly filtering the input stimulus with this feature , a nonlinear input-output curve that relates the magnitude of that feature in the input ( the filtered stimulus ) to the probability of firing . We computed the spike-triggered average ( STA ) as the relevant feature of the input [18] , [24] , and then constructed the nonlinear response function as the probability of firing given the stimulus linearly filtered by the STA . Repeating this procedure for noise stimuli with a range of standard deviations ( ) produces a family of curves for both neuron types ( Figure 1A ) . While the linear feature is relatively constant as a function of the magnitude of the rapid fluctuations , , the nonlinear input-output curves change , similar to experimental observations in single neurons in cortical slices [8] . When the input is normalized by , the mature neurons have a common input-output curve with respect to the normalized stimulus ( Figure 1B , red ) [8] over a wide range of input DC . In contrast , the input-output curves of immature neurons have a different slope when compared in units of the normalized stimulus ( Figure 1B , blue ) . Gain scaling has previously been shown to support a high rate of information transmission about stimulus fluctuations in the face of changing stimulus amplitude [1] . Indeed , these GS neurons have higher output entropy , and therefore transmit more information , than NGS neurons ( Figure 1E ) . The output entropy is approximately constant regardless of for a range of mean ( DC ) inputs – this is a hallmark of their gain-scaling ability . The changing shape of the input-output curve for the NGS neurons results in an increasing output entropy as a function of ( Figure 1E ) . With the addition of DC , the output entropy of the NGS neurons' firing eventually approaches that of the GS neurons; this is accompanied with a simultaneous decrease in the distance between rest and threshold membrane potential of the NGS neurons as shown previously [8] . Thus , GS neurons are better at encoding fast fluctuations , a property which might enable efficient local computation independent of the background signal amplitude in more mature circuits after waves disappear . The response of a neuron to slow input variations may be described in terms of its firing rate as a function of the mean input through a frequency-current ( – ) curve . This description averages over the details of the rapid fluctuations . The shape of this – curve can be modulated by the standard deviation ( ) of the background noise [20] , [21] . Here , the "background noise'' is a rapidly-varying input that is not considered to convey specific stimulus information but rather , provides a statistical context that modulates the signaled information assumed to be contained in the slow-varying mean input . Thus , a neuron's slow-varying responses can be characterized in terms of a family of – curves parameterized by . Comparing the – curves for the two neuron types using the same conductance-based models reveals substantial differences in their firing thresholds and also in their modulability by ( Figure 1C , D ) . NGS neurons have a relatively high threshold at low , and the – curves are significantly modulated by the addition of noise , i . e . with increasing ( Figure 1C ) . In contrast , the – curves of GS neurons have lower thresholds , and show minimal modulation with the level of noise ( Figure 1D ) . This behavior is reflected in the information that each neuron type transmits about firing rate for a range of ( Figure 1F ) . This information quantification determines how well a distribution of input DC can be distinguished at the level of the neuron's output firing rate while averaging out the fast fluctuations . The information would be low for neurons whose output firing rates are indistinguishable for a range of DC inputs , and high for neurons whose output firing rates unambiguously differ for different DC inputs . The two neuron types convey similar information for large where the – curves are almost invariant to noise magnitude . For GS neurons , most information is conveyed about the input rate at low where the – curve encodes the largest range of firing rates ( 0 to 30 Hz ) . The information encoded by NGS neurons is non-monotonic: at low these neurons transmit less information because of their high thresholds , compressing the range of inputs being encoded . Information transmission is maximized at for which the – curve approaches linearity , simultaneously maximizing the range of inputs and outputs encoded by the neuron . For both neuron types , the general trend of decreasing information as increases is the result of compressing the range of outputs ( 10 to 30 Hz ) . These two descriptions characterize the different processing abilities of the two neuron types . GS neurons with their -invariant input-output relations of the LN model are better suited to efficiently encode fast current fluctuations because information transmission is independent of . However , NGS neurons with their -modulatable – curves are better at representing a range of mean inputs , as illustrated by their ability to preserve the range of input currents in the range of output firing rates . To characterize the spectrum of intrinsic properties that might arise as a result of different maximal conductances , and , we determined the – curves for a range of maximal conductances in the conductance-based model neurons ( Figure 2 ) . Mease et al . [8] previously classified neurons as spontaneously active , excitable or silent , and based on the neurons' LN models determined gain-scaling ability as a function of the individual and for excitable neurons . Models with low had nonlinear input-output relations that did not scale completely with , while models with high had almost identical nonlinear input-output relations for all [8] . Therefore , gain scaling ability increased with increasing ratio , independent of each individual conductance . We examined the modulability of – curves by in excitable model neurons while independently varying and ( Figure 2 ) . Like gain scaling , the modulability by also depended only on the ratio , rather than either conductance alone , with larger modulability observed for smaller ratios . To further explore the implications of such modulability by , we computed the mutual information that each model neuron transmits about mean inputs for a range of ( Figure 2 ) . Neurons with behaved like GS neurons in Figure 1F , while neurons with behaved like NGS neurons . These results suggest that the ability of single neurons to represent a distribution of mean input currents by their distribution of output firing rates can be captured only by changing the ratio of and . Therefore , we focused on studying two neuron types with in the two extremes of the conductance range of excitable neurons: GS neurons with and NGS neurons with . Upon characterizing single neuron responses of the two neuron types to fast-varying information via the LN models and to slow-varying information via the – curves , we compared their population responses to stimuli with fast and slow timescales . A population of uncoupled neurons of each type was stimulated with a common slow ramp of input current , and superimposed fast-varying noise inputs , generated independently for each neuron ( Figure 3A ) . The population of NGS neurons fired synchronously with respect to the ramp input and only during the peak of the ramp ( Figure 3B ) , while the GS neurons were more sensitive to the background noise and fired asynchronously during the ramp ( Figure 3C ) with a firing rate that was continuously modulated by the ramp input . This suggests that the sensitivity to noise fluctuations of the GS neurons at the single neuron level allows them to better encode slower variations in the common signal at the population level [25]–[27] , in contrast to the NGS population which only responds to events of large amplitude independent of the background noise . During cortical development , wave-like activity on longer timescales occurs in the midst of fast-varying random synaptic fluctuations [13] , [14] , [28] , [29] . Therefore , we compared the population responses of GS and NGS neurons to a slow-varying input ( 500 ms correlation time constant ) common to all neurons with fast-varying background noise input ( 1 ms correlation time constant ) independent for all neurons ( Figure 3D ) . The distinction between the two neuron types is evident in the mean population responses ( peristimulus time histogram , i . e . PSTH ) . The NGS population only captured the stimulus peaks ( Figure 3E ) while the GS population faithfully captured the temporal fluctuations of the common signal , aided by each neuron's temporal jitter caused by the independent noise fluctuations ( Figure 3F ) . Although not an exact model of cortical wave development , this comparison supports the hypothesis that the intrinsic properties of single neurons can lead to different information transmission capabilities of cortical networks at different developmental time points , and the transition from wave propagation to wave cessation . The observed difference between the population responses of the GS and NGS neurons to the slow-varying stimulus in the presence of fast background fluctuations ( Figure 3D–F ) suggested that the two neuron types differ in their ability to transmit information at slow timescales . Therefore , we next examined how the identified single neuron properties affect information transmission across multiple layers in feedforward networks . Networks consisted of 10 layers of 2000 identical neurons of the two different types ( Figure 4A ) . The neurons in the first layer receive a common temporally fluctuating stimulus with a long correlation time constant ( 1 s , see Methods ) ; neurons in deeper layers receive synaptic input from neurons in the previous layer via conductance-based synapses . Each neuron in the network also receives a rapidly varying independent noise input ( with a correlation time constant of 1 ms ) to simulate fast-varying synaptic fluctuations . The noise input here is a rapidly-varying input that sets the statistical context for the slow-varying information; it does not transmit specific stimulus information itself . The GS and NGS networks have strikingly different spiking dynamics ( Figure 4B ) . The GS network responds with higher mean firing rates in each layer , as would be expected from the – curves characterizing intrinsic neuronal properties ( Figure 1C , D ) . While the GS neurons have a baseline firing rate even at zero input current , the NGS neurons only fire for large input currents , with a threshold dependent on the level of intrinsic noise; thus , the two neuron types have different firing rates . To evaluate how the networks transmit fluctuations of the slow-varying common input signal , independent of the overall firing rates , we evaluated the averaged population ( PSTH ) response of each layer , normalized to have a mean equal to 0 and a variance equal to 1 ( Figure 4C ) . The first few layers of the GS network robustly propagate the slow-varying signal as a result of the temporally jittered response produced by the sensitivity to fast fluctuations at the single neuron level , consistent with the population response in Figure 3F . However , due to the effects of these same noise fluctuations , this population response degrades in deeper layers ( Figure 4C , left , see also Figure S1 for ) . In contrast , the NGS network is insensitive to the fast fluctuations and thresholds the slow-varying input at the first layer , as in Figure 3E . Despite the presence of fast-varying background noise , the NGS network robustly transmits the large peaks of this stimulus to deeper layers without distortion ( Figure 4C , right ) . This difference in the transmission of information through the two network types is captured in the information between the population response and the slow-varying stimulus in Figure 4D . The GS network initially carries more information about the slow-varying stimulus than the NGS network; however , this information degrades in deeper layers when virtually all the input structure is lost , and drops below the NGS network beyond layer four ( Figure 4D , bottom ) . While the information carried by the NGS network is initially lower than the GS network ( due to signal thresholding ) , this information is preserved across layers and eventually exceeds the GS information . The observed differences in the propagation of slow-varying inputs between the two network types resemble changes in wave propagation during development . While spontaneous waves cross cortex in stereotyped activity events that simultaneously activate large populations of neurons at birth , these waves disappear after the first postnatal week [13] , [16] . We have demonstrated that immature neurons lacking the gain-scaling ability can indeed propagate slow-varying wave-like input of large amplitude as population activity across many layers . As these same neurons acquire the ability to locally scale the gain of their inputs and efficiently encode fast fluctuations , they lose the ability to propagate large amplitude events at the population level , consistent with the disappearance of waves in the second postnatal week [13] . While many parameters regulate the propagation of waves [14] , [29] , our network models demonstrate that varying the intrinsic properties of single neurons can capture substantial differences in the ability of networks to propagate slow-varying information . Thus , changes in single neuron properties can contribute to both spontaneous wave generation and propagation early in development and the waves' disappearance later in development . The layer-by-layer propagation of a slow-varying signal through the population responses of the two networks can be qualitatively predicted using a mean field approach that bridges descriptions of single neuron and network properties . Since network dynamics varies on faster timescales than the correlation timescale of the slow-varying signal , the propagation of a slow-varying signal can be studied by considering how a range of mean inputs propagate through each network . The intrinsic response of the neuron to a mean ( DC ) current input is quantified by the – curve which averages over the details of the fast background fluctuations; yet , the magnitude of background noise , , can change the shape and gain of this curve [20] , [21] . Thus , for a given neuron type , there is a different – curve depending on the level of noise , ( Figure 1C , D ) . One can approximate the mean current input to a neuron in a given layer , , from the firing rate in the previous layer through a linear input-output relationship , with a slope dependent on network properties ( connection probability and synaptic strength , see Eq . 15 ) . Given the estimated mean input current for a given neuron in layer , , the resulting firing rate of layer , , can then be computed by evaluating the appropriate – curve , , which characterizes the neuron's intrinsic computation ( 1 ) Thus , these two curves serve as an iterated map whereby an estimate of the firing rate in the Lth layer , , is converted into a mean input current to the next layer , , which can be further converted into , propagating mean activity across multiple layers in the network ( Figures 5 , 6 ) . While for neurons in the first layer , the selected – curve is the one corresponding to the level of intrinsic noise injected into the first layer , , for neurons in deeper layers , the choice of – curve depends not only on the magnitude of the independent noise fluctuations injected into each neuron , but also on the fluctuations arising from the input from the previous layer ( see Eq . 16 in Methods ) . The behavior of this iterated map is shaped by its fixed points , the points of intersection of the – curve with the input-output line , which organize the way in which signals are propagated from layer to layer . The number , location and stability of these fixed points depend on the curvature of and on ( Figure 5 ) . When the slope of at the fixed point is less than , the fixed point is stable . This implies that the entire range of initial DC inputs ( into layer 1 ) will tend to iterate toward the value at the fixed point as the mean current is propagated through downstream layers in the network ( Figure 5 , left ) . Therefore , all downstream layers will converge to the same population firing rate that corresponds to the fixed point . In the interesting case that becomes tangent to the linear input-output relation , i . e . the – curve has a slope equal to , the map exhibits a line attractor: there appears an entire line of stable fixed points ( Figure 5 , middle ) . This ensures the robust propagation of many input currents and population rates across the network . Interestingly , the – curves of the GS and NGS neurons for different values of fall into one of the regimes illustrated in Figure 5: GS neurons with their -invariant – curves have a single stable fixed point ( Figure 5 , left ) , while the NGS neurons have line attractors with exact details depending on ( Figure 5 , middle and right ) . The mechanics of generating a line attractor have been most extensively explored in the context of oculomotor control ( where persistent activity has been interpreted as a short-term memory of eye position that keeps the eyes still between saccades ) and decision making in primates ( where persistent neural activity has been interpreted as the basis of working memory ) [30] . Indeed , Figure 6A , B shows that the – curves for GS neurons at two values of , one low and one high , are very similar . The mean field analysis predicts that all initial DC inputs applied to layer 1 will converge to the same stable fixed point during propagation to downstream layers . Numerical simulations corroborate these predictions ( Figure 6A , B , bottom ) . A combination of single neuron and network properties determine the steady state firing rate through ( Eq . 15 ) . Activity in the GS networks can propagate from one layer onto the next with relatively weak synaptic strength even when the networks are sparsely connected ( 5% connection probability ) , as a result of the low thresholds of these neurons ( Figure 1D ) . The specific synaptic strength in Figure 6A , B was chosen arbitrarily so that the – curve intersects the input-output line with slope , but choosing different synaptic strength produces qualitatively similar network behavior ( Figure S2 ) . The parameter can be modulated by changing either the connectivity probability or the synaptic strength in the network; as long as their product is preserved , remains constant and the resulting network dynamics does not change ( Figure S2 ) . Furthermore , as a result of the lack of modulability of GS – curves by ( Figure 1D ) , the network dynamics remains largely invariant to the amplitude of background noise . In contrast , the amplitude of background noise fluctuations , , has a much larger impact on the shape of NGS – curves ( Figure 1C ) and on the resulting network dynamics ( Figure 5 ) . When the combination of sparse connection probability and weak synaptic strength leads to the slope being too steep ( weak connectivity in GS networks , Figure 6A , B ) , there may be no point of intersection with the NGS – curves: all DC inputs are mapped below threshold and activity does not propagate to downstream layers . Keeping the same sparse connection probability of and increasing synaptic strength enables the propagation of neuronal activity initiated in the first layer to subsequent layers in NGS networks . For a particular value of , there is an entire line of stable fixed points in the network dynamics ( Figure 5 , middle ) , so that a large range of input currents are robustly transmitted through the network . More commonly , however , the map has three fixed points: stable fixed points at a high value and at zero , and an intermediate unstable fixed point ( Figure 6C , D ) . In this case , mean field theory predicts that DC inputs above the unstable fixed point should flow toward the high value , while inputs below it should iterate toward zero , causing the network to stop firing . However , the map still behaves as though the – curve and the input-output transformation are effectively tangent to one another over a wide range of input rates ( green box in Figure 6C , D ) , creating an effective line of fixed points for which a large range of DC inputs is stably propagated through the network; this is generically true for a wide range of noise values , although the exact region of stable propagation depends on the value of ( Figure 5 , middle and right , Figure S3 ) . The best input signal transmission is observed when the network noise selects the most linear – curve that simultaneously maximizes the range of DC inputs and population firing rates of the neurons ( Figure 5 , middle ) . This is approximately the noise value selected in Figure 6C , D . We call this a stable region of propagation for the network since a large range of mean DC inputs can be propagated across the network layers so that the population firing rates at each layer remain distinct . Our results resemble those of van Rossum et al . [31] where regimes of stable signal propagation were observed in networks of integrate-and-fire neurons by varying the DC input and an additional background noise . The best regime for stable signal propagation occurred for additive noise that was large enough to ensure that the population of neurons independently estimated the stimulus , as in our NGS networks ( Figure 5 , middle and right , Figure S3 ) . The emergence of extended regions of stable rate propagation implies that the NGS mean field predictions ( Figure 6C , D , bottom ) are less accurate than for the GS networks where the convergence to the stable fixed points is exact ( Figure 6A , B ) . However , the NGS mean field predictions show qualitative agreement with the simulation results , in particular in the initial network layers where the approach to the nonzero stable fixed point is much slower than in the GS networks , i . e . occurs over a larger number of layers . Along with the slow convergence of firing rates toward a single population firing rate , the ability of network noise to modulate the NGS – curves suggests that multiple – curves can be used to predict network dynamics by combining added and intrinsically generated noise ( see Eq . 16 ) . As a result , for some input currents ( e . g . arrow in Figure 6C ) the firing rate goes down in the first three layers where network dynamics predicts convergence to the zero stable fixed point . The initial decrease of firing rate is due to the disappearance of weak synaptic inputs that cannot trigger the cells to spike . Network noise then selects a different – curve that shifts the dynamics into the rate stabilization region ( Figure 6C , green box ) where firing rates are stably propagated . The onset of synchronous firing of the neuronal population in each layer also contributes to rate stabilization . Population firing rates in deeper layers increase to a saturating value lower than the mean field predicted value . Similar results have been observed experimentally [32] and in networks of Hodgkin-Huxley neurons [33] . We find similar network dynamics for a more weakly connected NGS network using the smallest possible synaptic strength that allows activity to propagate through the network ( Figure S2 ) . As for the GS networks , as long as the product of connection probability and synaptic strength is constant , the slope of the input-output linear relationship , and the network dynamics remain unchanged , even if these network parameters change individually ( Figure S2 ) . An exception to this result is observed at very sparse connectivity ( 2% ) , where network behavior is more similar to the GS networks ( Figure S2 , bottom right ) . At this sparse connectivity , independent noise reduces the common input across different neurons and synchrony is less pronounced . This argues that the emergence of synchrony plays a fundamental role in achieving reliable propagation of a range of DC inputs ( and correspondingly population firing rates ) in the NGS networks . Although experimental measurements of the connectivity probability in developing cortical networks are lacking , calcium imaging of single neurons demonstrates that activity across many neurons during wave propagation is synchronous [34] . Intracellular recordings of adult cultured cortical networks also demonstrate that synchronous neuronal firing activity is transmitted in multiple layers [32] . To examine network behavior for comparable connectivity strength , we repeated the network simulations and mean field predictions of mean DC input propagation in GS networks with the same increased synaptic strength needed for propagation of activity in the NGS networks . We found that the behavior was similar to the weakly connected GS network: Regardless of the initial input current , the network output converged to a single output firing rate by layer 5 ( Figure 6E , F ) , making these networks incapable of robustly propagating slow-varying signals without distortion . As for the strongly connected NGS networks , neurons across the different layers in these strongly connected GS networks developed synchronous firing . This synchrony led to a small difference ( several Hz ) between the final firing rate approached by each network compared with the firing rate predicted from the mean field analysis . Although both the strongly connected GS and NGS networks developed synchronous firing , the behavior of the two types of networks remained different ( Figure 6 ) . The results in this section indicate that firing rate transmission depends on the details of single neuron properties , including their sensitivity to fast fluctuations as characterized by the LN models ( Figure 1A , B ) . Firing rate transmission also depends on the modulability of the – curves by the noise amplitude ( Figure 1C , D ) . Because of these differences in intrinsic computation , the GS and NGS networks show distinct patterns of information transmission ( Figure 5 ) : firing rate convergence to a unique fixed point , or a line of fixed points ensuring stable propagation of firing rates which can be reliably distinguished at the output , respectively . In the latter case , even when a line of fixed point is not precisely realized as in Figure 5 ( middle ) , competition between the slow convergence of firing rates to the mean field fixed point and the emergence of synchrony enable the propagation of firing rates through the different network layers , aided by the range of – curves sampled by network noise with amplitude . Given the predicted signal propagation dynamics , we now directly compute the mutual information between the mean DC input injected into layer 1 and the population firing rates at a given layer for each magnitude of the independent noise ( Figure 7 ) . This measures how distinguishable network firing rate outputs at each layer are for different initial mean inputs . The convergence of population firing rates across layers to a single value in the GS networks leads to a drop in information towards zero for both the weakly ( Figure 6A , B ) and strongly connected GS networks ( Figure 6E , F ) as a function of layer number and for a wide range of network noise ( Figure 7A , C ) . NGS networks can transmit a range of mean DC inputs without distortion ( Figure 6C , D ) ; thus , the information between input DC and population firing rate remains relatively constant in subsequent layers ( Figure 7B ) . The information slightly increases in deeper layers due to the emergence of synchronization , which locks the network output into a specific distribution of population firing rates . As noise amplitude increases , the selected – curve becomes tangent to the linear input-output relationship over a larger range of input firing rates ( Figure 6C , D ) ; hence , a larger range of inputs is stably transmitted across network layers . Counterintuitively , this suggests that increasing noise in the NGS networks can serve to increase the information such networks carry about a distribution of mean inputs . The differential ability of GS and NGS networks to reliably propagate mean input signals is predicted by the modulability of the – curves by the network noise . To understand the dynamical origins of this difference , we analytically reduced the neuron model ( Eq . 2 ) to a system of two first order differential equations describing the dynamics of the membrane potential and an auxiliary slower-varying potential variable ( Methods ) [35] . We analyzed the dynamics in the phase plane by plotting vs . . The nullclines , curves along which the change in either or is 0 , organize the flows of and ( Figure 8 ) ; these lines intersect at the fixed points of the neuron's dynamics . We studied the fixed points at different ratios of and , with a particular focus on the values discussed above ( and ) . These exhibit substantial differences in the type and stability of the fixed points , as well as the emergent bifurcations where the fixed points change stability as one varies the mean DC input current into the neuron ( Figure 8 ) . For a large range of DC inputs , the NGS neuron ( ) has a single stable fixed point ( either a node or a focus ) ( Figure 8A ) . In this case , the only perturbation that can trigger the system to fire an action potential is a large-amplitude noise current fluctuation . The of the current then determines the number of action potentials that will be fired in a given trial and strongly modulates the firing rate of the neuron . We show two trajectories at pA and 50 pA and at two different DC values of 0 and 30 pA ( Figure 8A ) , at which the – curves are strongly noise-modulated ( Figure 1C ) . As the DC increases beyond 62 pA , the fixed point becomes unstable and a stable limit cycle emerges ( not shown ) . In this case , any will move the trajectories into the stable limit cycle and the neuron will continuously generate action potentials , with a firing rate independent of . Indeed , Figure 1C shows that the – curves become less effectively modulated by for DC values greater than 62 pA . As the conductance ratio increases , the range of DC values for which the system has a single fixed point decreases ( Figure 8B ) . Indeed , the GS neuron ( ) has a stable limit cycle for the majority of DC values ( Figure 8C ) . This implies that GS neurons are reliably driven to fire action potentials for any and their firing rate is not very sensitive to . For low DC values , the stable limit cycle coexists with a stable fixed point , so in this case of the noise can modulate the firing rate more effectively , as is seen in Figure 1D . This analysis highlights the origins for the differential modulability of firing rate in NGS and GS neurons . Although the model reduction sacrifices some of the accuracy of the original model , it retains the essential features of action potential generation: the sudden rise of the action potential which turns on a positive inward sodium current , and its termination by a slower decrease in membrane potential which shuts off the sodium current and initiates a positive outward potassium current hyperpolarizing the cell . Although simpler neuron models ( e . g . binary and integrate-and-fire [36]–[38] ) allow simple changes in firing thresholds , the dynamical features inherent in the conductance-based neurons studied here are needed to capture noise-dependent modulation .
The adult brain exhibits a diversity of cell types with a range of biophysical properties . Organized into intricate circuits , these cell types contribute to network computation , but the role of intrinsic properties is unclear . Recently , we have shown that during early development , single cortical neurons acquire the ability to represent fast-fluctuating inputs despite variability in input amplitudes by scaling the gain of their responses relative to the scale of the inputs they encounter [8] . Before these intrinsic properties shift , the developing cortex generates and propagates spontaneous waves of large-scale activity [13] , [22] , [39] , [40] , which regulate developmental changes in ion channel expression , synaptic growth and synaptic refinement processes [29] , [41] , [42] . How do experimentally observed biophysical properties affect ongoing network dynamics at this time ? Using model neurons with conductance properties chosen to reproduce this developmental change in gain scaling , we investigated the implications of this change on the ability of feedforward networks to robustly transmit slow-varying wave-like signals . The conductance-based models that we considered are not intended as an exact biophysical model for developing cortical neurons; rather they allow us to study the more fundamental question of the role of single neuron computation on network behavior in a case with a well-defined and physiologically relevant network level property . We add to previous studies by considering first , the fidelity of propagation of temporally varying patterns by biophysically realistic neurons , basing our work in a biological context where the brain naturally enters a state of wave propagation . Second , our work highlights a role of cellular processes in large-scale network behavior that has rarely been studied . Our results implicate intrinsic conductance change as a way to switch between global synchronization and local responsiveness , rather than synaptic plasticity , which is typically used to evoke such a global network change [17] . Related changes in excitability that accompany the cessation of spontaneous activity have been observed in the mouse embryonic hindbrain , where they have been ascribed to hyperpolarization of resting membrane potential and increased resting conductance of channels [43] . Finally , we analyze network information transmission on two different timescales ( local fluctuations and network-wide wave-like events ) and thereby generalize previous classification of feedforward network propagation into either synchrony-based coding [32] , [44] , and rate-based coding [31] , [45] . We use two different descriptions of neuronal properties to characterize the neuron's ability to propagate information at these different time- and lengthscales . The processing of fast input fluctuations can be characterized using LN models [8] , [46]–[48] . While single neuron properties affect the linear feature [46] , [48] , [49] , here we focus on the scaling of the nonlinearity in the LN model to stimuli of different amplitudes . Information about slowly modulated input is described using noise-modulated – curves [20] , [21] , [50] . This ability of developing neurons to transmit distinct information at two different timescales is an example of a temporally multiplexed code [3] , [51]–[53] . Here , GS neurons perform temporal multiplexing as they simultaneously convey distinct information about fast and slow fluctuations , reliably encoding slowly varying stimuli , albeit only for a few network layers . The NGS neurons also implement a multiplexed code because of their dual role to transmit firing rates while maintaining synchrony . The above characterizations predict the success of global information propagation across multiple network layers [49] , [50] . In integrate-and-fire network models with a fixed – curve , different network dynamics has been achieved by varying connectivity probability and synaptic strength [31] , [45] , [54] , [55] . Here , in addition we considered the modulation of the – curves by the combined effects of injected independent noise and measured correlated noise from network interactions , permitting a description of network responses dependent on the input statistics , intrinsic single neuron properties and network connectivity ( Figure 6 ) . The role of -modulated – curves has also been fundamental in understanding how intrinsic neuron properties affect correlation transfer and encoding of rate- and synchrony-based signals in reduced networks of two neurons stimulated with a common input signal and independent noise [48] , [49] , [52] , [53] , [56] . We expect that generalizations of these methods will enable improved theoretical predictions for firing rate and correlation transfer beyond mean field , by computing the effects of temporal correlations such as we observe . Firing rate transmission in our NGS networks co-occurs with the development of precise spike-time synchronization over a wide range of stimulus statistics and network connectivity ( Figure 6 ) . This synchronization might be a feature of biologically inspired networks because similar patterns were reported in experimentally simulated feedforward networks in vitro [32] and Hodgkin-Huxley-based simulations [33] , but not in networks of threshold binary neurons [36] , [57] , nor integrate-and-fire neurons [55] . Several manipulations to single neuron or network properties might reduce this synchrony . These include: introducing sparse connectivity with strong synapses [17] , [37] , increasing independent noise input [31] , [36] , or embedding the feedforward into recurrent networks with inhibition to generate asynchronous background activity [37] , [38] , [55] , [58]; but these typically result in signal degradation or implausible assumptions in our models . We did not find a regime supporting reliable asynchronous rate propagation , consistent with other studies [32] , [33] , [36] , [44] . We identified the biophysical basis of the single-unit properties that underlies our results . The change in gain scaling is accompanied by a difference in the distance from rest to threshold membrane potential [8]: GS neurons have a smaller distance to threshold and are more likely to fire driven by noise fluctuations , while NGS neurons have a larger distance to threshold and must integrate many coincident inputs to fire . Indeed , a change in spiking threshold in simpler model neurons has been shown to modulate the mode of signal transmission in a feedforward network [36] , [59] , [60] . However , our mean-field and phase-plane dynamical analyses together show that threshold is not the only factor at work: the nature of rate propagation is intimately connected with the bifurcation properties of the neuron model . While we focused on two representative contrasting cases , these properties vary systematically with the conductance ratio of the neuron and we have mapped out the spectrum of possible behaviors of this model . The robustness of information propagation across network layers is likely to have important implications for how developmental information contained in wave propagation patterns is transmitted across the cortex . We have previously shown that cortical waves are initiated in a pacemaker circuit contained within the piriform cortex [12]–[14] , which is likely to provide the strong input necessary to drive NGS neurons . The waves propagate dorsally across the neocortex so that throughout the developmental period of wave generation , the neocortex acts as a follower region in the sequence of wave propagation . The reliability with which firing patterns of piriform neurons are retained as waves propagate into the neocortex will determine the nature of developmental information that the neocortex receives from those waves during its development . As gain scaling develops , more mature neurons can support efficient coding of local fluctuations and discard information about network-wide events . Therefore , the alteration of a single developmentally regulated conductance parameter can shift cortical neurons from synchrony-based encoders of slow inputs to noise-sensitive units that respond with high fidelity to local fluctuations independent of the overall scale . The growing sensitivity to noise of cortical neurons in the first postnatal week might help to prevent large-scale wave activity from dominating adult neural circuits , thus discouraging epileptiform patterns of network activity . At the same time , the emergence of gain scaling supports a transition to a state in which cortical circuits , rather than participating in network-wide events , can respond optimally to appropriately scaled local information , breaking up the cortical sheet into smaller information-processing units . The mature cortex is also capable of generating spontaneous activity that propagates over large distances in the absence of sensory stimulation [61]–[63] . Such wave activity is postulated to be involved in short-term memory and the consolidation of recent transient sensory experience into long-lasting cortical modifications . For example , recent in vivo experiments proposed that synaptic plasticity is enforced by slow waves that occur during sleep [64 , 65] . Spontaneous propagation activity patterns emerge from the interplay of intrinsic cellular conductances and local circuit properties [63]; our results raise the possibility that modulation of intrinsic properties through slow Na+ inactivation or neuromodulation could have multiple short-term effects on cortical information processing . While we have examined the effect of gain scaling as a specific form of adaptation emerging during development , other adaptation mechanisms also likely play an important role in information transmission in feedforward networks . For instance , spike frequency adaptation has been shown to have effects that accumulate across multiple layers of feed-forward networks [31] . This widely observed form of adaptation can arise from calcium-dependent potassium conductances which generate AHPs [21] , [66] , [67] . Indeed , we and others have found that AHP-generating conductances can also support gain scaling behavior by single neurons [9] , [68] . Independent of AHP conductances , slow sodium channel inactivation can also contribute to spike frequency adaptation [69] , [70] . Incorporating such slow-timescale channel dynamics will require taking into account temporal aspects of the coding of mean ( or variance ) [71] that are presently ignored in our mean-field analysis based on modulated – curves . These slow dynamics may contribute to successive layers of filtering that affect information transmission [10] . An analytical characterization of the impact of slow neuronal dynamics on networks is likely to require novel theoretical approaches beyond those used here . Similarly , other factors beyond the specific changing intrinsic neuronal properties addressed here contribute to the generation of spontaneous cortical waves with complex spatio-temporal properties . During the same developmental time period , the cortex undergoes substantial changes in information processing capacity that are beyond the scope of the present study [72]–[74] . Activity-dependent modification of synaptic connections driven by developmental cues contained in spontaneous wave patterns are likely to refine cortical networks into their mature state [14] , [16] , [39] , [42] , [73] . Furthermore , the emergence of synaptic inhibition as GABA becomes more hyperpolarizing contributes to diminishing the wave-like activity generated by the immature excitatory network [14] , [73] . Thus , synaptic plasticity and intrinsic neuronal properties interact to modulate the emergence , propagation and the eventual disappearance of spontaneous waves in the developing cortex , and also to endow spatially-distinct regions at different time points with different information processing capabilities .
We studied a modified version of a Hodgkin-Huxley style model adapted by Mainen et al . [75] for spike initiation in neocortical pyramidal neurons . The model consists of a leak current , mammalian voltage-gated transient sodium and delayed-rectified potassium currents with maximal conductances , and , and reversal potentials mV , mV and mV: ( 2 ) where µF/cm is the specific membrane capacitance and is the input current with denoting the area of the membrane patch with radius of 30 µm . The leak conductance was set to pS/µm2 such that the membrane time constant at the resting potential was 40 ms ( any values between 25 and 50 ms were consistent with experimental data ) [8] . The active conductances can be expressed via the gating variables , and such that and . We used pS/µm2 and pS/µm2 for the maximal conductances of the GS neurons , so that their ratio was ; and pS/µm2 and pS/µm2 for the maximal conductances of the NGS neurons , so that their ratio was . We also studied a larger range of these maximal conductances in Figure 2 . The gating variables have the following kinetics: with where can be , or , and: ( 3 ) ( 4 ) ( 5 ) The rate coefficients , and are of the form and and the kinematic parameters are provided in Table 1 . The equations were numerically solved using a first-order Euler method with an integration time step of ms . We used a threshold of −20 mV to detect spikes , although our results did not depend on the exact value of this parameter . For spike-triggered characterization we injected Gaussian noise current , , with mean , , and standard deviation , , to elicit spike trains in ten 1000-second long trials . All input current traces were realizations of the Ornstein-Uhlenbeck process [76] expressed as: ( 6 ) where has unit variance and correlation time of 1 ms to match experimental conditions [8] . Intrinsic computation in these neuron types was previously characterized in experiments and model neurons [8] using a one-dimensional Linear-Nonlinear ( LN ) cascade model of output spike times to the input Gaussian current stimulus with standard deviation [23] . The first component of the LN model is a feature which linearly filters the stimulus producing the amplitude of the feature present in the input; the second component is a nonlinear function which gives the instantaneous firing rate for each value of the filtered stimulus . We take the feature to be the spike-triggered average ( STA ) [18] , [24] , and obtain the expression for the nonlinear response function from Bayes' law: ( 7 ) where is the mean firing rates for fixed input mean and standard deviation , is the prior distribution which is a Gaussian with mean zero and variance , is the spike-triggered stimulus distribution obtained from the histogram of filtered stimulus values when the spikes occur . We refer to the neurons with ratio equal to 1 . 5 as gain-scaling , because scaling the stimulus by produces a nonlinearity in the LN model that is independent of , i . e . for inputs with two different standard deviations and ( mean fixed to zero in Figure 1A , B , red ) [8] . The neurons with ratio equal to 0 . 6 are termed nongain-scaling , because nonlinearities in the LN model vary with different values of the standard deviation when the stimulus is scaled by ( Figure 1A , B , blue ) . The gain-scaling properties of single neurons hold for all [8] . We considered a feedforward network architecture with layers , each layer consisting of neurons ( Figure 4A ) . We considered networks of neurons ( the results remain the same as long as ) . A common temporally fluctuating input current was injected to all neurons in the first layer . The common input was generated using ( 8 ) where is a random phase in , and is the total length of the stimulus . The exact properties of this stimulus ( size of the window , the cutoff frequency of 1 Hz ) were not important , as long as the correlation timescale of this stimulus was much longer than the correlation timescale of the fast fluctuations ( ms ) independently injected into each neuron . Instead of , neurons in deeper layers ( beyond the first ) received synaptic input from neurons in the previous layer via conductance-based synapses . In contrast to current-based synapses , conductance-based synapses have been shown to support the stable propagation of synfire chains [38] and a larger range of firing rates [37] . The synaptic input current into a neuron in layer in the network ( which receives inputs from a subset of neurons in the previous layers ) is given by ( 9 ) where mV is the excitatory reversal potential and is the membrane potential of the neuron . The synaptic conductance is a continuous variable which increases with the spike times of each input by the excitatory postsynaptic potential ( EPSP ) scaled by the corresponding synaptic strength . We used exponentially decaying EPSPs with a time constant ms . Then we can write the synaptic conductance as ( 10 ) where is the delta spike train of the -th neuron in the previous layer with spikes at times and when is the EPSP . denotes a random subset of the 2000 neurons in the previous layer providing synaptic input into the given neuron . There were no recurrent connections among the neurons . Each neuron in the network also received an independent noise input with mean 0 and standard deviation that fluctuates on a timescale significantly shorter than the timescale of the common input to represent random synaptic input that cortical networks experience during early development [28] . In all models , the noise stimulus added to each neuron was independent from the mean stimulus and correlated with a correlation time of 1 ms . Note that for the mean field analysis ( see below ) , simulations were performed with a constant mean ( Figure 6 ) , rather than the time-dependent ( Equation 8 ) . The range of stimulus standard deviations was chosen to produce firing rates larger than 3 Hz and such that voltages were not hyperpolarized below mV to match the corresponding experiments [8] . Given an input current , the output firing rate can be expressed by the -dependent – curve: . We computed the – curves for the GS and NGS neurons for a range of mean inputs and fluctuation amplitudes ( Figure 1C , D ) from 100 second long simulations . The mean current ranged from 0 to 120 pA in steps of 2 . 5 pA and the standard deviation from 5 to 150 pA in steps of 2 . 5 pA . The mean field analysis was used to predict firing rate transmission across the network ( Figure 6 ) . Given the synaptic current into a neuron in layer in the network ( which receives inputs from a subset of neurons in the previous layers connected with weights of strengths ) , the average synaptic current received by a neuron in one layer from a subset ( or all ) of neurons in the previous layer can be written as: ( 11 ) where the angle brackets denote average over time . In the limit that and are uncorrelated , then ( 12 ) The average synaptic conductance can be written as ( 13 ) where is the average firing rate of neuron . We let denote the connection probability between neurons in two consecutive layers; therefore , the subset has approximately neurons . We examined connectivity probability ranging between 0 . 5% , 5% and 10% while keeping the product of the connectivity probability and synaptic strength fixed , and observed no differences in how effectively firing rates were propagated across different layers in the network ( Figure S2 ) . The main results use . For the two network types , we chose synaptic strength sufficiently strong to allow for activity to be maintained in each network . For the NGS network we used , while for the GS network we explored in addition weaker synaptic strength of ; although the exact values used were not too important as long as the iterated map dynamics predicting the mean firing rates across the network had the same structure ( for example , number of fixed points ) ( Figure S2 ) . Since all synapses in our network are identical to , we can approximate ; similarly , all the neurons in a given layer are identical so . Then the average synaptic current into a neuron in a given layer can be approximated as ( 14 ) From the – relationship , the firing rate in layer can be expressed as a function of the firing rate of the neurons in the previous layer ( see Eq . 1 ) where the scaling coefficient is given by ( 15 ) When computing we used only subthreshold voltage fluctuations . The input-output relationship plotted in Figure 6 ( black line ) corresponds to the line of slope . We also computed the standard deviation of the subthreshold voltage fluctuations and thus estimated where was obtained using Equation 15 with instead of . Figure 6 text shows this as a gray boundary around the line with slope , which was used further to interpret the variability of propagation of firing rates . Furthermore , we note that when predicting the propagation of firing rates across subsequent layers in this mean field analysis , the – curve in Equation 1 was chosen such that was obtained by combining the standard deviation of the independent noise fluctuations added in each layer , and the standard deviation of the synaptic current recorded in each layer , where ( 16 ) We first measured information transmission in the network about slow variations in the input ( Figure 4 ) . The mutual information of stimulus and response was computed by testing a particular encoding model ( Figure 4D ) . Typically , this method assumes a model for estimating the stimulus and provides a lower bound on the information transfer because the model does not capture all aspects of the information [77] . We chose the stimulus reconstruction to be a simple population average of the neuronal response ( the PSTH ) , so that the stimulus estimate in layer , , is given by the mean neuronal response obtained from many repetitions of the identical slow stimulus , but different realizations of the fast fluctuations . We computed the information in the -th layer using the equation for a dynamic Gaussian channel [24] ( 17 ) where the signal-to-noise ratio can be written as ( 18 ) Assuming Gaussian probability distributions , the noise is ( 19 ) This quantity computes the information between stimulus and response by taking into account how similar the response ( reconstructed stimulus ) is to the original stimulus . Due to the different firing rates evoked in the different networks , when computing the information we normalized the reconstructed stimulus ( the PSTH ) to have zero mean and unit variance . To quantify the information about fast fluctuations as a function of the mean and of the input current injected into single neurons ( Figure 1E ) , we used the output entropy of the predicted firing rate probability in the LN model , , using the nonlinear response function expression from Equation 7 . When examining the fidelity of firing rate transfer in networks composed of the two neuron types , we wanted a measure of how distinguishable is a discrete set of output firing rates in each layer given a set of input currents in the first layer ( see Figure 7 , note that Figure 1F is like the data in Figure 7 layer 1 ) . This was the information conveyed by the network response of each layer about a stationary mean input , in the presence of background noise ( Figure 7 ) . We obtained the firing rate response of strongly connected NGS and GS networks ( synaptic strength ) and weakly connected GS networks ( ) for different layers , noise conditions and ranges of input . For the strongly connected NGS and GS networks , we used a range of 28 input currents uniformly distributed between 0 and 70 pA , and for the weakly connected GS networks , the same number of input currents uniformly distributed in the range of 0 to 22 pA . The noise values that we examined spanned the range of from 15 to 75 pA–which produced biologically relevant output firing rates and subthreshold voltage fluctuations in a valid regime mV . The output firing rates were obtained using 2 second long bins ( total length of the trial was 20 , 000 seconds ) . Qualitative trends in the information curves were maintained for 1 , 5 and 10 second long bins . Then , given the set of firing rate responses of the neurons of the -th layer for the input currents , we constructed by computing histograms of the output firing rates binned into the same 28 bins . We computed the mutual information for each layer ( 20 ) where is the probability distribution of the output firing rates [77] . denotes the prior probability of input stimuli which we took to be a uniform distribution so that each stimulus had the same probability 1/28 of occurrence . Although the exact value of the information will depend on the binning choice ( here into 28 bins ) , the contrast in performance of the GS and NGS neurons ( which was our goal ) was preserved for other binning choices . To reduce the full conductance-based model ( Eq . 2 ) that depends on four variables , , , and , to a system of two first-order differential equations , we followed the procedure described by Abbott and Kepler [35] for the Hodgkin-Huxley model . Although the neuron's membrane potential is affected by the three dynamic variables , , and , these three do not directly couple to each other but only interact through . This property allows us to approximate their dynamics by introducing an auxiliary potential variable . Since the time constant that governs the behavior for is much smaller than the time constants for and , then will reach its asymptotic value more rapidly than other changes in the model . Therefore , we lose some accuracy in the generation of spikes , but can write . Because of their longer time constants , and lag behind and reach their asymptotic values more slowly . This can be implemented by introducing an auxiliary voltage variable and then replacing and by and , since the functions and are well separated as a function of the dependent variable , in this case . To choose , we ask for the time dependence of in and the time dependence that the slowly changing and induce into in the full model to match – this is achieved by equating the time derivatives of at constant in the full and reduced models . Hence , we convert the full model ( Eq . 2 ) into the following system of first-order differential equations: ( 21 ) ( 22 ) where ( 23 ) and where ( 24 ) where and are evaluated at and . To study the dynamics of this system in Figure 8 , we plotted the nullclines , i . e . the curves where and . The points where these two curves intersect are the fixed points of the two-dimensional dynamics . In Figure 8 we use arrows in the phase planes to denote the flows around the nullclines .
|
Differences in ion channel composition endow different neuronal types with distinct computational properties . Understanding how these biophysical differences affect network-level computation is an important frontier . We focus on a set of biophysical properties , experimentally observed in developing cortical neurons , that allow these neurons to efficiently encode their inputs despite time-varying changes in the statistical context . Large-scale propagating waves are autonomously generated by the developing brain even before the onset of sensory experience . Using multi-layered feedforward networks , we examine how changes in intrinsic properties can lead to changes in the network's ability to represent and transmit information on multiple timescales . We demonstrate that measured changes in the computational properties of immature single neurons enable the propagation of slow-varying wave-like inputs . In contrast , neurons with more mature properties are more sensitive to fast fluctuations , which modulate the slow-varying information . While slow events are transmitted with high fidelity in initial network layers , noise degrades transmission in downstream network layers . Our results show how short-term adaptation and modulation of the neurons' input-output firing curves by background synaptic noise determine the ability of neural networks to transmit information on multiple timescales .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"computational",
"neuroscience",
"single",
"neuron",
"function",
"biology",
"and",
"life",
"sciences",
"computational",
"biology",
"neuroscience",
"artificial",
"neural",
"networks"
] |
2014
|
Intrinsic Neuronal Properties Switch the Mode of Information Transmission in Networks
|
The pathogenic yeast Cryptococcus neoformans causes fungal meningitis in immune-compromised patients . Cell proliferation in the budding yeast form is required for C . neoformans to infect human hosts , and virulence factors such as capsule formation and melanin production are affected by cell-cycle perturbation . Thus , understanding cell-cycle regulation is critical for a full understanding of virulence factors for disease . Our group and others have demonstrated that a large fraction of genes in Saccharomyces cerevisiae is expressed periodically during the cell cycle , and that proper regulation of this transcriptional program is important for proper cell division . Despite the evolutionary divergence of the two budding yeasts , we found that a similar percentage of all genes ( ~20% ) is periodically expressed during the cell cycle in both yeasts . However , the temporal ordering of periodic expression has diverged for some orthologous cell-cycle genes , especially those related to bud emergence and bud growth . Genes regulating DNA replication and mitosis exhibited a conserved ordering in both yeasts , suggesting that essential cell-cycle processes are conserved in periodicity and in timing of expression ( i . e . duplication before division ) . In S . cerevisiae cells , we have proposed that an interconnected network of periodic transcription factors ( TFs ) controls the bulk of the cell-cycle transcriptional program . We found that temporal ordering of orthologous network TFs was not always maintained; however , the TF network topology at cell-cycle commitment appears to be conserved in C . neoformans . During the C . neoformans cell cycle , DNA replication genes , mitosis genes , and 40 genes involved in virulence are periodically expressed . Future work toward understanding the gene regulatory network that controls cell-cycle genes is critical for developing novel antifungals to inhibit pathogen proliferation .
About 500 million years of evolution separate the fungal phyla Ascomycota and Basidiomycota [1 , 2] . The cell cycle is an essential biological process driving cell division of these distantly related yeasts , and therefore may be under strong selective pressure for conservation . Both Saccharomyces cerevisiae ( Ascomycota ) and Cryptococcus neoformans ( Basidiomycota ) can grow and divide asymmetrically in a budding yeast form . C . neoformans is a causative agent of deadly fungal meningitis , primarily in immune-compromised patients [3 , 4] . Many groups studying C . neoformans focus on virulence factors for human infection , such as the yeast’s polysaccharide capsule , melanin production , Titan cell formation , and others [5–9] . We propose that the function of cell-cycle regulators , which are essential for proliferation in the host , merit further investigation as virulence factors . Furthermore , there is evidence that virulence pathways are perturbed when cell-cycle progression is slowed , which suggests direct connections between cell-cycle regulators and virulence pathways [10 , 11] . The cell cycle is the process by which a cell duplicates its contents and faithfully divides into two genetically identical cells . In eukaryotes , a biochemical oscillator drives sequential cell-cycle events , where the cyclin-dependent kinase ( CDK ) and its variety of cyclin binding partners initiate events by phosphorylation , followed by destruction of kinase activity in mitosis by the anaphase-promoting complex ( APC ) . Another common feature of the eukaryotic cell cycle is a temporally regulated program of transcription , which has been demonstrated in S . cerevisiae , Schizosaccharomyces pombe , Arabidopsis thaliana , mouse fibroblasts , and human tissue culture cells [12–22] . These programs of periodic genes include cyclin mRNAs , DNA replication factors , APC activators , and other cellular components that are utilized at specific times during the cell cycle . Our group and others have proposed that this “just-in-time transcription” mechanism is an important aspect of energy-efficient and faithful cell divisions [23 , 24] . In S . cerevisiae , an interconnected network of periodic transcription factors ( TFs ) is capable of driving the periodic program of cell-cycle gene expression [15 , 25–27] . Aspects of this yeast TF network are conserved in human cells; for example , G2/M genes are activated by a periodic forkhead domain-containing TF in both eukaryotes [22 , 28] . The topology of cell-cycle entry is also functionally conserved , where a repressor ( S . c . WHI5 , H . s . RB1 ) is removed by G1 cyclin/CDK phosphorylation to activate a G1/S transcription factor complex ( S . c . SBF/MBF , H . s . E2F-TFDP1 ) [29] . However , the genes involved in cell-cycle entry are not conserved at the sequence level between fungi and mammals [30] , suggesting that the fungal pathway could be targeted with drugs without affecting mammalian host cells . Sequence-specific DNA-binding TFs have been identified in C . neoformans and phenotypically profiled by single gene knockouts [6 , 31 , 32] . This TF deletion collection was profiled over many virulence factor-inducing conditions to discover pathways that regulate disease and drug response genes [32] . Serial activation of TFs during capsule production has also been studied to elucidate the order in which TFs control virulence gene products [31] . However , the cell cycle has not been investigated in synchronous populations of cells to date . Although the phenotypes of some single mutant cell-cycle TFs have been examined from asynchronous populations , these studies offer limited understanding of temporal aspects of gene expression during the cell cycle . Here we investigate transcriptional dynamics of the pathogenic yeast C . neoformans using cells synchronized in the cell cycle . We compare our findings to the cell-cycle transcriptional program in S . cerevisiae . We find that a similar percentage of all genes ( ~20% ) are periodically transcribed during the cell cycle , and we present a comprehensive periodicity analysis for all expressed genes in both yeasts . We show that S-phase gene orthologs are highly conserved and temporally precede M-phase gene orthologs in both yeasts . Additionally , we find that many TFs in the cell-cycle entry pathway are conserved in sequence homology , periodicity , and timing of expression in C . neoformans , while others , notably genes involved in budding , are not . We also identify 40 virulence genes that appear to be cell-cycle-regulated , along with nearly 100 orthologous fungal genes that are periodic in the same cell-cycle phase . Taken together , these cell-cycle genes represent candidates for further study and for novel antifungal drug development .
Identifying approaches for synchronizing populations of C . neoformans has been challenging . We succeeded in synchronizing by centrifugal elutriation , a method that has been very successful for S . cerevisiae cells [15 , 27 , 33] . For C . neoformans , we isolated early G1 daughter cells by centrifugal elutriation and released the population into rich media ( YEPD ) at 30°C to monitor cell-cycle progression , as described previously [34] . This size-gradient synchrony procedure is conceptually similar to the C . neoformans synchrony procedure presented by Raclavsky and colleagues [35] . For S . cerevisiae , we isolated G1 cells by alpha-factor mating pheromone treatment [36] . We utilized this synchrony technique to isolate larger S . cerevisiae cells and to offset some loss of synchrony over time due to asymmetric cell divisions . A functional mating pheromone peptide for C . neoformans has been described but is difficult to synthesize in suitable quantities [37] . After release from synchronization , bud formation and population doubling were counted for at least 200 cells over time ( Fig 1 ) . The period of bud emergence was about 75 minutes in both budding yeasts grown in rich media , although the synchrony of bud emergence after the first bud in C . neoformans appeared to be less robust ( Fig 1A and 1B ) . Each yeast population completed more than two population doublings over the course of the experiments . Total RNA was extracted from yeast cells at each time point ( every 5 minutes for S . cerevisiae , or every 10 minutes for C . neoformans ) and multiplexed for stranded RNA-Sequencing . Between 87–92% of reads mapped uniquely to the respective yeast genomes ( S1 File ) . To identify periodic genes , we applied periodicity algorithms to the time series gene expression datasets . Four algorithms were used to determine periodicity rankings for all genes in each yeast: de Lichtenberg , JTK-CYCLE , Lomb-Scargle , and persistent homology [38–42] . Since each algorithm favors slightly different periodic curve shapes [43] , we summed the periodicity rankings from each algorithm and ranked all yeast genes by cumulative scores for S . cerevisiae and for C . neoformans ( S1 Table and S2 Table , respectively ) . By visual inspection , the top 1600 ranked genes in both yeasts appeared periodically transcribed during the cell cycle ( S1 Fig ) . There was no clear “threshold” between periodic and non-periodic genes during the cell cycle—rather , we observed a distribution of gene expression shapes and signatures over time ( S1 Fig ) . Previous work on the S . cerevisiae cell cycle has reported lists ranging from 400–1200 periodic genes . To validate our RNA-Sequencing time series dataset for the S . cerevisiae cell cycle , we compared the top-ranked 1600 periodic genes to previously published cell-cycle gene lists and found a 57–89% range of overlap with previous periodic gene lists ( S2 Fig ) [12–15 , 33 , 41 , 44 , 45] . Three filters were applied to each budding yeast dataset to estimate and compare the number of periodic genes ( S1 File ) . First , we pruned noisy , low-expression genes from each dataset , leaving 5913 expressed genes in S . cerevisiae ( S1 Table ) and 6182 expressed genes in C . neoformans ( S2 Table ) . Next , we took the top 1600 expressed genes from the cumulative ranking of the four periodicity algorithms described above . Finally , we applied a score cutoff to each list of top 1600 genes using the Lomb-Scargle algorithm ( see S1 File ) [39 , 40 , 43] . We estimated that there are 1246 periodic genes in S . cerevisiae ( ~21% expressed genes ) and 1134 periodic genes in C . neoformans ( ~18% expressed genes ) ( Fig 2 ) . We also provided multiple criteria for evaluating the cell-cycle expression patterns of individual genes in each yeast ( S1 Table , S2 Table , S1 Fig ) . Cellular processes that contribute to virulence are a major focus of work in the C . neoformans field . We took advantage of the partial C . neoformans deletion collection and genetic screens for virulence factors [6] and searched for periodic virulence genes . We found that 40 genes ( about 16% of the virulence genes characterized by the Madhani group and many previous studies ) were periodically expressed in C . neoformans during the cell cycle ( S3 Table ) . These virulence genes are periodic during normal cycles in rich media , which suggests that some virulence processes are directly cell-cycle-regulated . For example , budding and cell wall synthesis are coupled to cell-cycle progression in S . cerevisiae . A subset of 14 periodic virulence genes in C . neoformans had capsule and/or cell wall phenotypes reported in previous studies ( S3 Table ) . We then asked if the 40 periodic virulence genes might be co-regulated during the C . neoformans cell cycle ( S3 Fig ) . Over half of the periodic virulence genes clustered together and peaked in a similar cell-cycle phase ( 20–30 minutes into cycle 1 ) . 11 of the 14 capsule / cell wall genes were contained in this cluster ( S3 Fig , S3 Table ) . Next , we wanted to ask if periodicity and temporal ordering of orthologous genes is evolutionarily conserved between the two budding yeasts . We compiled the largest list to date of putative sequence orthologs between C . neoformans and S . cerevisiae from the literature , databases , and additional BLAST searches ( S1 File , S4 Table ) [32 , 46–48] . About half of the periodic genes from each yeast ( Fig 2 ) had at least one sequence ortholog in the other species . However , there were only about 230 pairs of orthologous genes that were labeled periodic in both yeasts . Those pairs of periodic orthologs have diverged in temporal ordering between C . neoformans and S . cerevisiae ( Fig 3 , S5 Table ) . These results indicated that the programs of periodic gene expression , and possibly the regulatory pathway , have diverged to some degree between the two budding yeasts . This altered temporal ordering between S . cerevisiae and C . neoformans periodic orthologous genes was likely not due to the experimental synchrony procedure . We obtained transcriptome data from two previous studies on S . cerevisiae cell-cycle-regulated transcription ( which applied a different cell-cycle synchrony procedure , used different lab strains of S . cerevisiae , and/or measured gene expression on different platforms ) , and our list of periodic S . cerevisiae genes maintained temporal ordering during the cell cycle in all three datasets ( S4 Fig ) . Cell-cycle regulated gene expression has also been investigated in a species of pathogenic Ascomycota , Candida albicans [49] . To ask about common periodic gene expression in an evolutionarily intermediate budding yeast species , we further identified putative periodic orthologous genes shared between S . cerevisiae , C . neoformans , and C . albicans . A core set of almost 100 orthologs appeared to have both conserved periodicity and temporal ordering between all three budding yeasts ( S5 Fig , S5 Table ) . This fungal gene set was enriched for functions in mitotic cell cycle and cell-cycle processes , which suggested that core cell-cycle regulators are under strong selection for conservation at the sequence level and by timing of periodic gene expression . We reasoned that some cell-cycle events must be invariable in temporal ordering between fungi ( S5 Fig ) . DNA replication ( S-phase ) should be highly conserved across organisms because duplication of genetic material is essential for successful division . Segregation of genomic content during mitosis ( M-phase ) is also essential for division , and duplication must precede division . Using annotations for S . cerevisiae [50] we identified lists of genes known to be involved in regulating events in various cell-cycle phases including bud formation and growth [51 , 52] , DNA replication [53 , 54] , and spindle formation , mitosis , and mitotic exit [55–58] . We filtered the resulting gene lists by periodicity in S . cerevisiae ( Fig 2A , S6 Table ) . We then identified orthologous genes in C . neoformans without enforcing a periodicity filter . We have previously shown that expression timing of canonical cell-cycle orthologs in S . cerevisiae and S . pombe can vary—some gene pairs shared expression patterns while others diverged [59] . To temporally align orthologous gene plots between S . cerevisiae and C . neoformans , we used the algorithmic approach described previously with S . cerevisiae and S . pombe time series transcriptome data [59] . The first , most synchronous cycle of budding data from each yeast was fit using the CLOCCS algorithm ( Fig 1 , S6 Fig ) [59 , 60] . Time points in minutes were then transformed into cell-cycle lifeline points to visualize the data ( see S1 File ) . As observed previously , S . cerevisiae genes that regulate budding , S-phase , and mitosis were largely transcribed periodically in the proper phases ( Fig 4A , 4D and 4G ) [12–15] . Cell-cycle gene expression peak time patterns were examined to quantitatively compare cell-cycle phases ( S7 Fig ) . Bud assembly and growth genes peaked throughout the cell-cycle transcription program , and the temporal ordering of these genes repeated across cell cycles ( Fig 4A , S7A and S7B Fig ) . Similarly , spindle assembly and mitosis genes peaked in the mid-to-late phases of the transcription program ( Fig 4G ) . DNA replication genes peaked in a defined window in the middle phase of the transcription program ( Fig 4D ) . We observed analogous expression patterns for C . neoformans orthologs associated with S-phase and mitosis ( Fig 4E and 4H ) , but orthologs associated with budding appeared to be expressed with less restriction to a discrete cell-cycle phase or strict temporal order ( S7 Fig ) . This budding gene pattern can be observed qualitatively where the unrestricted expression timing creates a more “speckled” appearance in the C . neoformans heatmap ( Fig 4B ) and differentially timed gene expression peaks ( Fig 4C ) . We hypothesize that bud emergence and bud growth are not as tightly coordinated with cell-cycle progression in C . neoformans cells . Unlike S . cerevisiae where bud emergence occurs primarily at the G1/S transition , C . neoformans bud emergence can occur in a broad interval from G1 to G2 phases [61 , 62] . The difference in budding transcript behaviors between S . cerevisiae and C . neoformans orthologs could therefore reflect the difference in the cell biology of bud emergence and growth ( Fig 4A and 4B ) . Only about 33% of the orthologous budding gene pairs were periodically expressed in C . neoformans , compared to 53% DNA replication and 61% mitosis orthologs ( Fig 4B , 4E and 4H ) . Furthermore , budding orthologs that were periodic in both C . neoformans and S . cerevisiae showed some divergence in expression timing ( Fig 4C ) . We also observed that bud emergence of C . neoformans cells during the time series appeared less synchronous in second and third cycles than S . cerevisiae cells ( Fig 1A and 1B ) . Bud emergence in C . neoformans could be controlled by both stress pathways and TF inputs because the first budding cycle is highly synchronous after elutriation synchrony , which causes a transient stress response in released cells ( Fig 1B ) . However , our data do not rule out a model where some budding genes in C . neoformans are controlled post-transcriptionally by localization , phosphorylation , or other periodic mechanisms . It is also possible that budding orthologs are more difficult to identify than other cell-cycle genes due to sequence divergence or that novel budding genes have evolved in the C . neoformans lineage . We have previously shown that a network of periodically expressed TFs is capable of driving the program of periodic genes during the S . cerevisiae cell cycle [15 , 27] . We hypothesized that a network of periodic TFs could also function in C . neoformans to drive a similar fraction of cell-cycle genes . Thus , the temporal re-ordering of part of the C . neoformans gene expression program ( Fig 3 ) could be explained by two models: evolutionary re-wiring of shared network TFs with S . cerevisiae or novel TF network components arising in C . neoformans to drive cell-cycle genes . First , we asked if network TFs were conserved from S . cerevisiae to C . neoformans . Indeed , a majority of network TFs and key cell-cycle regulators have putative orthology between the two yeasts ( Table 1 ) [30] . As observed for other cell-cycle genes ( Fig 4 ) , orthologs of some network TFs were expressed in the same phase in both yeasts , while others were expressed at different times ( Table 1 ) . Second , we asked if there were any novel periodic TFs in C . neoformans ( i . e . TFs with no predicted ortholog in S . cerevisiae , or TFs with an ortholog in S . cerevisiae that is not known to function in the TF network ) . We constructed a list of periodic C . neoformans TFs by filtering a previously annotated transcription factor list [32] with our list of periodic genes ( Fig 5 , S7 Table ) . Indeed , 30 novel TF genes were periodic during the C . neoformans cell cycle ( Fig 5A ) . Taken together , results from Table 1 and Fig 5 suggested that both network TF re-wiring and novel periodic TFs in C . neoformans could explain the differential ordering of periodic genes during the cell cycle ( Fig 3 ) . Putative S-phase regulators in C . neoformans exhibited transcript behaviors that were very similar in periodicity and in ordering to their S . cerevisiae orthologs ( Fig 4D–4F ) . Thus , we predicted that the network motifs and TFs controlling the transcription of periodic S-phase genes could be conserved . Orthologous genes in the G1/S topology were largely conserved in periodic expression dynamics at cell-cycle entry ( Fig 6 ) . The expression timing of some genes had shifted earlier in the C . neoformans cell cycle ( Fig 6C and 6F , Table 1 ) , but this result does not refute the hypothesis that these genes are activated and functional at G1/S phase . Therefore , the network topology of cell-cycle entry appeared largely conserved in C . neoformans both by sequence and by gene expression dynamics . The prediction of this model is that a common G1/S transcriptional network drives a common set of S-phase periodic genes . To test this model , we examined promoter sequences from TF network genes in S . cerevisiae and C . neoformans , as well as the promoters of 38 periodic DNA replication ortholog pairs , and did an unbiased search for enriched TF binding sequences . The core motif “ACGCGT” for SBF/MBF transcription factors [63–65] was identified in both S . cerevisiae and C . neoformans promoters . The motif was not enriched in randomly selected periodic gene promoters , suggesting that SBF/MBF is functionally conserved in C . neoformans to drive TF network oscillations and DNA replication gene expression ( S8 Fig ) .
Here , we present the first RNA-Sequencing dataset of transcription dynamics during the cell cycle of C . neoformans . Despite evolutionary distance between Basidiomycota and Ascomycota , S . cerevisiae and its extensive genome annotation provided an excellent analytical benchmark to compare to cell-cycle transcription in C . neoformans . RNA-Sequencing has been shown to be more quantitative than microarray technology for lowly- and highly-expressed genes using asynchronous S . cerevisiae cells due to microarray background fluorescence and saturation of fluorescence , respectively [66] . We demonstrate that 20% or more of all genes in the budding yeast genomes are periodically transcribed during the cell cycle . A ranking of periodicity for transcript dynamics in C . neoformans is provided ( S2 Table ) . For the sake of comparison , we have presented gene sets of 1100–1200 periodic genes with the highest relative periodicity scores as “cell-cycle-regulated”; however , there is a continuum of periodic gene expression dynamics during the cell cycle in both yeasts ( S1 Fig ) . The four periodicity algorithms applied here yielded a range of periodicity scores with no clear distinction between “periodic” and “non-periodic” gene sets ( S1 and S2 Tables ) . These results suggest that yeast mRNAs fluctuate in expression with various degrees of cell-cycle periodicity . We propose that the top 20% periodic genes presented in this study are directly regulated by periodic cell-cycle TFs in C . neoformans and in S . cerevisiae . We also posit that some of the remaining 80% genes are weakly cell-cycle regulated . For example , some genes could be subject to complex regulation with one regulatory input from a cell-cycle periodic TF and another input from a constitutively expressed TF . We raise two important questions about the yeast periodic gene expression programs: is periodic expression of a core set ( s ) of genes required for the fungal cell cycle , and how are periodic gene dynamics controlled in each yeast ? In both yeasts , periodic transcription is a high dimensional cell-cycle phenotype because transcriptional state reflects the phase-specific biology of the cell cycle over repeated cycles ( Fig 2 and Fig 4 ) . In other words , G1- , S- , and M-phase genes follow a defined temporal ordering pattern . S . cerevisiae cells synchronized by different methods and/or grown in different conditions display similar ordering of periodic cell-cycle genes , despite different cell-cycle period lengths ( S4 Fig ) . Here , we examined the transcriptome of cycling C . neoformans cells at 30°C . Other groups have shown that C . neoformans cells spend more time in G1 phase at 24°C [67] . We predict that future studies examining cell-cycle transcription of C . neoformans cells grown in different conditions ( i . e . non-rich media or 37°C infection temperature ) would continue to display a similar temporal ordering of cell-cycle genes . These findings provide more evidence that “just-in-time transcription” is a conserved feature of eukaryotic cell cycles [23] . We show that some orthologous periodic genes have diverged in temporal ordering during the cell cycles of S . cerevisiae and C . neoformans over evolutionary time ( Fig 3 ) . We specifically investigated genes that play a role in bud emergence and bud growth , and we find that many budding gene orthologs are not controlled in a defined temporal order during the C . neoformans cell cycle ( Figs 1A , 1B , 4A and 4B ) . On the other hand , DNA replication and mitosis genes do appear to be conserved by sequence homology , periodic expression , and temporal ordering ( Fig 4D–4I ) . Lastly , we find that a set of about 100 orthologous genes is both periodic and expressed in proper cell-cycle phase in the budding yeasts S . cerevisiae , C . neoformans , and C . albicans ( S5 Fig ) [49] . These findings suggest that there may be a conserved set of fungal cell-cycle-control genes , which represent novel therapeutic targets for fungal infections . We posit that a network of periodic transcription factors ( TFs ) could control the periodic gene expression program in C . neoformans , which has been shown in S . cerevisiae and suggested in human cells [15 , 22 , 25 , 27] . Many orthologous genes to S . cerevisiae TF network components have diverged in expression timing in C . neoformans cells ( Table 1 ) . However , we show that the G1/S network topology is likely conserved between S . cerevisiae and C . neoformans because orthologous genes display similar expression dynamics ( Fig 6 ) . Furthermore , we find that the promoters of G1/S TF network orthologs and promoters of periodic DNA replication orthologs are enriched for an “ACGCGT” sequence motif , which matches the SBF/MBF binding site consensus in S . cerevisiae ( S8 Fig ) [63–65] . Therefore , we propose that the G1/S transcriptional motif—where a co-repressor is removed by G1 cyclin/CDK phosphorylation and a TF activator complex is de-repressed—is also conserved in C . neoformans ( Fig 6B–6D and 6G ) [29 , 30] . Downstream of the G1/S activator complex , the C . neoformans TF network may also contain a common forkhead domain S-phase activator and homeobox domain G1/S repressor ( Fig 6E , Table 1 ) [14 , 68 , 69] . This partially conserved TF network model in C . neoformans explains the common G1/S topology , on-time DNA replication gene transcription , as well as differential expression of budding and other cell-cycle genes by divergent parts of the TF network . The regulation of periodic transcription and the function of a putative TF network warrant further investigation as virulence factors of fungal meningitis caused by C . neoformans . It has been previously shown that fluconazole drug treatment can affect cell ploidy in C . neoformans [70] . More recently , polyploid Titan cells were shown to produce haploid and aneuploid daughter cells during C . neoformans infection [71] . Therefore , future work on proper regulation of DNA replication and the contribution of periodic gene products could greatly benefit our understanding of genome stability in C . neoformans . The C . neoformans TF deletion collection was recently phenotyped , and the potential of targeted TF therapies was discussed [32 , 72] . We have added to the C . neoformans genotype/phenotype map by documenting the functional outputs of cell-cycle TFs over synchronized cell cycles . We also propose that a conserved G1/S topology of cell-cycle TFs may initiate the cell-cycle transcription network in C . neoformans . It is possible that a multi-drug combination targeting cell-cycle regulators and previously characterized virulence pathways could yield more successful antifungal therapies [72] . For example , a combination therapy could target TFs at the conserved G1/S topology to slow cell-cycle entry and also target fungal cell wall or capsule growth . In the circadian rhythm field , it has been shown that drugs targeting Clock Controlled Genes are most potent when administered at the time of the target gene’s peak expression [73] . Interestingly , deletion of the known SBF/MBF ortholog , Mbs1 ( CNAG_07464 ) , is viable in C . neoformans [32 , 74] . These genetic results do not match S . cerevisiae , where swi4 mbp1 double mutants are inviable [75] . In fact , deletion of the single known G1 cyclin ortholog , CNAG_06092 , is also viable in C . neoformans [10] . Mbs1 and the G1 cyclin are likely important for cell-cycle progression in C . neoformans because mutant phenotypes are highly defective in capsule formation in G1 phase , melanin production , and response to Hydroxyurea treatment during S phase [10 , 11 , 32 , 74] . However , the genetics are inconsistent with findings in S . cerevisiae and warrant further investigation to characterize the G1/S TF network topology of C . neoformans . It is possible that uncharacterized , redundant genes exist in the C . neoformans G1/S network motif . We find that 40 candidate virulence genes are periodically expressed during the C . neoformans cell cycle ( S3 Table , S3 Fig ) . An important direction for future work is to identify the mechanistic links between cell-cycle regulators and virulence pathways . 14 periodic virulence genes have annotated phenotypes in capsule formation and/or cell wall secretion . Fungal cells must secrete new cell wall and capsule during growth , and the direct links between cell cycle and these virulence factors in C . neoformans warrants further study because the cell wall and capsule are not present in host cells . The ultimate goal of this work is to identify the regulatory mechanism of periodic gene expression in C . neoformans and to find optimal drug targets and combination therapies for disrupting the fungal cell cycle .
The wild-type Saccharomyces cerevisiae strain is a derivative of BF264-15D MATa bar1 [76 , 77] . The wild-type Cryptococcus neoformans var . grubii serotype A strain is a derivative of H99F [47] . Yeast cultures were grown in standard YEP medium ( 1% yeast extract , 2% peptone , 0 . 012% adenine , 0 . 006% uracil supplemented with 2% dextrose sugar ) . For centrifugal elutriation , cultures were grown in YEP-dextrose ( YEPD ) medium at 30°C overnight . Elutriated early G1 cells were then resuspended in fresh YEPD medium at 30°C for time series experiments . For α-factor arrest , cultures were grown in YEPD medium at 30°C and incubated with 30 ng/ml α-factor for about 110 minutes . Synchronized cultures were then resuspended in fresh YEPD medium at 30°C . Aliquots were taken at each time point and subsequently assayed by RNA-Sequencing . Total RNA was isolated by acid phenol extraction as described previously [34] . Samples were submitted to the Duke Sequencing Facility ( https://www . genome . duke . edu/cores-and-services/sequencing-and-genomic-technologies ) for stranded library preparation and sequencing . mRNA was amplified and barcoded ( Illumina TruSeq Stranded mRNA Library Preparation Kit for S . cerevisiae and KAPA Stranded mRNA-Seq Library Preparation Kit for C . neoformans ) and reads were sequenced in accordance with standard Illumina HiSeq protocols . For S . cerevisiae , libraries of 50 base-pair single-end reads were prepared , and 10 samples were multiplexed and sequenced together in each single lane . For C . neoformans , libraries of 125 base-pair paired-end reads were prepared ( due to larger and more complex yeast transcriptome with introns ) , and 12 samples were multiplexed and sequenced together in each single lane . Raw FASTQ files were aligned to the respective yeast genomes using STAR [78] . Aligned reads were assembled into transcripts , quantified , and normalized using Cufflinks2 [79] . Samples from each yeast time series were normalized together using the CuffNorm feature . The normalized output FPKM gene expression levels were used in the analyses presented . A detailed description of each analysis pipeline is presented in the S1 File . RNA-Sequencing gene expression data from this manuscript have been submitted to the NCBI Gene Expression Omnibus ( GEO; https://www . ncbi . nlm . nih . gov/geo/ ) under accession number GSE80474 .
|
The opportunistic fungal pathogen Cryptococcus neoformans infects immune-compromised humans and causes fungal meningitis by proliferating in the central nervous system . The cell cycle has not been studied at the whole transcriptome level in C . neoformans . Here , we present the expression dynamics of all genes from a synchronous population of C . neoformans cells over multiple cell cycles . Our study shows that almost 20% of all C . neoformans genes are periodically expressed during the cell cycle . We also compare the program of cell-cycle-regulated transcription in C . neoformans to the well-studied but evolutionary distant yeast , Saccharomyces cerevisiae . We find that many orthologous cell-cycle genes are highly conserved in expression pattern ( e . g . DNA replication and mitosis genes ) , while others , notably budding genes , have diverged in expression ordering . We also identify 40 virulence genes from previous studies that are periodically expressed during the C . neoformans cell cycle in rich media . Our findings indicate that a conserved set of fungal transcription factors ( TFs ) controls the expression of conserved cell-cycle genes , while other periodic transcripts are likely controlled by species-specific TFs .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"cryptococcus",
"neoformans",
"medicine",
"and",
"health",
"sciences",
"genetic",
"networks",
"cryptococcus",
"pathology",
"and",
"laboratory",
"medicine",
"pathogens",
"cell",
"cycle",
"and",
"cell",
"division",
"cell",
"processes",
"microbiology",
"fungi",
"model",
"organisms",
"dna",
"replication",
"network",
"analysis",
"genome",
"analysis",
"dna",
"fungal",
"pathogens",
"research",
"and",
"analysis",
"methods",
"saccharomyces",
"mycology",
"computer",
"and",
"information",
"sciences",
"genomics",
"medical",
"microbiology",
"gene",
"expression",
"microbial",
"pathogens",
"yeast",
"biochemistry",
"cell",
"biology",
"nucleic",
"acids",
"virulence",
"factors",
"genetics",
"biology",
"and",
"life",
"sciences",
"yeast",
"and",
"fungal",
"models",
"saccharomyces",
"cerevisiae",
"computational",
"biology",
"organisms"
] |
2016
|
Investigating Conservation of the Cell-Cycle-Regulated Transcriptional Program in the Fungal Pathogen, Cryptococcus neoformans
|
Crumbs family proteins are apical transmembrane proteins with ancient roles in cell polarity . Mouse Crumbs2 mutants arrest at midgestation with abnormal neural plate morphology and a deficit of mesoderm caused by defects in gastrulation . We identified an ENU-induced mutation , wsnp , that phenocopies the Crumbs2 null phenotype . We show that wsnp is a null allele of Protein O-glucosyltransferase 1 ( Poglut1 ) , which encodes an enzyme previously shown to add O-glucose to EGF repeats in the extracellular domain of Drosophila and mammalian Notch , but the role of POGLUT1 in mammalian gastrulation has not been investigated . As predicted , we find that POGLUT1 is essential for Notch signaling in the early mouse embryo . However , the loss of mouse POGLUT1 causes an earlier and more dramatic phenotype than does the loss of activity of the Notch pathway , indicating that POGLUT1 has additional biologically relevant substrates . Using mass spectrometry , we show that POGLUT1 modifies EGF repeats in the extracellular domain of full-length mouse CRUMBS2 . CRUMBS2 that lacks the O-glucose modification fails to be enriched on the apical plasma membrane and instead accumulates in the endoplasmic reticulum . The data demonstrate that CRUMBS2 is the target of POGLUT1 for the gastrulation epithelial-to-mesenchymal transitions ( EMT ) and that all activity of CRUMBS2 depends on modification by POGLUT1 . Mutations in human POGLUT1 cause Dowling-Degos Disease , POGLUT1 is overexpressed in a variety of tumor cells , and mutations in the EGF repeats of human CRUMBS proteins are associated with human congenital nephrosis , retinitis pigmentosa and retinal degeneration , suggesting that O-glucosylation of CRUMBS proteins has broad roles in human health .
Glycosylation can regulate protein stability and function by ensuring efficient protein folding and by altering the binding affinity to interacting partners [1–7] . The biological importance of this type of protein modification is highlighted by dozens of human diseases caused by congenital disorders of glycosylation ( CDGs ) , which are categorized based the chemical linkage , the added sugar , and the enzymes mutated in affected individuals [8–10] . Developmental defects in mice with mutations in glycosyltransferases have defined specific developmental roles for protein glycosylation in FGF and Notch signaling and in the composition of the extracellular matrix [11–18] . Epidermal Growth Factor ( EGF ) repeats are cysteine-containing motifs of about 40 amino acids found in many transmembrane and secreted proteins , including the Crumbs proteins and members of the Notch family receptors and ligands . EGF repeats can mediate ligand-receptor interactions and facilitate protein folding , and these interactions can be modified by glycosylation [4 , 5] . EGF repeats can be modified by three types of O-linked glycosylation: O-fucosylation , O-GlcNAcylation and O-glucosylation . O-Fucose is added to the serine or threonine in the consensus C2-X-X-X-X- ( S/T ) -C3 ( where Cx refers to the conserved cysteines in the EGF repeats ) [19] , O-GlcNAc to the serine or threonine in the putative consensus C5XXGX ( S/T ) GXXC6 [20] and O-glucose to the serine in the consensus sequence C1-X-S-X- ( P/A ) -C2 [21] . Specific enzymes add these sugars to EGF repeats: Protein O-fucosyltransferase 1 ( POFUT1 ) [22] , EGF-specific O-GlcNAc transferase ( EOGT ) [23] and Protein O-glucosyltransferase 1 ( POGLUT1 ) [24] , respectively . POFUT1 and POGLUT1 are essential for development in both Drosophila and mammals [13 , 24–26] . POGLUT1 ( also called KTELC1 or human CAP10-like protein 46KD ( hCLP46 ) [27 , 28] ) is the only mammalian enzyme known to add O-glucose to the EGF repeats of NOTCH [24] . Drosophila Poglut1 ( Rumi ) was identified in a genetic screen based on its role in Notch signaling and was shown to modify the EGF repeats on the extracellular domain of Notch receptor . Rumi mutants have a temperature dependent defect in Notch signaling: at high temperature , the Notch receptor is not cleaved after binding to its ligand , preventing formation of the active Notch intracellular domain ( NICD ) , whereas Notch proteolysis is normal at low temperature . O-glucosylation does not appear to affect ligand binding , and the data suggest it is required to couple binding of ligand to proteolytic activation of the receptor [24] . Mammalian POGLUT1 was identified based on homology to the Drosophila protein and was also shown to have O-glucosyltransferase activity [13 , 29] . Inactivation of mouse Poglut1 causes midgestation lethality with defects in neural tube development , somitogenesis , cardiogenesis , and vascular remodeling [13] . Knockdown of mouse Poglut1 decreases Notch signaling in C2C12 cells , but does not strongly affect the levels of cell surface Notch protein or the binding of NOTCH to its ligand; instead , like the Drosophila protein , POGLUT1-dependent modification appears to affect mouse Notch activity at a step between ligand binding and S3 cleavage of NOTCH [13] . A number of other proteins contain predicted sites of POGLUT1 modification [13] , including Drosophila Crumbs and mammalian CRUMBS1 and CRUMBS2 . Crumbs family proteins have essential , evolutionarily conserved roles in the organization and integrity of epithelia . Mammals have three members of the Crumbs family ( CRUMBS1 , CRUMBS2 and CRUMBS3 ) . CRUMBS1 and CRUMBS2 have large extracellular domains that include multiple EGF-like repeats , and CRUMBS3 is a short membrane-anchored protein that includes the conserved cytoplasmic domain . Humans and mice that lack Crumbs1 are viable but experience light-dependent retinal degeneration [30] . Mice lacking Crumbs3 die shortly after birth due to defects in lung and intestinal epithelia [31] . Mouse Crumbs2 mutants arrest at mid-gestation with a variety of morphological defects that have been attributed to a defective polarity in the epiblast [32] . We isolated an ENU-induced mutation called wsnp ( wing-shaped neural plate ) based on its striking abnormal morphology at midgestation , including a deficit of mesoderm and a completely open neural plate [33] . Here we show that wsnp is a null allele of Poglut1 and that Notch signaling is almost completely blocked in Poglut1wsnp embryos in vivo . However , the phenotype of Poglut1wsnp and Poglut1-/- mutant embryos is more severe than the phenotype caused by complete loss of Notch signaling , suggesting that POGLUT1 has additional targets during mouse embryonic development . We noted that the early phenotype of Poglut1 mutant embryos was similar to that described for Crumbs2 [32] and strongly resembled that of another ENU-induced mutant characterized in the gene encoding Erythrocyte protein band 4 . 1l5 ( Epb4 . 1l5 ) , which can interact with the intracellular domain of Crumbs proteins [34–36] . Here we show that CRUMBS2 must be O-glucosylated by POGLUT1 for its activity during the mammalian gastrulation epithelial-to-mesenchymal transition ( EMT ) . In the absence of POGLUT1 , CRUMBS2 is trapped in the endoplasmic reticulum and is not trafficked to the apical plasma membrane . The data show that this trafficking defect causes a complete loss of CRUMBS2 function and argue that the loss of apical CRUMBS2 is responsible for the gastrulation defects seen in Poglut1 null embryos .
We isolated the wing-shaped neural plate ( wsnp ) mutant in a screen for ENU ( N-ethyl N-nitrosourea ) -induced recessive mutations that disrupt the morphology of the mid-gestation embryo [33 , 37] . The mutants at E8 . 5 had a shortened anterior-posterior body axis , lacked Pax3+ somites ( Fig 1A ) and had a flat SOX2+ neural epithelium that failed to close and form a neural tube ( Fig 1B ) . Sonic hedgehog ( Shh ) is expressed along the midline notochordal plate of wild-type E8 . 5 embryos , whereas wsnp mutants had discontinuous patches of midline Shh expression , demonstrating a disruption in specification or morphogenesis of the mesendoderm ( Fig 1C ) . Despite the reduction in mesoderm-derived tissues in wsnp mutants , they formed a single primitive streak on the posterior side of the embryo , as seen by markers of Nodal and Wnt signaling ( S1 Fig ) . We mapped wsnp to a 407 kb interval between the D16Mit90 and D16Mit12 SSLP markers . Exonic sequences in this interval were captured and SOLiD sequencing identified a single nucleotide substitution in the interval ( see Methods ) , a T to C transition in the splice donor site of intron 3 of Poglut1 , an endoplasmic reticulum resident enzyme that adds O-glucose to EGF repeats of proteins [24] . The wsnp mutation resulted in abnormally spliced products that lacked most of the CAP10 catalytic domain and the KTEL ER-retention signal , and was therefore likely to be a strong loss-of-function allele ( Fig 1D ) . We generated a null allele of the gene from embryonic stem cells from the International Mouse Knockout Project ( Poglut1 Δ; Materials and Methods ) . Poglut1wsnp/Δ embryos showed the same midgestation lethality and abnormal morphology seen in wsnp/wsnp and Poglut1Δ /Poglut1Δ embryos , demonstrating that wsnp was a null allele of Poglut1 ( Fig 1E ) . The embryonic phenotype of the wsnp and Poglut1 Δ/Δ homozygotes resembled that reported previously for a gene trap allele of the gene [13] . Because of its ubiquitous expression in the embryo ( S2A Fig ) , we tested whether POGLUT1 activity was required in the epiblast-derived or extraembryonic tissues of the embryo . We generated embryos in which Poglut1 was deleted specifically in the epiblast using the conditional allele derived from the International Mouse Knockout Project allele and the Sox2-Cre transgene [38] . Poglut1 epiblast-deleted embryos died before E9 . 0 and were indistinguishable from wsnp mutants , with reduced and unsegmented paraxial mesoderm and a flat neural plate ( S2C Fig ) . Thus , POGLUT1 activity is required in the embryonic tissues for normal development . Knockdown of Poglut1 in mammalian cell lines decreases Notch signaling [13] but a direct effect of POGLUT1 on Notch signaling has not been studied in an in vivo context . To determine whether Notch signaling was altered in Poglut1wsnp embryos , we tested whether NOTCH was correctly processed to form NICD , the active transcription factor form of the protein . In wild-type E8 . 5 embryos , most of the Notch protein was active , as assayed by the presence of active NOTCH1 in whole embryo lysates probed with the antibody that specifically recognizes the ϒ-secretase cleaved ( S3 cleaved , Val1744 ) active NOTCH1 [39] ( Fig 2A ) . The amount of NICD was greatly reduced ( 6 . 9 ± 1 . 3 fold ) in Poglut1wsnp mutants ( Fig 2A ) . In parallel , an antibody to the intracellular domain that detects both full-length and active NOTCH1 confirmed that active NOTCH1 was decreased in Poglut1wsnp mutants , while the full-length unprocessed form was more abundant in Poglut1wsnp embryos than in wild type ( Fig 2A ) . By immunostaining , NICD was detected around the node ( Fig 2B ) and in the nascent mesoderm cells emerging from the primitive streak in wild-type embryos ( Fig 2C ) , similar to previous findings [40] . Active NOTCH1 was greatly reduced in Poglut1wsnp mutants by this assay , in either the distal view or in sections through the primitive streak ( Fig 2B and 2C ) . Consistent with the lack of active NOTCH1 protein , the expression of direct transcriptional targets of the Notch pathway , Hes5 , Lunatic Fringe and Hes7 , was strongly reduced in Poglut1wsnp mutants compared to wild type at E8 . 5 ( Fig 2D–2F ) . Although the data show that POGLUT1 is required for normal Notch signaling in the mouse embryo , the phenotypes of Poglut1 mutant embryos were more severe than the phenotypes of Notch pathway mutants . Mammals have four Notch proteins and multiple ligands , and compound mutants lacking all of the receptors have not been analyzed . However , mammals have only a single gene encoding the essential transcriptional NICD co-factor Rbpjk and two Presenilin genes that can cleave Notch proteins to their active form . Rbpjk mutants and Presenilin double mutants differentiate mesoderm-derived structures , including somites , and close their neural tubes [41 , 42] , unlike Poglut1wsnp embryos . Thus POGLUT1 appears to have additional , Notch-independent activities in the early mouse embryo . In addition to the Notch family of receptors and ligands , over 40 proteins in the mouse genome contain predicted POGLUT1 modification sites [13] . Among these proteins , CRUMBS1 and CRUMBS2 have extracellular EGF repeats that could be modified by POGLUT1 ( Fig 3A ) , and the Crumbs2-/- embryonic phenotype appeared to be similar to that of Poglut1 mutants [32] , making it a good candidate for a POGLUT1 target . Indeed , recent studies showed that Drosophila Crumbs can be modified by addition of O-glucose , although that modification is not required for the function of Drosophila Crumbs [43] . Consistent with a difference in post-translational modification , we found that CRUMBS2 protein from Poglut1wsnp mutant embryonic extracts at E8 . 5 migrated more rapidly than CRUMBS2 protein from wild-type embryos ( Fig 3B ) . To analyze whether CRUMBS2 was a direct target of POGLUT1 in vivo , we needed to purify microgram amounts of CRUMBS2 protein from wild type and mutants . We therefore generated wild-type and Poglut1wsnp mutant embryonic stem cells ( ES cells ) from mouse blastocysts under 2i+LIF conditions [44] and expressed full-length CRUMBS2 with a V5-tag in these cells . ES cells are not epithelial and therefore did not express high levels of endogenous CRUMBS2 . After differentiation , wild-type embryoid bodies ( EBs ) at day 6 , which is equivalent to E8 . 5 of mouse gestation , expressed endogenous CRUMBS2 ( Fig 3C ) . EBs derived from Poglut1wsnp mutant ES cells also expressed endogenous CRUMBS2 , but it migrated more rapidly than the protein derived from wild-type embryoid bodies ( Fig 3C ) , as seen in embryos ( Fig 3B ) . We therefore used wild-type and Poglut1wsnp EBs for biochemical analysis . We purified the full-length V5-tagged CRUMBS2 from wild-type and mutant EBs ( S3A Fig ) , and subjected the proteins to chymotrypic digestion and analyzed by nanoLC-MS/MS as previously described [45] . Analysis of a peptide derived from EGF repeat 6 of CRUMBS2 purified from wild-type embryoid bodies showed that it was modified by the addition of an O-glucose trisaccharide , in which the O-glucose was elongated by the addition of two xylose residues ( Fig 3D; S3B Fig ) . This trisaccharide glycoform is also found on the EGF repeats of mammalian NOTCH1 containing the POGLUT1 consensus sequence for O-glucosylation [13 , 21 , 46] . This sugar modification was absent in EGF repeat 6 of CRUMBS2 purified from Poglut1wsnp embryoid bodies . Instead , the unmodified peptide was the only species detectable in the Poglut1wsnp mutants ( Fig 3E; S3C Fig ) . These data demonstrate that EGF repeat 6 in the extracellular domain of the full-length CRUMBS2 protein was O-glucosylated by POGLUT1 in vivo . The observed difference in migration on SDS-PAGE between CRUMBS2 protein from Poglut1wsnp and wild-type embryos and embryoid bodies ( Fig 3B and 3C ) was greater than could be accounted for simply by loss of O-glycosylation and suggested that there might be differences in N-glycosylation that takes place in the ER and the Golgi . To test this , we treated the CRUMBS2 protein purified from embryoid bodies with Endoglycosidase H ( Endo H ) or Peptide N-glycosidase F ( PNGase F ) . Endo H removes high-mannose type N-glycans , which are found on immature proteins in the ER , while PNGase F removes all types of N-linked glycans , including those found on proteins that have moved through the Golgi to the cell surface . Treatment with PNGase F caused CRUMBS2 from both Poglut1wsnp and wild-type samples to shift to the same size ( Fig 4A ) , confirming that the difference in migration is due to differences in N-glycosylation . Endo H treatment increased the mobility of the CRUMBS2 protein from Poglut1wsnp embryoid bodies , suggesting this protein was modified with high mannose type N-glycans and was likely localized to the ER ( Fig 4A ) . In contrast , Endo H had no effect on the majority of CRUMBS2 from wild-type embryoid bodies , demonstrating that this protein had moved through the Golgi to the surface . A smaller portion of the wild-type protein was sensitive to Endo H , suggesting that a small pool of the wild-type protein was in the ER . These data demonstrate that mutant protein was modified with Endo H-sensitive N-glycans , while the majority of the wild-type protein was modified with complex-type N-glycans sensitive to only PNGase F . The results strongly suggest that the CRUMBS2 protein in Poglut1wsnp mutants is localized to the ER and that O-glucosylation of EGF repeats is required for CRUMBS2 transport to the Golgi , where subsequent processing to complex type N-glycan structures occurs . By immunostaining , CRUMBS2 was detected in the apical plasma membrane of the E8 . 5 wild-type neural plate and primitive streak ( Fig 4B and 4C ) . In contrast , there was no detectable apical CRUMBS2 in the Poglut1wsnp neural epithelium or primitive streak ( Fig 4D and 4E ) , consistent with the biochemical finding that Crumbs2 was trapped in the ER when it was not O-glycosylated by POGLUT1 . CRUMBS1 was localized to the Golgi in both wild type and Poglut1wsnp mutants at this stage ( S4A and S4B Fig ) , which suggested that trafficking of these two Crumbs proteins is differentially regulated in this tissue and that CRUMBS2 is the relevant target of POGLUT1 in the early mouse embryo . To test whether the loss of cell-surface CRUMBS2 could account for the phenotype of Poglut1 mutant embryos , we compared the phenotypes of null alleles of the two genes using marker analyses . Like Poglut1 mutants , E8 . 5 Crumbs2 mutants have much less paraxial mesoderm , marked by expression of Meox1 , than wild type ( Fig 5A ) . Both Poglut1wsnp and Crumbs2 mutants had a discontinuous midline , as marked by expression of Brachyury in axial mesoderm ( Fig 5B ) . Although cardiac mesoderm was specified in Crumbs2 and Poglut1 mutants , as assayed by Nkx2 . 5 expression , the E8 . 5 cardiac anlage was thinner and wider in the mutant than in wild type ( Fig 5C ) and the heart fields failed to fuse and form a single heart tube in both mutants . Despite these morphological similarities , the expression of a Notch target gene in the somites , Uncx4 . 1 , appeared normal in the paraxial mesoderm of Crumbs2 mutants ( S5 Fig ) , whereas it is not expressed in Poglut1wsnp embryos , confirming that the Notch pathway is blocked by the absence of POGLUT1 but not the absence of CRUMBS2 . The most prominent phenotype of both Poglut1wsnp and Crumbs2 mutants was the shortened body axis accompanied by a deficit of mesoderm . Mesoderm cells arise during gastrulation in an epithelial-to-mesenchymal transition ( EMT ) at the primitive streak , in which cells delaminate from the epithelial epiblast layer at the site of breakdown of the basement membrane , down-regulate E-cadherin , acquire mesenchymal characteristics and begin to migrate around the circumference of the embryo . The primitive streak regions of Poglut1wsnp and Crumbs2 mutants were comparable to wild-type at E7 . 5 ( Fig 6A–6C ) . However , by E8 . 0 both the mutants had a broader streak than wild type , marked by the region of basement membrane breakdown at the streak ( Fig 6D–6F ) . In both mutants , cells near the streak accumulated some ectopic laminin ( Fig 6E’ and 6F’ ) . In both Poglut1wsnp and Crumbs2-/- embryos , cells expressing E-cadherin accumulated at the primitive streak ( Fig 6E” and 6F” ) , although some E-cadherin-negative cells were present in thin mesodermal wings ( Fig 6E” and 6F” ) . The decreased number of mesoderm cells was apparent in laminin-stained transverse sections: in wild-type embryos , anterior head mesenchyme was present below the neural epithelium ( Fig 6D ) but there was very little anterior head mesenchyme present in both the mutants ( asterisks in Fig 6E and 6F ) .
Our data confirm previous findings in cultured cells and show that POGLUT1 is required for mouse NOTCH1 activity in vivo , as it is in Drosophila . As in Drosophila , POGLUT1 is required for the cleavage of NOTCH1 and activity of the Notch signaling pathway in vivo in the midgestation mouse embryo . The effects of POGLUT1 on Notch are apparent early in development: activated NOTCH1 accumulates in cells of the nascent mesoderm immediately after they exit the primitive streak , presumably setting the stage for segmentation of the presomitic mesoderm [47] . In contrast , our data indicate that the earlier developmental phenotype of Poglut1 mutants is caused by the loss of glucose modification of another EGF repeat-containing protein , the apical transmembrane protein CRUMBS2 . Notch extracellular domain is decorated with O-fucose , as well as O-glucose , and loss of mouse Pofut1 phenocopies Rbpjk mutants . CRUMBS2 and CRUMBS1 also have putative POFUT1 consensus sites ( 11 and 9 sites , respectively ) [11 , 26 , 48] . The stronger phenotype of mouse Poglut1 than Pofut1 mutants suggests that the two types of O-glycosylation have distinct effects on the activities of CRUMBS proteins . Previous studies that defined the activities of enzymes that glycosylate EGF repeats used short fragments of proteins that contained a few EGF repeats expressed in cell lines [13 , 21 , 24] . In the experiments presented here , we showed that full-length CRUMBS2 protein is modified by addition of a glucose-xylose-xylose trisaccharide to EGF repeat 6 in vivo . Although we assayed only a single repeat , we predict that most or all of the 8 EGF repeats of CRUMBS2 that include the consensus sequence for modification are likely to be O-glucosylated . In the absence of O-glucose modification , CRUMBS2 is trapped in the ER and fails to accumulate at the apical membrane of cells in the mouse embryo . This is different from the effect of Poglut1 mutants on NOTCH: in both Drosophila Rumi mutants and mammalian Poglut1 knockdown cell lines , O-glucosylation is not required for cell surface localization or ligand binding of NOTCH; instead , O-glucosylation is required for the proper conformation of NOTCH to allow efficient cleavage by metalloproteases ( S2 cleavage ) and presenilin ( S3 cleavage ) to generate the active NICD transcription factor [13 , 24] . As POGLUT1 localizes to the endoplasmic reticulum , the lack of cell surface localization of CRUMBS2 in Poglut1 mutants is likely to be due to a requirement for glycosylation of the EGF repeats of CRUMBS2 for its correct folding in the ER and subsequent trafficking to the Golgi and the cell surface . This hypothesis is supported by the accumulation of Endo H-sensitive forms of CRUMBS2 in Poglut1wsnp mutant embryoid bodies . The EGF repeats in Drosophila Crumbs can also be modified by Rumi/POGLUT1 [43] . However , a mutant form of Drosophila Crumbs in which alanine replaced serine in all seven potential Rumi/POGLUT1 modification sites did not produce a mutant phenotype: homozygous S-to-A mutants were viable and appeared normal , and most of the mutant Crumbs protein localized normally to the membrane [43] . Thus glucose modification of CRUMBS2 is essential for its function in mammals but not in Drosophila . In Drosophila , Rumi mutants exhibit a Notch-dependent phenotype only when raised at higher temperatures [24 , 43] . In contrast , POGLUT1-dependent modifications are essential for the activity of mammalian NOTCH . The more significant role of O-glucose addition in mammals may correlate with the more extensive POGLUT1-dependent modifications present on the mammalian proteins and may also reflect the higher body temperature of mammals . In all cases examined to date , O-glucose is extended to the trisaccharide form on all the EGF repeats of mammalian NOTCH [13 , 21] and CRUMBS2 ( this work ) . However , the predominant species of POGLUT1-modified EGF repeats in Drosophila Notch and Crumbs is glucose monosaccharide , which only occasionally is extended to the trisaccharide [24 , 43] . POGLUT1-dependent modification of CRUMBS2 proteins is likely to be important in a variety of human diseases . Mutations in EGF repeats of human CRUMBS1 , another likely substrate of POGLUT1 , are found in retinitis pigmentosa and Leber congenital amaurosis [49–52] , and could affect the glycosylation status of CRUMBS1 and its membrane localization in the eye . Mutations in human CRUMBS2 are associated with some cases of congenital nephrosis [53 , 54] , suggesting that POGLUT1 is also a candidate gene in this human syndrome . Whole exome sequencing has demonstrated that human POGLUT1 mutations are responsible for a subset of the cases of Dowling-Degos Disease , an autosomal dominant hyperpigmentation disorder [55] . POGLUT1 is overexpressed in several human leukemia , breast cancer and endometrial cancer cell lines [28 , 56–58] . It has been suggested that POGLUT1 mutations disrupt the activation of Notch signaling pathway in these diseases; our results suggest that POGLUT1-dependent Crumbs activity should also be considered .
This work was approved by the Memorial Sloan Kettering Cancer Center IACUC ( protocol number 02-06-013 ) and studies were conducted in accordance to their guidelines . The wsnp allele was generated by ENU mutagenesis of C57/BL6J mice [33] . This allele harbors a T to C transition in the splice donor site of intron 3 of Poglut1 , which creates an SfaNI restriction fragment length polymorphism . ES cells harboring the knock-out first allele were obtained from the International Mouse Knockout Project ( HEPD0700_1_A09 ) . They were injected into C57BL/6J blastocysts to generate chimeras , and chimeras were screened for transmission of gene trap allele ( Poglut1gt ) , which introduces the LacZ coding gene downstream of Exon 3 . The conditional allele ( Poglut1flox ) was generated by crossing Poglut1gt to actin-Flip mice [59] . The null allele ( Poglut1Δ ) was generated by crossing conditional allele to CAG-Cre [60] . The conditional allele of Crumbs2 [61] was crossed to CAG-Cre ( Jax ) to generate the null allele ( Crumbs2-/- ) . Crumbs1rd8 [62] and Sox2-Cre [38] have been described . Mice carrying the nodal-lacZ knock-in allele [63] was a gift from Elizabeth Robertson and the TOPGAL mice [64] were provided by Elaine Fuchs . The wsnp mutation was mapped to a 407 kb interval between D16Mit90 and D16Mit12 simple sequence length polymorphism ( SSLP ) markers by meiotic recombination . Genomic DNA was purified from three mutants pooled together , and exonic DNA across the genomic interval was enriched using Agilent SureSelect solution based technology with custom target capture . Samples were multiplexed and bar-coded for SOLiD sequencing . Sequencing reads were aligned to the C57BL/6 reference genome using SHRiMP [65] . Whole mount in situ hybridization and LacZ staining was performed as published [36] . For in situ hybridization embryos were dissected in ice-cold PBS-0 . 4%BSA and fixed in 4% paraformaldehyde overnight . Following a series of dehydration and rehydration in methanol series , embryos were hybridized with the corresponding in situ probes , washed and developed using BM-purple . For β-galactosidase activity , embryos were dissected in ice cold PBS and fixed in freshly prepared fixative solution ( 5mM EGTA , 0 . 2% glutaraldehyde and 2mM MgCl2 in 0 . 1M phosphate buffer ( pH 7 . 3 ) ) for 10–15 minutes . Following washes in detergent rinse ( 2mM MgCl2 , 0 . 02% Nonidet P-40 and 0 . 01% sodium deoxycholate in 0 . 1M phosphate buffer ) , the embryos were stained with X-GAL ( 1mg/ml X-GAL , 5 mM potassium ferricyanide ) in detergent rinse ) overnight . Following staining , the embryos were rinsed in PBS and fixed in 4%PFA for an hour before imaging . Embryos were dissected in ice cold PBS-BSA and fixed in 4% PFA for one hour at room temperature for immunostaining or over-night at 4°C for in-situ hybridization . The embryos were embedded in OCT ( optimal cutting temperature ) and cryosectioned at 10–12 μm thickness . Immunostaining on frozen sections was performed as published [36] . Primary antibodies were diluted in blocking buffer and incubated overnight at 4°C . The secondary antibodies were diluted in blocking buffer and incubated for 1 hour at room temperature . DAPI was included in the secondary incubation . For whole mount active NOTCH1 immunostaining , embryos were dissected in ice cold PBS-BSA and fixed overnight in 4% PFA/PBS at 4°C , dehydrated in methanol and stored at -20°C overnight . Following rehydration , antigen unmasking was performed in Vector unmasking solution ( H-3300 Vector labs ) at 98°C for 10 minutes . After reaching room temperature embryos were washed in MilliQ water and incubated in acetone at -20°C for 8 minutes . Next the embryos were washed and incubated in blocking buffer overnight at 4°C ( Blocking buffer = 10% goat serum , 5% BSA , 0 . 3% Triton-X100 in PBS ) . The embryos were incubated in anti-cleaved NOTCH1 antibody ( 1:1000 ) for 2 days . Following 4–5 washes with blocking buffer , the embryos were incubated in secondary antibody overnight at 4°C . The embryos were washed extensively before mounting in glass-bottom dishes for confocal imaging . The following antibodies were used: anti-CRUMBS1 and anti-CRUMBS2 were obtained from Jane McGlade ( Hospital for Sick Children , Toronto ) and were used at 1:100 and 1:50 respectively [66]; Anti-pan-CRUMBS was a kind gift from Ben Margolis ( University of Michigan , Ann Arbor ) and was used at 1:200 [67] . Commercially available antibodies used were: anti-SOX2 ( Santa Cruz 1:100 ) , anti-E-CADHERIN ( Sigma 1:200 ) , anti-LAMININ ( Sigma 1:200 ) and anti-active NOTCH1 ( Val 1744—Cell Signaling 1:1000 ) . Confocal microscopy was performed using a Leica-Inverted SP5 or Leica-Upright SP5 laser , point-scanning confocal microscope . Confocal datasets were analyzed using the Volocity software package ( Improvision ) . Embryos were dissected in ice cold PBS and snap frozen on dry ice . The embryos were lysed in RIPA buffer with protease inhibitor added . The lysates were equilibrated on ice for 20 minutes , after which they were sonicated ( 3 X 30 seconds ) . Lysates were incubated for 10 minutes in ice and then centrifuged to remove the cell debris . The supernatant was mixed with 2X SDS loading dye ( 1:1 ) and loaded on an SDS-PAGE for analysis . The dilutions for primary antibody were: anti-pan-CRUMBS ( Ben Margolis [64] , 1:2000 ) , anti-Cleaved-NOTCH1 ( Val 1744—Cell Signaling 1:1000 ) , anti-NOTCH1 ( Abcam 1:1000 ) and anti-V5 ( Invitrogen 1:5000 ) . The Crumbs2 cDNA was synthesized from wild-type embryos harvested at E8 . 5 . Full-length Crumbs2 was cloned into the Gateway vector pDEST40 to generate Crumbs2 with His and V5 tags at its C-terminal . This sequence was sub-cloned with the tags into the EcoR1 site of the pCAGGS vector to generate pCAGGS-Crumbs2 for high expression in embryonic stem cells . ES cells were derived using the 2i protocol from both wild type and Poglut1wsnp blastocysts [44] , were eventually weaned off iMEFs and were cultured in KnockOut DMEM ( Gibco ) supplemented with 15% fetal bovine serum ( HyClone ) , 2 mM L-glutamine ( Gibco ) , 0 . 1 mM β-mercaptoethanol ( Gibco ) , 0 . 1 mM non-essential amino acids ( Gibco ) , 1 mM sodium pyruvate ( Gibco ) , 1% v/v penicillin and streptomycin ( Gibco ) , 1 , 000 units LIF ( Millipore ) , 1 μM PD0325901 and 3 μM CHIR99021 ( Stemgent ) . To generate stable cells lines expressing tagged full-length CRUMBS2 , wild-type and Poglut1wsnp ES cells were electroporated with pCAGGS-Crumbs2 DNA linearized by cutting with ScaI and a circular PGK-Puro-pA plasmid [68] that confers a transient puromycin resistance . Stable lines were selected with puromycin ( Invitrogen ) using published protocols [69] . Fifteen independent colonies were screened for expression of tagged full length CRUMBS2 both by immunofluorescence and western blots analysis with the anti-V5 antibody . One cell line each for wild type and Poglut1wsnp was selected based on high expression levels and was used for protein purification . ES cells were trypsinized and re-suspended at 106 ES cells per 10 ml media in non-adherent conditions ( 10 cm bacterial petri dishes ( VWR ) coated with Sigmacote ) . EBs were cultured in the absence of LIF and 2i in DMEM medium containing 15% fetal bovine serum , 2 mM L-glutamine ( Gibco ) , 0 . 1 mM β-mercaptoethanol ( Gibco ) , 0 . 1 mM non-essential amino acids ( Gibco ) , 1 mM sodium pyruvate ( Gibco ) , 1% v/v penicillin and streptomycin ( Gibco ) . The medium was replaced every day . To determine the time course of endogenous CRUMBS2 expression , EBs were harvested daily from day 1 to day 6 . ES cells were cultured in large ( 6X15 cm ) plates and differentiated to embryoid bodies . Wild-type and Poglut1wsnp mutant embryoid bodies were harvested 2 days after induction of differentiation from ES cells and snap frozen . These were re-suspended in RIPA buffer with 2% DDM ( n-Dodecyl β-D-maltoside ) and Roche protease inhibitor and subjected to three cycles of freeze-thaw to increase the release of membrane protein , then sonicated for 4 times for 30 seconds each and incubated at 4°C for one hour with continuous shaking . The lysates were clarified by centrifugation for 20 minutes at 15 , 520 g . The supernatant was incubated with anti-V5 antibody ( 1:1000 ) for 4 hours at 4°C shaking . This mixture was incubated with magnetic nickel beads ( Invitrogen ) for 2 hours at 4°C shaking . The beads were rinsed four times in RIPA lysis buffer , and the beads were heated to 95°C in 2XSDS loading buffer to elute the protein . The eluted protein was run on a 7% SDS-PAGE . The gel was washed 3 times in HPLC purified water and then stained with Pierce gel code blue stain reagent for 5–10 minutes . Following staining , the bands were cut out , subjected to chymotrypsin digest and used for mass spectrometric analysis . After 5 days of differentiation , wild-type and Poglut1wsnp EBs were lysed in sample buffer containing SDS and β-mercaptoethanol , and denatured by boiling for 5 min . After cooling , NP-40 was added and samples were diluted in PBS to final concentrations of 1% NP-40 , <1%SDS and 1% β-mercaptoethanol . Samples were then incubated with 5U of Peptide-N-Glycosidase F ( PNGase F ) , prepared as described [70] or 25mU of Endoglycosidase H ( Endo H ) ( Roche ) overnight at 37°C . Western blotting was performed to detect the removal of N-glycans from V5-tagged CRUMBS2 protein . Bands containing purified mouse CRUMBS2-V5 were reduced , alkylated , and subjected to in-gel chymotryptic digest as described previously [45] . Peptides were analyzed by nanoLC-MS/MS as previously described [21] . Glycosylation of peptides was identified by using neutral loss searches , and Extracted Ion Chromatograms ( EICs ) were generated to compare relative amounts of each glycoform of the relevant peptides .
|
Post-translational addition of sugar chains is essential for normal activity of many secreted and transmembrane proteins and dozens of human genetic diseases are associated with congenital disorders of glycosylation . Protein O-glucosyltransferase 1 ( POGLUT1 ) , which is essential for early mouse development , catalyzes the addition of O-glucose to extracellular EGF repeats of proteins , including NOTCH1 . Here we show that mouse POGLUT1 modifies NOTCH1 in vivo; however , the essential role of POGLUT1 in gastrulation is due to POGLUT1-dependent glycosylation of EGF repeats in the apical polarity protein CRUMBS2 . In contrast to findings in Drosophila , where modification of Crumbs by POGLUT1 is not required , mouse POGLUT1 is required for the activity of CRUMBS2: the unmodified protein fails to localize to the apical membrane and the gastrulation defects of Poglut1 mutants are indistinguishable from those of Crumbs2 mutants . Human mutations in POGLUT1 cause Dowling-Degos Disease type 4; the hyperpigmentation associated with this autosomal dominant disease was previously attributed to altered Notch signaling , but our results suggest that this disease and other POGLUT1-associated phenotypes may be due to altered activity of CRUMBS proteins .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
Protein O-Glucosyltransferase 1 (POGLUT1) Promotes Mouse Gastrulation through Modification of the Apical Polarity Protein CRUMBS2
|
Stochastic fluctuations in signaling and gene expression limit the ability of cells to sense the state of their environment , transfer this information along cellular pathways , and respond to it with high precision . Mutual information is now often used to quantify the fidelity with which information is transmitted along a cellular pathway . Mutual information calculations from experimental data have mostly generated low values , suggesting that cells might have relatively low signal transmission fidelity . In this work , we demonstrate that mutual information calculations might be artificially lowered by cell-to-cell variability in both initial conditions and slowly fluctuating global factors across the population . We carry out our analysis computationally using a simple signaling pathway and demonstrate that in the presence of slow global fluctuations , every cell might have its own high information transmission capacity but that population averaging underestimates this value . We also construct a simple synthetic transcriptional network and demonstrate using experimental measurements coupled to computational modeling that its operation is dominated by slow global variability , and hence that its mutual information is underestimated by a population averaged calculation .
To survive in challenging conditions , cells need to detect , transduce , and process signals from their environment . A cell’s ability to precisely process environmental signals is limited by intrinsic fluctuations and variability of its cellular processes . This variability takes root in the stochastic nature of biochemical reactions . For a given pathway , this includes the stochastic steps involved in transcription and translation [1–4] as well as diffusion-reactions , dissociations , allosteric changes , and degradation of biological molecules . A signal propagates across cellular networks through molecules undergoing these various reactions , and gets distorted and altered by their probabilistic nature . Therefore , metrics for quantifying the limits of faithful information propagation ( signaling fidelity ) in biological pathways are crucial for understanding their information processing and transduction capabilities . Mutual information [5] ( MI ) is a natural metric for characterizing information transmission between the inputs of a stochastic network and its nodes . MI quantifies the level of precision with which a given node ( s ) in a network estimates and responds to an input ( s ) by accounting for both the mean and variability in the response . Recent studies have used MI to characterize information transmission between environmental inputs and transcription factors in a number of genetic circuits [6–10] . In these studies , steady-state MI was computed for a variety of in silico networks to assess their stationary response as a function of input dose . More recently , these ideas were extended to optimize time-dependent MI in delay circuits with binary inputs , and MI was used to discuss maximally informative network topologies in these contexts [11] . In addition , time dependent MI calculations were used to obtain fundamental limits on the suppression of molecular fluctuations for different network topologies [12] . Several experimental studies have also used MI to assess signaling fidelity . MI was used as a metric to argue that negative feedback enables dose-response alignment and enhances information transmission in the pheromone pathway in yeast [13] . Similarly , MI was used to estimate time-dependent information transfer in tumor necrosis factor ( TNF ) signaling , and to assess transmission bottlenecks in this system [14] . Recently , robustness and compensation of information transmission in different pathways and pharmacological perturbations were attempted in PC12 cells using similar measurements [15] . These experimental studies relied on driving isogenic cell populations with various inputs , and then calculating the mutual information based on the overall variability in the population response . Such calculations mostly found low MI values , suggesting that cellular pathways might have on average low information transmission capacity . In this work , we argue that these calculations often under-estimate MI of a pathway in a single cell , since they do not account for 1 ) variability in initial conditions and 2 ) variability that is extrinsic to the pathway . The overall effect of these two sources of variability is that any single cell has a quantitatively distinct input-output relationship [3 , 4 , 16] and that calculations that take this into account are needed for more accurate estimation of MI from experimental data . By assuming that extrinsic variability manifests as cell-to-cell differences in a global parameter , such as translation capacity , we demonstrate in a simple in silico circuit that mixing cells with different parameters sets ( and/or different initial conditions ) reduces the value of the computed MI . We also argue this point experimentally by building a simple synthetic circuit that exhibits strong extrinsic variability , and then demonstrating with the help of computational modeling that single cells within the population have a larger mutual information than that exhibited by the averaged population . These results indicate that cells might possess higher capacity for information transmission than previously appreciated .
We first assumed that this circuit is isolated from the rest of the cell , and that any stochasticity it exhibits is only the result of its chemical reactions ( intrinsic variability ) . When this system is unstimulated ( t ≤ 0 ) , its molecular species assume a joint steady-state distribution , p ( y 1 * , y 1 , M y 2 , y 2 , t ≤ 0 ) . As an example , we show the marginal distribution of Y 1 * in Fig 1c . This distribution represents the range of initial conditions in Y 1 * that a population containing this network would exhibit before any input is applied . We will first compute the MI of the network while ignoring this initial distribution of states , assuming that all cells in the population start from the same initial condition ( for state S1 , this is the mean of the initial joint distribution , see Fig 1c ) which we refer to as a homogeneous initial condition . This could be thought about as the mutual information of one cell in that population . We plot the time-dependent mutual information between the input and the different species of the circuit: I ( x+; y1 , t∣S1 ) , I ( x+; My2 , t∣S1 ) , I ( x+; y2 , t∣S1 ) ( Fig 1d ) . The MI from the input to y1 , I ( x+; y1 , t∣S1 ) , has rapid dynamics , peaking initially and decaying with time to a steady-state . The initial peak in this MI is due solely to the activation and inactivation of y1 , while the subsequent decrease to steady-state is due to the fluctuations in the synthesis and degradation of y1 . By contrast , the MI from the input to My2 ( I ( x+; My2 , t∣S1 ) ) is slower and on the order of tens of minutes , while that of the protein y2 ( I ( x+; y2 , t∣S1 ) ) is on the order of hours . This is not unexpected , as the mutual information signals for each species follow the causality of the circuit where y2 shows the largest delay . The increase of I ( x+; y2 , t∣S1 ) as a function of time has an intuitive explanation in terms of y2 dynamics . To visualize this , we plot y ‾ 2 ( n , t ) , the mean of y2 as well as y ‾ 2 ( n , t ) ± σ y 2 ( n , t ) versus X+ ( n ) for t = 0 , 75 , and 750 minutes ( Fig 1e , red lines ) , where σy2 ( n , t ) is the standard deviation in y2 . We will refer to these plots as the time-dependent dose response relationships . For t = 750 , more values of x+ are resolvable from measurement of y2 than at time t = 75 . For example , for x+ > 150 , steps in the dose response curves constrained between the standard deviations ( Fig 1e , black lines for t = 75 and 750 minutes ) approximate how well a measurement in y2 can infer the value of x+ . At time 750 , about 2 steps are resolvable allowing for two distinct ranges of x+ to be inferred . While for t = 75 minutes , only one distinct range of x+ is inferrable . The larger the number of resolved states , the higher the value of the mutual information . When mutual information is calculated between the input and a given node over the entire time duration of the signals , the mutual information between the input and each successive node has an upper bound equal to that of the prior node . This is known as the data processing inequality [5] . However , since we are evaluating the time-dependent MI at a given time t , the instantaneous value y2 can have more information about the input than y1 . Indeed , at t approximately greater than 150 minutes , we find that the MI I ( x+; y2 , t∣S1 ) is greater than I ( x+; y1 , t∣S1 ) or I ( x+; My2 , t∣S1 ) ( Fig 1d ) . This is because for the particular parameter set used in this example , the noise propagated from y1 onto y2 is averaged out , and the only variability in y2 stems from its own production and degradation . As a result , I ( x+; y2 , t∣S1 ) can be modulated to be higher or lower than I ( x+; y1 , t∣S1 ) by changing the rates of y2 production and degradation [17] . On the other hand , increasing the number of y1 molecules would increase its mutual information as this would reduce the noise in the y1 signal . Therefore , the mutual information at each node of this pathway can be modulated through choice of kinetic parameters . Similar observations that filtering can improve time-dependent MI between success nodes have been discussed in the context of other types of pathways [18] . Next , we examined mutual information while accounting for the fact that cells assume a distribution of initial states across the population upon receiving the input stimulus . We do so by incorporating the pre-stimulus steady state initial joint distribution into the MI calculations . This variability in initial conditions transiently reduces the MI ( Fig 1f ) . At steady-state , the mutual information curves computed for a single or a distribution of initial states eventually converge onto each other at approximately t = 2750 minutes ( Fig 1f , inset ) . This convergence at longer times occurs because a population in which every cell assumes the same exact initial conditions will eventually produce a heterogeneous distribution of states due to the intrinsic stochasticity of the biochemical reactions . For the values of parameters used in this example , the convergence of the two MI curves proceeds very slowly . Therefore , even when only intrinsic fluctuations are present , with no extrinsic contributions to variability , and for a given distribution of initial conditions , a single cell still transiently assumes , on average , a higher time-dependent mutual information than the whole population . In our case , this difference is very modest . Thus far , in our MI calculations , we have only accounted for variability in initial conditions given a single parameter set for the pathway . More realistically , any given pathway in a cell is subjected to variability through coupling to other cellular activities . This is known as extrinsic noise to distinguish it from intrinsic noise generated by the pathway itself . There are many extrinsic sources of variability that cellular pathways experience . For example , different cells may contain different numbers of polymerases or ribosomes , and hence have different capacities for transcription and translation [3] . This extrinsic variability can be accounted for in many ways , the simplest is to assume that the transcription or translation rate constants themselves can assume different values in different cells across the population . To demonstrate the contribution of extrinsic variability to MI calculations , we consider a simple case where cells in the population have different translation rates . To do so , we add a stochastic global variable , G , which affects the protein creation rates such that β ^ y 1 * = β y 1 * G / G ‾ and β ^ y 2 = β y 2 G / G ‾ , where β y 1 * and βy2 are the nominal values for the parameters used above . In this way , the protein creation rates keep their mean value , but fluctuate because of their coupling to G . For this example , G follows a memoryless birth/death process such that the mean of G is G ‾ = β g / γ g ( βg , γg are the birth and death rates ) . It follows that G has a coefficient of variation given by η g = 1 / G ‾ . First , setting G ‾ = 50 , we chose βg = 1 . 5 × 10−6 mol-s−1 and γg = 3 × 10−8 s−1 . These values establish a stationary distribution of states , which we use as an initial distribution for the MI calculations . Fluctuations in the translation rate induce extra variability in the pathway components ( compare the initial distributions of Y 1 * in Fig 2a to Fig 1c ) . As a result , mutual information calculations with this added extrinsic variability ( and using the population distribution of initial conditions ) show that I ( x+; y2 , t ) is now drastically reduced compared to the case when a single parameter set is used to represent the lack of global variability ( compare black line in Fig 2b with value in Fig 1d ) . Here also , as expected , MI calculations from a single initial state corresponding to parameters G = G ‾ ( state S1 ) , G = G ‾ − G ‾ ( state S2 ) and G = G ‾ + G ‾ ( state S3 ) , generate high transient values ( red ( S1 ) , blue ( S2 ) and green ( S3 ) curves in Fig 2b ) . This discrepancy between single cell and population MI is further highlighted by examining the time-dependent dose response relationship between y2 and x+ at t = 750 ( Fig 2d ( full population ) and Fig 2c ( S1 , S2 , and S3 ) . Again , the sub-populations generated from S1 , S2 , S3 each have little variability ( high mutual information ) relative to the full population . While constraining G ‾ = 50 , we investigated the time-dependent MI for different values of βg and γg . Our original choice of γg = 3 × 10−8 forces G , and hence the translation rates β ^ y 1 * and β ^ y 2 , to fluctuate very slowly . Therefore , the convergence of the MI values computed from a single initial condition versus the full distribution also proceeds slowly . As γg increases , this convergence proceeds faster ( Fig 2e ) . Therefore , these results indicate that the mutual information of a pathway can be severely underestimated by population-based measurements if the pathway is subjected to global fluctuations that proceed on a slower timescale than the pathway itself . Next , we sought to probe the major determinants of mutual information for a simple synthetic transcriptional circuit ( Fig 3a ) . In this circuit , a constitutively expressed transcription factor Y 1 * interacts with a small molecule X+ , leading to the activation of the transcription factor . The active transcription factor Y1 translocates into the nucleus and activates expression of a gene Y2 . In our implementation , Y 1 * is an estradiol ( input X+ ) responsive chimeric transcription regulator ( TR ) consisting of three fused elements: an activation domain ( from MSN2 ) , a lipid-binding domain ( from the human estradiol receptor , hER-LBD ) , and a DNA binding domain ( from GAL4 ) . When estradiol binds to the LBD , the activated TR Y1 translocates to the nucleus and controls the expression of promoters containing Gal4-binding-sites . Therefore , the protein Y2 ( in this case a fluorescent protein ) is produced from a Gal4-responsive GAL10 promoter ( See Materials and Methods for more details ) . At the same time , Y 1 * is produced from an altered version of the promoter of the alcohol dehydrogenase 1 gene ( ADH1 ) . We constructed two strains for measurement purposes . Strain 1 contains the circuit in addition to two copies of the GAL10 promoter , one driving YFP and the other driving mCherry . Strain 2 contains the circuit , but this time with two copies of the ADH1 promoter , one driving the production of Y 1 * and the other driving the production of YFP ( which we will refer to as Y1r ) . The same strain also contains a GAL10 promoter driving the production of the mCherry ( Y2 ) protein . These strains were useful for two reasons . First , we wanted to establish how mutual information computations depend on the ability to simultaneously measure different quantities in a circuit ( e . g . Y1 and Y2 versus Y2 alone ) . Given that this necessitates the use of two fluorescent proteins , in this case YFP and mCherry , we wanted to ascertain that the results we obtain are qualitatively independent of the choice of fluorophores , given that mCherry has lower dynamic range than YFP with higher background fluorescence and hence increased noise at low concentrations . For each strain , we subjected 12 exponentially growing populations ( wells ) of cells cultured in non-repressive media to input concentrations of estradiol ( x+ ) log-sampled between 0 and 100 nM . The 12 measurement points sufficiently sampled the dose response relationships . The number of cells measured from each well was greater than 3000 , ensuring good statistics for approximating the MI [14] . All cultures were started from zero estradiol concentrations . Samples were taken at t = 0 , 65 , 165 , 330 and 580 minutes . Fig 3b shows the time-dependent dose-response relationships of estradiol versus y2 ( in this case YFP , strain 1 ) for these timepoints , where fluorescence values were normalized with respect to side scatter in order to minimize the effects of cellular volume and shape dependent differences . The dose-response relationships of y2 normalized by their respective maximum mean values ( Fig 3d ) exhibit an interesting trend: for the last 3 time points , the traces for the mean and variability are very similar to each other . The only outlier to this trend is the time point at t = 65 minutes after stimulation ( Fig 3d ) . For this timepoint , fluorescence is weak and strongly overlaps with autofluorescence and folding delays , and therefore the true signal cannot be accurately estimated . Autofluorescence and folding delay also contributes , albeit less dramatically , to the measurement at the t = 165 minutes timepoint ( Fig 3d ) . The mCherry measurements ( strain 1 or strain 2 ) generated the same trend ( S1a and S1b Fig ) albeit with a noisier outcome than YFP due to the limited dynamic range of mCherry . As a consequence , the y2 ( YFP ) and y1r ( YFP ) experimental measurements from the two strains can be used in combination for comparison of modeling with data . The fact that variability in the y2 data irrespective of the fluorescent protein does not decrease with increasing mean values suggests that dominant fluctuations are unlikely to be intrinsic to the pathway . Fig 3c plots measurements of y1r ( YFP , strain 2 ) . Unexpectedly , despite the common assumption that the ADH1 promoter has constitutive and constant expression , we found that it exhibits a modest dependence on estradiol . We do not know the root of this dependence , but it is likely to reflect the influence of the circuit itself on the metabolic state of the cell , hence affecting ADH1 promoter activity . Overall , the growth rate of these strains is independent of estradiol for concentrations under 100 nM over the duration of the experiment ( S1c and S1d Fig ) , and therefore this effect can be compensated for in the mutual information calculations . It is worth noting here that we are making the assumption that despite the fact that YFP ( Y1r ) and Y 1 * are different proteins sharing only the same transcription rate ( since both are driven by the ADH1 promoter ) , they share the same dominant noise characteristics . This would be the case if their intrinsic noise , which can be different , is insignificant compared to a dominant source of extrinsic noise affecting both . Next , we present data and modeling demonstrating that , indeed , noise in both Y1 and Y2 is most likely dominated by the same extrinsic global component . Since the measured distributions are approximately gaussian for the majority of estradiol concentrations ( S2 Fig ) and the synthetic circuit ( Fig 3a ) follows the same basic chemical equations as the simple pathway we have studied in Fig 1b , we used this already established model to computationally explore different noise scenarios ( see Materials and Methods for a more technical justification of the model ) . Specifically , we simulated the model ( parameter values listed in Table 2 ) with both intrinsic variability and added global extrinsic parameter variability as sources of stochasticity . The data we collected are in fluorescent units , therefore we set our model to arbitrarily yield maximum y2 protein expression levels of about 2500 molecules , likely an underestimation of the actual system . However , this choice constitutes a scaling factor and does not affect any of our results . We also accounted for the estradiol dependence of ADH1 ( Fig 3c ) by adding to the model a term depicting the modest estradiol dependent repression of this promoter . For global parameter variability , we again chose to focus on the parameters affecting protein expression . We potentially could model the global parameter variability with cell-to-cell heterogeneity in the protein degradation rates . However , given that our experimental data does not measure the expression of genes involved in either of these processes , i . e . no way to experimentally distinguish the source ( s ) of global parameter variability , we chose to model global variability in the protein creation rates . Following the same procedure as in the previous section , we added a stochastic global parameter , G , which affects the protein creation rates for Y1* and Y2 such that β ^ y 1 * = β y 1 * G / G ‾ and β ^ y 2 = β y 2 G / G ‾ . The noise in the experimental Y1r data is approximately . 155 , therefore modeling Y1 and Y1r using Poissonian statistics sets the mean of the global noise variable to G ‾ ≈ 42 . We first assumed that global parameter fluctuations are slow relative to the circuit timescales ( γg = 3 × 10−6 ) . Simulating the model with this slow global source of fluctuations ( SGF model , ( Fig 3e–3g ) ) generated profiles for normalized y2 ( Fig 3g ) that recapitulated the highly similar variance envelopes of the experimental time-dependent dose responses ( Fig 3d ) . This behavior was a characteristic feature of the model for any γg < = 3 × 10−6 . By contrast , as the global fluctuating variable assumes a faster timescale ( γg > 3 × 10−6 ) , the variability envelopes in the normalized time-dependent dose response of y2 started to diverge from each other ( S3a and S3b Fig , γg = 3 × 10−4 ) . As expected , the system modeled with intrinsic variability only ( G ‾ = ∞ , parameter values listed in Table 3 ) shows a normalized time-dependent dose response in which variability decreases as a function of time as the protein levels increase ( S3c and S3d Fig ) . Given the data in Fig 3b , if the fluctuations were purely intrinsic , the ratio of the standard deviation to the mean between times 165 and 580 minutes would decrease by a factor of approximately 1 . 7 . This is a change we should be able to detect in our data since for the number of cells sampled , the error in estimating the means and standard deviations in the dose response relationships are . 5 percent and 2 percent , respectively . However , as previously discussed , the experimental data shows that this ratio is relatively invariant for the last 3 time points ( Fig 3d , inset ) while increasing for the both the fast global fluctuations model ( S3b Fig , inset ) and the intrinsic variability model ( S3d Fig , inset ) . Our argument is further strengthened by the fact that in order to capture the noise observed in y1r with the intrinsic variability model for the first timepoint , we had to set the y1r mean copy number in the model to an unrealistically low value for a strong promoter such as ADH1 ( approximately 40 proteins ) , further indicating that variability is unlikely to be intrinsic . The results for the SGF model for γg < = 3 × 10−6 are not an artifact of the estradiol dependence of ADH1 since an SGF model without this effect yields indistinguishable results ( S3e and S3f Fig ) . We therefore conclude that the dominant source of variability in this synthetic circuit is likely to be due to a globally slow fluctuating variable . This is consistent with previous results , which also indicated that global parameters play a dominant role in cell to cell variability and that these parameters exhibit fluctuations at a slower timescale than fluctuations of processes involved in gene expression [4] . In terms of mutual information , the fact that the normalized time-dependent dose responses coincide in terms of their variability ( Fig 3d ) implies that the experimentally computed mutual information I ( x+ , y2 , t ) at t = 165 , 330 and 580 minutes should be similar . This is indeed the case ( Fig 4a , solid black ) . Importantly , I ( x+ , y2 , t ) peaks and reaches a plateau at approximately 1 bit , at an earlier time than when y2 reaches its steady-sate . This further lends credence to the idea that the variability in the population is dominated by global parameter variability . Gratifyingly , the model with ‘slow’ global parameter fluctuations ( with γg = 3 × 10−6 ) also captures the time-dependent mutual information seen in the data without any further parameter tuning ( Fig 4a , solid blue ) . Since slow global fluctuations seem to dominate in this circuit , our analysis above indicates that the population mutual information might be under-estimating the fidelity of a single cell . To illustrate this point , we used the model to computationally isolate and compute the mutual information I ( x+; y2 , t∣Si ) for single cells S1 , S2 and S3 as defined in the computational example above . These calculations yield a substantially higher MI value than the population MI for the time span simulated ( Fig 4a ) . Evidently , and as explained above , the MI for S1 , S2 , or S3 will eventually converge back to the whole population MI , but here it will do so on a much slower timescale than that of the system . For example , the dose response and distribution of y2 at time 580 minutes when the system is started from S1 , S2 and S3 ( Fig 4b ) still shows tighter variability than that of the full population . Allowing for intrinsic variability in the initial conditions , i . e . starting with cells with G = G ‾ at time zero ( state Sg ) yields a similar MI value to that of I ( x+; y2 , t∣S1 ) for γg = 3 × 10−6 ( Fig 4a ) . Finally , we explored how simultaneous measurement of y1r and y2 affects mutual information calculations . Calculations using the model indicate that as expected , knowledge of y1r improves the estimate of mutual information . For a slow globally fluctuating variable ( γg = 3 × 10−6 ) , the joint mutual information I ( x+; [y1r y2] , t ) is larger than I ( x+; y2 , t ) . It can be shown that I ( x+; [y1r y2] , t ) = I ( x+; y1r , t ) + E[I ( x+; y2 , t∣y1r ) ] where E[I ( x+; y2 , t∣y1r ) ] is the expected value of I ( x+; y2 , t∣y1r ) . Since the influence of estradiol on y1r adds ( albeit very slightly ( Fig 3c ( data ) and Fig 3f ( model ) ) to the mutual information , i . e . I ( x+; y1r , t ) > 0 , we normalized for this effect . To do so , at a given time , we set the y1r mean at each estradiol value to the value of the y1r mean at zero estradiol while adjusting the variance to preserve the noise in y1r at each estradiol value . Importantly , this operation does not affect correlation between y2 and y1r at each estradiol value , but enforces I ( x+; y1r , t ) = 0 . We confirm that this does not change our conclusions that knowledge of y1r improves the estimate of mutual information ( compare Fig 4c blue , dashed black and dashed magenta ) . For comparison , we can carry out mutual information from the data obtained using Strain 2 in which both y1r and y2 are measured . In this strain , y2 is the fluorescent protein mCherry which has a limited dynamic range . Importantly , at the highest estradiol values and peak mCherry signal ( time 580 minutes ) , the measured correlation between y1r and y2 ( greater than . 79 ) is less than ten percent below the model predictions . Even at these signal levels the noise in the mCherry signal still deteriorates the correlation . For decreasing values of estradiol the correlations become increasingly inaccurate . Therefore , the values of the MI cannot be quantitatively compared to the model which was fitted to YFP data . However , the qualitative trend of increased MI due to measurement of y1r relative to computing the MI with no knowledge of y1r should also hold . This is indeed seen to be the case ( Fig 4d ) . This insight is in agreement with recent work [19] that studied mutual information in the RAS/ERK pathway . Nuclear ERK ( erknuc ) was used as a readout of pathway information transmission . The MI at time t between this readout and the input x was conditioned for single cell ERK levels , using measurement of total ERK ( erktot ) . It was also shown that I ( x; [erktot erknuc] , t ) is greater than I ( x; erknuc , t ) . Therefore , simultaneous measurements of different cellular variables improve estimates of mutual information capabilities of single cells .
In this work , we illustrated how variability in initial conditions across a population , as well as slow-fluctuating extrinsic ( global ) variables can generate low values for the population mutual information in response to an input . We also demonstrated that when subpopulations of cells that have similar parameters or initial states are isolated , their mutual information values are transiently much higher than those of the whole population . These findings are important in light of the fact that many previous studies have found that extrinsic variability is a substantial contributor to pathway fluctuations . Indeed , our own experimental data using a synthetic circuit also implicated extrinsic fluctuations as a major source of variability . As a result , cells in a population cannot be considered to be the same noisy channel for mutual information calculations . Rather , each cell is a different noisy channel possessing its own parameters . Recent work [20] using light-inducible input signals [21 , 22] to a mammalian RAS/MAPK pathway observed that different isogenic single cells have quantitatively different dose-response relationships . Interestingly , for the RAS/MAPK mammalian system , the dose-repsonse relationships were repeatable for hours within a given single cell [20] , suggesting slow global parameters that affect that pathway for that duration . A direct assessment of mutual information requires repeated time-resolved measurements in single cells . Another strategy to better approximate mutual information is to simultaneously measure a large number of interconnected variables , including global states . This might be increasingly feasible with breakthrough technologies such as mass-cytometry ( a . k . a . CyTOF ) [23] as well as improvements in fluorescent reporter technologies . In the mean time , however , we have demonstrated that computational modeling , especially with respect to the patterns of time-dependent variability , can generate valuable insights into whether intrinsic or extrinsic fluctuations dominate variability in a circuit . These results produce a more accurate quantification of mutual information , and therefore promise to generate a more realistic assessment of signaling fidelity in cellular circuits . Our results and those from [20] support a view in which individual cells have distinct transfer functions over relevant signaling timescales , and have superior signaling fidelity ( > 1 . 5 bits ) than estimated from pooled measurements of a population . From this perspective , it could be the case that a diversity of high fidelity but different single cell signaling transfer functions across the population is a beneficial trait . However , some situations might arise where variability in population signal transmission capacity is not desirable . In this case , cells might use strategies such as negative feedback to constrain this variability . In either case , cells might also capitalize on the integration of signals from many pathways that respond to a given input ( s ) in order to generate a desired population response . In this view , each such pathway will add to the mutual information of the desired cellular output ( e . g . level or activity of a transcription factor ) , allowing the population to further circumvent in this way any information fidelity bottleneck . Researchers of the subject are likely to encounter both situations , and perhaps a revised form of population mutual information might be needed to quantify these effects , along with the formulation of new information theoretic metrics . As an example , for any given input x ( t ) , the mutual information I ( y 2 ; y 1 r * , t ∣ x ( t ) ) gives us a sense of the diversity ( or spread ) in responses in y2 given the cell-to-cell variability encoded in y 1 * . S4 Fig shows the results of this metric applied to the simple signaling cascade ( Fig 1b ) for different input step function amplitudes x+ and for different times . We envision these kind of metrics to reflect the different subpopulations with similar parameters within a given population and to serve as a potential tool to quantify how cell-to-cell variability across a population might change in structure due to various time-dependent inputs . Finally , most studies to date have focused on variability in populations of non-communicating cells . Information fidelity in cells that communicate , for example through quorum sensing for bacterial communities or cell-to-cell mechanical coupling for tissues , is still largely unstudied . How cell-to-cell communication modulates global variability and variability in initial conditions across a population , and hence mutual information of cellular pathways , is a topic that should be explored in order to determine whether and when multicellularity offers a beneficial strategy in terms of signaling fidelity .
Because we are using a finite number of experiments , the input distribution pu ( x+ ) is sampled with N discrete points . In practice , these points are spaced to accurately sample the input-output transfer function p ( y , t∣x+ ) for x+ ranging from 0 to Xmax . The time-dependent mutual information is then calculated with this data . For values of x+ between the sampled values , p ( y , t∣x+ ) is approximated by linearly interpolating the moments of the adjacent sampled distributions . Since the distributions generated by systems in this paper are approximately gaussian ( and approximately negative binomial at very low x+ for the synthetic circuit data ) , only the means and covariances are required . A larger number of experiments ( N ) generates a more accurate approximation of mutual information . However , we observed that convergence to accurate MI values does not increase monotonically with N for the logarithmic sampling of the doses response that we have adopted . Rather , convergence proceeds exponentially , followed by marginal gains in accuracy as N increases . Therefore , for every N , we examine the last three sample number values , N , N−1 and N−2 . Given their measured convergence rates , we can extrapolate an upper bound on the MI at an infinite number samples . We choose N whose calculated MI at N is within 1 percent of the extrapolated upper bound . The chemical equations for the circuit in Fig 1b are ∅ ⟶βy1* Y1* ⟶γy1*Y1* ∅ ( 2a ) X + Y 1 * ⟶ θ x X Y 1 * Y 1 ⟶ θ y 1 Y 1 Y 1 * ( 2b ) Y 1 ⟶ γ y 1 Y 1 ∅ ( 2c ) ∅ ⟶ f 1 ( y 1 ) M y 2 ⟶ γ m 2 M y 2 ∅ ( 2d ) ∅ ⟶ β y 2 M y 2 Y 2 ⟶ γ y 2 Y 2 ∅ ( 2e ) where f 1 ( y 1 ) = β m 2 * + β m 2 y 1 n 1 y 1 n 1 + y 1 0 n 1 . The propensities of the reactions appear above the reaction arrows . The system is a simple cascade of reactions where the input X activates Y1 , and subsequently the Y1-dependent transcription of Y2 . The parameter values are tabulated in Table 1 . Here the mean total number of Y1 molecules , active and inactive , is β y 1 * / γ y 1 * = 1500 . The max mean numbers for Y2 mRNA and Y2 protein are βm2/γm2 = 200 and β m 2 β y 2 γ m 2 β y 2 = 2000 , respectively . This system has only a single stationary solution . This allows us to approximate and efficiently calculate the master equation with a local affine assumption using the first two moments Eqs ( 6 ) and ( 7 ) taken from [24] . The formulation that we assume in our model and data consists of a system of well-stirred chemical reactions with N molecular species . For some environmental input X ( t ) , we define the pathway state Y ( t ) to denote the vector whose integer elements Yi ( t ) are the number of molecules of the ith species at time t . If there are M elementary chemical reactions that can occur among these N species , then we associate with each reaction rj ( j = 1 , … , M ) a non-negative propensity function defined such that aj ( Y ( t ) ) τ+o ( τ ) is the probability that reaction rj will happen in the next small time interval [t , t+τ] , as τ → 0 . The polynomial form of the propensities aj ( y ) may be derived from fundamental principles under certain assumptions [25] . The occurrence of a reaction rj leads to a change of νj ∈ ZN ( the set of nonnegative integers ) for the state Y . νj is therefore a stoichiometric vector that reflects the integer change in reactant species due to a reaction rj . This set of well-stirred chemical reactions can be represented by the joint probability density function P ( y , t∣ X ( t ) ) which describes the probability of the system being in state y at time t , given the environmental signal X ( t ) . The evolution of P ( y , t∣X ( t ) ) is given by ∂ P ( y , t | X ( t ) ) ∂ t = ∑ j = 1 M [ a j ( y - ν j ) P ( y - ν j , t | X ( t ) ) - a j ( y ) P ( y , t | X ( t ) ) ] ( 4 ) Eq ( 4 ) is the so-called chemical master equation ( CME ) [26 , 27] . To approximate the CME with moment equations , we approximate the propensity function aj ( y ) with a locally affine Taylor series expansion [24] about the mean of the distribution , z ( t ) , to get a j ( y ) ≈ a j ( z ( t ) ) + ∑ i = 1 N ∂ a j ( y ) ∂ y i | y = z ( t ) [ y i - z i ( t ) ] ( 5 ) From the time dependent mean equation for the kth species is ∂ z k ( t ) ∂ t = ∑ j = 1 K ν j k a j ( z ( t ) ) ( 6 ) and the time dependent covariance equation for the kth and k′th species is ∂ C k k ′ ( t ) ∂ t = ∑ j = 1 K ( ν j k ∑ i ∂ a j ( y ) ∂ y i | y = z ( t ) C i k ′ + ν j k ′ ∑ i ′ ∂ a j ( y ) ∂ y i ′ | y = z ( t ) C k i ′ + ν j k ν j k ′ a j ( z ( t ) ) ) ( 7 ) The calculation of the mutual information requires probability distributions . Given that we solve the first two moments , we constrain our distributions to be either a negative binomial distribution or a normal distribution . For cases when μ k < 3 ( C k k ) , we apply the negative binomial distribution since it only requires the first two moments and is non-negative . The negative binomial is very close to a normal distribution for μ k > 3 * ( C k k ) and we therefore apply the normal distribution in these regions . The value of 3 used is heuristic , but the tail of the normal distribution at negative values is negligible at this point . For linear transcriptional systems , the negative binomial is a natural steady state solution [28] which was our motivation for applying it . Importantly , our data never violated any constraints required by the negative binomial distribution , for example , μk ≤ Ckk . Note that the negative binomial distribution is only required for our modeling of the synthetic circuit data . Our theoretical example in the first half of the paper has large enough basal levels at zero input which always keeps it in the normal distribution regime . As a demonstration of the validity of the moment approach , S5 Fig shows very good agreement in the distributions derived from stochastic simulations ( SSA ) and the moment equations for the synthetic circuit model ( γg = 3 × 10−6 , initial condition S1 ) . Here we discuss how multi-variate MI measurements relates to MI measurements from particular initial conditions: We start with the distribution p ( ym , ys , t∣x ( t ) ) where ym are the dynamic cellular pathway/network signals , ys are the slowly fluctuating pathway component quantities relative to the timescale of a given experiment , and x ( t ) is the input signal ( s ) . The time dependent mutual information is I ( x ( t ) ; [ y m y s ] , t ) = ∑ x ( t ) ∑ y s ∑ y m p ( y m , y s , t | x ( t ) ) p ( x ( t ) ) log 2 p ( y m , y s , t | x ( t ) ) p ( y m , y s , t ) = ∑ x ( t ) ∑ y s ∑ y m p ( y m , t | y s , x ( t ) ) p ( y s , t | x ( t ) ) p ( x ( t ) ) log 2 p ( y m , t | y s , x ( t ) ) p ( y s , t | x ( t ) ) p ( y m , t | y s ) p ( y s , t ) ( 8 ) where the second line is simply a chain-rule representation . In addition to the assumption that the quantities of ys are fluctuating extremely slowly , we will also impose that the quantities in ys are independent of x ( t ) . This results in p ( ys , t∣x ( t ) ) = p ( ys , t ) ≈ p ( ys ) . The time dependent MI is approximated as I ( x ( t ) ; [ y m y s ] , t ) ≈ ∑ x ( t ) ∑ y s ∑ y m p ( y m , t | y s , x ( t ) ) p ( y s ) p ( x ( t ) ) log 2 p ( y m , t | y s , x ( t ) ) p ( y s ) p ( y m , t | y s ) p ( y s ) = ∑ y s p ( y s ) ∑ x ( t ) ∑ y m p ( y m , t | y s , x ( t ) ) p ( y s ) p ( x ( t ) ) log 2 p ( y m , t | y s , x ( t ) ) p ( y m , t | y s ) = ∑ y s p ( y s ) I ( x ( t ) ; y m , t | y s ) = E [ I ( x ( t ) ; y m , t | y s ) ] ( 9 ) Finally , we can examine the mutual information between ys and ym for a given input signal ( s ) x ( t ) using the formula I ( y m ; y s , t | x ( t ) ) = ∑ y s ∑ y m p ( y m , t | y s , x ( t ) ) p ( y s , t | x ( t ) ) log 2 p ( y m , t | y s , x ( t ) ) p ( y m , t | x ( t ) ) ( 10 )
|
This work demonstrates how different sources of variability within biochemical networks impact the interpretation of information transmission . These sources are the intrinsic noise generated within the pathway of a single cell , variability due to initial conditions and/or global parameters across the population . A theoretical analysis of a simple signaling pathway and experimental exploration of a synthetic circuit are used to discuss the contributions of these sources of variability to information transmission using mutual information as a metric .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
The Impact of Different Sources of Fluctuations on Mutual Information in Biochemical Networks
|
Pch2 is a widely conserved protein that is required in baker's yeast for the organization of meiotic chromosome axes into specific domains . We provide four lines of evidence suggesting that it regulates the formation and distribution of crossover events required to promote chromosome segregation at Meiosis I . First , pch2Δ mutants display wild-type crossover levels on a small ( III ) chromosome , but increased levels on larger ( VII , VIII , XV ) chromosomes . Second , pch2Δ mutants show defects in crossover interference . Third , crossovers observed in pch2Δ require both Msh4-Msh5 and Mms4-Mus81 functions . Lastly , the pch2Δ mutation decreases spore viability and disrupts crossover interference in spo11 hypomorph strains that have reduced levels of meiosis-induced double-strand breaks . Based on these and previous observations , we propose a model in which Pch2 functions at an early step in crossover control to ensure that every homolog pair receives an obligate crossover .
Meiosis generates haploid gametes from diploid progenitor cells . The reduction in ploidy results from the segregation of homologous chromosome pairs in the first meiotic division ( MI , [1] ) . Prior to MI , each chromosome is joined to its homolog at chiasmata , which serve to tether homologs to each other . This interaction promotes the tension between homologs needed to form a bipolar spindle that facilitates homolog segregation . Homologous chromosome pairs lacking chiasmata connections often fail to segregate properly at MI . Chromosome nondisjunction can also result if chiasmata are present , but not properly placed on chromosomes , or if sister chromatid cohesion is disrupted [2]–[5] . Regardless of the cause , chromosome missegregation produces aneuploid gametes that lead to infertility or conditions like Down syndrome in humans [6] . Chiasmata form at sites where programmed Spo11-catalyzed DNA double-stranded breaks ( DSBs ) , induced early in meiotic prophase , are repaired to form crossovers [1] . In baker's yeast , crossovers ( COs ) are formed via two main pathways . The first pathway , by which the majority of COs are made , involves Msh4-Msh5 and Mlh1-Mlh3 [7]–[15] . In this pathway , DSBs are processed and acted upon by strand exchange enzymes to form single-end invasion intermediates ( SEIs ) that are converted into double Holliday junctions ( dHJs ) . The latter are resolved into crossovers which display interference; the COs are more uniformly spaced than if placed at random ( see below; [16]–[22] ) . The COs formed via the second major pathway , which require Mms4-Mus81 , are not subject to CO interference [10] , [11] , [23] . Little is known about the intermediates that form in this latter pathway . The recombination steps that lead to CO formation occur in meiotic prophase . In leptotene , when meiotic DSB formation initiates recombination , an axial element containing Hop1 and Red1 proteins assembles along each pair of sister chromatids . In zygotene , when SEIs are detected , mature tripartite synaptonemal complex ( SC ) starts to form when the Zip1-containing central element connects the axial elements , which are now termed “lateral elements . ” Mature SC initiation begins at centromeres and later at CO-designated sites . These SC initiation events then spread outward until synapsis is completed in pachytene [24] , [25] . Hop1/Red1 and Zip1 are enriched in separate domains on the mature SC . This organization is Pch2-dependent because in pch2Δ mutants , Zip1 and Hop1 appear to be more uniformly distributed along the chromosome axes [26] , [27] . At the end of pachytene , recombination intermediates are resolved ( reviewed in [28] ) . In yeast , ∼40% of the ∼140–170 meiotic DSBs are repaired to generate noncrossover ( NCO ) products [29] , [30] . These NCO products are thought to form by a synthesis-dependent strand annealing mechanism ( SDSA , [31] ) , separate from the interfering CO mechanism , and do not result in MI disjunction-promoting chiasmata . Martini et al . [32] found that when meiotic programmed DSBs are decreased in spo11 hypomorphic strains , COs are favored at the expense of NCOs [24] . This CO homeostasis phenomenon may be an additional manifestation of CO interference [32] , [33] . The above studies indicate that DSBs are subject to a CO vs . NCO decision step , which is regulated by interference . Interference regulates this decision by ensuring that CO designation for a given DSB inhibits nearby DSBs from receiving this designation , thereby relegating them to a NCO fate . It is not clear whether non-interfering COs are formed through such a decision process; these COs are thought to form through a parallel pathway [10] , [23] . For this paper , the CO vs . NCO decision refers solely to COs that are subject to interference . The interference-regulated CO vs . NCO decision likely occurs very early in recombination , roughly at the time of SEI formation ( late leptotene-early zygotene ) and does not appear to be controlled by domains or sequences contained within the chromosome [17] , [18] , [20] , [21] , [34] , [35] . CO interference is strongest near a CO event and weakens with distance along the chromosome , although interference can act over large distances , up to ∼150 kb in yeast and ∼60 Mb in mice [30] , [33] , [36] , [37] . In addition , interference between COs appears stronger on longer chromosomes compared to shorter chromosomes [35] , [38]–[40] , but see [41] . However , smaller chromosomes have relatively high DSB density and may also have a higher density of non-interfering COs [39] , [41] , [42] . The mechanisms underlying interference regulation of the CO vs . NCO decision are unknown , despite the fact that numerous mutants showing defects in CO interference have been identified in baker's yeast . For at least a subset of these mutants , the defects in CO interference likely reflect problems in CO formation and not in the early CO vs . NCO decision ( reviewed in [28] ) . For example , mutants defective in either the SC central element protein Zip1 or the CO-promoting factor Msh4 have reduced CO levels and the remaining COs show reduced or no interference [7] , [33] , [43]–[45] . However , Zip2 foci , which mark the early CO designated sites , still display interference in zip1 and msh4 mutants [35] . This result , combined with the fact that NCOs form in zip1Δ and msh4Δ , suggests that interference regulation of the CO vs . NCO decision requires neither these factors nor the mature SC [20] , [22] , [35] , [46] . Rather , Zip1 and Msh4 are needed downstream of the decision to ensure CO formation [22] , [35] . It is likely that these results are applicable to other members of the ZMM ( Zip , Msh , Mer ) class of proteins . We examined the PCH2 gene for a role in interference regulation of the CO vs . NCO decision . PCH2 is a putative AAA-ATPase widely conserved in organisms that construct a synaptonemal complex in meiosis [26] , [47] . PCH2 was first identified in S . cerevisiae as a meiotic checkpoint factor due to the ability of pch2Δ to suppress the meiotic arrest of zip1Δ mutants [26] . This observation was extended by Wu and Burgess [47]; they proposed that Pch2 and Rad17 comprise separate branches of a checkpoint that ensures proper timing of the MI division , with the Pch2-dependent branch monitoring synaptonemal complex formation and the Rad17-dependent branch monitoring recombination events . Other checkpoint roles for Pch2 were reported in C . elegans , where it is required for apoptosis in response to unsynapsed pairing centers and in Drosophila , where it is required to delay meiotic progression in certain CO formation mutants [48] , [49] . Recent studies indicate that PCH2 is not solely a checkpoint factor; it is essential for proper meiotic axis organization and timely meiotic progression in baker's yeast , and complete DSB repair and fertility in mice [27] , [47] , [50] . Here we report that pch2 mutants display increased meiotic CO levels on larger chromosomes and are defective in CO interference . We also show that mutation of PCH2 reduces spore viability in spo11 hypomorphic strains . These data support an early role for Pch2 in DSB repair and a model in which Pch2-promoted meiotic axis organization controls CO levels and their distribution .
Previous work indicated that pch2Δ mutants show delays in meiotic DSB repair; thus , a time course comparison of DSB levels in meiotic prophase between pch2Δ and wild-type could be misleading [27] , [47] , [61] . Wu and Burgess [47] assayed DSB formation at the well-characterized HIS4LEU2 hotspot in wild-type and pch2Δ in a sae2Δ strain background where DSBs are formed but not resected or repaired . They reported that wild-type and pch2Δ strains displayed similar DSB levels . More recently , the Hochwagen group , using microarray analysis , observed increases in DSB formation in pch2Δ surrounding the rDNA on chromosome XII , but nowhere else in the genome ( A . Hochwagen personal communication ) . We assayed DSB formation in pch2Δ mutants at the YCR048W hotspot on chromosome III and near the centromere on chromosome XV [42] , [62] , [63] . These experiments were performed in a dmc1Δ background where DSBs are formed at wild-type levels and resected ( eventually hyperresected ) , but not repaired [29] , [42] , [64] . This approach allowed us to assay total DSB at loci other than HIS4LEU2 , where DSBs are thought to occur at saturating levels , and avoid the use of the sae2Δ background where maximal DSB levels may not be reached [29] , [32] , [42] , [65] . One concern with performing this analysis in the dmc1Δ background is that two reports [26] , [66] indicated that the checkpoint arrest seen in dmc1 mutants is bypassed in pch2 dmc1 strains; however , a more recent report [61] indicated that it is not . Our pch2Δ dmc1Δ mutants displayed a meiotic arrest as measured by a failure to form spores ( < 0 . 6% spore formation for pch2Δ dmc1Δ vs . ∼90% for wild-type at T = 24 hrs ) . However as shown below , we observed a significant bypass of the dmc1 arrest in pch2Δ spo11-HA dmc1Δ strains . Quantification of DSB levels in the dmc1Δ background is difficult due to the extensive resection of the breaks . We therefore analyzed five independent cultures of dmc1Δ and pch2Δ dmc1Δ strains . Similar to previous work ( [47]; A . Hochwagen personal communication ) , we saw no difference in DSB levels ( % of total DNA ) between dmc1Δ and pch2Δ dmc1Δ strains at the YCR048W ( 5 and 6 kb DSB bands; 19±6% for dmc1Δ , 18±5% for pch2Δ dmc1Δ ) and CEN15 ( 8 kb DSB band; 4 . 7±1 . 2% for dmc1Δ , 4 . 5±1 . 0% for pch2Δ dmc1Δ ) hotspots ( Figure 5A and 5B; T = 7 hrs in meiosis ) . It is important to note that Hochwagen et al . [61] reported that pch2Δ dmc1Δ mutants do not resect DSB ends as rapidly as dmc1Δ; however , such a difference in resection rate could only result in an overestimation of the level of DSBs in pch2Δ dmc1Δ . These data , together with previous work , suggest that the pch2 mutation does not disrupt DSB levels in a SPO11 background . As shown above , the pch2Δ mutation severely compromised the spore viability of spo11 hypomorph strains . Because some spo11 mutations confer semi-dominant and conditional phenotypes , as well as alter DSB patterns [60] , we assayed DSB levels at YCR048W in spo11-HA dmc1Δ strains in the presence or absence of the pch2Δ mutation ( Figure 5C ) . At T = 3 . 5 hrs in meiosis , similar DSB levels were observed in pch2Δ spo11-HA dmc1Δ ( 16% ) and spo11-HA dmc1Δ ( 15% ) strains . However , at T = 7 hrs , lower levels were observed in pch2Δ spo11-HA dmc1Δ ( 13±6 %; seven independent cultures ) compared to spo11-HA dmc1Δ ( 18±6%; seven independent cultures ) . In time courses performed side by side , pch2Δ spo11-HA dmc1Δ strains displayed 30 to 90% of the spo11-HA dmc1Δ levels at T = 7 hrs . Such variability was not observed in side-by-side experiments involving pch2Δ dmc1Δ and dmc1Δ strains . As shown below and analyzed in the Discussion , we attribute the variability in DSB levels to the bypass of the dmc1Δ arrest in pch2Δ spo11-HA dmc1Δ . This was determined by measuring the completion of the MI division in spo11-HA dmc1Δ and pch2Δ spo11-HA dmc1Δ strains . At T = 28 hrs in meiosis , only 1-2% of spo11-HA dmc1Δ strains completed MI; this indicates that the dmc1Δ arrest is maintained in these strains . For pch2Δ spo11-HA dmc1Δ , at T = 4 . 5 hrs , no cells ( n>200 ) had completed the MI division . However , at T = 6 . 5 hrs , 8 to 30% of the cells completed MI , and these values increased to 54 to 60% ( with similar spore formation levels ) at T = 28 hrs . As predicted for a dmc1Δ mutant , the spores produced by pch2Δ spo11-HA dmc1Δ were inviable .
Martinez-Perez [70] recently reported a link between meiotic axis protein organization and CO interference in C . elegans . They analyzed the distribution patterns of the central element protein SYP-1 and the axial element proteins HTP-1 and HTP-2 , which , like Hop1 , are HORMA domain proteins . Analogous to observations made for Hop1 and Zip1 in yeast , Martinez-Perez et al . [70] found that the HTP axial element and the SYP-1 central element proteins sort into reciprocal domains on late pachytene chromosome axes . Based on the above , the finding that HTP1/2 is depleted at COs , the fact that Spo11 and Msh5 are required for domain formation , and the correlation seen between HTP1/2 depletion sites and chiasmata , Martinez-Perez et al . [70] suggest that HTP/SYP-1 domain boundaries mark CO sites . This information suggests that Hop1/Zip1 boundaries indicate where the CO/NCO decision marks subsequent CO sites . Such a model takes into account the finding that C . elegans displays only one domain of each type whereas S . cerevisiae contains a large number of alternating Zip1/Hop1 domains . This pattern is consistent with the fact that each chromosome pair in C . elegans typically enjoys a single CO whereas chromosome pairs in S . cerevisiae enjoy multiple COs ( ∼80–90 total COs in S . cerevisiae [30] , [35] vs . six in C . elegans [21] , [70] ) . Based on observations presented in Martinez-Perez et al . [70] we suggest that the altered pattern of Hop1 and Zip1 localization on the chromosome axis seen in pch2Δ mutants results from , but is not the cause of , the increase in COs . In this interpretation , the defect in CO control in pch2Δ mutants leads to additional COs , reflected by a greater number of domains , thus making the axis distribution of Hop1 and Zip1 appear more uniform . This model fits with respect to the known timing of the CO vs . NCO decision [17] , [18] , [20] , [21] , [34] , [35] , and the finding that early Hop1 organization appears normal in pch2Δ mutants [27] . Testing such a model will require an examination of Hop1 and Zip1 localization patterns in strains ( e . g . pch2Δ spo11 hypomorphs ) containing decreased levels of DSBs; our model predicts that the Hop1 and Zip1 domains would become more distinct due to fewer COs , although not completely like wild-type due to defects in CO interference . The wild-type spore viability seen in pch2Δ mutants suggests that Pch2-mediated CO control is not required to maintain the viability of yeast grown in lab conditions . We offer two explanations for this finding: 1 ) COs are present in excess ( ∼80–90 per cell ) of the number needed for all homologs to receive an obligate CO ( 16 per cell ) . 2 ) The reduction in interference in pch2Δ is accompanied by , and likely causes , an increase in the overall number of COs . This increase in crossing over could compensate for distribution failures that jeopardized obligate CO formation ( [30] , [33]; Figure 3 , Figure 4; Table 4 , Table 5 , Table S2 ) . Our results and those of Martini et al . [32] demonstrate a buffered system in baker's yeast in which excess DSBs and COs lessen the need for interference to ensure obligate CO formation . Because of this buffer , obligate CO formation can be maintained if interference or DSBs are reduced , but not both ( Figure 3; [32] ) . Such buffering may exist because the consequences of having too many COs are less severe than too few . For example , pch2Δ mutants have dramatic increases in CO levels , but show wild-type spore viability , whereas mutants that significantly decrease CO levels like mlh3Δ , have reduced spore viabilities due to MI nondisjunction [9] , [11] . Future searches for mutants that disrupt the CO vs . NCO decision must be broadened to include genes with high spore viability or synthetic phenotypes with spo11 hypomorphs . Although the role of Pch2 in limiting CO levels , after the requisite number required for ensuring obligate CO formation is reached , is not required , it is likely to be advantageous . Too many COs , especially closely spaced ones , have been suggested to disrupt the sister chromatid cohesion required to create tension on the MI spindles and ensure proper homolog disjunction at MI [71] , [72] . In addition , our data suggests that the CO limiting role of Pch2 also promotes timely meiotic progression , which could also be advantageous to cells ( Figure 6 ) . What causes the loss in spore viability seen in pch2Δ spo11 hypomorphs ? pch2Δ/pch2Δ spo11-HA/spo11-HA strains displayed an excess of tetrads with 4 , 2 , and 0 viable spores , a high percentage of two-spore viable tetrads containing sisters , and an increased frequency of chromosome III nondisjunction . Our data are consistent with MI chromosome nondisjunction being a major component of the spore death phenotype , perhaps due to a failure to ensure obligate CO formation on all chromosomes . In such a model , when DSBs become limiting , the proper distribution of COs becomes even more critical to ensure obligate CO formation . Similar DSB levels were seen at YCRO48W at 3 . 5 hours in meiosis in spo11-HA dmc1Δ and pch2Δ spo11-HA dmc1Δ; however , by 7 hrs , fewer breaks were observed in the triple mutant ( Figure 5C and 5D ) . Our DSB level measurements are not definitive due to the checkpoint bypass observed in the triple mutant . We provide two explanations for the triple mutant phenotype . In one scenario , early forming DSBs appear at wild-type levels while later-forming DSBs form at lower levels that are insufficient for sustained recombination checkpoint activation . In a second scenario , DSBs form normally , but undergo some level of Dmc1-independent , possibly intersister , repair that permits a bypass of the checkpoint . Such repair would not lead to MI disjunction-promoting chiasmata . Both of these scenarios are sufficient to explain the spore inviability seen in pch2Δ spo11 hypomorphs ( Figure 3 ) . Future experiments to distinguish these hypotheses should include an analysis of meiotic Rad51foci in spo11-HA dmc1Δ and pch2Δ spo11-HA dmc1Δ strains [73] . We cannot rule out that other cellular defects contribute to the MI non-disjunction phenotype seen in pch2Δ spo11 mutants . For example , both pch2Δ and spo11 hypomorphs have SC defects , which could lead to CO control-independent synthetic phenotypes in the double mutants [24] , [27] . It is also possible that Pch2 promotes MI disjunction by regulating sister chromatid cohesion establishment and/or removal , or by preventing/resolving chromosome entanglements [2]–[4] , [68] , or that some spore death in pch2Δ spo11 hypomorphs is independent of MI non-disjunction . Interference mutants have been proposed to act downstream of the CO vs . NCO decision ( e . g . zip1 , msh4; Introduction; [22] , [35] ) , or display an apparent defect in interference due to an increase in non-interfering COs ( ndj1 , csm4; [15] , [35] ) . The only other yeast interference mutants that appear similar to pch2Δ are tid1Δ and dmc1Δ-2µRad54 [34; but see 58] . We will focus on tid1Δ , because its CO phenotype is better characterized . Tid1/Rdh54 is a member of the Swi2/Snf2 family , and thus may act in meiotic chromatin axis remodeling , though this has yet to be tested [74] . tid1Δ mutants display moderate levels of spore viability ( 58% 4-spore viable tetrads ) , and Tid1 has been shown to be involved in the strand exchange step of recombination [57] . Similar to pch2Δ , tid1Δ mutants display a defect in interference and increased gene conversion . Also , like pch2Δ , CO levels in tid1Δ appear similar to wild-type on a small chromosome ( III ) . On a medium-sized chromosome ( V ) , tid1Δ mutants displayed wild-type CO levels in two intervals , but a significant ( 2 . 4-fold ) increase in a third [34] . These data suggest that tid1Δ and pch2Δ have similar CO patterns . We are eager to test this hypothesis in the strain sets used in this study . Furthermore , we are intrigued by the idea that strand exchange and meiotic chromatin axis components are both required/involved in interference-regulation of the CO vs . NCO decision .
Yeast strains are listed in Table S1 . All strains were grown at 30°C on standard YPD ( yeast peptone dextrose; [75] ) . The sporulation media was described previously [11] , [68] . For tetrad genotyping , synthetic minimal selective media , synthetic complete media with 5 µM Cu , and YPD supplemented with complete amino acid mix and 3 mg/L cycloheximide were used [75] . When required , Geneticin ( Invitrogen ) , nourseothricin ( Hans-Knoll Institute fur Naturstoff-Forschung ) , and hygromycin B ( Calbiochem ) were added to YPD media as described [76] , [77] . The EAY1108/EAY1112 SK1 congenic strain set is described in Argueso et al . [11] , and the NHY942/NHY943 SK1 isogenic strain set is described in de los Santos et al . [10] . The spo11 hypomorphic mutants were described by Diaz et al . [60] and Henderson and Keeney [24] although the NHY942/NHY943 strains containing these alleles , which are used in this work , are described in Martini et al . [32] . As in Martini et al . [32] , we refer to spo11-HA3His6 as spo11-HA , spo11 ( D290A ) -HA3His6 as spo11da-HA , and spo11 ( Y135F ) -HA3His6 as spo11yf-HA . Strains EAY2562-2565 are derivatives of a cross between EAY2260 and SKY633 . The msh5Δ , mms4Δ , and dmc1Δ alleles used in this work were all complete open reading frame ( ORF ) deletions . The pch2Δ allele contains a deletion of amino acids 17–587 ( in the 603 amino acid ORF ) . All deleted regions were replaced with HPHMX4 , KANMX4 , or NATMX4 as shown in Table S1 [76] , [77] . The deletion cassettes were made via PCR and integrated into the genome using standard techniques [78] . Details on strain construction and primer sequences are available on request . Diploids for tetrad analysis were all made using the zero growth mating protocol [79] . The haploid parental strains were patched together on YPD for 4 hours and then spread on sporulation plates . The plates were incubated at 30°C for 2 days , after which tetrads were dissected . Tetrads from the EAY1108/EAY1112 strain background were dissected on synthetic complete media , whereas tetrads from the NHY942/NHY943 strain background were dissected on YPD media supplemented with complete amino acids . All tetrads were incubated 3–4 days at 30°C and then replica-plated to various selective media . The replica plates were scored after one day of incubation at 30°C . In the EAY strain background , the data for wild-type , mms4Δ , and msh5Δ were originally published in Argueso et al . [11] . In the NHY strain background , a subset of the wild-type data was originally published in Wanat et al . [68] . The distributions of each tetrad type were calculated using RANA software [11] . Genetic map distances±the standard error were calculated using the Stahl Laboratory Online Tools ( http://www . molbio . uoregon . edu/~fstahl/ ) which utilizes the formula of Perkins [54] . The G-test spreadsheet , available from The Online Handbook of Biological Statistics ( http://udel . edu/~mcdonald/statintro . html ) , was used to compare tetrad distribution patterns between strains . The Dunn-Sidak correction ( p value of 0 . 05/ number of comparisons ) was applied when multiple comparisons per data set were performed [80] . Recombination frequencies from spore data were calculated as described previously ( RANA software; [11] ) , with p-values determined as above ( http://udel . edu/~mcdonald/statintro . html ) . Three different analyses were performed to measure interference . The NPD ratio ( Table 4 ) was determined using the “Better Way” calculator ( http://www . molbio . uoregon . edu/~fstahl/ ) . This method compares the number of each tetrad type observed to the numbers expected if CO distribution was random and calculates a chi square value , which was converted to a p value using VassarStats ( http://faculty . vassar . edu/lowry/VassarStats . html ) . Coefficients of coincidence ( Table 5 ) were determined as described previously [11] , [68] . Tetrads were sorted using Mactetrad 6 . 9 software to calculate interference via the Malkova et al . method ( [37] , Figure 4; Table S2 ) . For all time courses , a saturated YPD overnight culture from each strain to be analyzed was diluted in 200 ml YPA ( 2% potassium acetate ) and grown for 17 hours . The YPA culture was then spun down , washed once in 1% potassium acetate and resuspended in 100 ml 1% potassium acetate ( similar to [81] ) . All strains were grown in the same batches of media and treated identically . DAPI staining to analyze progression past MI ( MI + MII ) was performed as described [81] . Cells were visualized using an Olympus BX60 microscope and at least 200 cells were counted for each time point . DNA was isolated from meiotic cultures as described [29] . Southern blotting was performed using standard techniques [82] . The percent of DSB formation for four to six independent time courses ( % of hybridizing bands±standard deviation , SD ) was calculated using Image Quant software .
|
During meiosis , cells that ultimately become gametes ( such as eggs or sperm ) undergo a single round of DNA replication followed by two consecutive divisions . In most organisms , the segregation of chromosomes at the first meiotic division is dependent upon genetic exchange , or crossing over , at homologous sites along chromosomes . Crossing over must therefore be regulated to ensure that every pair of matched chromosomes receives at least one crossover . Matched chromosomes that do not receive a crossover frequently undergo missegregation at the first meiotic division , yielding gametes that do not contain the normal chromosome number . Such missegregation events have been linked to human infertility syndromes . We used a genetic approach to study meiotic crossover control in baker's yeast . Our work suggests that Pch2 is required in crossover control during meiosis; mutants lacking Pch2 display altered crossover levels and distribution . Furthermore , pch2 mutations cause enhanced gamete inviability in strains that are mildly defective in initiating recombination . Based on these observations , we hypothesize that Pch2 acts early in crossover control , in steps that occur prior to those proposed for previously characterized crossover-promoting factors .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"cell",
"biology/nuclear",
"structure",
"and",
"function",
"molecular",
"biology/recombination",
"genetics",
"and",
"genomics/nuclear",
"structure",
"and",
"function",
"molecular",
"biology/chromosome",
"structure",
"genetics",
"and",
"genomics/chromosome",
"biology",
"molecular",
"biology/chromatin",
"structure",
"molecular",
"biology/dna",
"repair"
] |
2009
|
The pch2Δ Mutation in Baker's Yeast Alters Meiotic Crossover Levels and Confers a Defect in Crossover Interference
|
The generation of distinct neuronal subtypes at different axial levels relies upon both anteroposterior and temporal cues . However , the integration between these cues is poorly understood . In the Drosophila central nervous system , the segmentally repeated neuroblast 5–6 generates a unique group of neurons , the Apterous ( Ap ) cluster , only in thoracic segments . Recent studies have identified elaborate genetic pathways acting to control the generation of these neurons . These insights , combined with novel markers , provide a unique opportunity for addressing how anteroposterior and temporal cues are integrated to generate segment-specific neuronal subtypes . We find that Pbx/Meis , Hox , and temporal genes act in three different ways . Posteriorly , Pbx/Meis and posterior Hox genes block lineage progression within an early temporal window , by triggering cell cycle exit . Because Ap neurons are generated late in the thoracic 5–6 lineage , this prevents generation of Ap cluster cells in the abdomen . Thoracically , Pbx/Meis and anterior Hox genes integrate with late temporal genes to specify Ap clusters , via activation of a specific feed-forward loop . In brain segments , “Ap cluster cells” are present but lack both proper Hox and temporal coding . Only by simultaneously altering Hox and temporal gene activity in all segments can Ap clusters be generated throughout the neuroaxis . This study provides the first detailed analysis , to our knowledge , of an identified neuroblast lineage along the entire neuroaxis , and confirms the concept that lineal homologs of truncal neuroblasts exist throughout the developing brain . We furthermore provide the first insight into how Hox/Pbx/Meis anteroposterior and temporal cues are integrated within a defined lineage , to specify unique neuronal identities only in thoracic segments . This study reveals a surprisingly restricted , yet multifaceted , function of both anteroposterior and temporal cues with respect to lineage control and cell fate specification .
The generation of distinct neuronal cell types at different axial levels represents a crucial feature of nervous system development . This segment-specific neuronal subtype specification relies upon both anteroposterior and temporal cues , and significant progress has been made in understanding each of these two processes . Along the anteroposterior axis , a number of studies have revealed that the Hox homeotic genes play key roles , acting in several different ways to control segment-specific nervous system development ( reviewed in [1]–[6] ) . In particular , studies of mammalian motoneuron development have revealed critical input from Hox genes , acting in surprisingly restricted ways to specify unique motoneuron identities [7]–[9] . Hox genes often act in tight interplay with Hox cofactors of the Pbx and Meis families [10] , [11] , and although less studied in the nervous system , these factors have also been found to play important roles during segment-specific cell fate determination [9] . However , our understanding of Hox/Pbx/Meis function in the nervous system is still rudimentary , in particular with respect to how these cues are integrated with lineage progression and with respect to their specific targets in the different settings . Along the temporal “axis” studies have revealed that neural progenitor cells undergo stereotypic temporal transitions in competence , which result in the generation of distinct cell types at different time points ( reviewed in [12] ) . In Drosophila , a well-defined cascade of transcription factors , the temporal gene cascade of hunchback-Kruppel-Pdm-castor-grainyhead , is expressed in sequential fashion by most central nervous system ( CNS ) progenitors ( neuroblasts ) , and control distinct “competence windows” in neuroblasts ( reviewed in [13] ) . Despite progress with respect to anteroposterior control of nervous system development on the one hand , and to temporal changes in neuroblasts on the other , little is known regarding how these two fundamental developmental axes are integrated to establish distinct neuronal cell types at different axial levels . The developing Drosophila CNS is generated from a stereotyped set of some 1 , 000 neuroblasts ( reviewed in [14] ) . They are organized into 18 segments: three brain segments ( B1–B3 ) , three subesophageal segments ( S1–S3 ) , three thoracic segments ( T1–T3 ) , and nine abdominal segments ( A1–A9 ) ( Figure 1A and 1B ) . These segments are typically referred to as the brain ( B1–B3 through S1–S3 ) and the ventral nerve cord ( VNC; T1–T3 through A1–A9 ) ( Figure 1B ) . Neuroblasts undergo series of asymmetric cell divisions , “budding” off secondary progenitor cells denoted ganglion mother cells ( GMCs ) , that in turn typically divide one final time to generate neurons and/or glia [15]–[18] . Each neuroblast has a unique and stereotypic identity , as revealed by the size of its lineage—ranging from two to 40—and by the types of neurons and glia generated [19]–[21] . Each thoracic and abdominal hemisegment contains 30 neuroblasts that delaminate from the ectoderm in seven distinct rows [22] . In each of the six thoracic hemisegments , the lateral-most thoracic row 5 neuroblast , NB 5-6T , generates a unique lateral cluster of four neurons—the Ap cluster—that specifically expresses the LIM-HD transcription factor Apterous ( Ap ) and the Eyes absent ( Eya ) cofactor ( Figure 1A–1D ) [23] , [24] . Two of the Ap cluster neurons can be further identified by the specific expression of two neuropeptides—FMRFamide ( FMRFa ) and Nplp1 ( Figure 1D ) [25]–[27]—and the four Ap cluster neurons thus represent at least three distinct cell types: Ap1/Nplp1 , Ap2/3 ( ipsilaterally projecting interneurons ) , and Ap4/FMRFa ( Figure 1D ) . Studies have identified several genes acting to ensure proper Ap cluster specification and to activate the cell-specific expression of Nplp1 and FMRFa [23]–[34] . Moreover , to better understand the genetic mechanisms of Ap cluster specification , we recently resolved the entire NB 5-6T lineage , finding that Ap neurons are born at the end of this large lineage . We furthermore identified the temporal transitions that control generation of the three distinct Ap cluster neuronal cell types at the end of this lineage [35] . These studies revealed critical input from the two late temporal genes castor ( cas ) and grainyhead ( grh ) . cas plays multiple roles to specify Ap neurons , one of which is to trigger a critical feed-forward loop involving the COE/Ebf family member collier/knot ( col ) [25] . In contrast , grh acts selectively to specify the Ap4/FMRFa neuron . Several combined elements presented us with a unique opportunity for addressing how an identifiable neural lineage is modified along the entire anteroposterior axis to generate segment-specific cell types , including; 1 ) the development of an NB 5–6–specific reporter and Gal4 “driver” 2 ) the characterization of the NB 5-6T lineage , 3 ) the identification of a unique thoracic-specific group of cells generated by this lineage ( the Ap cluster ) , 4 ) the highly restricted expression of the FMRFa and Nplp1 neuropeptides within two of the four Ap cluster neurons , and 5 ) the elucidation of an elaborate progenitor and postmitotic genetic pathway specifying the Ap cluster neurons . We find that Ap cluster neurons exclusively appear in thoracic segments as a result of several distinct mechanisms , acting at the different axial levels . In the abdomen , the Hox genes of the bithorax complex ( Bx-C ) —Ultrabithorax ( Ubx ) , abdominal-A ( abd-A ) , and Abdominal-B ( Abd-B ) —act with the Pbx/Meis Hox cofactors encoded by homothorax ( hth ) and extradenticle ( exd ) genes , to terminate progression of the NB 5-6A lineage , via neuroblast cell cycle exit . This occurs within an early ( Pdm ) temporal window , thereby preventing the actual generation of Ap cluster neurons , as well as the progression into late temporal windows specified by cas and grh . In the thorax , the thoracic Hox gene Antennapedia ( Antp ) acts with hth and exd to specify Ap cluster neurons within NB 5-6T . Of the many possible ways in which Pbx/Meis and Hox input could control this event , we find that Antp , hth , and exd integrate with the temporal gene cas to specifically activate col and the col-mediated critical feed-forward loop . Intriguingly , we find that the actual levels of Hth expression acts in an instructive manner , acting at low levels to trigger neuroblast cell cycle exit in NB 5-6A , and at high levels to trigger col expression in NB 5-6T . In more anterior segments , equivalents of “Ap cluster cells” are generated , but fail to differentiate into Ap cluster neurons , not only due to the absence of Antp expression , but also due to absent or low-level expression of the temporal factor Grh , which is critical for specifying the Ap4/FMRFa cell fate . Co-misexpression of Antp with grh specifies Ap cluster neurons , with expression of the neuropeptides Nplp1 and FMRFa in anterior brain segments . By co-misexpressing Antp and grh in a Bx-C triple mutant background ( Ubx , abd-A , Abd-B ) , a “thoracic CNS” is generated with Ap clusters emerging throughout the neuroaxis . In summary , the dynamic and restricted expression of Hox , Pbx/Meis , and temporal genes , coupled with their unique functions , act to modify an equivalent CNS lineage along the neuroaxis by three different mechanisms: 1 ) abdominal lineage size control , 2 ) thoracic integration upon a specific feed-forward loop , and 3 ) the anterior absence of proper Hox and temporal expression .
In Drosophila , the control of segment identity is in part controlled by the homeotic ( Hox ) genes and the Pbx/Meis Hox cofactors , encoded by the homothorax ( hth ) and extradenticle ( exd ) genes [11] , [41] . Mutations in these genes strongly affect both the abdominal and thoracic NB 5–6 lineage ( see below ) . We thus mapped the expression of the relevant Hox factors; Antennapedia ( Antp ) , the bithorax Hox complex ( Bx-C ) factors , Ultrabithorax ( Ubx ) , Abdominal-A ( Abd-A ) , and Abdominal-B ( Abd-B ) , as well as Hth and Exd , in the NB 5–6 lineage ( Figure S1 and S2 ) . We find that expression of Hth and Exd commences in NB 5–6 at stage 11 in abdominal and thoracic lineages , is found in all cells within the lineages at stage 13 , and is maintained throughout the lineages during subsequent stages ( Figure S1; Figure 2B ) . Both genes are also expressed by more anterior NB 5–6 lineages ( unpublished data ) . Hth is expressed at low levels initially , but increases rapidly at stage 13 , in thoracic and anterior segments in general [42] , [43] ( unpublished data ) , as well as in thoracic and more anterior NB 5–6 lineages specifically ( Figure 3 ) . Antp is expressed in a gradient in the VNC , high anterior and low posterior , with the anterior limit in T1 ( Figure S2A–S2C ) [44] . In both NB 5-6A and NB 5-6T , Antp expression commences at stage 12 , and is maintained in all cells born after this stage . Ubx expression commences within NB 5-6A at stage 11 , and is subsequently expressed in earlier-born cells in this lineage in segments A1 to A7 . Abd-A and Abd-B are expressed similarly to Ubx , with Abd-A in segments A2 to A9 , and Abd-B in segments A5 to A9 ( Figures S1 and S2; Figure 2A and 2B ) . Thus , Bx-C gene expression fits with a potentially suppressive role on Ap cluster formation , Antp expression with a potentially positive role , and hth/exd expression with dual roles . We recently mapped the complete outline of the NB 5-6T lineage [35] . These studies revealed that the four Ap cluster neurons are the last-born cells within the NB 5-6T lineage , and that they are born within a Cas/Grh late temporal window ( Figure 2B ) . We conducted a similar analysis of the NB 5-6A lineage ( Figure S3 ) . We find that NB 5-6A stops dividing at stage 12 , within an earlier temporal window specified by Pdm , and thus ends up generating a smaller lineage when compared to the NB 5-6T ( Figure 2B ) . These findings are in line with previous studies of the 5–6 lineage [19] , [20] . As anticipated from these findings , there is no expression evident of the critical Ap cluster determinant Col ( see below ) . We find apoptosis of four to five cells within the NB 5-6A lineage , but are unable to identify cleaved Caspase 3 staining unequivocally in the neuroblast ( Figure S3L and S3M ) . Thus , the truncation of the NB 5-6A lineage could either result from an earlier cell cycle exit in the neuroblast , or from neuroblast apoptosis . To distinguish between these two possibilities , we analyzed NB 5-6A lineage progression in the H99 deletion , a mutation that removes the three critical RHG-domain cell death genes reaper , head involution defective , and grim , and is well established not to display any embryonic apoptosis [45] . In H99 mutants , Ap clusters do not appear in abdominal segments ( Figure 4C and 4D ) . As anticipated from the apoptosis of four to five cells within the wild-type NB 5-6A lineage ( Figure S3L and S3M ) , we find that four to five additional cells are present in H99 ( Figures 4A , 4B , and 5E ) . However , we do not find evidence of additional rounds of mitosis past stage 13 ( Figure S3F and S3G; n = 12 hemisegments ) . In addition , the four to five additional cells observed in H99 are observed already at stage 13 ( Figure 4H ) . In the wild type , the neuroblast cannot be identified using Dpn ( n = 14 hemisegments ) , but in contrast in H99 , we are able to identify a ventral Dpn-positive cell ( Figure 4F and 4G; nine out of 11 hemisegments ) . These results demonstrate that NB 5-6A generates a truncated lineage , when compared to NB 5-6T , not due to apoptosis of the neuroblast , but rather due to an earlier cell cycle exit , within the Pdm window , followed thereafter by apoptosis . Thus , the lack of Ap clusters in abdominal segments represents the logical outcome of a truncated NB 5–6 lineage , since it never generates Ap cluster cells and never progresses into the late competence window specified by the Cas and Grh temporal factors , both of which are critical for Ap cluster specification . The NB 5-6A lineage is smaller in size when compared to NB 5-6T . The Bx-C Hox genes are expressed at the proper time and place to be involved in this lineage truncation ( Figure 2 ) . Indeed , we find that mutations in Bx-C lead to the appearance of bona fide Ap clusters in more posterior regions , with the anticipated complexity due to their overlapping segmental expression levels and functions ( Figure 5 ) . Focusing on Ubx and the A1 segment , we utilized the lbe ( K ) -Gal4 marker to address cell numbers in the NB 5-6A lineage , and found that the lineage contains a larger number of cells—equivalent in size to NB 5-6T ( Figure 6A , 6B , and 6E ) . The temporal gene cas and the Ap cluster determinant col are both expressed at the end of the NB 5-6T lineage , but are not normally expressed in the smaller NB 5-6A lineage ( Figure 6A; Figure 2 ) . As anticipated from the larger NB 5-6A lineage observed in Ubx mutants , we also find ectopic expression of Cas and Col ( Figure 6A , 6B , 6E , and 6F ) . Conversely , we find that when we misexpress Ubx early in NB 5-6T lineage , using the lbe ( K ) -Gal4 driver—a driver that will ensure strong Ubx expression specifically in NB 5–6 already at stage 11 ( Figure 1A ) —Ubx is sufficient to suppress thoracic lineage progression , resulting in an abdominal-sized lineage , and loss of Cas and Col expression ( Figure S4A–S4C ) . Similar results were obtained misexpressing abd-A ( unpublished data ) . In contrast , late postmitotic misexpression of Ubx in Ap cluster neurons , driven from the apGal4 driver , revealed no effect upon Ap cluster specification ( Figure S4D–S4F ) . Thus , Bx-C genes are necessary and sufficient to terminate the NB 5–6 lineage within the Pdm temporal window , and serve this function rapidly after onset of their expression . In Drosophila , the thoracic segments , and in particular T2 , have sometimes been viewed as a “ground state” of development , i . e . , in the absence of all Hox gene input , abdominal and thoracic segments develop into a rudimentary T2 segment [46] . On this note , it was interesting to address how the NB 5-6T lineage would develop in an Antp mutant . To our surprise , we found a complete absence of Ap clusters in Antp mutants , as evident by the complete loss of the determinants Col , aplacZ , Eya , Dac , and Dimm , as well as terminal identity markers: the neuropeptides Nplp1 and FMRFa ( Figure S5 ) . However , the lbe ( K ) -lacZ marker revealed that the lineage still progressed , and the two Ap neuron determinants squeeze ( sqz ) and Nab were not down-regulated ( Figure S5I , S5J , S5Q , S5R , S5X , and S5Z ) . Hox genes often act genetically and physically with the two Hox cofactors Hth and Exd [11] , [41] , and we therefore anticipated similar effects in these two mutants when compared to Antp . Indeed , we find that both hth and exd mutants fail to properly specify Ap neurons , as evident by the complete loss of Nplp1 , FMRFa , aplacZ , Eya , and Dimm , as well as the partial loss of Dac and Col expression ( Figure S5 ) . Similar to Antp mutants , the lbe ( K ) -lacZ marker revealed that the lineage still progressed , and sqz and Nab were not down-regulated in hth or exd mutants ( Figure S5K , S5L , S5S , S5T , S5X , and S5Z ) . The loss of the key Ap neuron determinant Col in Antp , hth , and exd mutants prompted us to ask the question of whether or not the primary function of Antp , hth , and exd may be to activate col . If so , it should be possible to rescue Antp , hth , and exd with col . This experiment was not technically feasible for exd , due to its maternal contribution , but was conducted for Antp and hth . For hth , we indeed find a restricted role , and cross-rescue of hth with col restores Ap clusters ( Figure 7A , 7C–7E , and 7H ) . In contrast , Antp is not rescued by col ( Figure 7A , 7B , and 7F–7H ) . We thus find that Antp , hth , and exd play some common roles during NB 5-6T development , such as the regulation of col . But whereas hth can be cross-rescued by col , Antp cannot . The failure of col to rescue Antp suggests that Antp plays additional roles during Ap neuron specification . The temporal gene cas plays a key role in regulating many Ap neuron determinants , including col . However , the complete loss of Ap cluster determinants in cas mutants can largely be cross-rescued by re-expression of col [35] . But Cas is expressed already at stage early 11 , and generates 6 cells prior to activating the Ap window . Moreover , whereas the initiation of the Ap window coincides with Grh expression ( Figure 2B ) , grh mutants still generate Ap clusters with normal Ap1/Nplp1 and Ap2/3 neurons [35] . This indicates the existence of an unknown critical cue , acting within the cas window to trigger the Ap window , i . e . , activating col . In hth mutants , there is a failure of Ap neuron specification , evident from the loss or reduction of Col , aplacZ , Eya , Dimm , Nplp1 , and FMRFa expression ( Figure S5 ) . However , similar to cas , Ap clusters can be rescued simply by re-expressing col , indicating that the primary role of hth is to activate col . Is hth then the critical trigger , acting in the large Cas window to trigger the Ap window by activating col ? Our expression analysis argued against this idea , since we found that Hth is indeed present in the NB 5-6T lineage already at stage 11 ( Figure 2B ) , two stages prior to onset of Col expression . However , the answer seems to lie in the actual levels of Hth—as mentioned earlier , we noticed that its expression was weak at stage 11 , with a sharp increase at stages 12–13 , preceding the onset of col expression ( Figure 3 ) . To test whether increasing levels of Hth is sufficient to trigger the Ap window , we overexpressed hth using the lbe ( K ) -Gal4 driver—a driver that will ensure strong hth expression specifically in NB 5–6 , already at stage 11 ( Figure 1A ) . Strikingly , hth overexpression triggered premature Col expression in the NB 5-6T lineage ( Figure 8A , 8B , and 8E ) . As anticipated from this effect , we also noticed a robust increase in the number of Ap neurons specified within the NB 5-6T lineage , as evident by ectopic expression of Eya , Nplp1 and FMRFa ( Figure 8C , 8D , and 8F ) . Quantification of NB 5-6T cell numbers , at stage 14 and 18 h after egg laying ( AEL ) , revealed no increase in cell numbers , either at stage 14—showing that neither GMCs nor Ap neurons are dividing erroneously—or at late stages—showing that the neuroblast is not continuing to divide past stage 15 ( Figure 8E and 8F ) . Thus , hth overexpression results in Ap neuron specification of cells born in the early Cas window , but does not trigger extra cell divisions in any part of the lineage . Within the large Cas window , a switch from low- to high-level Hth appears to function as a critical temporal switch , which together with Cas , Antp , and Exd , acts to trigger Col expression . Col in turn specifies Ap neurons by activating a critical feed-forward loop [25] . Thus , within this particular CNS lineage , one critical integration point between anteroposterior and temporal cues is the activation of the COE/Ebf regulator col and its feed-forward loop . Are the roles of the Bx-C genes in lineage termination in the NB 5-6A lineage also dependent upon the Pbx/Meis factors ? That appears to be the case: similar to Ubx , both hth and exd mutants display an increase in NB 5-6A lineage cell numbers , approaching those normally found in NB 5-6T , as well as ectopic expression of Cas and Col ( Figure S6C–S6F ) . But why do the Bx-C and Pbx/Meis mutants display such different phenotypes—when assayed using late Ap neuron markers , Bx-C mutants display striking homeotic transformations , with bona fide Ap clusters generated throughout the VNC ( Figure 5 ) . In contrast , hth and exd mutants display a complete loss of Ap cluster specification ( Figure S5 ) . The answer to this paradox comes from the dual role of hth and exd outlined above—these genes not only control lineage termination of the NB 5-6A lineage , but also specify Ap neurons in the larger NB 5-6T lineage . Thus , we reasoned that in hth and exd mutants , “Ap cluster cells” are likely present in abdominal segments , but are not properly specified into Ap cluster neurons due to the second and later role of hth and exd . To reveal this dual role , we focused on hth and attempted to rescue hth with itself , but at different stages of NB 5–6 lineage progression . Specifically , because NB 5-6A exits the cell cycle at stage 12 , we sought to reintroduce hth expression before versus after this exit point . To this end , we used the stage 11 driver lbe ( K ) -Gal4 versus the stage 12 driver elav-Gal4 ( Figure 1A; Figure S6 ) . The prediction was that if hth was rescued by itself at a later stage , then posterior , ectopic Ap cluster cells would be triggered to differentiate into bona fide Ap cluster neurons , and this rescue would therefore phenocopy Bx-C mutants . This is indeed what we find: late rescue of hth not only restores Ap clusters in thoracic segments , but results in ectopic Ap clusters also in the majority of abdominal segments ( Figure 9A–9C and 9E ) . In contrast , if hth was rescued earlier in the lineage , prior to the neuroblast cell cycle exit point , hth would be able to play both its early role—blocking NB 5-6A cell cycle in the abdomen—and its late role—specifying Ap neurons in the thorax . Thus , we predicted that early rescue would reveal a more complete rescue of hth , with Ap clusters only in the thoracic segments . As anticipated , this is what we find , as evident from robust rescue of Ap clusters in thoracic segments , but with reduced prevalence of ectopic abdominal clusters ( Figure 9D and 9E ) . Thus , low-level Hth is essential in NB 5-6A prior to stage 12 to ensure cell cycle exit . If this critical stop-point is bypassed , subsequent reintroduction of Hth is not able to halt the lineage progression at any later point , but will , however , allow Hth to act in its late cell specification role , i . e . , activating col and thereby specifying generic Ap neurons . This dual role of hth—acting early with Bx-C genes in abdominal segments to restrict lineage size , and with Antp in thoracic segments to specify Ap cluster neurons—is revealed by Gal4/UAS-mediated rescue at different stages of NB 5–6 lineage progression . How is the NB 5–6 lineage modified in more anterior segments ? Analyzing anterior NB 5–6 lineages , we found a significant degree of variation with respect to lineage size . However , expression of the temporal factor Cas was observed in all segments ( Figure S7 ) . Cas is expressed at the end of the thoracic NB 5–6 lineage ( Figure 2B ) , and plays a critical role to activate Ap neuron determinants [35] . Cas is not expressed in the abdominal lineage , since the abdominal NB 5–6 lineage terminates just prior of progression into the Cas window ( Figure 2B ) . However , the presence of NB 5–6 lineages anteriorly , containing a Cas window , suggested that Ap cluster neuron equivalents may indeed be present in anterior segments . Antp plays a critical role for Ap neuron specification , but its expression stops at the T1 segment ( Figure 2A ) . Therefore , we postulated that anterior misexpression of Antp may be sufficient to specify ectopic anterior Ap clusters . We confirmed this notion , as evident by the appearance of aplacZ , Eya , and Nplp1 expression in anterior segments ( Figures 10A , 10B , and 11E ) . To verify that these ectopic Ap clusters indeed were generated from anterior NB 5–6 equivalents , we utilized the NB 5–6–specific driver lbe ( K ) -Gal4 to misexpress Antp , and could again identify ectopic anterior Ap clusters ( Figure S8 ) . In the anterior-most segments , B1 and B2 , there is added complexity due to more extensive expression of both aplacZ and Eya already in the wild type , and the presence of aplacZ/Eya coexpressing cells ( Figures 10A and 11E ) . However , these cells do not coexpress Nplp1 , nor do they stem from anterior NB 5–6 lineages ( Figures 10A; unpublished data ) . Thus , we were able to identify ectopic Ap clusters in the B2 segment ( Figure 10B ) . However , we found no ectopic Ap clusters in the B1 segment ( Figure 10B ) . Our analysis of the function of Antp and col in the NB 5-6T , demonstrated that while Antp plays a key role in activating col , Antp also plays additional roles to specify Ap cluster neurons ( Figure 7 ) . In line with these findings , we find that whereas misexpression of col is able to act in the non-Antp domain to trigger ectopic anterior expression of both aplacZ and Eya , it is not able to trigger formation of bona fide anterior Ap clusters , as evident by the failure to activate Nplp1 and FMRFa ( Figures 10C and 11E; unpublished data ) . Similarly , Antp is unable to trigger ectopic , anterior Ap clusters in a col mutant background ( Figure 10D ) . Thus , the regulatory interplay observed between Antp and col in the NB 5-6T lineage is recapitulated anteriorly , in the ectopic setting . We found no evidence of regulatory interplay between Antp , hth , exd , and the late temporal genes cas and grh ( Figure S9 ) . In line with this notion , we do not find any evidence of a complete homeotic transformation by Antp of anterior NB 5–6 lineages , since the total number of cells in the lineage , as well as the number of Cas and Grh cells , are unaffected by Antp misexpression ( Figure S10 ) . Thus , our results support the notion that these effects of Antp misexpression occur postmitotically and are not due to complete homeotic transformation of anterior NB 5–6 lineages . Misexpression of Antp triggers ectopic Ap clusters in anterior NB 5–6 lineages , with the expression of the Nplp1 neuropeptide . However , we failed to detect FMRFa neuropeptide in these ectopic Ap clusters ( Figure 11A , 11B , and 11E ) . Ap neurons are generated at the end of the NB 5-6T lineage , in a temporal window that in addition to cas , also expresses the grh temporal gene , i . e . , in a Cas/Grh coexpressing window ( Figure 2B ) . Whereas cas plays a global role at the end of the NB 5-6T lineage , regulating most Ap neuron determinants , grh plays a more selective role , and at high levels , is necessary and sufficient to specify the last-born cell , the FMRFa neuron [35] . The failure of Antp to trigger FMRFa expression in the anterior ectopic Ap clusters prompted us to examine the expression of Grh in anterior NB 5–6 lineages . This analysis revealed that there is indeed weak or no expression of Grh in anterior NB 5–6 lineages ( Figure S7 ) . We next tested whether or not ectopic expression of grh alone could trigger ectopic anterior Ap cluster neurons , with Nplp1 and FMRFa expression . However , given the lack of Antp expression in anterior segments , we were not surprised to find that grh misexpression did not to trigger ectopic Ap clusters ( Figure 11C and 11E ) . Therefore , we postulated that by co-misexpressing Antp with grh , we should be able to trigger the appearance of ectopic Ap clusters with a more complete identity , i . e . , with expression not only of Nplp1 , but also of FMRFa . This is indeed what we find ( Figure 11D and 11E ) . To verify that these ectopic Ap clusters indeed were generated from anterior NB 5–6 equivalents , we utilized the NB 5–6–specific driver lbe ( K ) -Gal4 to misexpress Antp and grh , and could again identify ectopic anterior Ap clusters ( Figure S8 ) . However , we were again unable to trigger Ap clusters in the B1 segment ( unpublished data ) . Whereas many of the posterior Drosophila CNS segments , such as A2–A7 , are generally viewed as identical , repetitive units , all brain segments ( B1–B3 through S1–S3 ) are considered unique [47] . Our lineage analysis of anterior wild-type NB 5–6 lineages confirmed this notion , revealing that both lineage size , as well as the expression of Cas and Grh , is different between segments ( Figure S10 ) . Intriguingly , we also find that the effects of Antp misexpression , as well as Antp/grh co-misexpression , resulted in different types of ectopic Ap clusters in different brain segments , with reproducibly distinct numbers of Eya , Nplp1 , and FMRFa neurons ( Figure 7F ) . These findings suggest that Antp/grh co-misexpression is not able to override all aspects of segment specificity within anterior NB 5–6 lineages . In Bx-C mutants , we find homeotic transformation of abdominal segments into a thoracic identity , with ectopic Ap clusters in each segment . When we co-misexpress Antp and grh , we find ectopic Ap clusters in anterior segments . We reasoned that by performing both of these genetic manipulations simultaneously , we would be able to trigger formation of a “thoracic CNS , ” i . e . , a CNS containing Ap clusters along the entire neuroaxis . This is indeed what we found: co-misexpression of Antp and grh , in a Ubx , abd-A , Abd-B triple mutant background , resulted in ectopic Ap clusters along the neuroaxis , evident by expression of Eya , Nplp1 , and FMRFa in all segments ( Figure 11G and 11H ) . Again , the anterior-most segment , B1 , did not display ectopic Ap clusters .
In the developing Drosophila CNS , each abdominal and thoracic hemisegment contains an identifiable set of 30 neuroblasts , which divide asymmetrically in a stem-cell fashion to generate distinct lineages . However , they generate differently sized lineages—from two to 40 cells [19] , [20]—indicating the existence of elaborate and precise mechanisms for controlling lineage progression . Moreover , about one third of these lineages show reproducible anteroposterior differences in size , typically being smaller in abdominal segments when compared to thoracic segments [19]–[21] , [48] . Thus , neuroblast-specific lineage size control mechanisms are often modified along the anteroposterior axis . Previous studies have shown that Hox input plays a key role in modulating segment-specific behaviors of neuroblast lineages [49] . Recent studies have resulted in mechanistic insight into these events . For instance , in the embryonic CNS , Bx-C acts to modify the NB 6-4 lineage , preventing formation of thoracic-specific neurons in the abdominal segments . This is controlled , at least in part , by Bx-C genes suppressing the expression of the Cyclin E cell cycle gene in NB 6-4a [50]–[52] . Detailed studies of another neuroblast , NB 7-3 , revealed that cell death played an important role in controlling lineage size in this lineage: when cell death is genetically blocked , lineage size increased from four up to 10 cells [53] , [54] . Similarly , in postembryonic neuroblasts , both of these mechanisms have been identified . In one class of neuroblasts , denoted type I , an important final step involves nuclear accumulation of the Prospero regulator [55] , a key regulator both of cell cycle and differentiation genes [56] . In “type II” neuroblasts , grh acts with the Bx-C gene Abd-A to activate cell death genes of the RHG family , and thereby terminates lineage progression by apoptosis of the neuroblast . This set of studies demonstrates that lineage progression , in both embryonic and postembryonic neuroblasts , can be terminated either by neuroblast cell cycle exit or by neuroblast apoptosis . In the abdominal segments , we find that the absence of Ap clusters results from a truncation of the NB 5–6 lineage , terminating it within the Pdm early temporal window , and therefore Ap cluster cells are never generated . Our studies reveal that this truncation results from neuroblast cell cycle exit , controlled by Bx-C , hth , and exd , thereafter followed by apoptosis . In Bx-C/hth/exd mutants , the neuroblast cell cycle exit point is bypassed , and a thoracic sized lineage is generated , indicating that these genes may control both cell cycle exit and apoptosis . However , it is also possible that cell cycle exit is necessary for apoptosis to commence , and that Bx-C/hth/exd in fact only control cell cycle exit . Insight into the precise mechanisms of the cell cycle exit and apoptosis in NB 5-6A may help shed light on this issue . Whichever mechanism is used to terminate any given neuroblast lineage—cell cycle exit or cell death—the existence in the Drosophila CNS of stereotyped lineages progressing through defined temporal competence windows allows for the generation of segment-specific cell types simply by regulation of cell cycle and/or cell death genes by developmental patterning genes . Specifically , neuronal subtypes born at the end of a specific neuroblast lineage can be generated in a segment-specific fashion “simply” by segmentally controlling lineage size . This mechanism is different in its logic when compared to a more traditional view , where developmental patterning genes act upon cell fate determinants . But as increasing evidence points to stereotypic temporal changes also in vertebrate neural progenitor cells [12] , this mechanism may well turn out to be frequently used to generate segment-specific cell types also in the vertebrate CNS . Our findings of Hox , Pbx/Meis , and temporal gene input during Ap cluster formation is not surprising—generation and specification of most neurons and glia will , of course , depend upon some aspect or another of these fundamental cues . Importantly however , the detailed analysis of the NB 5-6T lineage , and of the complex genetic pathways acting to specify Ap cluster neurons , has allowed us to pin-point critical integration points between anteroposterior and temporal input . Specifically , cas , Antp , hth , and exd mutants show striking effects upon Ap cluster specification , with effects upon expression of many determinants , including the critical determinant col . Whereas Antp plays additional feed-forward roles , and exd was not tested due to its maternal load , we found that both cas and hth mutants can be rescued by simply re-expressing col ( [35]; this study ) . This demonstrates that among a number of possible regulatory roles for cas , hth , Antp , and exd , one critical integration point for these anteroposterior and temporal cues is the activation of the COE/Ebf gene col , and the col-mediated feed-forward loop . Both col and ap play important roles during Drosophila muscle development , acting to control development of different muscle subsets [57] , [58] . Their restricted expression in developing muscles has been shown to be under control of both Antp and Bx-C genes [59] , [60] . Molecular analysis has revealed that this regulation is direct , as Hox proteins bind to key regulatory elements within the col and ap muscle enhancers [59] , [60] . The regulatory elements controlling the CNS expression of col and ap are distinct from the muscle enhancers [23] , [59]–[61] ( unpublished data ) , and it will be interesting to learn whether Hox , as well as Pbx/Meis and temporal regulatory input , acts directly also upon the col and ap CNS enhancers . One particularly surprising finding pertains to the instructive role of Hth levels in NB 5-6T . At low levels , Hth acts in NB 5-6A to block lineage progression , whereas at higher levels , it acts in NB 5-6T to trigger expression of col within the large cas window . It is interesting to note that the hth mRNA and Hth protein expression levels increase rapidly in the entire anterior CNS ( T3 and onward ) ( this study ) [42] , [43] . In addition , studies reveal that thoracic and anterior neuroblast lineages in general tend to generate larger lineages [19] , [20] , [49] and thus remain mitotically active for a longer period than abdominal lineages . On this note , it is tempting to speculate that high levels of Hth may play instructive roles in many anterior neuroblast lineages . In zebrafish , Meis3 acts to modulate Hox gene function , and intriguingly , different Hox genes require different levels of Meis3 expression [62] . In the Drosophila peripheral nervous system , expression levels of the Cut homeodomain protein play instructive roles , acting at different levels to dictate different dendritic branching patterns in different sensory neuron subclasses [63] . Although the underlying mechanisms behind the levels-specific roles of Cut [63] , Meis3 [62] or Hth ( this study ) are unknown , it is tempting to speculate that they may involve alterations in transcription factor binding sites , leading to levels-sensitive binding and gene activation of different target genes . The vertebrate members of the Meis family ( Meis1/2/3 , Prep1/2 ) are expressed within the CNS , and play key roles in modulating Hox gene function . Intriguingly , studies in both zebrafish [64] , [65] and Xenopus [66]–[68] reveal that subsequent to their early broad expression , several members are expressed more strongly or exclusively in anterior parts of the CNS , in particular , in the anterior spinal cord and hindbrain . Here , functional studies reveal complex roles of the Meis family with respect to Hox gene function and CNS development . However , in several cases , studies reveal that they are indeed important for specification , or perhaps generation , of cell types found in the anterior spinal cord and/or hindbrain , i . e . , anteroposterior intermediate neural cell fates [62] , [66]–[70] . As we learn more about vertebrate neural lineages , it will be interesting to learn which Meis functions may pertain to postmitotic neuronal subtype specification , and which may pertain to progenitor cell cycle control . In anterior segments—subesophageal ( S1–S3 ) and brain ( B1–B3 ) —a more complex picture emerges where both the overall lineage size and temporal coding is altered , when compared to the thoracic segments . Specially , whereas all anterior NB 5–6 lineages do contain Cas expressing cells , expression of Grh is weak or absent from many Cas cells . The importance of this weaker Grh expression is apparent from the effects of co-misexpressing grh with Antp—misexpression of Antp alone is unable to trigger FMRFa expression , whereas co-misexpression with grh potently does so . It is unclear why anterior 5–6 lineages would express lower levels of Grh , since Grh expression is robust in some other anterior lineages ( unpublished data ) . In the B1 segment , we and others identify not one , but two NB 5–6 equivalents [14] . However , the finding of two NB 5–6 equivalents is perhaps not surprising , since the B1 segment contains more than twice as many neuroblasts as posterior segments [36]–[38] . Due to weaker lbe ( K ) -lacZ and -Gal4 reporter gene expression , and cell migration , we were unable to map out these lineages . However , irrespective of the features of the B1 NB 5–6 lineages , we were unable to trigger bona fide Ap cluster formation by Antp/grh co-misexpression in B1 . Together , these findings suggest that the B1 segment develops using a different modus operandi , a notion that is similar to development of the anterior-most part of the vertebrate neuroaxis , where patterning and segmentation is still debated [71] , [72] . On that note , it is noteworthy that although Hox genes play key roles in specifying unique neuronal cell fates in more posterior parts of the vertebrate CNS [8] , [9] , [73] , [74] , and can indeed alter cell fates when misexpressed , the sufficiency of Hox genes to alter neuronal cell fates in the anterior-most CNS has not been reported—for instance , Hox misexpression has not been reported to trigger motoneuron specification in the vertebrate forebrain . Thus , in line with our findings that Antp is not sufficient to trigger Ap cluster neuronal fate in the B1 anterior parts , the anterior-most part of both the insect and vertebrate neuroaxis appears to be “off limits” for Hox genes . The Hox , Pbx/Meis , and temporal genes are necessary , and in part sufficient , to dictate Ap cluster neuronal cell fate . However , they only do so within the limited context of NB 5–6 identity . Within each abdominal and thoracic hemisegment , each of the 30 neuroblasts acquires a unique identity , determined by the interplay of segment-polarity and columnar genes [75] , [76] . In the periphery , recent studies demonstrate that anteroposterior cues , mediated by Hox and Pbx/Meis genes , are integrated with segment-polarity cues by means of physical interaction and binding to regulatory regions of specific target genes [77] . It is tempting to speculate that similar mechanisms may act inside the CNS as well , and may not only involve anteroposterior and segment-polarity integration , but also extend into columnar and temporal integration .
The fly stocks used were as follows: y w exdB108 f FRT18D/FM7 [78] and X∧X y f/ovoD2 FRT18D/Y;F38/F38 ( both provided by R . White ) . UAS-exd-GFP [79] ( provided by R . Mann ) . exd1 [80] . Antp25and Antp14 [81] . AntpNs-rvC12 [82] . UAS-Antp , UAS-Ubx , UAS-abd-A [83] ( obtained from F . Hirth ) . UAS-Abd-Bm [84] ( obtained from J . Castelli-Gair ) . Abd-BM1 and Abd-BM2 [85] . Df ( 3R ) Ubx109/Dp ( 3;3 ) P5 [86] . Dp ( 3;1 ) P68; ss1 Ubx1 abd-AD24 Abd-BD18/In ( 3LR ) UbxU , Sbsbd-2 ss1 Ubxbx-34e UbxU [87] . C155-Gal4 ( elav ) [88] . abd-AMX1 [89] . Ubx1 [90] . abd-AP10 and Ubx9 . 22 [91] . hth5E04 [92] . Df ( 3R ) Exel6158 ( referred to as hthDf3R ) [93] . UAS-hth [94] ( provided by A . Salzberg ) . col1 , col3 [95] and UAS-col [96] ( provided by A . Vincent ) . Df ( 3L ) H99 [97] . lbe ( K ) -Gal4 and UAS-grh [35] . ladybird early fragment K driving lacZ ( referred to as lbe ( K ) -lacZ ) ( provided by K . Jagla ) [39] . UAS-nls-myc-EGFP ( referred to as UAS-nmEGFP ) , UAS-myc-EGFP–farnesylation , sqzGal4 [29] . apmd544 ( referred to as apGal4 ) [98] . aprK568 ( referred to as aplacZ ) [99] . gsb01155 ( referred to as gsblacZ ) [100] . elav-Gal4 [101] ( provided by A . DiAntonio ) . Mutants were maintained over GFP- or YFP-marked balancer chromosomes . As wild type , w1118 was often used . Staging of embryos was performed according to Campos-Ortega and Hartenstein [47] . Unless otherwise stated , flies were obtained from the Bloomington Drosophila Stock Center . Primary antibodies used were: Guinea pig α-Col ( 1∶1 , 000 ) , guinea pig α-Dimm ( 1∶1 , 000 ) , chicken α-proNplp1 ( 1∶1 , 000 ) , and rabbit α-proFMRFa ( 1∶1 , 000 ) [25] . Chicken α-proFMRFa ( 1∶1 , 000 ) , chicken α-myc ( 1∶5 , 000 ) and rat α-Grh ( 1∶1000 ) [35] . Rabbit α-Nab ( 1∶1 , 000 ) [32] ( provided by F . Díaz-Benjumea ) . Rabbit α-Cas ( 1∶250 ) [102] ( provided by W . Odenwald ) . Mouse monoclonal antibody ( mAb ) α-Col ( 1∶250 ) ( provided by M . Crozatier and A . Vincent ) . Guinea pig α-Deadpan ( 1∶1 , 000 ) ( provided by J . Skeath ) . Rat monoclonal α-Gsbn ( 1∶10 ) ( provided by R . Holmgren ) . Rabbit α-Hunchback ( 1∶1 , 000 ) and rabbit α-Krüppel ( 1∶500 ) ( provided by R . Pflanz ) . mAb α-Nubbin/Pdm1 ( 1∶10 ) ( provided by S . Cohen ) . mAb α-Exd ( B11M; 1∶5 ) and mAb α-Ubx ( FP3 . 38; 1∶10 ) ( provided by R . White ) . mAb α-Abd-A ( 1∶400 ) ( provided by I . Duncan ) . Rabbit α-Hth ( 1∶500 ) ( provided by A . Salzberg ) . Rabbit α-phospho-histone H3-Ser10 ( pH3 ) ( 1∶250 ) and mAb α-myc ( 1∶2 , 000 ) . Rabbit α-ß-Gal ( 1∶5 , 000 ) . Rabbit α-cleaved caspase-3 ( 1∶100 ) . Rat monoclonal α-BrdU ( 1∶100 ) . Chicken α-ß-Gal ( 1∶1 , 000 ) . Rabbit α-GFP ( 1∶500 ) . mAb α-Dac dac2–3 ( 1∶25 ) , mAb α-Antp ( 1∶10 ) , mAb α-Abd-B ( 1∶10 ) , mAb α-Pros MR1A ( 1∶10 ) , and mAb α-Eya 10H6 ( 1∶250 ) . All polyclonal sera were preabsorbed against pools of early embryos . Immunostaining was performed according to [35] . Zeiss LSM 5 or Zeiss META 510 confocal microscopes were used to collect data for all fluorescent images; confocal stacks were merged using LSM software or Adobe Photoshop . Where immunolabeling was compared for levels of expression , wild-type and mutant tissue were stained and analyzed on the same slide . Statistical analysis was performed using Microsoft Excel , and bar graphs generated using GraphPad Prism software . Where appropriate , images were false colored to facilitate for color-blind readers .
|
An animal's nervous system contains a wide variety of neuronal subtypes generated from neural progenitor ( “stem” ) cells , which generate different types of neurons at different axial positions and time points . Hence , the generation and specification of unique neuronal subtypes is dependent upon the integration of both spatial and temporal cues within distinct stem cells . The nature of this integration is poorly understood . We have addressed this issue in the Drosophila neuroblast 5–6 lineage . This stem cell is generated in all 18 segments of the central nervous system , stretching from the brain down to the abdomen of the fly , but a larger lineage containing a well-defined set of cells—the Apterous ( Ap ) cluster—is generated only in thoracic segments . We show that segment-specific generation of the Ap cluster neurons is achieved by the integration of the anteroposterior and temporal cues in several different ways . Generation of the Ap neurons in abdominal segments is prevented by anteroposterior cues stopping the cell cycle in the stem cell at an early stage . In brain segments , late-born neurons are generated , but are differently specified due to the presence of different anteroposterior and temporal cues . Finally , in thoracic segments , the temporal and spatial cues integrate on a highly limited set of target genes to specify the Ap cluster neurons .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"Methods"
] |
[
"developmental",
"biology",
"developmental",
"biology/stem",
"cells",
"developmental",
"biology/cell",
"differentiation",
"neuroscience/neurodevelopment",
"developmental",
"biology/neurodevelopment",
"developmental",
"biology/molecular",
"development",
"developmental",
"biology/developmental",
"molecular",
"mechanisms"
] |
2010
|
Segment-Specific Neuronal Subtype Specification by the Integration of Anteroposterior and Temporal Cues
|
Starvation of cells for the DNA building block dTTP is strikingly lethal ( thymineless death , TLD ) , and this effect is observed in all organisms . The phenomenon , discovered some 60 years ago , is widely used to kill cells in anticancer therapies , but many questions regarding the precise underlying mechanisms have remained . Here , we show for the first time that starvation for the DNA precursor dGTP can kill E . coli cells in a manner sharing many features with TLD . dGTP starvation is accomplished by combining up-regulation of a cellular dGTPase with a deficiency of the guanine salvage enzyme guanine- ( hypoxanthine ) -phosphoribosyltransferase . These cells , when grown in medium without an exogenous purine source like hypoxanthine or adenine , display a specific collapse of the dGTP pool , slow-down of chromosomal replication , the generation of multi-branched nucleoids , induction of the SOS system , and cell death . We conclude that starvation for a single DNA building block is sufficient to bring about cell death .
Starvation of cells for the DNA precursor dTTP can cause rapid cell death in all domains of life [1] . This phenomenon , called thymineless death ( TLD ) , was first discovered in 1954 in E . coli upon exposing thymine-requiring ( thyA ) strains to medium lacking thymine [2] . As TLD can be promoted in cells from bacteria to man , it has been widely employed for therapeutic purposes . Methotrexate and trimethoprim , both antifolates , and 5-fluorouracil are antitumor and antibacterial agents based on their ability to block thymidylate ( dTMP ) synthesis [3] , [4] leading to low dTTP levels that kill or prevent proliferation of actively-dividing cells . However , despite decades of interest , our understanding of the TLD phenomenon is still incomplete , particularly with regard to the primary initiating events that cause cell death . Recent progress has revealed a complexity of participating and contributing events , and has led to models centered on the impairment of DNA replication and resulting stalling of replication forks [5]–[9] . Such stalled forks give rise to DNA breakage if not repaired by homologous recombination . Importantly , despite the stalling of existing replication forks , initiation of new replication forks at the oriC chromosomal origin can continue [9] , [10] , causing increased complexity of the chromosome , which becomes a major determinant of cell death [7] , [9] . Recombinational processes play an important role throughout TLD in at least a dual fashion: they can rescue starving cells from early stages of TLD , but ultimately contribute actively to death at later stages by creating unresolvable or unrepairable intermediates and DNA breaks [5] , [8] . Notably , significant breakage and disappearance of origin-containing DNA is observed , consistent with the importance of ongoing DNA initiation in TLD [6] , [9] . TLD is also accompanied by persistent SOS induction , which contributes to cell death by initiating lethal filamentation [5] . However , a critical unanswered question is whether the phenomenon is truly thymine-specific or can be , likewise , imposed by starvation for other DNA precursors . Obviously , models based on stalled DNA replication would apply equally if stalling were mediated by starvation for any other dNTP . However , selective manipulation of the concentration of individual dNTPs is experimentally difficult , because their joint de novo synthesis is regulated by feedback on the enzyme ribonucleotide reductase [11] . Until now , dTTP was the only nucleotide for which the phenomenon could be demonstrated , as its pool can be manipulated separately through the thymine salvage pathway [12] . Nevertheless , long-sought conditions by which cells can be starved specifically for a dNTP other than dTTP were found , serendipitously , in our studies of the optA1 allele of the E . coli dgt gene . The dgt gene encodes a dGTPase with an unusual activity , hydrolyzing dGTP into deoxyguanosine and triphosphate ( PPPi ) [13] , [14] . Its deletion was found to result in a spontaneous mutator effect [15] , which was attributed to possible changes in the cellular dNTP levels , particularly dGTP . Indeed , an approximately 2-fold increase in the dGTP level of a dgt mutant had been described [14] . The optA1 allele of dgt is a promoter-up mutation , which increases gene expression by as much as 50-fold [16] . Consistent with this up-regulation a decrease in dGTP level was reported [17] , although modest in view of the 50-fold gene overexpression . This presumably reflects the ability of cells to adjust their dNTP levels through feedback regulation on the ribonucleotide reductase . In the present study we report that a more dramatic and specific dGTP decrease can be achieved by combining the optA1 allele with a defect in the gpt gene . The gpt gene functions in purine salvage by converting guanine into guanosine monophosphate ( GMP ) . In the present study , we show that an optA1 gpt strain grown in minimal medium with casamino acids ( CAA ) in the absence of an external purine source , like hypoxanthine ( Hx ) or adenine , dies in a manner sharing many of the features associated with TLD . Certain differences with TLD are also noted , which we argue reflect different kinetic manifestations of the same intrinsic mechanism . We propose to term this phenomenon dGTP starvation .
The initial observation that triggered our interest was that the optA1 allele of dgt caused impaired growth when the strain also contained the large ( 120-kb ) Δ ( pro-lac ) X111 chromosomal deletion [18] . When stationary cultures of such a strain were diluted by at least 5 , 000-fold in minimal glucose medium enriched with CAA ( 1% ) , the cultures failed to grow beyond OD630 nm = 0 . 1 . Complementation of the deletion by an F'prolac covering the deleted region reverted the cells to normal growth . Upon reconstruction of this defect in the widely-used MG1655 strain background , we found that the growth impairment was attributable to the combination of optA1 with the lack of gpt , a gene located inside the boundaries of the Δ ( pro-lac ) X111 deletion . The gpt gene encodes Guanine Phosphoribosyltransferase , a purine salvage enzyme responsible for salvaging guanine , hypoxanthine , and xanthine via their conversion to the corresponding NMP , with guanine being the preferred substrate due to the lowest Km [19] . The growth curves displayed in Fig . 1A show the growth defect of the optA1 gpt strain . While the single optA1 or gpt strains show normal growth in the minimal medium with casamino acids , the optA1 gpt double mutant fails to reach beyond OD630 nm = 0 . 1 for at least 10 hrs . In contrast , the double mutant strain grows normally in the presence of the added purine sources hypoxanthine ( Hx ) or adenine ( Ade ) , as shown in Fig . 1B . Addition of guanine ( Gua ) as purine source has no such effect; in fact , it exacerbates the growth defect . The deleteriousness of the optA1 gpt combination may be understood based on the activities of the corresponding Dgt and Gpt enzymes within the salvage and de novo purine biosynthesis pathways . Diagram 1C shows how enhanced dGTPase activity resulting from optA1 leads to increased breakdown of dGTP , yielding deoxyguanosine ( G-dRib ) and deoxyribose-1-phosphate ( dRib-1-P ) and , subsequently , guanine ( Gua ) upon further metabolism by the DeoD purine nucleoside phosphorylase . In the gpt background , this guanine cannot be readily returned back , via GMP , to the guanine nucleotide pool; therefore , the expected result is a limitation for purine nucleotides , presumably most acutely for dGTP . Consistent with this model are the observed alleviation of the growth inhibition by addition to the growth medium of exogenous purines like hypoxantine ( Hx ) or adenine ( Ade ) , which do not require gpt action ( Fig . 1C ) , but not guanine ( Gua ) , which is , in fact , inhibitory ( see Fig . 1B ) . Any accumulated guanine is expected to contribute further to the starvation , as it is a corepressor for the PurR repressor [20] , [21] controlling de novo purine biosynthesis ( Fig . 1C ) . Indeed , we found that one alternative way of circumventing the gpt block ( see Fig . 1C ) is the stimulation of de novo IMP production by inactivation of the purR repressor [21] , as shown in Fig . 1A . Cell and nucleoid morphology of optA1 gpt strains were followed by microscopy , as shown in Fig . 2 . The starved cells developed progressively extensive filamentation with swollen regions ( bulges ) ( Fig . 2D and E ) in the middle of the cells . At the 7 hr time point , DAPI staining revealed disturbed nucleoid shapes within the filaments , which became fused and compacted ( Fig . 2D ) . The enlarged nucleoids coincided with the filament bulges , thus accounting for the distortion of the cell envelopes . Use of Live/Dead staining indicated extensive death of the filaments ( Fig . 2F ) . Subsequent experiments were designed to measure more carefully the physiology of the starving cells . It soon became clear that the growth defect of the optA1 gpt strain was dependent on the cell density: only strongly diluted cultures ( for example 1/5 , 000 from overnight cultures ) showed the growth restriction . This suggested to us that maintenance of an active growth phase was required; reduced growth experienced at higher biomass values would allow the cells to escape . On that basis , we developed a standardized protocol in which inoculates from overnight cultures were first grown for a sufficient number of generations in the presence of hypoxanthine to assure exponential growth , and then cells were transferred to medium without hypoxanthine . When necessary , subsequent dilutions were made in fresh medium to keep the OD630 nm at or below 0 . 2 . Performed in this manner , the experiment of Fig . 3A shows that non-starved cells are able to continue in exponential growth indefinitely , as expected . However , cells growing without hypoxanthine display a slowdown in biomass growth ( OD ) rate compared to the control and an eventual complete growth arrest at 5–6 hrs . More importantly , the viable count of the starved culture increased initially only by about 4-fold overall and then declined by 40- to 50-fold , indicating extensive death of the cells . Interestingly , after the 6 hr time point the culture appeared to recover ( Fig . 3A ) . As discussed later in more detail , this recovery reflects the accumulation of suppressor mutants that have become resistant to the starvation condition . However , we will first report on the properties of the starving and dying optA1 gpt cells prior to the appearance of the suppressors . The Supplementary Fig . S1 shows the importance of the 50-fold dilution at the 150 min time point; lower dilutions are not sufficient . In addition to turbidity and viable cell count , we also monitored the extent of DNA synthesis in the starving cells . The results in Fig . 3B show that the starved cells synthesized DNA at a reduced rate . Much of this reduction may reflect a slow down of ongoing replication forks due to the shortage of dGTP , as further explored in the following three experiments . In the experiment of Fig . 3C we followed the status of the bacterial chromosomal DNA by determining the ratio of chromosomal origins to termini ( ori/ter ) by Q-PCR . The ratio was around 2 . 6 in the non-starved cells , but it increased progressively to a value above seven during hypoxanthine starvation . Such increase is consistent with a slowed-down progression of the replication forks ( increase in chromosomal replication time or C-time ) along with a continued production of new forks at the oriC origin . This is also one of the hallmarks of thymine starvation , at least in the early stages of the process [7] , [9] , [10] . Informatively , the ori/ter ratio can be used to predict the branching ( complexity ) of the nucleoid [22] under these conditions . As shown in Fig . 3C ( right side ) , the chromosomes of the starved cells are predicted to become increasingly complex . We also investigated the progress of existing replication forks in the absence of new initiations . For this , we used the antibiotic rifampicin , which permits , in principle , continuation of DNA synthesis from existing forks but prevents new initiations [23] . The results in Fig . 3D show that the non-starved cultures display the saturation kinetics of DNA synthesis typically observed in this kind of experiment ( replication fork run-out ) . In contrast , the hypoxanthine-starved cultures show a dramatic loss of DNA synthesis capacity ( to roughly only 15% of the control ) , consistent with the predicted sudden limitation in dGTP upon hypoxanthine removal . It appears that , in the absence of transcription , even existing forks cannot be completed . The simplest explanation would be a near complete loss of dGTP upon hypoxanthine deprivation under these conditions of the inhibited transcription . To more directly investigate the rates of DNA synthesis in starved vs . non-starved cells , we conducted pulse-labeling experiments using [methyl-3H]-thymidine . The results shown in Fig . 3E indicate that the hypoxanthine-starved culture suffers from an immediate about 20-fold reduction in the DNA synthesis rate . Interestingly , within the next 40 minutes , DNA synthesis capacity recovers slowly to approximately 50% of the control value , presumably due to transcriptional adaptation that may recover dGTP levels , at least in part . The near-complete loss of DNA synthesis capacity at the earliest time point is fully consistent with the inability of the cells to complete their ongoing forks in the presence of rifampicin ( see Fig . 3D ) . These results are also supported by the changes in dGTP level as described in a subsequent section . The Fig . 3E also shows the appearance of a spike of thymidine incorporation about 50 minutes after the start of starvation . This spike does not represent an occasional fluctuation within the measurements , as it was observed reproducibly in at least four repeated experiments . The simplest explanation for this spike may be an initiation event occurring at this time point . If correct , the interesting question arises as to how the present circumstances can lead to a coordinated culture-wide event . Following the spike , DNA synthesis capacity quickly drops to the pre-spike level . This inability to sustain the new higher rate of [methyl-3H]-thymidine incorporation indicates that , regardless of the number of forks , the total amount of DNA synthesis is severely constrained , likely due to the amount of available dGTP ( see section below ) . To investigate the effects of the starvation on the dNTP DNA precursors , we analyzed the intracellular dNTP pools after 2-hr and 4-hr incubations in the purineless medium . A clear ∼6-fold reduction , was seen in the dGTP concentration at 2 hrs ( Fig . 4A , Table S1 ) , while the concentration of the other dNTPs was not significantly affected ( except for a possible small increase in the dCTP and dTTP pools ) . At times later than 2 h , dGTP could no longer be detected , although this was due in part to the appearance of another , as yet unidentified peak in the HPLC profile nearby the dGTP peak ( See Fig . 4C ) . No changes in dGTP level were noted in the single optA1 or gpt mutants . In contrast , only a slight reduction in the rGTP pool was noted ( 2-fold or less , see Fig . 4B , Table S1 ) . To investigate the time course of dGTP depletion upon Hx removal we withdrew samples at a series of earlier time points and analyzed them by HPLC by a slightly different protocol designed to improve the resolution of the dGTP peak . During the first 30 minutes after hypoxanthine withdrawal , the dGTP pool was dramatically reduced ( Fig . 4D ) ; in fact , no dGTP could be detected above the ( 0 . 001 ) detection limit , indicating that the dGTP level was decreased by 10-fold or more . Interestingly , the dGTP level temporarily recovered at the 45-min time point but declined afterwards . We note that the kinetics of dGTP reduction of Fig . 4D are fully consistent with the DNA synthesis rate changes ( Fig . 3E ) . Thus , DNA synthesis in the optA1 gpt mutant appears strictly governed by dGTP availability . One of the hallmarks of TLD is the occurrence of significant DNA damage and induction of the SOS response [5] , [8] . To investigate whether cell death during dGTP starvation is also accompanied by SOS induction , we assayed the expression level of an umuC::lacZ reporter construct , which can be used as a diagnostic for induction of the damage-inducible SOS response [24] . The results revealed SOS induction in the optA1 gpt strain starting at 4 hours after purine removal ( Fig . 5A ) . This time coincides with the culture reaching its maximum value of nucleoid complexity ( see Fig . 3C ) and the start of the decline in culture viability ( Fig . 3A ) . We also studied the effect of recA and sulA deficiencies on the survival of the optA1 gpt strain . These mutations produced opposite effects ( Fig . 5B ) : the recA defect dramatically sensitized the cells , leading to an immediate viability loss after Hx removal , very similar to the rapid early death of recA strains during TLD [5] , [8] . In contrast , the sulA defect alleviated lethality ( Fig . 5B ) , indicating that SOS-mediated lethal filamentation is a contributing factor to cell death [5] . Another informative aspect of dGTP starvation was revealed from plating efficiency determinations on solid media . In these experiments , the optA1 gpt strain was first grown to saturation in medium containing hypoxanthine , followed by plating on medium lacking hypoxanthine . It was observed that the optA1 gpt strain was able to form colonies on these plates with normal efficiency , at least when the plates were placed at 37° ( Fig . 6A ) . In contrast , a strongly reduced plating efficiency ( 10−4 or less ) was observed when the plates were placed at 42° ( Fig . 6C ) . This loss of plating efficiency required the presence of casamino acids ( CAA ) ( Fig . 6D ) . The more sensitive optA1 gpt recA triple mutant strain was not able to produce colonies , even at 37° ( Fig . 6B ) . Overall , these results are consistent with the requirement for maintenance of an active growth status for the optA1 gpt strains . Presumably , in developing colonies on the plate the growth rate can be sufficiently slowed down , perhaps due to nutrient limitation in the colony environment , to permit survival . Such escape from death is apparently not possible at 42°C , probably due to increased origin firings at this higher temperature [25] . No escape is possible at any temperature for the ΔrecA strain . The surviving fraction of cells ( 10−5 to 10−4 ) observed on these plates also reflects the appearance of suppressor mutants , as described below . The starvation experiments in the liquid media show an apparent recovery of the optA1 gpt and optA1 gpt recA cultures after some 6–8 hrs ( Figs . 3A and 5B ) . In fact , fully-grown cultures can be obtained in many cases after overnight growth . When these fully-grown cultures were diluted and subjected to a repeat starvation procedure , the cells proved resistant . We concluded that they had incurred suppressor mutations rendering the cells resistant . We also investigated the colonies that appeared on the 42°C glucose+CAA solid media plates of Fig . 6 . When testing several of these colonies , it was found that they , too , were resistant to subsequent starvation , indicating that they had acquired mutations that allowed them to escape death . The plating experiment of Fig . 6C proved a convenient avenue for obtaining a large number of resistant clones for further analysis . A total 108 colonies ( from 10 independently grown cultures ) were picked from the restrictive plates , purified and subjected to a repeat growth and plating cycle . Out of the 108 clones , 106 proved fully resistant . One obvious way by which resistance could be acquired is loss of the OptA1 phenotype through inactivation of the dgt gene . We tested for the possible loss of the OptA1 phenotype using the bacteriophage T4 assay ( see Materials and Methods ) . This revealed that 20 out of the 106 clones ( ∼20% ) had lost the OptA1 phenotype , likely due to loss of dgt function . The remaining 80% presumably represent mutants that lost some other functions involved in nucleotide metabolism ( for example purR , see Fig . 1A ) or in aspects of DNA replication and/or recombination . Further study of these suppressors may prove informative regarding the underlying mechanisms . One additional experiment ( Fig . 7 ) provided useful insight into the emergence of suppressors along with confirming the critical importance of cell or biomass density ( biomass density is a more accurate term here because cells progressively filament during starvation ) . In this experiment , a non-starved stationary culture was diluted to different extents - over a range of three orders of magnitude - and inoculated into media with or without hypoxanthine for an overnight growth attempt . The results showed that all cultures were able to reach full-grown density after overnight growth . However , when analyzed for the presence of suppressors , the cultures differed dramatically depending on their starting dilution . For cultures started at low to modest dilutions ( 100- to 800-fold ) the majority of cells in the final culture had remained sensitive to dGTP starvation . On the other hand , dilutions of 3 , 000-fold or larger yielded cultures composed entirely of resistant clones ( Fig . 7 ) . Thus , relatively high densities permit survival without mutation , likely by reduction in growth rate , whereas low densities produce apparent survival only by mutation .
Our investigation of the hypoxanthine-starved optA1 gpt strains can be summarized by the following findings: Each of the findings ( a-h ) above can find their counterpoint in the phenomenology of TLD , although quantitative differences are certainly noted . This similarity of observations supports our contention that TLD and dGTP starvation are , in fact , two manifestations of the same underlying phenomenon , in which cells deprived of a single DNA precursor suffer from chromosomal distress that ultimately kills them . We conclude that in dGTP-starved cells , like in TLD , ongoing replication forks are slowed down and even stall ( Fig . 3E ) due to the deprivation of one of the DNA precursors . The deprivation of dGTP is seen in dramatic fashion immediately upon hypoxanthine withdrawal , where both dGTP level and DNA synthesis rate are reduced to near zero ( Figs . 3E , 4D ) , but also later when , despite a modest recovery , dGTP levels remain reduced and continue to decline . At the same time , the nutritional status of the cells remains high , leading to continued high levels of RNA and protein synthesis , as is also clear from the about 100-fold increase in biomass during the first few hours of starvation ( Fig . 3A ) . Continued biomass growth permits continued initiations of new forks at the chromosomal origin ( oriC ) . The slowdown of DNA synthesis along with continued new initiations is supported by the observed increase in the ori/ter ratio ( Fig . 3C ) . The resulting build-up of chromosomal complexity ( Fig . 3C ) then leads to a series of secondary consequences , such as replication forks collisions , DNA bulges reflecting unresolved complex chromosomes ( Fig . 2D and E ) , double-stranded breaks , SOS induction ( Fig . 5A ) , and lethal filamentation ( Fig . 2 ) , as also occurring in TLD . The occurrence of double-strand breaks and the need for their repair is clearly suggested by the exquisite sensitivity of the recA-deficient optA1 gpt strain to dGTP starvation ( Figs . 5B , 6B ) . Survival of the cells during the early cessation of DNA synthesis ( Figs . 3E , 4D ) appears to be critically dependent on recombinational repair , indicating that double-strand breaks are occurring at this stage [27] . This fully resembles the sensitization of recA mutants to the early stages of the TLD process [8] , [9] . While in rec+ cells the damaged forks can be repaired , their repair does not solve the stalling of the forks as long as the dGTP concentration stays limiting . It is likely that this early stalling is an important contributor to the build-up of chromosome complexity ( ori/ter ratio ) that takes place over the next several hours , culminating in SOS induction , filamentation , and cell death at the later time ( 4–6 h ) . At the same time , differences between dGTP starvation and TLD can be noted . Quantitatively , TLD appears to be a more destructive phenomenon , causing a more immediate loss of colony-forming ability ( three orders of magnitude during three hours [2] , [5] , [8] , [9] . During dGTP starvation , we observed only an ∼1 . 5 order of magnitude decline after the initial phase of continued cell divisions ( Fig . 3A ) . Thus the effect of dGTP starvation is milder than that of TLD . The important distinction is likely that in TLD the necessary precursor thymine is experimentally totally absent [26] , [28] , while the cells have no avenue to synthesize dTTP by alternative means . In contrast in dGTP starvation , dGTP can still be produced , albeit at limited levels . It might be suggested that dGTP starvation is more comparable to the phenomenon of thymine limitation , where thyA cells are grown at low , rate-limiting thymine concentrations while displaying an increased ori/ter ratio [29] . However , dGTP starvation differs from this type of limitation in one critical aspect: thymine-limited strains grow indefinitely in a steady state [30] , i . e . , the optical density , DNA concentration , and colony forming units increase exponentially at the same rate provided that the external thymine concentration is above the minimal required according to strain's background . In the case of dGTP starvation , cells are definitely not in steady state . For example , the increase in biomass until the arrest ( about 100-fold ) is not matched by that of the viable cell count ( about 4-fold ) ( Fig . 3A ) . This discrepancy between biomass and cell count is consistent with the observed filamentation of the dGTP-starved cells ( Fig . 2 ) . Also , during thymine limitation no cell death is observed . Kinetically , the initial rapid disappearance of dGTP ( Fig . 4D ) along with the associated cessation of replication fork movement ( Fig . 3E ) places dGTP starvation phenomenologically closer to dTTP starvation than to thymine limitation . dGTP starvation differs from TLD in that our experiments show a recovery of dGTP concentration at later time points . This recovery is likely a transcriptional response to the hypoxanthine withdrawal , which has no equivalent for dTTP production in TLD . Nevertheless , the recovery of dGTP is insufficient and dGTP levels continue to decline from that point on . Also note that the reduced DNA synthesis rate measured at later times is to be distributed over an increased number of forks . Thus , the rate of progression for each individual fork will be reduced accordingly . A rough estimate suggests that the rate per fork may be down at least one order of magnitude . At such reduced rates , lethal chromosomal complexity may not be avoidable . A third difference between dGTP starvation and TLD is that we did not find evidence for extensive origin destruction . Origin destruction has been discovered as one of the later aspects of TLD , purportedly by RecA-mediated ‘repair’ of multi-forked oriC [6] , [9] . Presumably , in the presence of low dGTP concentrations , enough DNA synthesis can take place to avoid this step of origin destruction , although it does not prevent cell death . Interestingly , during dGTP starvation , ΔrecA strains die at an accelerated rate that is very similar to the death rate of ΔrecA strain during TLD ( ∼5% survival after 2 h of starvation ) ( Fig . 5B and [8] , [9] ) . In both examples , no origin destruction occurs [9] . In this case , efficient killing occurs by events away from the origin , likely by lack of repair of stalled and broken replication forks . Nucleoids with increased complexity occur normally in bacteria under conditions of fast growth . Under these conditions , the generation time ( τ ) is shorter ( faster ) than the C-time ( time needed to complete a round of chromosomal synthesis ) , which is achieved by the firing of newly replicated origins prior to completion of the previous rounds , resulting in more complex chromosomes [31] . On the other hand , it appears that in wild-type cells the C-time never is never greater than twice the lowest achievable doubling time ( C≤2τ ) , so that the number of ongoing replication rounds ( or ‘fork positions’ , n = C/τ [32] ) , a quantitative measure of nucleoid complexity , rarely exceeds 2 . Conditions of n>2 have been obtained experimentally upon severe thymine limitation of thyA strains at very low extracellular thymine concentrations [29] , [33] . Under such conditions , cells continue to increase their size , culminating in distorted , monstrous shapes [34] , resembling the ones observed in our study ( Fig . 2 ) . Due to the deviation from steady-state growth , a physiological characterization of such distorted cells is difficult , and the same reason prevents a precise calculation of C and τ during dGTP starvation . However , based on our measured ori/ter ratios it is possible to estimate n using the equation [22] . During dGTP starvation , the ori/ter ratio reaches a value of near 8 , indicating a value of n near 3 . 0 ( Fig . 3C ) . This would indicate the presence of a total of 14 ( 2+4+8 ) active forks per chromosome ( see Fig . 3C ) . Extreme chromosome complexity due to overinitiation in a dnaA overexpressor strain has been shown to lead to collapse of replication forks , collisions between adjacent forks , and lethal chromosomal damage [35] . The existence of a limit to the extent of nucleoid complexity was proposed by Zaritsky [36] in his Eclipse model ( this term was adopted from the Nordström studies on plasmid R1 replication [37] ) , which states that , due to structural constraints , scheduled initiations normally fire only if the previous fork has moved away some minimal distance from the origin . If this critical condition is not met , the scheduled initiation is postponed to avoid collisions of replication forks that may endanger the integrity of DNA [36] . One might speculate that under the starvation conditions discussed here this minimal distance limit is breached . Another , informative distinction between dGTP starvation and TLD is the recovery of the dGTP-starved liquid cultures after about 7 h ( Fig . 3A ) . As described in the Results , recovery in this experiment is due to growth of mutants that have become resistant to the starvation condition . No production of resistant mutants occurs during TLD . This distinction between the two starvation procedures undoubtedly results from the fact that in TLD there is a zero provided supply of dTTP precursors , combined with the fact that the thyA cells do not have access to any alternative pathway by which dTTP might be synthesized , while during dGTP starvation there is a restricted , but not necessarily zero dGTP supply . Investigation of the suppressors of dGTP starvation may provide new insights into the various metabolic processes that can affect the ability of the cells to survive the dGTP-restricted conditions . We already noted that one category of suppressors is deduced to map in the dgt gene ( ∼20% ) , which is expected to restore dGTP levels . The majority of suppressors however reside at other loci , and their nature remains to be determined . It is an interesting question whether the suppressors represent preexisting mutants present in the population prior to initiation of the starvation conditions , or whether they are generated during the limited growth permitted during the starvation conditions . In view of the dNTP pool imbalances generated due to the dGTP drop , DNA replication during these conditions is likely to have reduced fidelity [38] , [39] . Possibly such reduced fidelity may account for the relatively high frequency ( 10−4 to 10−5 ) of suppressors found in the plating experiments of Fig . 6 . As an alternative to genetic mutation , cells can survive dGTP starvation by entering a slower-growth phase in which origin firing is reduced to be compatible with the newly established slow DNA synthesis rate ( Figs . 6 , S1 ) . This is clearly evidenced by the dilution experiments ( Fig . S1 ) , where increased cell densities slow down growth and permit survival . A further example of this type of survival is provided by the plating-efficiency of the optA1 gpt strain on solid media lacking hypoxanthine ( Fig . 6 ) . When plated at 37°C , the cells are able to develop into colonies , but at 42°C this is not the case; instead , suppressor colonies appear , at a frequency of 10−5 to 10−4 . Inside developing colonies , growth may be slowed down sufficiently to allow cells to reach an adaptive phase , and this may be the case at 37°C , but not at 42°C , where increased metabolic activity , such as increased origin firings [25] may push the cells over the edge of sustainability . Conversely , reduction in origin firings by omitting the casamino acid supplement ( CAA ) from the plates allows cells to survive even at 42°C ( Fig . 6D ) . These growth-dependent effects described above are not unique to dGTP starvation . For example , even in TLD , growth-dependent effects have been described . While in TLD no growth on plates lacking thymine occurs ( as thymine is an essential compound and thyA strains do not posses any alternative ways for dTTP synthesis ) , the extent of TLD can , nevertheless , be moderated by changes in growth conditions . For example , immunity to TLD can be provided by inhibition of protein or RNA synthesis [40] , [41] or by the silencing of new initiations in dnaA ( Ts ) strains [42] . In fact , nutritional shift-up of certain thyA strains can promote death even in the presence of thymine , presumably due to a newly created imbalance between the rates of origin firing and DNA synthesis [43] . The discovery and analysis of the phenomenon of dGTP starvation solves one of the outstanding questions regarding TLD since it was discovered some six decades ago: whether the phenomenon is thymine-specific or whether it can be provoked by starvation for other DNA building blocks . Indeed , even though the proposed mechanisms for the inactivation of thyA strains during TLD have become increasingly substantiated , it is still important to clearly separate the killing process from the thymine specificity . The phenomenon of dGTP starvation solves this basic issue: starvation for other DNA precursors ( like dGTP ) should be equally lethal . The finding that a critical starvation for the DNA precursor dGTP can cause cell death may lead to additional avenues for therapeutic applications . The dGTP model as described here may present a particularly realistic model system for cell death in such applications . In such cases the affected nucleotide may become critically restricted as presented here but not completely absent as in the TLD model system . dGTP may also be an attractive target as it has typically the lowest concentration among the four DNA precursors [44] . Interestingly , human cells have been found to contain a novel dGTPase , termed SAMHD1 , that has properties similar to the bacterial Dgt enzyme: it likewise hydrolyzes dGTP to yield deoxyguanosine and triphosphate [45] , [46] . SAMHD1 activity has been shown to act like a viral restriction factor in cells where it is expressed at elevated levels by lowering the dNTP concentrations sufficiently so that viral entities like HIV-1 cannot replicate [45] , [46] . SAMHD1 also protects the cells against autoimmune responses , such as the Aicardi-Goutieres syndrome [45] . While in those cases SAMHD1 acts like a restriction factor by inhibiting viral replication and creating conditions that lead to the breakdown of a variety of RNA and DNA substrates , it is imaginable that this activity under certain physiological conditions in actively growing cells could also be directed towards the cellular DNA .
All shown experiments used E . coli strain MG1655 and its derivatives . Genetic deficiencies were introduced into MG1655 by P1 transduction using P1virA . The optA1 allele of dgt was introduced linked to transposon zad-220::Tn10 as described [43] . The gpt::kan allele was obtained from the National BioResource Project ( NIG ) of Japan ( http://www . shigen . nig . ac . jp/ecoli/strain/top/top . jsp ) . The purR::cat , recA::cat and sulA::cat mutants were generated by the method of Datsenko and Wanner [47] using primers described in Table 1 . For testing SOS induction , the relevant strains were transformed with plasmid pSK1002 , which contains the lacZ reporter gene fused to the umuDC promoter [24] . For strain construction , maintenance , and determination of viable counts , LB medium was used with supplementation of the following antibiotics , where appropriate: tetracycline ( 15 µg/ml ) for optA1 linked with zad-22::Tn10 , kanamycin ( 50 µg/ml ) for gpt::kan , chloramphenicol ( 25 µg/ml ) for the purR::cat , recA::cat and sulA::cat alleles , and ampicilin ( 100 µg/ml ) for pSK1002 transformants . For experiments relating to starvation , cells were grown at 37°C in minimal medium containing Vogel-Bonner salts [48] containing glucose ( 0 . 4% ) , casamino acids ( 1% ) ( Becton-Dickinson ) , D-pantothenic acid ( 5 µM ) , and hypoxanthine ( 50 µg/ml ) . To assay the differential responses in media with or without purine source , two aliquots were filtered through a 25 , 47 or 90-mm diameter polycarbonate membrane filter ( 0 . 4 µm pore size; Millipore ) and diluted up to 10-fold in the identical medium with or without hypoxanthine ( 50 µg/ml ) . Aliquots for different assays were withdrawn at densities not exceeding 0 . 2 OD630 nm . Culture aliquots ( 300–350 ml ) harvested at OD630 nm = 0 . 2 were filtered through a 90-mm diameter polycarbonate membrane filter ( 0 . 4 µm pore size; Millipore ) . The filter was transferred to a Petri dish lid containing 10 ml of 60% aqueous methanol at −20°C . After 2 h at −20°C the filter was removed and the liquid suspension boiled for 5 min , followed by centrifugation for 15 min at 17 , 000× g and lyophilization of the supernatant . The residue was dissolved in 1 ml of sterile water , filtered through syringe filter ( Millipore , 0 . 22 µm pore size ) and lyophilized again . The final residue was dissolved 50 µl sterile water . HPLC analysis of the extracted dNTPs was performed by reversed-phase chromatography on an Agilent 1100 high-pressure liquid chromatography instrument with UV detection at 254 nm . Nucleotides were separated on a Zorbax Eclipse XDBC18 3 . 5 µM ( 150 by 4 . 6 mm ) column equipped with a Zorbax Eclipse XDBC18 guard column , adapting a prior method used for the separation of nucleotides [38] . At a flow rate of 0 . 8 ml/min , a linear gradient of 70∶30 buffer A to buffer B was run to 40∶60 over 30 min . The gradient was then changed over 60 min from 40∶60∶0 to 0∶87 . 5∶12 . 5 for buffer A - buffer B - buffer C . To wash the column between samples the gradient was first changed from 0∶87 . 5∶12 . 5 to 0∶70∶30 over 10 minutes with a final stepwise change to 70∶30∶0 for an additional 20 min . In a later set of experiments aimed at quantifying specifically dGTP during an extended starvation time course ( Fig . 4D ) , a modified protocol was used , as follows . At a flow rate of 1 ml/min , a linear gradient of 75∶25 buffer A to buffer B was changed to 52∶48 over 23 min . The gradient was then changed over 12 min from 52∶48 to 49∶51 and for an additional 10 min from 49∶51 to 40∶60 . To wash the column between samples the gradient was first changed from 40∶60∶0 to 0∶77 . 5∶22 . 5 for buffer A - buffer B - buffer C over 15 minutes and for an additional 10 min from 0∶77 . 5∶22 . 5 to 0∶70∶30 with a final stepwise change to 70∶30∶0 for an additional 10 min . Buffer A consisted of 5 mM tetrabutyl ammonium phosphate ( PicA Reagent; Waters ) , 10 mM KH2PO4 , and 0 . 25% methanol adjusted to pH 6 . 9 . Buffer B consisted of 5 mM tetrabutyl ammonium phosphate , 50 mM KH2PO4 , and 30% methanol ( pH 7 . 0 ) . Buffer C was acetonitrile . Nucleotide standards were obtained from Sigma . For chromosomal DNA extraction , 7- or 13-ml culture aliquots were harvested at various time points into the same volume of ice-cold PBS solution containing 20 mM NaN3 . DNA extraction was performed with the Easy-DNA kit ( Invitrogen ) and quantitated by staining with Picogreen ( Invitrogen ) . For determination of run-out DNA synthesis ( Fig . 3D ) , rifampicin was added at time zero at a concentration 300 µg/ml . For ori/ter determination , the DNA was digested with EcoRI and subjected to Quantitative PCR ( Stratagene Mx 3000 ) with SIBR Green detection using primers ( Table 2 ) specific to the origin and the terminus regions of the E . coli chromosome [49] . For determination of the DNA synthesis rate by pulse labeling , optA1 gpt cultures were grown with hypoxanthine to OD630 nm 0 . 1 , filtered , resuspended in the identical media with or without hypoxanthine , and brought to the same turbidity . Every 10–15 minutes 0 . 5 ml samples were withdrawn and pulse-labeled with 1 µCi of [methyl-3H]-thymidine at specific activity 0 . 5 µCi/nmole for 3 minutes . Samples were quenched with 0 . 5 ml of cold trichloroacetic acid ( TCA 10% ) containing 500 µg/ml of unlabeled thymidine to a final concentration 5% TCA and 250 µg/ml of unlabeled thymidine and kept on ice bath at least for 30 minutes . The entire samples were then collected on pre-wet 25-mm glass microfibre filters ( Whatman ) , and washed with cold TCA ( 5% with 250 µg/ml of unlabeled thymidine ) and 100% ethanol . The radioactivity on the filters was determined in a LS6500 liquid scintillation counter ( Beckman ) with Ecolume liquid scintillation cocktail ( MP Biomedicals ) . Aliquots of growing cultures were fixed with 0 . 25% formaldehyde and stained with DAPI . The cells were visualized by Nomarsky and DAPI fluorescence microscopy ( NIKON eclipse E600 ) and photographed using a Micropublisher CCD color Camera ( QImaging ) . Live/Dead stain ( Invitrogen ) was used to test viability of the cells . 5 . 0-ml samples were removed , and the cells were pelleted in a microcentrifuge . β-Galactosidase assays were performed essentially as described by Miller [18] . Cell pellets were resuspended in 1 ml of Z-buffer ( 60 mM Na2HPO4 , 40 mM NaH2PO4 , 10 mM KCl , 1 mM MgSO4 , 10 mM dithiothreitol ) . 80 µl of chloroform and 40 µl of 0 . 1% sodium dodecyl sulfate ( SDS ) were added to the cell suspension , which was then vortexed vigorously for 10 s . To start the reactions , 200 µl of ONPG ( 4 mM ) was added , and the reaction mixtures were incubated at 30°C for 4 . 5 min . The reactions were stopped with 0 . 5 ml of 1 M sodium bicarbonate , and the cellular debris was pelleted . The optical density was recorded with a BECKMAN DU 640 spectrophotometer with a 405-nm filter . Miller units were calculated as follows: units = 1 , 000[ ( OD405/ ( t×v×OD630 nm ) ] , where OD405 nm denotes the optical density at 405 nm . OD630 nm reflects the cell density at 630 nm , t is the reaction time in minutes , and v is the volume of culture used in the assay . To test whether suppression of the sensitivity of the optA1 gpt strains to hypoxanthine deprivation resulted from loss of the optA1 allele , a large number of clones that survived on the starvation plates at 42°C ( see Fig . 6C ) were tested for their ability to support growth of the bacteriophage T4 tsL141 mutant [50] , as described [43] . Clones that restrict growth of this phage at 30°C are optA1 [51] . Wild-type phage T4D was used as a positive control . The T4 phages were obtained from Dr J . W . Drake , NIEHS .
|
Starvation of cells for DNA precursor dTTP is strikingly lethal in many organisms , like bacteria , yeast , and human cells . This type of death is unusual in that starvation for other nutritional requirements generally results in growth arrest , but not in death . The phenomenon is called thymineless death ( TLD ) , because it was first observed some 60 years ago when a thymine-requiring ( thyA ) E . coli strain was exposed to growth medium lacking thymine . The TLD phenomenon is of significant interest as it is the basis for several chemotherapeutic ( anticancer ) treatments in which rapidly growing cells are selectively killed by depletion of the cellular dTTP pool . The precise mechanisms by which cells succumb to dTTP depletion are of significant interest , but have remained elusive for a long time . In the present work , we demonstrate for the first time that the effect is not specific for dTTP starvation . We show that an E . coli strain starved for the DNA precursor dGTP dies in a manner similar to dTTP-starved cells . The effect , which we have termed dGTP starvation , might be exploited - like TLD - therapeutically .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"cell",
"death",
"cellular",
"stress",
"responses",
"cell",
"cycle",
"and",
"cell",
"division",
"cell",
"processes",
"microbiology",
"nucleotides",
"escherichia",
"coli",
"mutation",
"prokaryotic",
"models",
"model",
"organisms",
"nucleic",
"acid",
"components",
"dna",
"bacterial",
"pathogens",
"research",
"and",
"analysis",
"methods",
"medical",
"microbiology",
"microbial",
"pathogens",
"biochemistry",
"cell",
"biology",
"nucleic",
"acids",
"genetics",
"biology",
"and",
"life",
"sciences",
"molecular",
"cell",
"biology"
] |
2014
|
dGTP Starvation in Escherichia coli Provides New Insights into the Thymineless-Death Phenomenon
|
Influenza A virus ( IAV ) infection is normally controlled by adaptive immune responses initiated by dendritic cells ( DCs ) . We investigated the consequences of IAV infection of human primary DCs on their ability to function as antigen-presenting cells . IAV was internalized by both myeloid DCs ( mDCs ) and plasmacytoid DCs but only mDCs supported viral replication . Although infected mDCs efficiently presented endogenous IAV antigens on MHC class II , this was not the case for presentation on MHC class I . Indeed , cross-presentation by uninfected cells of minute amounts of endocytosed , exogenous IAV was ∼300-fold more efficient than presentation of IAV antigens synthesized by infected cells and resulted in a statistically significant increase in expansion of IAV-specific CD8 T cells . Furthermore , IAV infection also impaired cross-presentation of other exogenous antigens , indicating that IAV infection broadly attenuates presentation on MHC class I molecules . Our results suggest that cross-presentation by uninfected mDCs is a preferred mechanism of antigen-presentation for the activation and expansion of CD8 T cells during IAV infection .
Influenza A virus ( IAV ) infection is one of the oldest and most common diseases known to mankind , estimated to cause 500 , 000 deaths per year , primarily in infants and elderly [1] . In healthy humans , IAV infection typically causes brief but often severe illness . Normally , IAV infection is confined to the airways where the virus replicates in respiratory epithelial cells [2] . Rapidly , alveolar macrophages produce pro-inflammatory cytokines and chemokines , which promote infiltration of peripheral blood leukocytes to the site of infection [3] . While influx of neutrophils and secretion of cytokines and chemokines in the lung is a fundamental defense during the initial stage of infection , the resulting “cytokine storm” may also contribute to pathogenesis [4] . However , control and clearance of IAV infection depend on pathogen-specific adaptive immune responses [5] . The initiation of adaptive immunity relies on dendritic cells ( DCs ) , professional antigen-presenting cells ( APCs ) with the capacity to activate naïve T cells [6] . The two major subsets of human DCs , myeloid and plasmacytoid DCs ( mDCs and pDCs , respectively ) both have antigen-presenting capacity although mDCs are generally considered to be superior . pDCs are of central importance in virus infections since they respond rapidly to viruses and secrete high levels of anti-viral type I interferons [7] . DCs reside in the epithelia of the upper respiratory tract , the site of entry for IAV , and are also rapidly mobilized to this site following inhalation of microbial agents [8]–[9] . Since there is little evidence of viral replication in lymphoid tissue , the main source of IAV antigen is thought to be DCs that exit the respiratory tract and travel to lymphoid tissue where immune responses are initiated [10]–[11] . During acute viral infections , activation and expansion of antigen-specific CD8 T cells are crucial for control and clearance of infection [5] , [12]–[13] . In general , MHC class I molecules ( MHCI ) present peptides derived from endogenously synthesized proteins . Viruses that replicate in DCs can therefore be detected by the immune system by direct presentation of viral antigens . Since not all viruses infect DCs , antigen-presentation by uninfected DCs is thought to occur via cross-presentation , a poorly understood process unique to DCs where exogenous antigen is loaded on MHCI in the ER or possibly other intracellular compartments [14] . Mice lacking CD8α+ DCs are deficient in their capacity to mount an anti-viral immune response [15] suggesting that cross-presentation is crucial for a CD8 T cell response against viruses . On the other hand , it could also suggest that direct presentation by virus infected CD8α+ DCs is required for CD8 T cell responses . The relative contributions of direct versus cross-presentation for the induction of anti-viral CD8 T cell responses have been a topic of discussion for several years [16]–[19] , and have been compared in mouse models [19]–[22] . However , the efficiency of direct versus cross-presentation of IAV and the potential IAV-mediated suppression of antigen-presentation in human DCs remains an unresolved topic . IAV infection often predisposes individuals to secondary infections , usually bacterial , with higher lethal outcomes than either infection alone , suggesting that the initial infection affects the host's ability to respond to a second pathogen . While the connections between viral and bacterial infections have been known for decades [23] , the mechanism ( s ) and modes of interaction contributing to these effects are poorly understood . IAV infection clearly suppresses innate immune responses [24]–[27] , but the extent to which adaptive immune responses are also affected , and why , remains unclear . While IAV infection has been studied extensively in animal models , relatively little is known about how IAV infection affects the function of human DC subsets due in part to their limited availability , making such experiments challenging . In mice , mDCs , rather than pDCs , appear to be responsible for presenting IAV antigen to CD8 T cells for the priming of anti-IAV immune responses [28]–[31] . In humans , it is much less clear what subset ( s ) of DCs are important in antigen-presentation during IAV infection . Human DCs infected with infectious IAV or exposed to inactivated IAV can activate IAV-specific T cells [32]–[36] , however , it remains unclear if or how IAV infection of human DCs affects their function . Here , we investigated the consequences of IAV infection on the ability of DCs to present IAV antigen , or other antigens , to autologous T cells using relevant subsets of primary human DCs .
pDCs are known to be more resistant to the cytopathic effect of IAV than mDCs , suggesting that pDCs are resistant to infection [36] , [37] . To extend this observation and to determine any possible consequences for antigen-presentation , primary human mDCs and pDCs were exposed to IAV and the frequency of IAV+ DCs was analyzed . While the frequency of IAV+ mDCs increased over time , infection in pDCs remained undetectable ( Figure 1A ) . The IAV+ mDCs reflected the production of newly synthesized viral proteins rather than enhanced virion uptake since adding the virus at 4°C , or blocking virus endosomal egress with NH4Cl , inhibited the appearance of IAV+ DCs ( Figure 1B ) . Interestingly , infectious IAV was not detected in the supernatant even after 24 hr , indicating that despite high viral protein production , mDCs did not support generation of infectious particles ( Figure S1A ) . These observations were confirmed for several IAV strains ( Figure S1B ) . To assess whether the lack of infection in pDCs reflected poor endocytosis of virus , human mDCs and pDCs were exposed to IAV and analyzed by confocal microscopy . After 1 hr , the majority of both DC subsets displayed internalized virus ( Figures 1C–D ) , suggesting that other factors blocked pDC infection , such as the pDCs' constitutive expression of the interferon-inducible antiviral protein MxA ( Figures 1E and S2 ) [37] . It has previously been shown that expression of MxA renders cells resistant to IAV infection [38] . We were unable to knockdown MxA in pDCs using siRNA while maintaining pDC viability ( data not shown ) . Still , the constitutive high expression of MxA in pDCs suggests that this protein could aid in the observed resistance to IAV infection , despite efficient IAV internalization by pDCs . pDCs , like mDCs , were nevertheless found to respond to the presence of IAV even in the absence of the synthesis of virus-encoded proteins . This was illustrated by comparing the ability of infectious IAV and non-infectious heat-inactivated ( HI ) IAV to trigger DC maturation . Both replicating and HI IAV were internalized equivalently . In addition , both could fuse with the endosomal membrane at low pH , as indicated by agglutination and acid-dependent lysis of chicken red blood cells ( data not shown ) . As expected , HI IAV did not infect DCs , and infection of mDCs by replication-competent IAV was blocked by NH4Cl ( Figures 2A–B ) . Yet , both pDCs and mDCs upregulated MHCI and MHCII in response to infectious and HI IAV ( Figure 2C ) . Furthermore , pDCs responded by secreting large amounts of IFNα ( Figure 2D ) . This was true for several IAV strains ( Figure S3 ) . mDCs also secreted IFNα in response to IAV , although the levels were 100–1000 fold lower than for pDCs ( Figure 2E ) . pDCs recognize IAV via TLR [39] while most cells respond to single stranded RNA viruses via the RIG-I pathway [40] . Human primary mDCs express TLR7 and TLR8 that recognize single-stranded RNA . The virus-related TLRs can be stimulated by inactivated viruses , however , the barely detectable amount of IFNα secreted by mDCs in response to TLR7/8L suggests that this was a consequence of signaling via cytoplasmic receptors rather than via TLRs ( Figure 2E ) . In addition , pDCs secreted TNFα , IL-6 and MIP-1α in response to IAV , while mDCs required stimulation with purified TLR7/8 ligand ( TLR7/8L ) to secrete significant amounts of cytokines and chemokines ( Figure S4 ) . Thus , IAV enters both mDCs and pDCs and triggers their maturation , but only mDCs support viral protein synthesis . To determine if infected human mDCs and pDCs could present antigen to and activate IAV-specific CD8 T cells , we exposed DCs from HLA-A2+ donors to either infectious IAV or non-infectious HI IAV and co-cultured them with autologous CFSE-labeled CD8 T cells . After 10 days , the frequency of memory CD8 T cells specific to the immunodominant influenza M1 ( 58–66 ) epitope was determined and the overall CD8 T cell response assessed by CFSE dilution . While both mDCs and pDCs could expand IAV-specific CD8 T cells , mDCs were superior ( Figures 3A–B ) . This difference likely reflected different capacities for antigen-processing since both subsets presented pre-processed peptide ( that does not require cellular processing ) similarly ( Figure 3C ) . IAV presentation to CD8 T cells by pDCs was the same for both infectious and non-infectious virus , strongly suggesting that pDCs were only capable of cross-presenting IAV antigens , albeit inefficiently from virions internalized by endocytosis . Presentation on MHCI is a hallmark of viral immunity since infected cells express virus-derived peptides recognized for elimination by cytotoxic CD8 T cells . In DCs , it is unclear if the generation of CD8 T cell responses reflects the formation of peptide-MHCI complexes from endogenously synthesized viral proteins or the cross-presentation of antigens from exogenous sources ( e . g . internalized virions or infected cells ) [41] . Indeed , DCs are well known to have an enhanced capacity for cross-presentation , which requires that internalized antigens exit the endosomal pathway for peptide cleavage in the cytosol and subsequent loading onto MHCI molecules in the ER or elsewhere [14] . Although HI IAV cannot infect DCs , it retains the capacity to fuse with the endosomal membrane [32] thus providing an intrinsic capacity to reach the cytosol , which is possibly the rate-limiting step in the cross-presentation [42] . Strikingly , mDCs exposed to HI IAV induced a statistically significant , two-fold more effective expansion of IAV-specific CD8 T cells than mDCs infected with IAV ( Figures 3A–B ) . This was surprising because it is generally assumed that presentation of peptides from endogenously synthesized proteins is more efficient than cross-presentation [19] . In addition , IAV infected DCs expressed far greater amounts of IAV proteins than DCs exposed to HI IAV . This was readily apparent ( Figures 3D and 2A ) ; although infected cells stained heavily for IAV proteins , cells exposed to HI IAV contained only 4–5 virions ( defined as IAV+ puncta ) per cell ( Figure S5 ) . We determined the mean fluorescence intensity ( MFI ) of the IAV staining in IAV+ mDCs and found that mDCs infected with infectious IAV displayed 10-fold more IAV staining than mDCs exposed to HI IAV ( Figure 3E ) . To more accurately determine the relative content of IAV proteins , we next analyzed cell lysates of mDCs exposed to infectious or HI IAV by Western blot . Due to continued synthesis in the infected cells , and continued degradation of virions in the HI IAV exposed cells , a quantitative comparison was no more than an estimate . Yet , influenza proteins were present at 100–1000 fold higher amounts in infected cells as compared to cells that had internalized HI IAV ( Figure 3F ) . The 28 kDa band most likely corresponds to the M1 protein , ∼3000 copies of which are contained within each virion , or 15 , 000 copies per cell after endocytosis of 5 virions . Assuming all of the HI IAV fuse with the endosome membrane releasing all of the incoming M1 into the cytosol , we estimate that the uninfected DCs are at least 300-fold more efficient at stimulating M1-specific CD8 T cells than infected DCs: i . e . despite vastly greater amounts of cytosolic M1 , IAV infected DCs process and present M1-derived peptides to CD8 T cells less well . This difference in antigen processing and presentation translates into the observed difference in frequency of proliferating IAV-specific CD8 T cells depicted in Figure 3B . On the other hand , pDCs exposed to either infectious IAV or HI IAV were comparable in their ability to expand IAV-specific CD8 T cells ( Figures 3A–B ) . Comparing presentation of HI IAV by mDC and pDCs , pDCs were 10–20 fold less effective at cross-presentation than mDCs ( Figures 3A–B ) . This difference was consistent over a range of IAV concentrations and DC∶T cell ratios ( Figure S6 ) . Contrary to CD8 T cell responses , CD4 T cells responded comparably to mDCs exposed to either infectious or HI IAV . As with presentation on MHCI , mDCs were superior to pDCs for MHCII-restricted presentation ( Figure 4 ) . Thus , IAV infection did not diminish the efficiency of presentation to CD4 T cells , showing that the effect on the MHCI pathway was selective . IAV infection often predisposes individuals to secondary infections , suggesting that infection history affects the ability to mount adaptive responses to new pathogens . Therefore , we investigated if uninfected and IAV infected DCs were comparable in their ability to present a second antigen to CD8 T cells and support their activation and expansion . The most common secondary infection in IAV infected individuals is Streptococcus pneumoniae . However , in the absence of tools to look at potential T cell responses against S . pneumoniae , we made use of existing immunodominant CD8 T cell memory responses against EBV and CMV in HLA-A2+ donors as model antigens . Since pDCs were not susceptible to IAV infection , we focused on mDCs . Uninfected and HI IAV exposed mDCs had similar capacities to cross-present CMV to antigen-specific CD8 T memory cells . In contrast , IAV infected mDCs consistently produced several-fold lower frequency of proliferating CMV pp65-specific CD8 T cells ( Figure 5A ) . The impaired ability of IAV infected mDCs to cross-present inactivated CMV to CD8 T cells was apparent over a range of CMV concentrations ( Figure 5B ) and DC∶T cell ratios ( Figure 5C ) . Analysis of the CMV pentamer-negative , CFSElow population in the absence of exogenous CMV also showed that the overall T cell response was more pronounced to HI IAV ( 48 . 6% ) than to infectious IAV ( 15 . 9% ) ( Figure 5A , left panel ) . Furthermore , IAV infected and HI IAV stimulated mDCs loaded with pre-processed CMV peptide were comparable or superior to uninfected mDCs in their ability to expand CMV-specific CD8 T cells , consistent with a defect in antigen-processing capacity rather than in antigen-presentation ( Figure 5D ) . Similar results were observed for cross-presentation of HI EBV or EBV infected cell extract ( Figure S7 ) . This suggests that IAV infected mDCs have an impaired capacity to cross-present both different sources of antigen ( CMV and EBV ) as well as different forms of antigen ( inactivated virus and virus infected cells ) to CD8 T cells as compared to uninfected mDCs . We also compared the ability of uninfected , IAV infected , and HI IAV stimulated mDCs to present CMV to autologous CD4 T cells . Unlike cross-presentation to CD8 T cells , IAV infected and HI IAV exposed mDCs stimulated CMV-specific CD4 responses similarly ( Figure 5E ) . This again indicates that IAV infected mDCs can function as APCs in general but that IAV infection selectively impairs the ability to cross-present antigen on MHCI to CD8 T cells . One explanation for the decreased ability of IAV infected mDCs to cross-present could be IAV-induced DC death . This explanation appeared unlikely since presentation of pre-processed peptide was similar between uninfected and IAV infected mDCs ( Figure 5D ) , and IAV infected mDCs could activate CD4 T cells ( Figures 4A and 5A ) . Assessing the viability of the mDCs in the co-cultures after 10 days , the time at which we measure T cell activation , is a challenge since DCs are in great minority and the majority of T cells that have not seen their cognate antigen have died . To investigate the role of IAV mediated cell death , mDCs were exposed to infectious IAV or HI IAV and the frequency of dead mDCs was compared to untreated mDCs by Annexin V staining . The viability of all mDCs was comparable at 2 hr and 6 hr after virus exposure ( Figure 6A ) . At later time points ( days ) , the viability of IAV infected mDCs was reduced compared to untreated or HI IAV exposed mDCs ( Figure 6A ) , in line with the cytopathic effects of IAV and similar to what has been described by others [36]–[37] . However , as most of the antigen processing and presentation to CD8 T cells probably occurs within the first 24 hr of antigen capture [43] , these data suggest that IAV induced cell death alone is not a likely explanation for the difference in cross-presentation observed . Another way to address the role of DC viability in the observed difference in cross-presentation is to stop antigen processing after 8 hr by fixing the DCs with paraformaldehyde ( PFA ) and then analyze CD8 T cell expansion . Since fixed cells do not present antigen as effectively as viable cells , the overall frequency of expanded CMV pp65-specific CD8 T cells was lower than in the cultures with live DCs ( Figure 6B ) . Nevertheless , the relative difference remained the same: fixed IAV infected mDCs were poorer at expanding CD8 T cells than fixed uninfected mDCs . In contrast , fixed IAV infected mDCs loaded with pre-processed peptide were better than fixed uninfected mDCs at expanding CD8 T cells ( Figure 6C ) . This is in agreement with IAV induced mDC maturation and MHCI upregulation ( Figure 2C ) . Another explanation for the reduced ability of IAV infected mDCs to expand CMV-specific CD8 T cells could be poorer uptake of CMV particles . To address this possibility , we exposed uninfected , IAV infected and HI IAV stimulated mDCs to inactivated CMV particles and quantified the frequency of CMV-containing DCs by immunofluorescence ( Figure 7A ) . The frequency of mDCs containing CMV particles was very similar over a range of CMV doses , irrespective of IAV infectivity ( Figure 7B ) . We counted the number of CMV pp65-positive puncta per mDC in each condition . Again , uninfected , IAV infected and HI IAV stimulated mDCs were comparable , although there was a slight tendency for IAV infected mDCs to exhibit a higher average number of CMV pp65+ particles per cell ( Figure 7C ) . These results suggested that there was no difference in antigen load between uninfected and IAV infected mDCs . Taken together , we conclude the reduced ability of IAV infected mDCs to cross-present likely depends on interference downstream of antigen uptake induced by viral replication and independent of DC maturation .
We have investigated the susceptibility of human primary DC subsets to IAV infection , including the functional consequences of IAV infection on their ability to process and present antigens to activate T cells . Although both mDCs and pDCs could internalize IAV , only mDCs supported IAV protein synthesis . Yet , IAV infected mDCs were less efficient at stimulating CD8 T cell responses in vitro when compared to uninfected mDCs . Indeed , cross-presentation of exogenous IAV M1 derived from internalized HI IAV was ∼300-fold more efficient when normalized to the total amount of mDC-associated antigen in infected vs . uninfected DCs . In addition , IAV infected mDCs also had a reduced capacity to present a second exogenous antigen , either inactivated CMV or EBV virions or infected cells , to CD8 T cells as compared to uninfected mDCs . Furthermore , we found that mDCs were more efficient at activating and expanding IAV-specific CD8 T cells than pDCs after exposure to either infectious or HI IAV . Although previous studies have demonstrated that human DCs present infectious and inactivated IAV to CD4 and CD8 T cells [32]–[33] , [36] , [44] , the relative efficiency of these two processes and the functional implication of any difference have not been studied in detail . In mouse models , the importance of CD8α+ DCs and cross-presentation in control of virus infection is well-established [15] . Furthermore , different virus infections in mouse including vaccinia [20] , MCMV [22] and HSV-1 [21] have been reported to rely on cross-presentation rather than direct presentation for CD8 T cell responses and clearance of infection , although this conclusion has recently been challenged [19] . Our observation that cross-presentation can be more efficient than direct presentation on MHCI to CD8 T cells also in a human system extends earlier findings and underscores their relevance . The fusogenic capacity of IAV is required for optimal CD8 T cell activation [32] , and may be crucial for efficient cross-presentation , since it likely removes the rate limiting step of endosomal egress . Our work further shows that IAV infected mDCs are impaired in their ability to cross-present a second antigen to CD8 T cells when compared to uninfected mDCs or mDCs exposed to replication-incompetent HI IAV . Importantly , it seems that replicating IAV , rather than the mere presence of IAV and subsequent DC maturation , affects the ability of mDCs to process and present antigen on MHCI for activation and expansion of CD8 T cells . To pinpoint which viral protein ( s ) and what specific cellular target ( s ) is affected will be key in future studies . We have also demonstrated that in physiologically relevant DCs , cross-presentation can provide a more effective strategy for CD8 T cell stimulation than presentation of endogenous antigen by infected DCs , similar to what has been observed for other viruses in mouse experimental models [20]–[22] . Thus , not only is there no need for DCs to be infected by the viruses whose antigens they present , but also infection may actually suppress the initiation of adaptive responses . In future studies , it will be important to verify the relevance of our data using human lung DCs and/or in clinical studies that take into account the complex interaction of cells during IAV infection in vivo . We found that pDCs were resistant to IAV infection despite significant virus internalization , confirming and extending previous reports [33]–[34] , [36]–[37] , [45]–[46] . pDCs responded to IAV exposure by secreting large amounts of IFNα , but showed only modest upregulation of co-stimulatory molecules compared to TLR7/8L stimulation . In our hands , pDCs were less potent than mDCs at inducing CD8 T cell activation after acquiring a large antigen that requires processing into peptides before loading onto MHCI . This was probably due to a lower expression of co-stimulatory molecules and therefore weaker DC-T cell interaction and/or a reduced capacity to process large antigens compared to mDCs , rather than a lack of viral antigen available for presentation , as pDCs carrying abundant IAV NP are unable to activate IAV-specific T cells [30] . Previous studies have reported that human pDCs are similar [36] or superior [35] in their ability to present antigen to CD8 T cells as compared to human mDCs . The lack of clear consensus may partly be explained by differences in maturation/phenotype of the pDCs as well as length of exposure , and dose of IAV , as it has recently been reported that the timing of pDC stimulation and route of antigen uptake affect the ability of pDCs to present antigens [47] . It is well documented that IAV infection renders infected individuals more prone to secondary bacterial infections , but the importance of CD8 T cell response to control and clear extracellular bacterial infections is unclear . IAV infection has an immunomodulatory effect that is thought to promote an increased susceptibility to secondary infections [24] , [26] . The impact of IAV induced immunomodulation combined with an impaired ability to cross-present subsequently encountered antigens might act together to compromise a proper immune response to secondary pathogens . Systemic injection of TLR ligands results in reduced cross-presentation of a subsequently encountered antigen [18] . While this was suggested to be a consequence of systemic DC maturation , we recently showed that reduced antigen-presentation in vivo after systemic TLR injection could also be a consequence of the antigen not reaching DCs in the spleen due to alterations in splenic blood flow [48] . Previous studies using monocyte-derived DCs have shown that IAV infection induces suboptimal maturation of the cells with respect to upregulation of co-stimulatory molecules and secretion of cytokines as compared to LPS stimulation [49] . Using recombinant IAV that did not encode the multifunctional viral protein NS1 , the authors found that NS1 has an inhibitory effect on expression of several genes involved in monocyte-derived DC maturation and migration , including the pro-inflammatory cytokines IL-6 and TNFα [49] . In our hands , primary mDCs do not show a defect in their ability to upregulate co-stimulatory molecules in response to IAV as compared to TLR stimulation , but did indeed show lower secretion of pro-inflammatory cytokines . In any event , altered DC maturation seems unlikely to fully explain the defect in mDC cross presentation of a second antigen , since mDCs stimulated with HI IAV , which mature to the same extent as IAV infected mDCs , were found to retain their ability to cross-present . In future studies , it will be important to use recombinant IAV strains in which different viral proteins have been mutated or deleted to study their potential impact on DC maturation and antigen-presentation on the protein level , as well as in functional assays as outlined in the present study . Finally , our findings shed light on how uninfected human DCs , rather than IAV infected DCs , may be crucial for processing and presentation of IAV antigen to initiate anti-viral immunity . Even if infected DCs can present antigen , cross-presentation facilitated by uninfected DCs may be sufficient or even required for induction of anti-viral immune responses . While the in vivo situation is much more complicated , the in vitro results presented here do create the conceptual possibility that the same situation applies in vivo . Indeed , it was unexpected that uninfected mDCs cross-present viral antigens more efficiently than IAV infected DCs present endogenously synthesized antigens . Thus , IAV infection not only inhibits cross-presentation of subsequently encountered antigens , but also acts to diminish direct presentation . As a result , it is now of interest to determine the mechanism of both forms of inhibition . Besides DC death , other potential contributors may include the partial reduction in host cell protein synthesis following IAV infection or a direct inactivation of the antigen processing machinery , as observed for medium to large DNA viruses that cause chronic infections [50]–[51] . As discussed above , the multifunctional IAV protein NS1 is an important virulence factor associated with the suppression of innate immunity [52]–[54] . The major function of NS1 is to antagonize the type I IFN mediated host response . Current evidence suggests that NS1 can limit IFNβ production both on the pre- and post-transcriptional level . While most IAV strains can utilize both strategies , some strains may have lost one of these mechanisms naturally or as a consequence of passage in the laboratory [53] . NS1 not only prevents the activation of IRF3 , a transcription factor involved in IFNβ induction ( pre-transcriptional ) , but can also block the expression of cellular genes such as MxA at the post-transcriptional level , and thereby IFN gene expression . In contrast to more recent human strains of IAV like A/TX/91 ( TX ) , NS1 expressed by A/PR/8 ( PR8 ) , a widely used laboratory IAV strain , can only limit pre-transcriptional events of IFNβ induction [54] . Monocyte-derived DCs infected with IAV/TX displayed higher viral replication but reduced capacity to induce IFNγ secretion in allogeneic naive CD4 T cells compared to monocyte-derived DCs infected with IAV/PR8 [55] . Monocyte-derived DCs infected with NS1 deleted versions of the two virus strains were comparable in their ability to induce IFNγ secretion in allogeneic naive CD4 T cells [55] , suggesting that a more recent human isolate of IAV ( TX ) is a more potent inhibitor of DC function than a laboratory adapted strain ( PR8 ) . In addition , recent data using human lung epithelial cells indicate that NS1 specifically suppresses the expression of several genes involved in IFN-stimulated MHCI antigen presentation and immune-proteasome activation during IAV infection [56] . Another potential viral protein to consider in this context is the most recently discovered IAV protein , PB1-F2 [57] . PB1-F2 is a virulence factor described to contribute to pathogenesis of IAV as well as secondary bacterial infections [58]–[60] . Taken together , these studies have contributed significantly to our initial understanding of how individual IAV proteins may impact the immune response to IAV and they also highlight the importance of studying a wider selection of IAV strains . Whether NS1 and/or PB1-F2 also affect the ability of IAV infected primary DCs to cross-present is a relevant question that merits further investigation . A deeper understanding of how IAV infection of human DCs impairs their function may prove to be useful for improved vaccine design or therapeutic approaches to enhance endogenous responses .
This study was approved by the Genentech Institutional Review Board . Written informed consent was obtained from all human participants . Our procedures for isolation of subsets of DCs and T cells from blood have been described previously [61] . Briefly , healthy blood donors underwent automated leukapheresis and enriched populations of lymphocytes and monocytes were obtained by counterflow centrifugal elutriation . DCs were isolated from elutriated monocytes using magnetic bead isolation followed by sequential separation on AutoMacs ( Miltenyi Biotec ) . The BDCA-4 and the CD1c isolation kits were used for isolation of pDCs and mDCs , respectively . pDCs and mDCs were cultured at 1×106 cells/ml in complete medium ( RPMI 1640 Glutamax supplemented with 1% streptomycin and penicillin , 1% HEPES ( all Invitrogen ) , 10% fetal bovine serum ( Gibco ) ) in the presence of recombinant human IL-3 ( 10 ng/ml , R&D Systems ) or GM-CSF ( 2 ng/ml , PeproTech ) . T cells were isolated from elutriated lymphocytes by negative selection and separation on AutoMacs . T cells were cultured at 10×106 cells/ml in complete medium and rested overnight before use . Influenza A/NWS/33 and Influenza A/PR/8/34 strains ( ATCC ) were propagated in MDCK cells . Supernatants were concentrated by ultracentrifugation and resuspended in RPMI . Influenza A/X31 was propagated in chicken eggs , purified and concentrated on sucrose gradients ( Virapur ) . Mock infected supernatants and allantoic fluid were processed in the same manner and used as controls to exclude any non-specific activation of DCs ( data not shown ) . TCID50 for all IAV strains was determined by infecting a light monolayer of MDCKs in the presence of trypsin and monitoring the cytopathic effect . DCs were infected with 600 , 000 infectious particles ( assessed in MDCK plaque assay ) of IAV per 1 , 000 , 000 DCs ( 0 . 6 MOI ) . This dose of IAV resulted in 50–95% IAV+ mDCs after 24 hr of exposure . Virus was replication incompetent after heat-inactivation at 56°C for 30 min . Unless otherwise stated in the text , IAV refers to IAV/X31 . DCs were exposed to IAV , washed twice in RPMI and infection was monitored using an anti-IAV rabbit polyclonal ( Pinda , Dr . Ari Helenius , ETH Zurich , Switzerland ) or anti-nucleoprotein ( NP ) antibody ( clone A3 , Chemicon ) and flow cytometry ( FACSCanto II , BD Biosciences ) . Alternatively , infected DCs were allowed to adhere to alcian blue ( Sigma ) coated glass coverslips for 20 min at 37°C , fixed with 4% paraformaldehyde ( PFA ) ( Electron microscopy sciences ) for 20 min at room temperature and permeabilized with 0 . 05% saponin ( Sigma ) , stained with antibodies and analyzed by immunofluorescence confocal microscopy ( Leica TCS SP5 , Leica Microsystems ) . To prevent IAV infection , 20 mM NH4Cl was added before IAV . After IAV infection , DCs were harvested , washed twice and surface stained with antibodies against ( CD14 ( MφP9 ) , CD11c ( B-ly6 ) , CD123 ( 9F5 ) , CD86 ( FUN-1 ) , CD40 ( 5C3 ) , HLA-ABC ( W6/32 ) all BD Biosciences ) or HLA-DR ( L243 , Biolegend ) . DCs were washed , fixed and analyzed by flow cytometry . Supernatants were harvested and cytokines were measured by ELISA ( IFNα; PBL Interferon Source ) or Luminex ( Biorad ) . MxA expression was determined using a mouse anti-MxA monoclonal antibody ( clone M143 , Dr . Otto Haller , University of Freiburg , Germany ) and flow cytometry or immunofluorescence confocal microscopy . After 4 hr of IAV exposure , DCs were washed and co-cultured with autologous CD4 T cells at different DC∶T cell ratios . After 1 hr , GolgiPlug containing Brefeldin A ( BD Biosciences ) was added and the cells were further incubated overnight . Cells were harvested and stained with surface antibodies against CD4 ( SK3 ) , CD3 ( SK7 ) , CD8 ( SK1 ) , CD14 ( all BD Biosciences ) and HLA-DR , followed by fixation and permeabilization for 10 min using BD cytofix/cytoperm ( BD Biosciences ) . Cells were stained intracellularly with antibodies against IFNγ ( B27 , BD ) , TNFα ( MAb11 , BD ) and IL-2 ( MQ1-17H12 , Caltag laboratories ) and analyzed by flow cytometry . DCs isolated from HLA-A2+ donors were exposed to IAV or loaded with 0 . 25–250 ng/mL pre-processed peptide for 4 hr , washed and co-cultured with autologous CD8 T cells labeled with 0 . 25 µM CFSE ( Molecular Probes ) . As a positive control , the TCR superantigen Staphylococcal enterotoxin B ( 1 µg/ml , Sigma ) was used . HLA-A2 restricted HIV-1 gag pre-processed peptide ( SLYNTVATL ) was used as an irrelevant pre-processed peptide control ( ProImmune ) . After 10 days , cells were harvested and stained with HLA-A2 Influenza M1 ( GILGFVFTL ) pentamer ( ProImmune ) for 15 min at room temperature followed by labeling with antibodies against CD3 , CD8 , CD14 , CD19 ( SJ25C1 ) , CD11c ( B-ly6 ) ( BD Biosciences ) , fixation and analysis by flow cytometry . HLA-A2+ mDCs were exposed to IAV for 4 hr , washed and loaded with 7–700 µg/mL of total protein whole , inactivated CMV ( Microbix ) or 0 . 25–250 ng/mL pre-processed HLA-A2 CMV pp65 peptide ( NLVPMVATV ) for an additional 3 hr . DCs were washed and co-cultured with autologous CD8 T cells labeled with CFSE . After 10 days , cells were harvested and stained with HLA-A2 CMV pp65 ( NLVPMVATV ) pentamer ( ProImmune ) followed by labeling with antibodies against CD3 , CD8 , CD19 , CD11c , CD14 , fixation and analysis by flow cytometry . Alternatively , mDCs were loaded with 200 µg/mL total protein from whole , heat-inactivated EBV ( Virusys ) or 200 µg/mL total protein cell extract from EBV infected or control cells ( Virusys ) or 0 . 25–250 ng/mL pre-processed HLA-A2 EBV BMLF-1 peptide ( GLCTLVAML ) for 3 hr , washed and co-cultured with CD8 T cells . After 10 days , cells were harvested and stained with HLA-A2 EBV BMLF-1 ( GLCTLVAML ) pentamer ( ProImmune ) and surface antibodies as described above . The CMV and EBV antigen preparations were titrated to find a dose that was not toxic to the cells yet adequate to activate memory T cells . After 4 hr IAV exposure , mDCs were washed and pulsed with 7–700 µg/mL of total protein from whole , inactivated CMV or overlapping pre-processed peptides to CMV pp65 , 15-mers overlapping by 11 ( 2 . 5 µg of peptide per mL , ProImmune ) for 3 hr . DCs were washed and co-cultured with autologous CD4 T cells at a 1∶30 DC∶T cell ratio . After 2 hr , GolgiPlug was added and the cells were incubated overnight . Cells were harvested and stained with surface antibodies against CD4 , CD3 , HLA-DR , CD14 , CD8 , followed by fixation and permeabilization . Cells were subsequently stained with antibodies against IFNγ , TNFα and IL-2 , and analyzed by flow cytometry . After 8 hr of IAV exposure , mDCs were harvested and lysed in SDS lysis buffer ( 1% SDS , 20 mM Tris pH 7 . 5 and protease inhibitors ( Roche ) ) . DNA was shed mechanically and lysates were snap frozen on dry ice . Lysates were run on a 4–12% Bis-Tris reducing gel , transferred to a PVDF membrane and blotted for viral proteins with the anti-IAV polyclonal Pinda . GAPDH was used as loading control . mDCs were exposed to infectious IAV or HI IAV or left untreated . DCs were harvested , washed twice in ice-cold PBS , resuspended in 1× binding buffer and stained with Annexin V and propidium iodide ( BD Biosciences ) and analyzed by flow cytometry within one hour of processing . After 4 hr of IAV exposure , DCs were washed and pulsed with 7–700 µg/mL of total protein whole , inactivated CMV ( Microbix ) for 3 hr . DCs were washed twice in complete medium , adhered to coverslips , fixed and permeabilized . DCs were stained with antibodies against IAV ( Pinda ) , CMV pp65 ( clones 2+6 , Leica ) and HLA-DR and mounted with Prolong Gold containing DAPI ( Molecular Probes ) . Samples were analyzed by immunofluorescence confocal microscopy . Statistical significance was assessed using paired t test and considered significant at P value less than 0 . 05 .
|
Although the interactions between viruses and dendritic cells ( DCs ) have been studied for many years , surprisingly little is known on the functional relationship between infection and antigen presentation in primary human DCs . Here , we asked specifically whether Influenza A virus ( IAV ) infection of human primary plasmacytoid DCs and myeloid DCs ( pDCs and mDCs , respectively ) affected their ability to function as antigen-presenting cells and activate T cells specific to IAV or other antigens . Our data confirm that pDCs are poorly infected and also present IAV antigens poorly . mDCs , on the other hand , are readily susceptible to IAV infection and present IAV antigen to T cells . However , we found that MHC class I presentation by mDCs infected with IAV are ∼300-fold less efficient relative to what mDCs are capable of achieving by cross-presentation following the endocytosis of only a very few non-infectious virions . Importantly , IAV infection of mDCs not only reduces the efficiency of IAV presentation but also reduces their ability to cross-present antigens from other viruses encountered subsequently . The reduced overall antigen processing capacity of mDCs describes a mechanism that may contribute to the suppression of immunity to secondary pathogens that appear during the course of IAV infection .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"medicine",
"infectious",
"diseases",
"immune",
"cells",
"clinical",
"immunology",
"immunity",
"virology",
"viral",
"transmission",
"and",
"infection",
"immunology",
"biology",
"immunologic",
"techniques",
"microbiology",
"viral",
"diseases"
] |
2012
|
Influenza A Virus Infection of Human Primary Dendritic Cells Impairs Their Ability to Cross-Present Antigen to CD8 T Cells
|
The unconventional prefoldin URI/RMP , in humans , and its orthologue in yeast , Bud27 , have been proposed to participate in the biogenesis of the RNA polymerases . However , this role of Bud27 has not been confirmed and is poorly elucidated . Our data help clarify the mechanisms governing biogenesis of the three eukaryotic RNA pols . We show evidence that Bud27 is the first example of a protein that participates in the biogenesis of the three eukaryotic RNA polymerases and the first example of a protein modulating their assembly instead of their nuclear transport . In addition we demonstrate that the role of Bud27 in RNA pols biogenesis depends on Rpb5 . In fact , lack of BUD27 affects growth and leads to a substantial accumulation of the three RNA polymerases in the cytoplasm , defects offset by the overexpression of RPB5 . Supporting this , our data demonstrate that the lack of Bud27 affects the correct assembly of Rpb5 and Rpb6 to the three RNA polymerases , suggesting that this process occurs in the cytoplasm and is a required step prior to nuclear import . Also , our data support the view that Rpb5 and Rpb6 assemble somewhat later than the rest of the complexes . Furthermore , Bud27 Rpb5-binding but not PFD-binding domain is necessary for RNA polymerases biogenesis . In agreement , we also demonstrate genetic interactions between BUD27 , RPB5 , and RPB6 . Bud27 shuttles between the nucleus and the cytoplasm in an Xpo1-independent manner , and also independently of microtubule polarization and possibly independently of its association with the RNA pols . Our data also suggest that the role of Bud27 in RNA pols biogenesis is independent of the chaperone prefoldin ( PFD ) complex and of Iwr1 . Finally , the role of URI seems to be conserved in humans , suggesting conserved mechanisms in RNA pols biogenesis .
Eukaryotic RNA polymerases are a family of multimeric enzymes , RNA pol I , II , and III , responsible for the specific synthesis of different RNAs . RNA pol I is specialized in the synthesis of the pre-rRNA precursor of the three largest rRNA and typically account for about 75% of the entire transcription output in fast-growing yeast cells . RNA pol III transcribes mostly tRNAs and 5S rRNA , together with several short non-translated RNAs , while transcription corresponds to about 15% of the total RNA . RNA pol II , the enzyme that produces all mRNAs and many non-coding ones , transcribes most of the nuclear genome but nevertheless contributes to less than 10% of total RNA in growing cells . RNA pol I , II , and III are composed of 14 , 12 , and 17 subunits respectively , with a catalytic core formed by the two largest subunits highly conserved through evolution and five common subunits to the three enzymes [1]–[3] . Despite intensive studies concerning the structure and the transcriptional regulation of the three RNA polymerases [4] , [5] , little is known about the mechanisms governing their assembly and their nuclear import . Noteworthy findings in both human and yeast demonstrate the participation of different proteins in the transport of the RNA pol II to the nucleus , Iwr1 and Npa3 in yeast , and GPN1 ( RPAP4 ) and GPN3 in humans [6]–[10] . It has also been suggested that RPAP2 plays a role in import on the basis that it is cytoplasmic , binds fully assembled enzyme and shuttles in a CRM1-dependent manner [11] . However , no data concerning proteins involved in the nuclear transport of the RNA pol I or III are available . In addition , proteomic analysis in humans cells seek to decipher the mechanisms of RNA pol II biogenesis and assembly identifying a number of polymerase-associated factors . Among these , HSP90 and its R2TP/Prefoldin-like chaperone , including hSpagh ( RPAP3 ) , are clearly involved in these processes [8] , [12] . In humans , R2TP/Prefoldin-like complex contains Rpb5 , a common subunit to the three eukaryotic RNA polymerases [2] , as well as the unconventional prefoldin Rpb5 interactor ( URI/RMP ) , a member of the prefoldin ( PFD ) family of ATP-independent molecular chaperones [8] , [13] . URI physically binds Rpb5 , other nuclear proteins involved in transcription , including the general transcription factor TFIIF [14]–[16] and components of the Paf-1 complex that promotes RNA pol II CTD phosphorylation and histone modification during transcription elongation [17] . Notably , its yeast homologue Bud27 also binds Rpb5 [18] . URI was originally characterized in human and yeast cells as regulator of gene expression controlled by TOR ( for target of Rapamycin ) pathway [18] . Furthermore , URI has been linked to translation initiation [19] , transcription regulation , chromatin stability or DNA damage response [13] , [20] . URI is located mainly in the cytoplasm , although nuclear and perinuclear localization has also been observed in different organisms [20]–[22] . However , in Saccharomyces cerevisiae , only a cytoplasmic localization has been detected [19] . In addition , URI is believed to function as a scaffold protein able to assemble additional members of PFD family ( through its PFD- and Rpb5-binding domains ) in both human and yeast [13] , [23] and different authors have proposed a role in the cytoplasmic assembly of the human RNA pol II [8] , [12] , [13] . Despite that some experiments , using immunoprecipitation and mass-spectrometry , have shown that URI interacts with RNA polymerases and other intermediary subcomplexes involved in RNA pol II cytoplasmic assembly , very little is known about the role of URI in RNA polymerases biogenesis . In this report , we investigate the role of the URI yeast homologue Bud27 in RNA polymerases assembly and its relationship with Rpb5 . Bud27 interacts with different subunits of the three RNA polymerases , with the common subunit Rpb5 and Rpb10 , as well as with a component of the prefoldin-like complex , Yke2 ( Pfd6 ) . Deletion of BUD27 leads to a substantial accumulation of the three RNA polymerases in the cytoplasm , a defect offset by the overexpression of RPB5 . Supporting this , our data demonstrate that the lack of Bud27 affects the correct assembly of Rpb5 and Rpb6 to the three RNA polymerases , suggesting that this process occurs in the cytoplasm and is a required step prior to nuclear import . As previously proposed in human [12] , Rpb6 appears to assemble rather late after Rpb5 assembly . In agreement , we observed a genetic interaction between BUD27 , RPB5 and RPB6 . In addition , Bud27 PFD- or Rpb5-binding domains are not necessary for RNA polymerases biogenesis . Finally , our data demonstrate that Bud27 shuttles between the cytoplasm and nucleus , as shown by the deletion of the NES domain , contrary to what has been previously described , a mechanism that is independent of the Xpo1-mediated pathway . URI silencing suggests that similar role for this protein accounts in humans .
Immunoprecipitation studies and protein identification by mass spectrometry have shown that URI interacts with components of the RNA pol II in human cells [8] , [12]–[14] . In addition , work in yeast [24] , [25] , using systematic characterization of complexes by TAP and mass spectrometry predicted the association of the URI orthologue , Bud27 , with multiple components of the RNA pol II machinery . However , only the physical interaction between Bud27 and Rpb5 in yeast has been further demonstrated and no clear data concerning the association of Bud27 with the other two RNA polymerases have been reported . To gain insights into the association of Bud27 with the three RNA pols and to identify proteins that associate with Bud27 in yeast , we used the BUD27 gene TAP-tagged at its 3′ end in TAP purifications . As shown , the Bud27-TAP cells grow normally ( Figure 1A ) . Bud27-TAP was affinity-purified from a whole-cell lysate by two consecutive affinity columns ( IgG-Sepharose and Calmodulin-Sepharose ) . After the second purification , 10 proteins were specifically enriched . It bears noting that most of these proteins are members of the three RNA polymerases: Rpa190 , Rpa135 and Rpa49 ( RNA pol I ) ; Rpc160 and Rpc128 ( pol III ) ; Rpc40 ( pol I and III ) ; Rpb1 ( pol II ) ; Rpb10 and Rpb5 ( pol I , II and III ) ( Figure 1B ) . In addition , we also identified Yke2 ( Pfd6 ) , a member of the Gim/prefoldin protein complex involved in the folding of alfa-tubulin , beta-tubulin , and actin and also a member of the RPAP3/R2TP/prefoldin-like complex participating as intermediary of the RNA pol II assembly in humans [13] , [26] . Curiously beta-tubulin and Ssb1 , a chaperone member of the HSP70 family [27] , were present in the affinity-purified Bud27 preparation . To confirm the interaction between Bud27 and the three RNA polymerases , we purified Bud27-TAP from a strain containing also Rpa190-HA tagged ( RNA pol I ) and Rpc25-Myc tagged ( RNA pol III ) proteins . As shown in Figure 1C , an Rpa190-HA reacting band was revealed . No such band was detected when the TAP purification was performed in control strain BY4741 or in control strain YFN229 containing Rpb3 TAP but also tagged forms of the other two RNA polymerases ( Rpa190-HA , and Rpc25-Myc ) , indicating that Rpa190-HA does not interact with the TAP module and that no anti-Rpa190-HA-reacting material was adsorbed nonspecifically to the beads . Similarly , an Rpc25-Myc reacting band ( by using anti-Myc antibodies ) was noted when Bud27-TAP was purified , but not in the control strain BY4741 or in a control strain containing Rpb3-TAP . Finally , no reacting band was revealed for Rpb1 , although it clearly co-purified in a control strain containing Rpb3-TAP . Thus , we cannot rule out that Bud27 and RNA pol II could associate only transiently in vivo or less efficiently . These observations indicate that interactions between Bud27 and the RNA pol I and III are specific , and also suggest that this could similarly account for RNA pol II . Our data demonstrating a physical interaction between Bud27 and the three RNA polymerases , as well as the reported localization of Bud27 in the cytoplasm [19] , suggest a role for this prefoldin in the cytoplasmic biogenesis of the RNA pols . To clarify the localization of Bud27 in the cell and to explore if this protein shuttles between cytoplasm and nucleus , as is the case in humans and Drosophila [20] , [21] , we used a functional BUD27-GFP gene fusion , cloned in a centromeric plasmid , expressed from a Tet-repressible promoter [25] . The functionality of this Bud27-GFP fusion protein was confirmed by its ability to complement the temperature sensitivity of a Δbud27 mutant strain ( Figure S1 ) . As shown in Figure 2A , Bud27-GFP was preferentially localized at the cytoplasm . A more detailed analysis of Bud27 amino acid sequence using the NetNES 1 . 1 and cNLS mapper servers [28] , [29] predicted a possible leucine-rich nuclear export signal ( NES ) between positions 686 and 695 ( LRDEIRDFQL ) and a nuclear localization signal ( NLS; amino acids 562 to 595 ) . Proteins with a NES signal are actively translocated to the cytoplasm via the action of nuclear export pathway mediated by an evolutionarily conserved CRM1/exportin protein Xpo1 , suggesting that Bud27 could shuttle between the nucleus and cytoplasm in an Xpo1-dependent manner . To verify this possibility , we examined the localization of Bud27 in xpo1-1 cells transformed with the above-mentioned plasmid after shifting from 30°C to 37°C for up to 5 h , a condition under which the Xpo1-dependent protein export was blocked [30] . However , no nuclear Bud27-GFP accumulation was found following a shift to 37°C for up to 5 h ( Figure 2B ) . On the other hand , and as the existence of NES and NLS signals in Bud27 suggest that this is a shuttling protein , we deleted the NES sequence from Bud27-GFP . Our results demonstrated that this led to a nuclear accumulation of Bud27 ( Figure 2C ) . In addition , the deletion of this sequence did not affect the ability of the fusion protein to complement the growth defect caused by the bud27 null mutation ( Figure S1 ) . To gain insight into the mechanism by which Bud27 translocates to the nucleus , we analysed Bud27ΔNES-GFP localization after adding benomyl , a drug demonstrated to promote depolarization of microtubules and accumulation or the largest subunit of the RNA pol II , Rpb1 , in the cytoplasm [8] . As expected , benomyl led to the accumulation of RNA pols in the cytoplasm , as shown by monitoring Rpb8-ECFP in vivo ( Figure 2D upper panel ) . Contrary , nuclear localization of Bud27ΔNES-GFP is not significantly altered by the addition of benomyl ( Figure 2D lower panel ) . These data together demonstrate that Bud27 shuttles between nucleus and cytoplasm and suggest that the NES sequence is required for the nuclear export of Bud27 in an Xpo1-independent manner . In addition , the data recorded using benomyl suggest that the translocation of Bud27 to the nucleus is independent of microtubule polarization . To help elucidate the function of Bud27 and of its association with the RNA pols , and to investigate the effect of Bud27 in the assembly and/or transport of these enzymes to the nucleus , we tested the hypothesis that Bud27 is needed for correct localization of the three RNA pols . The reason to suspect a role for this prefoldin in the cytoplasmic biogenesis of the three RNA pols came from the observation that Bud27 is localized mainly in the cytoplasm of yeast cells while the three RNA pols in the nucleus . Then we performed immunocytochemistry experiments in a wild-type and a Δbud27 mutant strain containing an Rpa190-HA tagged ( RNA pol I ) and Rpc160-Myc tagged ( RNA pol III ) proteins , using anti-HA , anti-Rpb1 ( 8WG16 ) and anti-Myc antibodies , to analyse the intracellular localization of the largest subunits of the RNA pol I , II , and III , respectively . In a wild-type strain ( Figure 3A ) , fluorescence for Rpb1 and Rpc160 was restricted to the nucleus , while for Rpa190 it was mainly nucleolar , indicating that , as expected , RNA pol II and III were localized in the nucleus and RNA pol I , mainly in the nucleolus . However , deletion of BUD27 resulted in the accumulation of the three RNA pols in the cytoplasm , although nuclear and nucleolar , but more diffuse , staining was also observed . To corroborate these results and to monitor the localization of another subunit shared by the three RNA polymerases , we genomically tagged Rpb8 with a C-terminal ECFP tag and we monitored its localization by live cell imaging ( Figure 3B ) . As in the case for Rpa190 , Rpb1 , and Rpc160 , fluorescence for Rpb8 was detected mainly in the nucleus of a wild-type strain , while a clear cytoplasmic accumulation was found in a Δbud27 mutant strain . These results indicate that the three RNA enzymes are partially mislocalized , according to the differences in the amount of Rpb1 associated with chromatin fractions ( Figure 3C ) . In addition , nuclear localization of the three RNA pols seems not to depend on Bud27 localization , since Bud27 variant that lacks NES sequence and that is accumulated in the nucleus did not impair the nuclear localization of the RNA pol I , II or III ( Figure S2A ) . Also , we investigated whether cells resumed growth and RNA pols nuclear localization when a Δbud27 mutant strain was complemented with a BUD27-TAP gene fusion into a centromeric plasmid expressed from a Tet-repressible promoter [25] . As expected , BUD27-TAP fully restored growth of the Δbud27 mutant strain at the restrictive temperature of 37°C ( Figure S1 ) . In addition , Rpb8-ECFP signal was again restricted to the nucleus , suggesting that the three RNA pols again became nuclear ( Figure S2B ) . Furthermore , these data correlate with an increase in the amount of Rpb1 associated with chromatin fractions in the Δbud27 mutant strain overexpressing the BUD27-TAP gene fusion ( Figure 3C ) . All together , these data indicate that Bud27 is necessary for correct nuclear localization of RNA pols in the S . cerevisiae nucleus , and suggest that it may play a role in assembly and/or nuclear transport of the three enzymes . URI is an evolutionarily conserved member of the prefoldin family among eukaryotes [18] , [22] . Then , to start elucidating if in humans URI IS also involved in the biogenesis of the RNA pols , we performed siRNA silencing experiments in human pulmonary fibroblast and monitored the effect of URI depletion on Rpb1 intracellular localization . As shown by q-RT PCR , URI mRNA expression decreased to 40% at 100 nM of siURI ( Figure S3 ) . Furthermore , silencing of URI resulted in the accumulation of Rpb1 in the cytoplasm of treated cells ( Figure 3D ) , as revealed by immunocytochemistry experiments using 8WG16 antibodies . Contrary , control experiment did not affect nuclear localization of Rpb1 . These data suggest that , as it is the case for Bud27 , URI modulates the translocation of RNA pol II to the nucleus , pointing to a common role for these conserved proteins in the biogenesis of , at least , the RNA pol II . To investigate the effect of Bud27 in RNA pols complex assembly , we immunoprecipitated RNA pol II from a wild-type and a Δbud27 mutant strain containing functional tagged versions of different RNA pol II subunits ( Rpb2-TAP , Rpb3-HA and Rpb4-Myc ) . We performed immunoprecipitation experiments using anti-Rpb1 antibodies ( 8WG16 ) and analysed different RNA pol II subunits corresponding to the different assembly intermediate previously described , Rpb1 together with Rpb4/5/7/8/9 and Rpb2 together with Rpb3/10/11/12 [10] , [12] , [31] . Our results revealed no significant differences in the yield or subunit composition between the two polymerases for Rpb1 , Rpb2 , Rpb3 , or Rpb4 ( Figure 4A ) . It is worth noting that Rpb2 purification gave also similar amount of Rpb1/Rpb2 ( Figure 4B ) . Also , we analysed Rpb5 , a common subunit shared by the three RNA polymerases and demonstrated by us and others [18] to physically interact with Bud27 . Surprisingly , the amount of Rpb5 clearly decreased in RNA pol II immunoprecipitated from a Δbud27 mutant strain ( Figure 4A ) , indicating that lack of Bud27 affects RNA pol II assembly . To corroborate these results and to investigate whether Bud27 also participates in the assembly of the other two RNA pols ( I and III ) , we immunoprecipitated the three RNA pols from a wild-type and a Δbud27 mutant strain containing functional tagged versions of different RNA pols subunits ( Rpa190-HA and Rpc160-Myc ) . The immunoprecipitation of the three largest RNA pol subunits , using anti-HA , anti-Rpb1 ( 8WG16 ) , and anti-Myc antibodies , again revealed significant differences in yield between Rpb5 and the largest subunits of the three RNA pols between wild-type and mutant polymerase complexes ( Figure 4C , lines 5 and 6 ) . Rpb6 , another common subunit to the three RNA pols seems to assemble rather late in humans although this mechanism is unclear [12] . Thus , we also analysed the amount of Rpb6 in immunoprecipitated RNA pol II between the wild-type and Δbud27 mutant strains . Again , surprisingly , the amount of Rpb6 clearly decreased in RNA pol II mutant complex ( Figure 4A and 4C , lines 5 and 6 ) , but also in RNA pol I and III ( Figure 4C , lines 5 and 6 ) . These data indicate that Bud27 has a role in the assembly of the three RNA pols , by interfering with the correct assembly of Rpb5 and Rpb6 in the complexes . Furthermore , the fact that no significant differences were detected between the wild-type and the mutant strain in whole-cell extracts for any of the RNA pols subunits analysed suggests that when Bud27 lacks , the non-assembled subunits are not rapidly degraded . To extend our analysis , we performed Rpb3-TAP purification from wild type and Δbud27 mutant containing functional tagged version of Rpb3 ( Rpb3-TAP ) , as Rpb3-TAP purification has been largely and successfully used to purify RNA pol II [32]–[34] . The protein mixture obtained in each case was subjected to multidimensional protein identification technology ( MudPIT ) [35] and to separation by gel electrophoresis . From our analysis we can conclude that all RNAPII subunits are associated to Rpb3 in absence of Bud27 ( Table S1 ) , but however , the yield of Rpb3 recovery drops abruptly in Δbud27 cells . It is also significant the reduction of Rpb3 co-purifying proteins . Taking together , our results suggest that lack of Bud27 led to an instable RNA pol II complex and then , that Bud27 is also necessary to maintain stability of the enzyme . The data above strongly indicate that Bud27 is required for assembly of the three RNA pols and that this mechanism depends on correct assembly or stabilization of Rpb5 . Because Rpb5 was identified as an interactor with Bud27 , we considered the possibility that these proteins are functionally linked . In an attempt to clarify the relationship between Rpb5 and Bud27 , we explored whether RPB5 overexpression corrects the temperature sensitivity of the Δbud27 mutant strain . Notably , increasing dosage of Rpb5 suffices to rescue the growth defect of the strain lacking Bud27 at a restrictive temperature ( Figure 5A ) , as is the case for BUD27 overexpression ( Figure S1 ) . Bud27 in yeast and URI in human cells were originally characterized as regulators of gene expression controlled by TOR ( for target of Rapamycin ) pathway [18] . As opposed to a S . cerevisiae wild-type strain , the Δbud27 mutant was slightly sensitive to Rapamycin ( Figure 5A ) . Again , RPB5 overexpression corrected the sensitivity of the Δbud27 mutant cells to this drug ( Figure 5A ) . Considering the functional connection between BUD27 and RPB5 , we investigated whether increasing Rpb5 dosage could also correct Bud27-dependent RNA pols localisation . Then we monitored the localization of the three RNA pols by analysing Rpb8-ECFP by live cell imaging . Rpb8 was observed mainly in the nucleus when RPB5 was overexpressed , as compared to the same strain containing a void plasmid as a control ( Figure 5B ) , as in the case of BUD27 overexpression ( Figure S1 ) . Similar results were found by performing immunolocalization experiments using anti-HA , anti-Rpb1 ( 8WG16 ) and anti-Myc antibodies in a Δbud27 mutant strain containing functional tagged versions of different RNA pols subunits ( Rpa190-HA and Rpc160-Myc; Figure S4 ) . Taken together , our data showing that lack of Bud27 does not impair the nuclear localization of RNA pols under RPB5 overexpression , suggest that Bud27 is not the main requisite for their nuclear import . Also , we investigated whether RPB5 overexpression was sufficient to rescue the defects in RNA pols assembly observed in the Δbud27 mutant strain that could also explain the correction in RNA pols mislocalization . We immunoprecipitated the three RNA pols from a wild-type and a Δbud27 mutant strain containing functional tagged versions of different RNA pols subunits ( Rpa190-HA and Rpc160-Myc ) under RPB5 overexpression . Immunoprecipitation of the three largest RNA pols subunits , using anti-HA , anti-Rpb1 ( 8WG16 ) and anti-Myc antibodies , revealed no significant differences in the yield between Rpb5 and the largest subunits of the three RNA pols between wild-type and mutant polymerase complexes ( Figure 5B ) with respect to the same strains harbouring a void plasmid as a control . In addition , similar results were found for Rpb6 . These results demonstrate that increasing dosage of Rpb5 is sufficient to correct the RNA pols assembly defects caused by the lack of Bud27 , suggesting that it accounts for the rescue of nuclear RNA pols localization . To test which conserved domains of Bud27 are important for growth , we transformed Δbud27 cells with plasmids expressing full-length Bud27 or proteins deleted for the indicated domains , and tested their ability to complement the temperature-sensitive phenotype ( Figure S1 ) . Surprisingly , neither the PFD or the Rpb5-binding domain nor both were functionally important in vivo , as previously indicated [19] . To assess whether the above-mentioned conserved domains participated in the nuclear localization of the three RNA pols , we monitored Rpb8-ECFP and Rpb1-GFP by live cell imaging in a Δbud27cells with plasmids containing full-length BUD27 or BUD27 deleted for the PFD , Rpb5-binding domains , or both . As shown in Figure 6 , deletion of Rpb5-binding domain led to a clear Rpb1-GFP cytoplasmic accumulation . Similarly , a more diffuse cytoplasmic Rpb8-ECFP signal was observed . However , PFD-binding domain was not crucial for the correct nuclear localization of the three RNA pols . Thus , we conclude that nuclear localization of the three RNA pols mediated by Bud27 is dependent on Rpb5-binding domain . URI ( Bud27 ) is believed to function as a scaffold protein able to assemble additional members of chaperone prefoldin ( PFD ) family through its PFD and Rpb5-binding domains in both human and yeast . Human and yeast PFD is a complex composed of six different subunits , PFD1-PFD6 , referred to as the prefoldin/GimC complex , which functions as a molecular chaperone and delivers newly synthesised unfolded proteins to cytosolic chaperonin TRiC/CCT to facilitate the folding of proteins [13] , [23] . In humans , PFDs and URI are constituents of an 11-subunit complex associated to the RNA pol II , namely the RPAP3/R2TP/prefoldin-like complex , suggested to participate in RNA pol II assembly [12] , [13] . In addition , data from yeast and mammalian cells have shown that the functions of the prefoldin and CCT chaperone complexes are eliminated by removal of individual subunits [36] . To evaluate whether the deletion of other components of the prefoldin complex in yeast also participate in the RNA pols nuclear localization , we analysed Rpb8-ECFP and Rpb1-GFP localization by live cell imaging in strains deleted for different prefoldins: YKE2 ( PFD6 ) , shown by us and others [23] to physically interact with Bud27 , with Rpb5 [24] and genetically with other components of the RNA pol I and III machinery [37] , [38]; GIM6 ( PFD1 ) , which genetically interacts with BUD27 [19] and with subunits of RNA pol I and III machinery [37] , [39] , [40]; GIM4 ( PFD2 ) , which interacts neither with BUD27 nor with the RNA pols . Surprisingly , in contrast to Bud27 , these three components of the prefoldin complex are dispensable for the nuclear localization of the three RNA pols at 30°C or even at 37°C ( Figure 7 , for 30°C ) , indicating that the role of Bud27 in the biogenesis of the RNA pols seems to be independent from the rest of the prefoldin complex . The conserved protein Iwr1 was originally identified as a protein that co-purified with almost every subunit of RNA pol II and that interacts with the basal transcription machinery and regulates the transcription of specific genes [30] . Recently , it has been shown that Iwr1 specifically binds RNA pol II between Rpb1 and Rpb2 and directs its nuclear import , although it is not involved in RNA pol I or III transport [6] , [10] . Curiously , IWR1 genetically interacts with BUD27 [41] and RPB5 [42] and physically with Bud27 [24] although we could not reproduce this interaction by our TAP purification . Based on these data , and to rule out the possibility that accumulation of RNA pol II in the cytoplasm in cells lacking Bud27 is not an indirect effect due to the mislocation of Iwr1 , we analysed localization of Iwr1 in Δbud27 cells . For this experiment , cells were transformed with a plasmid expressing an Iwr1 protein lacking its NES domain and shown to accumulate in the nucleus , instead of a plasmid with entire IWR1 , since it leads to a diffuse signal and no clear nuclear localization [30] . As shown ( Figure 8 ) , deletion of BUD27 does not affect the nuclear localization of Iwr1-ΔNES . These data suggest that nuclear localization of Iwr1 does not require interaction with Bud27 .
Human URI/RMP and its orthologue in yeast , Bud27 , is the most studied member of the prefoldin-like ( PFD ) family of ATP-independent molecular chaperones , also called unconventional prefoldin Rpb5 interactor [18] . Based on recent data of large-scale proteomic screen of RNA polymerases in human cells , URI has been shown to be a component of the HSP90/R2TP complex and has been proposed to participate in the biogenesis of RNA polymerases [8] , [12] . However , this role has not been confirmed . In this work , we show evidence that Bud27 in S . cerevisiae is the first example of a protein that participates in the biogenesis of all RNA polymerases , and mediates the assembly of the three transcriptional complexes . Furthermore , Bud27 seems not to play a role in the nuclear transport of the RNA pols . Our data also suggest similar role for human URI . Consistent with a role for Bud27 in RNA pols biogenesis , we demonstrated by TAP purification and immunoprecipitation experiments that Bud27 interacts with different subunits of the three RNA pols . Our data agrees with those of Krogan et al . showing physical interactions between Bud27 and RNA pol II [24] or between URI or Bud27 and Rpb5 , a common subunit shared by the three RNA pols [14] , [18] . Moreover other authors identified , in human cells , URI as a component of the R2TP/prefoldin-like complex which binds the largest subunits of the RNA pol II [12] , [43] . Finally , in accordance with these physical associations , genetic interactions between BUD27 and different components of the three transcriptional machineries have been found in yeast by us ( Mirón-García , unpublished data ) and others [44] . Biochemical and structural studies of RNA pols have proposed a detailed model of these enzymes , but however , little is known on how they assemble into the complexes or how they are transported from the cytoplasm to the nucleus . Our work provides new data to elucidate the mechanisms governing RNA pols biogenesis and localization and complement those in human and yeast [6]–[8] , [12] . Bud27 is the first protein so far demonstrated to participate in the biogenesis of the three RNA pols , modulating their assembly . Accumulation of the three RNA pols in the cytoplasm , as a consequence of BUD27 deletion accounts for a defect in nuclear transport . However , lack of Bud27 does not impair RNA pols nuclear localization under RPB5 overexpression , suggesting that Bud27 is dispensable for their nuclear import . Only two proteins in S . cerevisiae had been shown to participate in RNA pol II biogenesis so far , Iwr1 and Npa3 , and only the Npa3 homologue RPAP4/GPN1 in humans . Notably , these proteins are necessary for nuclear transport but none of them are involved in pol assembly [6]–[8] . Our immunoprecipitation experiments in Δbud27 mutant strain and the analysis of different subunits corresponding to the different assembly intermediate previously described [12] , [31] , demonstrate that Bud27 mediates the assembly of the three RNA pols . In fact , the lack of Bud27 alters the correct assembly of Rpb5 and Rpb6 in the three RNA complexes and led to a more instable enzyme . Results are also consistent with RNA pols assembly in the cytoplasm as a prerequisite for their nuclear import and agree with recent observations in yeast and humans [6] , [8] , [12] . In addition , co-purification and gel filtration analysis ( unpublished data ) suggest that RNA pols intermediaries do not appear in yeast in the absence of Bud27 . These data , together with the fact that only differences in the yield of Rpb5 and Rpb6 were observed and that RPB5 overexpression corrects not only assembly but also nuclear RNA pols transport , point to the fact that Bud27 mediates RNA pols assembly in an Rpb5-dependent manner , in agreement with data confirming its role in protein folding [23] . Our data point to a role of Bud27 in the correct folding of Rpb5 to the rest of the complex , since lack of Bud27 leads to transcription defects not related to a decrease in nuclear RNA pols amount but to RPB5-dependent processes ( Mirón-García , in preparation ) . Furthermore , we cannot rule out that Bud27 could act to stabilize the interaction between Rpb5 and the rest of the RNA pols complexes , and hence their integrity . Interestingly , the Rpb5-binding domain but not the conserved PFD-binding domain of Bud27 is essential for this role , since only Δbud27 cells expressing a version of Bud27 lacking the Rpb5-binding domain show cytoplasmic RNA pols localization . In addition , the defect in Rpb6 assembly is also consistent with the fact that RPB6 overexpression partially corrects the temperature sensitivity of the Δbud27 mutant ( our unpublished data ) and with the physical contact between Rpb5 and Rpb6 on the RNA pol II structure [45] . Furthermore , as BUD27 is not essential and its deletion seems only to affect part of the RNA pols complexes , it appears that Bud27 participates in coordination with other proteins to address its role in RNA pols biogenesis . These results also provide information concerning the mechanisms governing the assembly of Rpb5 and Rpb6 into the rest of the complexes . Our data suggest that Rpb6 could assemble rather late once Rpb5 is assembled , as previously proposed in human [12] . However , we cannot disregard the possibility that it assembles before and Rpb5 would be necessary to stabilize its contact with the other components of the RNA pols . Alternatively , we cannot rule out that Rpb5 could be assembled in early steps of the RNA pols biogenesis participating in maintaining the integrity of the RNA pols , although this possibility seems less unlikely . As is the case for Bud27 in yeast , HSP90 and its cochaperone RPAP3 in human cells have been shown to coordinate the assembly of RNA pol II and its involvement in RNA pol I and III assembly has been suggested [43] . Curiously , HSP90 and RPAP3 are members of the RTP2/prefoldin-like complex , a molecular machine dedicated to the assembly of multi-molecular protein complexes , such as the RNA pol II . It contains 11 components , among them URI and Rpb5 . On the contrary , in yeast the R2TP complex contains five proteins and Rpb5 and Bud27 have not been found as bona-fide constituents [43] . Interestingly , prefoldin 6 ( Yke2 ) , which is part of the human RTP2 complex also appears as a Bud27 interactor when purified via TAP . Thus is temptating to speculate that in yeast , also Bud27 and Rpb5 could associate with the RT2P complex . Moreover , physical interactions between Bud27 and a member of the Hsp70 family chaperones ( Ssb1 ) , as well as with the beta-tubulin chain of the microtubules ( Tub2 ) have been identified ( Figure 1B ) . This is consistent with previous data from two-hybrid analysis [19] , [23] . Notably , polymerization of tubulins into microtubules requires prefoldins and chaperonin CCT complex , which has been shown to interact with RNA pol II subunits [8] , [36] . Impairing microtubule assembly , both in humans and yeast , leads to a RNA pol II mislocalization in the cytoplasm [8] . Thus it is possible that these interactions are functional , althought further work will be necessary to address this issue . As shown here RNA pols nuclear localization is dependent on Rpb5-binding domain . However , Bud27 PFD and Rpb5-binding conserved domains are not required for growth , neither for its role in translation [19] . Then , these domains could be also important for other roles of Bud27 in the nucleus , such as for its interaction with other transcriptional regulators [14] , [16] , [20] . Consistent with this possibility Bud27 shuttles between the cytoplasm and nucleus via an Xpo1-independent pathway . Discrepant results have been reported concerning Bud27 localization . According to Desplaces et al . , Bud27 is excluded from the nucleus in yeast [19] . However , in humans and Drosophila [20] , [21] nuclear localization for Bud27 has also been reported . Moreover , physical association between Rpb5 , Bud27 , and transcription factors TFIIF , as well as between Rpb5 and TFIIB [14] , [16] , let us to propose that Bud27 could compete with these transcription factors to bind Rpb5 . Finally , silencing experiments in human cells , account for a conserved role of URI in RNA pols biogenesis , suggesting similar mechanisms that must be deciphered .
Common yeast media , growth conditions , and genetic techniques were used as described elsewhere [46] . Rapamycin ( LCLAbs , USA ) and Benomyl ( Sigma-Aldrich ) was used at the indicated concentrations . Strains , plasmids and primers are listed in Table 1 , Table 2 , and Table 3 . Rpb8-ECFP and Rpb1-GFP tagging was performed by yeast recombination of a PCR fragment as described in Longtine et al . [47] amplified from plasmid pKT210 [48] or from chromosomal DNA from strain FY86 , using Rpb8ECFP-501 and Rpb8ECFP-301 or Rpb1-308 and Rpb1-310 primers , respectively ( see Table 3 ) . 400 ml cells growing exponentially ( A600∼0 . 6–0 . 8 ) in yeast extract-peptone-dextrose ( YPD ) medium or synthetic minimal ( SD ) were washed twice with ultrapure water and lysis buffer ( 50 mM HEPES [pH 7 . 5] , 120 mM NaCl , 1 mM EDTA , 0 . 3% Chaps 50% ) . Cells were resuspended in 1 ml lysis buffer supplemented with 1x protease inhibitor cocktail ( Complete , Roche ) , 0 . 5 mM PMSF , 2 mM sodium orthovanadate and 1 mM sodium fluoride and whole-cell extracts were prepared using a MixerMill MM400 RETSCH ( 3 min 30 Hz ) . Immunoprecipitations were carried out as described elsewhere [49] with some modifications: 150 µl of whole-cell extract ( 2000 µg ) and lysis buffer for all washes were used . 35 µl of Dynabeads M-280 Sheep anti-Mouse IgG ( Invitrogen ) were used with 9E10 anti C-Myc antibody ( 1 µg , Santa Cruz Biotechnology ) , 8WG16 anti-Rpb1 antibody ( 1 . 5 µg , Covance ) and 12CA5 anti-HA antibodies ( 0 . 4 µg , ROCHE ) . For TAP purification , the same protocol was used with Dynabeads Pan Mouse IgG ( Invitrogen ) . The affinity-purified proteins were released from the beads by boiling for 10 min . Eluted proteins were analysed by Western blot with different antibodies: 9E10 anti-C-Myc , 12CA5 anti-HA , 8WG16 anti-Rpb1 , PAP , anti-POLR2C ( 1Y26 , Abcam ) , and anti-Rpb6 or anti-Rpb5 ( a gift from M . Werner ) . TAP purification for protein identification by mass spectrometry was performed as described previously [50] . Bud27-TAP fusion protein and associated proteins were recovered from cell extracts by affinity selection on an IgG matrix . After washing , the TEV protease is added to release the bound material . The eluate is incubated with calmodulin-coated beads in the presence of calcium . After washing , the bound material is released with EGTA . This enriched final fraction was analyzed by mass spectrometry using the MudPIT approach as described in [35] . Cells were grown at 30°C in YPD or SD medium ( A600∼0 . 5–0 . 7 ) , fixed with 37% w/v formaldehyde at room temperature for 2 h with slow shaking , and then centrifuged and washed twice with PBS . Cells were resuspended in spheroplasting buffer ( 1 . 2 M sorbitol , 0 . 1 M K-phosphate buffer pH 6 . 5 ) and cell wall digested with 125 µg/ml zymolyase 20T ( USBiological ) and 22 . 7 mM 2-mercaptoethanol ( SIGMA ) by incubation for 1 h at 37°C without shaking . The spheroplasts were washed twice with PBST ( PBS with 0 . 05% Tween 20 ) and then resuspended in the same solution . Cell suspension was added to an AAS ( 3-aminopropyltriethoxysilane , Sigma ) slide , incubated at room temperature until slide was dry and washed twice with PBST . Then , 50 µl of PBS-BSA ( 1 mg/ml BSA ) were added , and the slides . After incubation for 30 min in a humid chamber were washed three times with PBS . Next , 50 µl of 1∶100 dilution of the primary antibodies ( 8WG16 , anti C-Myc or anti-HA ) in PBS-BSA ( 1 mg/ml BSA ) were added and incubated 2 h at room temperature in a humid chamber . Slides were then washed three times with PBS , and incubated for 1 h , in the dark , at room temperature in a humid chamber with 50 µl of 1∶100 dilution of secondary antibody ( Cy2 antimouse; Jackson Labs ) . The slides were washed three times with PBS and incubated for 5 min with 50 µl of 1 µg/ml DAPI ( in PBS ) . After washing three times with PBS , slides were finally covered with a Vectashield ( Vector Laboratories ) mounting solution . Human cells were fixed for 15 min at room temperature with 4% ( v/v ) paraformaldehyde in PBS . Following fixation , cells were washed three times for 15 min with PBS , treated with 50 mM ammonium chloride for 30 min and permeabilised for 20 min with PBS containing 0 . 1% ( v/v ) Triton X-100 ( wash solution ) . Blocking solution ( wash solution with 5% ( w/v ) BSA ) was then added for 30 min and then , cells incubated for 1 h with 1∶100 dilution of the primary antibodies ( 8WG16 ) in blocking solution . Cells were washed three times for 15 min with wash solution and then incubated with 1∶100 dilution of secondary antibody ( Cy2 antimouse; Jackson Labs ) for 1 h in blocking solution . After washing three times with wash solution , 15 min each , slides were finally covered with a Vectashield ( Vector Laboratories ) mounting solution containing DAPI . The fluorescence intensity was scored with a fluorescence microscope ( Olympus BX51 ) . Chromatin isolation was performed as previously described [51] with some modifications . Briefly , about 5×108 cells growing exponentially ( A600∼0 . 6–0 . 8 ) were resuspended in 3 ml of 100 mM PIPES/KOH ( pH 9 . 4 ) containing 10 mM DTT and 0 . 1% sodium azide and then incubated at room temperature for 10 min . Cells were spun down , resuspended in 2 ml of 50 mM phosphate buffer ( pH 7 . 5 ) , containing 0 . 6 M Sorbitol , 10 mM DTT , and 4 µl of 20 mg/ml zymoliase and incubated 10 min at 37°C in a water bath to spheroplast formation . Spheroplasts were then pelleted at 4°C , washed with 50 mM HEPES-HOK buffer ( pH 7 . 5 ) containing 100 mM KCl , 2 . 5 mM MgCl2 and 0 . 4 M Sorbitol , resuspended in equal volume ( ∼80 µl ) of EBX buffer ( 50 mM HEPES/KOH ( pH 7 . 5 ) , 100 mM KCl , 2 . 5 mM MgCl2 , 0 . 25% Tritón-X100 , 0 . 5 mM PMSF , 0 . 5 mM DTT , cocktail protease inhibitors Complete Roche 1x ) and incubated for 3 min on ice . This whole cell extract was laid onto 400 µl of EBX-S buffer ( EBX with sucrose 30% ) and centrifuged at 12000 rpm for 10 min . After the sucrose gradient a chromatin pellet became visible and was washed with 400 µl of EBX buffer and finally resuspended in 100 µl of the same solution . A 1/10 dilution of chromatin pellet was used for SDS-PAGE and Western blotted with antibodies against Rpb1 ( 8WG16 ) , alfa-tubulin ( T5168; Sigma-Aldrich ) and Nop1 ( 28F2; Abcam ) . URI , prefoldin-like chaperone ( Gene ID: 8725 ) gene was silenced by transfection with the siRNA heteroduplex hsURI1_02 and hsURI1_02_as ( Table S1 ) . Human pulmonary fibroblast ( HPF ) were cultured in growth medium ( GM ) , consisting of DMEM supplemented with 10% fetal bovine serum , 2 mM l-glutamine , and 50 U/ml penicillin–streptomycin . Cells were seeded in twenty four-well plates ( 30 , 000 cells per well ) and transfected in triplicate with 40–200 nM heteroduplex oligonucleotides using the Lipofectamine 2000 Transfection Reagent ( Invitrogen ) , following the manufacturer's protocol , and incubated for 24 h at 37°C . Control cells were treated in the same conditions without siRNA heteroduplex . The experiments were performed three times ( three replicates ) . After indicated time , cells were either harvested for RNA extraction or fixed with PFA 4% in PBS for inmunolocalisation analysis . Total RNA was isolated from URI silenced HPF cells using the SV Total RNA Isolation System ( Promega ) , according to the manufacturers . Retrotranscription was performed from 200 ng of total RNA using the Maxima First Strand cDNA Synthesis Kit ( Fermentas ) in a final volume of 20 µl , according to the manufacturer's protocol . As a control , each sample was subjected to the same process without reverse transcriptase . URI mRNA accumulation was analyzed by q-RT-PCR with oligonucleotides hsURI1E9-f01 and hsURI1E10-r01 using cDNA corresponding to 10 ng . Human cyclophilin A ( PPIA ) was used as an internal control . Each PCR reaction was performed at least three times , with three independent samples . All oligonucleotides used are indicated in Table 3 .
|
The mechanisms governing the assembly and the transport of the three eukaryotic RNA polymerases to the nucleus are in discussion . Interesting papers have demonstrated the participation of some proteins in the assembly of the nuclear RNA polymerases and in their transport to the nucleus , but the mechanisms involved are poorly understood . Our data help clarify the mechanisms governing biogenesis of the three eukaryotic RNA pols and demonstrate that the prefoldin Bud27 of Saccharomyces cerevisiae mediates the correct assembly of the three complexes prior to their translocation to the nucleus , in a process which is dependent on Rpb5 . In addition , our data support the view that , during the assembly of the RNA pols , Rpb5 and Rpb6 assemble rather late compared to the rest of the complexes . Furthermore , this role of Bud27 seems to be specific , as it is not extended to other prefoldin members . Finally , the role of Bud27 seems to be conserved in humans , suggesting conserved mechanisms in RNA pols biogenesis .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"spectrometric",
"identification",
"of",
"proteins",
"cellular",
"structures",
"protein",
"interactions",
"macromolecular",
"assemblies",
"microbiology",
"gene",
"function",
"immunochemistry",
"model",
"organisms",
"chaperone",
"proteins",
"cell",
"nucleus",
"proteins",
"biology",
"proteomics",
"cytoplasm",
"biochemistry",
"cell",
"biology",
"genetics",
"yeast",
"and",
"fungal",
"models",
"saccharomyces",
"cerevisiae",
"molecular",
"cell",
"biology",
"genetics",
"and",
"genomics"
] |
2013
|
The Prefoldin Bud27 Mediates the Assembly of the Eukaryotic RNA Polymerases in an Rpb5-Dependent Manner
|
Mutations that alter signaling of RAS/MAPK-family proteins give rise to a group of Mendelian diseases known as RASopathies . However , among RASopathies , the matrix of genotype-phenotype relationships is still incomplete , in part because there are many RAS-related proteins and in part because the phenotypic consequences may be variable and/or pleiotropic . Here , we describe a cohort of ten cases , drawn from six clinical sites and over 16 , 000 sequenced probands , with de novo protein-altering variation in RALA , a RAS-like small GTPase . All probands present with speech and motor delays , and most have intellectual disability , low weight , short stature , and facial dysmorphism . The observed rate of de novo RALA variants in affected probands is significantly higher ( p = 4 . 93 x 10−11 ) than expected from the estimated random mutation rate . Further , all de novo variants described here affect residues within the GTP/GDP-binding region of RALA; in fact , six alleles arose at only two codons , Val25 and Lys128 . The affected residues are highly conserved across both RAL- and RAS-family genes , are devoid of variation in large human population datasets , and several are homologous to positions at which disease-associated variants have been observed in other GTPase genes . We directly assayed GTP hydrolysis and RALA effector-protein binding of the observed variants , and found that all but one tested variant significantly reduced both activities compared to wild-type . The one exception , S157A , reduced GTP hydrolysis but significantly increased RALA-effector binding , an observation similar to that seen for oncogenic RAS variants . These results show the power of data sharing for the interpretation and analysis of rare variation , expand the spectrum of molecular causes of developmental disability to include RALA , and provide additional insight into the pathogenesis of human disease caused by mutations in small GTPases .
Developmental delay and intellectual disability ( DD/ID ) affect about 1–2% of individuals worldwide [1] . Many highly penetrant genetic variants underlying DD/ID have been identified , but a large fraction of disease risk remains unexplained [2 , 3] . While some DD/ID-cases may result from environmental factors and small-effect common variants [4] , it is likely that many probands harbor pathogenic , highly penetrant variation in as-yet-unknown disease-associated genes . The RASopathies are a group of genetic conditions often associated with developmental disorders [5] , having in common mutational disruption of genes in the RAS/MAPK pathway that alter patterns of signal transduction . RASopathies are individually rare and pleiotropic but are collectively one of the most common causes of developmental disorders . Associated features include neurocognitive impairment , craniofacial dysmorphology , anomalies of the cardiovascular and musculoskeletal systems , cutaneous lesions , and increased risk of tumor formation [6] . For example , variation in HRAS is associated with Costello Syndrome ( MIM:218040 ) , variation in KRAS is associated with Noonan Syndrome 3 ( MIM:609942 ) and Cardiofaciocutaneous syndrome 2 ( MIM:615278 ) , and variation in NRAS has been observed in probands with Noonan syndrome 6 ( MIM:613224 ) and other RASopathy-associated phenotypes [7] . Given the genetic and phenotypic heterogeneity among DD/ID in general and RASopathies in particular , collaboration and data sharing among clinicians , researchers , and sequencing centers is necessary to enable , or accelerate , discoveries of new forms of disease . One tool to facilitate such collaborations is GeneMatcher , launched in 2013 as a way to connect researchers and clinicians with interests in specific genes [8] . Here , we present details of a cohort , assembled via GeneMatcher , of eleven total probands ( including one set of monozygotic twins ) with protein-altering variation in RALA , which encodes a RAS-like small GTPase; the variants arose de novo in ten of these probands . All probands present with developmental delay . Detailed phenotyping , computational analyses of observed variation , and functional studies lead to the conclusion that missense variation affecting the GTPase activity and downstream signaling of RALA underlies a new neurodevelopmental RASopathy-like disorder .
All eleven probands presented with speech problems , including absent speech in seven and speech delay in the remaining four . Ten of the eleven probands are reported to have hypotonia , with eight unable to walk . Intellectual disability was specifically noted for 8 of 11 , ( but not ruled out for the remaining three , see Table 1 ) . Birth measurements were available for nine probands and three ( 33% ) reported either length or weight ( or both ) at less than the tenth percentile . Height and weight measurements at last examination were available for all probands ( except for height in one ) . Six of ten probands ( 60% ) were reported to have heights less than the 10th percentile at last examination , while eight of eleven ( 73% ) were reported to have weights less than the 10th percentile . Three probands had head circumference measurements greater than or equal to the 90th percentile at last evaluation . Nine of eleven probands were reported to have dysmorphic facial features . Several consistent features were observed , including a broad , prominent forehead , horizontal eyebrows , epicanthus , mild ptosis , slightly anteverted nares , wide nasal bridge , short philtrum , thin upper lip vermillion with an exaggerated Cupid’s bow , pointed chin , and low-set ears with increased posterior angulation ( Fig 1 ) . Additional common but variable features were observed: seizures were present in most probands ( 6/11 ) , as were structural brain abnormalities detected by MRI ( 9/11 ) . Six of eleven probands were reported to have skeletal anomalies such as clinodactyly ( 3 of 6 ) and/or 2/3 toe syndactyly ( 2 of 6 ) . None of the probands are reported to have had cancer . Clinical summaries with additional details are available in Supporting Information ( S2 Text ) . Genetic variation within this cohort includes eight de novo heterozygous missense variants in nine probands ( including the monozygotic twin pair ) , one de novo heterozygous in-frame deletion of one amino acid , and one heterozygous premature stop of unknown inheritance ( Table 1 , Fig 2A ) . Except for R176X ( see below ) , all observed variants are absent from gnomAD [10] and TopMed genomes ( “Bravo” ) [11] . These variants have scaled CADD scores ranging from 22 . 1 to 41 suggesting they are highly deleterious , similar to the majority of mutations previously reported to cause Mendelian diseases [9] . Seven probands ( 1–7 ) , including the monozygotic twin pair , harbor recurrent de novo variants affecting one of only two codons , those encoding residues Val25 and Lys128 , while the remaining three de novo variants affect Asp130 , Ser157 , and Ala158 . All of these residues are computationally annotated as one of 24 residues , within a total protein length of 206 amino acids , that form the GTP/GDP-binding region of the RALA protein ( Fig 2 , Methods ) . While Val25 does not directly interact with GTP/GDP , variation observed at this position ( Val25Met and Val25Leu ) would likely result in distortion of the structure of the GTP/GDP-binding pocket ( Fig 2B and 2C; S1 Fig ) . Lys128 , Asp130 , and Ser157 all form hydrogen bonds with GTP/GDP in the wild type protein ( Fig 2B and 2C; S2 Fig , S3 Fig , S4 Fig ) . Although Lys128Arg would retain the positive charge of the side chain , steric hindrance resulting from the larger size of the Arg side chain would likely result in disruption of this binding pocket ( S2 Fig ) . Both Asp130Gly and Ser157Ala are predicted to result in loss of hydrogen bond formation ( Fig 2B and 2C; S3 Fig , S4 Fig ) . The remaining de novo variant , an in-frame deletion of Ala158 , results in a shift of Lys159 into the GTP/GDP binding region of RALA , which likely hinders GTP/GDP binding ( S5 Fig ) . Variation at all five of these residues is thus structurally predicted to alter GTP/GDP binding . This conclusion is consistent with the high degree of conservation at these residues throughout evolution of RALA ( S6 Fig ) as well as in other related genes including HRAS , KRAS , and NRAS ( S7 Fig ) and RAP1A/B and RHOA[12] . The predicted nonsense variant Arg176X in proband 11 lies within the last exon of RALA , and thus may not result in nonsense-mediated decay ( NMD ) of the transcript . This would yield a protein that lacks the 29 C-terminal residues ( S8 Fig ) , which are known to contain at least two critical regulatory regions . Phosphorylation of Ser194 by Aurora kinase A ( AURKA ) activates RALA , affects its localization , and results in activation of downstream effectors like RALBP1 [13 , 14] . Additionally , the C-terminal CAAX motif ( CCIL in the case of RALA ) is essential for proper localization and activation of RALA via prenylation of Cys203 [15 , 16] . We next assessed whether the de novo variants in our cohort were enriched compared to that which would be expected in the absence of a disease association . The background frequency of de novo missense or loss-of-function variation in RALA ( 6 . 16 x 10−6 per chromosome ) was taken from Samocha et al . [17] , in which the authors estimated the expected rates of de novo mutation based on gene length and trinucleotide sequence context . In our study , eight unrelated individuals were drawn from cohorts of at least 400 proband-parent trios , collectively spanning 16 , 086 probands ( S1 Table ) . When comparing the frequency of observed de novo variation to the expected background frequency of de novo missense or loss-of-function variation in RALA ( 6 . 16 x 10−6 per chromosome ) [17] , we find a highly significant enrichment for de novo variants in affected probands ( 8 observed de novo variants in 32172 screened alleles vs . 0 . 198 expected , Exact Binomial test p = 4 . 93 x 10−11 ) . We note that this p-value is likely conservative , as it results from comparison of the observed rate to the expected frequency of de novo variation over the entire gene . However , six of the nine de novo alleles affect only two codons , and all observed de novo variants are within the GTP-interacting space of 24 residues ( 11 . 7% of the 206-aa protein , Fig 2A ) . This clustering likely reflects a mechanism of disease that depends specifically on alterations to GTP/GDP binding and , subsequently , RALA signaling . Population genetic data also support pathogenicity of the observed de novo variants . RALA has a pLI score of 0 . 95 in ExAC [10] , suggesting that it is intolerant to loss-of-function variation . While RALA has an RVIS score rank [18] of 50 . 45% , it also has an observed/expected ratio percentile of 0 . 92% , a score that has been suggested to be more accurate for small proteins wherein observed and expected allele counts are relatively small [19] . Furthermore , population genetic data also support the likely special relevance of mutations in the GTP/GDP-binding pocket . No high-quality ( “PASS” only ) missense variants are observed at any frequency at any of the 24 GTP/GDP-coordinating residues in either gnomAD [10] or BRAVO [11]; in contrast , there are missense variants observed at 34 of the 182 RALA residues outside the GTP/GDP-interaction region ( S2 Table ) . This distribution across RALA is likely non-random ( Fisher’s exact test p = 0 . 017 ) and suggestive of especially high variation intolerance in this region of RALA . RALA and RAS-family genes have a high degree of similarity , and germline variation in several RAS-family GTPases is known to be associated with developmental disorders [5] . Comparisons of phenotypes observed here to those reported in these RASopathies suggest considerable overlap , including DD/ID , growth retardation , macrocephaly , high broad forehead , and mildly dysplastic dorsally rotated ears . Further , we compared the specific variants observed here to variants in HRAS , KRAS , or NRAS previously reported as pathogenic for RASopathies ( S3 Table , S7 Fig ) . De novo heterozygous missense variation at Val14 of KRAS , the homologous equivalent of Val25 in RALA , was previously reported in four unrelated individuals with Noonan syndrome [20 , 21] . Functional studies showed that this variant may alter intrinsic and stimulated GTPase activity and may increase the rate of GDP release [20 , 21] . A de novo variant in HRAS at Lys117 , the homologous equivalent of Lys128 in RALA , was found in two unrelated probands with Costello Syndrome [22] . Lastly , a de novo HRAS variant at Ala146 , the homologous equivalent of Ala158 in RALA , was reported in at least three patients with Costello Syndrome [23] . Variation at this residue has also been reported as a recurrent somatic variant in colorectal cancers [24] . Additionally , a recent study identified three probands with brain malformations and de novo missense variants in ARF1 , encoding a small GTP-binding protein [25] . The location of one of the observed missense variants , ARF1 K127E , is analogous to RALA K128 , a residue affected by variation in our cohort . This previous study also found that ARF1 missense variation is depleted from the general human population , particularly for missense variation affecting the GTP/GDP-binding region; that conclusion is consistent with the observation here that missense variants are not observed in the GTP/GDP-binding residues of RALA in large population databases . We investigated the functional consequences of the variants described above by expressing and purifying recombinant RALA proteins , and then measuring their abilities to hydrolyze GTP ( see Methods ) . While wild-type RALA showed robust GTPase activity under these experimental conditions , all mutants tested here exhibited a dramatic reduction in GTPase activity , including a mutant RALA that was not observed in probands but which carries a missense substitution , G23D , homologous to the G12D KRAS or HRAS variant commonly observed in tumors ( Fig 3A ) . As GTPase activity of mutant RAS family proteins alone is not always a clear indication of downstream effects [20 , 21] , we also assessed binding of these mutants to a RALA effector protein using an ELISA-based method ( see Methods ) . In this assay , recombinant G23D RALA protein exhibited approximately two-fold increased binding ( p < 0 . 0001 , Fig 3B ) , as anticipated for a constitutively active gain-of-function alteration [20 , 26] . V25L , V25M , D130G and R176X each showed a 2-5-fold reduction in effector binding compared to wild-type ( each p < 0 . 0001 , Fig 3B ) . In contrast , the S157A mutant exhibited increased binding compared to wild-type , suggesting that it may act in a constitutively-active manner similar to G23D ( p < 0 . 0001 , Fig 3B ) . We note that while there is some variation among mutants in the efficiency of protein production and purification ( Methods , S9 Fig ) , whether or not one normalizes to relative band intensity from Western blots of purified protein does not qualitatively affect these conclusions ( S10 Fig ) . In this and other cases of rare disease sequencing , it is important to consider other variation in any given patient that may be pathogenic . In six of the eleven cases presented here , the RALA variant was found to be the only plausible candidate . In five cases , other variants were discovered that were also initially considered as potential disease-causing mutations ( Table 1 , S2 Text ) . Proband 2 has a hemizygous variant in FLNA ( p . V606L ) , inherited from his unaffected heterozygous mother . Phenotype comparison , consultation with a filaminopathy disease expert , and application of the ACMG variant interpretation guidelines [27] resulted in the scoring of this variant as a VUS . FLNA is an interaction partner of RALA [28] , but the disease relevance of this variant is unclear . Proband 7 has a de novo variant in SHANK2 ( p . A1101T ) ; however , this allele is present in gnomAD three times and thus is not likely to be a highly penetrant allele resulting in DD/ID . Proband 8 has a variant in SCN1A , p . R187Q; however , this variant was inherited from an unaffected father , is present in gnomAD in one heterozygote , and , according to the referring clinician , the phenotype observed in the proband is not consistent with Dravet syndrome . Finally , proband 10 carries a paternally-inherited 1 . 349 Mb duplication of 1q21 . 1-q21 . 2 . This duplication has been reported to be associated with mild to moderate DD/ID , autism spectrum disorders , ADHD and behavioral problems , and other variable features [29] . While the patient may have some phenotypic features of this duplication , the patient’s MRI findings and severity of delays are not likely explained by this inherited duplication . Proband 11 carries a nonsense variant , R176X , which is unusual given the apparent specificity for the GTP/GDP-binding region of RALA observed in the other cases in our cohort . Clinically , we consider the R176X to be a variant of uncertain significance for several reasons . The R176X allele has been observed twice in the Bravo genome database , and parental DNA for this proband was not available , so we do not know whether the variant is de novo or inherited . In addition , the proband has microcephaly and more profound delays than others in the cohort , and also has large regions of homozygosity consistent with parental consanguinity . These regions of homozygosity suggest an additional and/or more complex molecular pathogenesis .
The application of genome sequencing to clinical settings is rapidly expanding our knowledge of mutations that cause rare disease , and has engendered new strategies for analysis , new rubrics for molecular pathology , and new platforms for collaboration . Here we apply these advances to show that mutations in the GTP/GDP-binding region of RALA cause developmental and speech delay , together with minor dysmorphic features . Mutations in RAS family members and RAS signaling pathways are well-recognized causes of several dysmorphic syndromes and cancer , but germline mutations in RALA have not to our knowledge been previously associated with disease . Our results add to basic knowledge about the biology and function of RAS super-family members , raise new questions about the molecular pathogenesis of mutations that affect small GTPases , and have important implications for clinical genomics . Among the RAS super-family of small GTPases , RALA and RALB are among the most closely related to the RAS subfamily ( ~50% amino acid similarity ) , and function as a third arm of the RAS effector pathway in addition to RAF and PI3K activation [5] . RALA and RALB have different expression patterns—RALA is broadly expressed whereas expression of RALB is enriched in endocrine tissues [30]—but also exhibit some degree of genetic redundancy: in gene-targeted mice , loss of function for RALA causes a severe neural tube defect that is exacerbated by simultaneous loss of RALB [31] . In neuronal culture systems , RALA has been implicated in the development , plasticity , polarization , migration , branching , and spine growth of neurons [32–36] , as well as the renewal of synaptic vesicles and trafficking of NMDA , AMPA , and dopamine receptors to the postsynaptic membrane [28 , 35 , 37] . Previous studies have evaluated the effects of RALA in multiple ways , including through loss of function studies ( e . g . , mouse knockouts , RNA interference , etc . ) , and designed mutational alterations to GTP/GDP hydrolysis , suggesting that multiple types of RALA perturbation have molecular and cellular consequences . Several aspects of our results suggest that developmental delay in humans is not caused by a simple loss-of-function of RALA . First , no clearly pathogenic loss-of-function ( i . e . , nonsense , frameshift , splice-site ) alleles were reported to GeneMatcher or exist in the literature to our knowledge . Second , in mice , heterozygosity for loss of function does not obviously affect development or viability [31] . Third , the de novo missense alleles described here are clearly enriched in GTP/GDP-binding residues , with six alleles recurring at only two codons . Fourth , missense RALA alleles in the general population all lie outside of the GTP/GDP-binding residues , suggesting selective depletion similar to that seen for other small GTP-binding proteins [25] . Finally , multiple disease-associated RALA positions observed here are homologous to positions at which mutations in other small GTPases have been shown to alter GTPase activity and thereby lead to disease . Combined , these observations suggest that reduced RALA dosage is not per se pathogenic , but instead that disease results from a mechanism that depends specifically on alterations to GTP/GDP binding dynamics . In functional assays , we found that all of the proband alleles exhibited reduced GTPase activity , similar to most oncogenic RAS alleles . However , they exhibited variability in their ability to bind RALA effector protein , with one showing increased effector binding , as is typically observed in tumor-associated RAS alleles , and the others all reducing effector binding . Similar variability of in vitro functional effects were reported for KRAS GTP/GDP-binding domain mutations observed in patients with developmental disorders [21] . Thus , while a commonality of altered GTP/GDP-binding is apparent , our results are consistent with multiple potentially relevant molecular mechanisms , including 1 ) altered levels of GTPase activity , 2 ) altered GTP/GDP release; and 3 ) altered regulatory and/or effector protein binding . We note that the phenotypes of individuals with the same allele were not qualitatively more similar to one another than phenotypes of individuals with distinct alleles ( see Table 1 ) . For example , seizures were observed for five of 10 probands with a de novo missense variant , including only one of three probands with V25M and one of two probands with K128R . Thus , genetic , environmental , or stochastic variability beyond the specific molecular effects of any given RALA mutation also contribute to disease manifestation . Additional functional assessments of RALA variants and larger cohorts of affected individuals are needed to clarify the relevant molecular mechanisms and the extent of genotype-phenotype correlations within the umbrella of RALA-associated disease . In summary , we show that de novo missense variation disrupting the GTP/GDP-binding functions of RALA lead to developmental delay , intellectual disability , and related phenotypes . These observations add to the diverse and pleiotriopic group of Mendelian disorders caused by variation in RAS-family GTPases and related RAS pathways .
Informed consent for participation in research was obtained from all families described here . Further , written informed consent to publish clinical photographs was also obtained for all probands pictured in Fig 1 . Additional research approval details are as follows . Proband 1 enrolled in a study approved by review boards at Western ( 20130675 ) and the University of Alabama at Birmingham ( X130201001 ) . Proband 2 enrolled in a research study approved by the Ethics Committee of University Hospital Motol , Prague ( 20120627 ) . Proband 3 was sequenced and analyzed in a diagnostic setting , and oral consent was obtained for research purposes approved by Kaiser Permanente Hawaii . Probands 4 and 5 were sequenced and analyzed in a diagnostic setting , and written consent was obtained for research purposes approved by the Agence Régionale de Santé , Île de France . Proband 6 enrolled in a research study approved by the Institute for Genomic Medicine at Columbia University ( AAAO8410 ) . Probands 7 and 10 enrolled in a study approved by the Western Institutional Review Board ( 1175206 ) . Proband 8 enrolled in a study approved by review boards at the University of Tennessee Health Science Center ( UTHSC201801 ) . Proband 9 was sequenced and analyzed in a diagnostic setting , and written consent was obtained for research purposes approved by Orlando Health . Proband 11 enrolled in a research study approved by a review board at University of Alabama at Birmingham ( F170303004 ) . Exome sequencing ( ES ) or genome sequencing ( GS ) was performed at one of six sites , in either a research or clinical setting . Named sites and additional details , including cohort sizes used in p-value calculations , are provided in Supporting Information ( S1 Text , S1 Table ) . The protein structure determined by Holbourn et al . [38] was used for the assessment of the potential effect of the mutations on RALA activity ( PDB ID: 2BOV ) . The structure was visualized using PyMOL 0 . 99rc6 [39] . Additional protein modeling was performed as previously described [40] . The GTP/GDP-binding residues of RALA were defined as those in which any atom of a residue ( side chain or backbone ) lies within 1 . 5 angstroms of an atom of the ligand . RALA cDNA was synthesized ( Integrated DNA Technologies , Skokie , IL , USA ) based on the coding sequence of NM_005402 . 3 , with substitutions identified in patients described here ( probands 1–9 , 11; see Note below ) used to represent variation . Following PCR amplification , coding sequences were cloned into Champio pET302/NT-His ( ThermoFisher Scientific , Waltham , MA , USA , # K630203 ) using Gibson Assembly Master Mix ( New England BioLabs , Ipswich , MA , USA , #E2611 ) . All RALA coding sequences were Sanger sequenced and compared to NM_005402 . 3 . The only differences within the coding regions of RALA were those observed in the probands . Single Step ( KRX ) Competent Cells ( #L3002 , Promega Corporation , Madison WI , USA ) were transformed with plasmids , and bacteria were grown overnight at 37°C in LB plus ampicillin . Bacteria were diluted 1:100 in fresh LB plus 0 . 05% glucose and 0 . 1% rhamnose to induce a 6-His-tagged recombinant RALA protein . Bacteria were collected after 8 h incubation at 25°C , and snap-frozen on dry ice . 6-His-tagged proteins were purified using Dynabeads His-Tag Isolation and Pulldown ( #10103D , ThermoFisher Scientific , Waltham , MA , USA ) according to the manufacturer’s protocol . Protein purity was assessed using standard SDS-PAGE and Coomassie Blue staining . Protein concentration was quantified using a Take3 microplate reader ( BioTek , Winooski , VT , USA ) by assessing absorbance at 280 nm . Protein amounts were normalized among samples in Dynabead elution buffer prior to use in assays . GTPase activity of 0 . 95 μg of purified , recombinant proteins was assessed using the GTPase-Glo Assay ( #V7681 , Promega Corporation , Madison WI , USA ) . Luminescence was quantified using an LMax II 384 Microplate Reader ( Molecular Devices , San Jose , CA , USA ) . Binding of purified , recombinant proteins to a proprietary Ral effector protein was assessed using the RalA G-LISA Activation Assay Kit ( #BK129 , Cytoskeleton , Inc . Denver , CO ) , as per the manufacturer’s protocol . Briefly , purified RALA protein was incubated in the presence or absence of 15 μM GTP ( #P115A , Promega ) for 1 . 5 h at 25°C , then 23 . 75 ng of purified RALA/GTP mixture was applied to the Ral-BP binding plate . A Take3 microplate reader was used for quantification of this colorimetric assay . Purified proteins were detected using a polyclonal RALA Antibody ( #3526S , Cell Signaling Technology , Danvers , MA , USA ) at a dilution of 1:1000 , and an anti-rabbit IgG secondary antibody ( #926–32211 , IRDye 800CW Goat anti-Rabbit IgG , Li-cor , Lincoln , NB , USA ) at a dilution of 1:20 , 000 . Proteins were also detected using a 6x-His Tag monoclonal antibody ( #MA1-21315 , ThermoFisher Scientific , Waltham , MA , USA ) at a dilution of 1:1000 , and an anti-mouse IgG secondary antibody ( #102673–408 , VWR , Radnor , PA , USA ) at a dilution of 1:20 , 000 . These antibodies were used to confirm protein levels and determine appropriate binding of the RALA antibody ( See panel B of S9 Fig ) . An Odyssey CLx Imaging System ( Li-cor , Lincoln , NB , USA ) was used to visualize the Western . Relative quantification of the image was performed using Image J ( https://imagej . net/ ) . We note that while we attempted to study the effects of all variation observed here , Proband 10 was identified after functional validation began , and the recombinant protein with the K128R variant ( observed in probands 6 and 7 ) was not able to be expressed and purified consistently . Thus GTPase and G-LISA experiments were not performed using K128R or A158del mutants .
|
While many causes of developmental disabilities have been identified , a large number of affected children cannot be diagnosed despite extensive medical testing . Previously unknown genetic factors are likely to be the culprits in many of these cases . Using DNA sequencing , and by sharing information among many doctors and researchers , we have identified a set of individuals with developmental problems who all have changes to the same gene , RALA . The affected individuals all have similar symptoms , including intellectual disability , speech delay ( or no speech ) , and problems with motor skills like walking . In nearly all of these cases ( 10 of 11 ) , the genetic change found in the child was not inherited from either parent . The locations and biological properties of these changes suggest that they are likely to disrupt the normal functions of RALA . Functional experiments also show that the genetic changes found in these individuals alter two key functions of RALA . Together , we have provided evidence that genetic changes in RALA can cause developmental disabilities . These results will allow doctors and researchers to identify additional children with the same condition , providing a clinical diagnosis to these families and leading to new research opportunities .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] |
[
"sequencing",
"techniques",
"medicine",
"and",
"health",
"sciences",
"face",
"disabilities",
"enzymes",
"enzymology",
"alleles",
"genome",
"sequencing",
"molecular",
"biology",
"techniques",
"research",
"and",
"analysis",
"methods",
"public",
"and",
"occupational",
"health",
"proteins",
"guanosine",
"triphosphatase",
"recombinant",
"proteins",
"head",
"molecular",
"biology",
"genetic",
"loci",
"biochemistry",
"hydrolases",
"anatomy",
"genetics",
"biology",
"and",
"life",
"sciences",
"human",
"genetics",
"genetics",
"of",
"disease"
] |
2018
|
De novo mutations in the GTP/GDP-binding region of RALA, a RAS-like small GTPase, cause intellectual disability and developmental delay
|
Gene expression can be highly heterogeneous in isogenic cell populations . An extreme type of heterogeneity is the so-called bistable or bimodal expression , whereby a cell can differentiate into two alternative expression states . Stochastic fluctuations of protein levels , also referred to as noise , provide the necessary source of heterogeneity that must be amplified by specific genetic circuits in order to obtain a bimodal response . A classical model of bimodal differentiation is the activation of genetic competence in Bacillus subtilis . The competence transcription factor ComK activates transcription of its own gene , and an intricate regulatory network controls the switch to competence and ensures its reversibility . However , it is noise in ComK expression that determines which cells activate the ComK autostimulatory loop and become competent for genetic transformation . Despite its important role in bimodal gene expression , noise remains difficult to investigate due to its inherent stochastic nature . We adapted an artificial autostimulatory loop that bypasses all known ComK regulators to screen for possible factors that affect noise . This led to the identification of a novel protein Kre ( YkyB ) that controls the bimodal regulation of ComK . Interestingly , Kre appears to modulate the induction of ComK by affecting the stability of comK mRNA . The protein influences the expression of many genes , however , Kre is only found in bacteria that contain a ComK homologue and , importantly , kre expression itself is downregulated by ComK . The evolutionary significance of this new feedback loop for the reduction of transcriptional noise in comK expression is discussed . Our findings show the importance of mRNA stability in bimodal regulation , a factor that requires more attention when studying and modelling this non-deterministic developmental mechanism .
Cellular differentiation is guided by complex gene regulatory networks that integrate different intra- and extracellular signals . This deterministic view has been challenged by the discovery of so-called bistable or bimodal regulation , whereby the decision to differentiate is stochastic . A classic example is the development of genetic competence in Bacillus subtilis [1] . Despite the fact that all cells are genetically identical , and are exposed to the same environmental conditions , only a minor fraction of a B . subtilis culture will develop into genetically transformable cells . Thus , a competent culture is composed of two different cell types . In essence , this bimodal distribution is the result of the positive feedback loop that regulates expression of the competence transcription factor ComK ( Fig 1A ) [2] . ComK is responsible for the expression of proteins required for DNA uptake and integration , but it also activates its own transcription [3–6] . If the cellular levels of ComK exceed a certain threshold , the auto-stimulatory loop is triggered and this leads to a rapid accumulation of ComK , which causes entry into the competent state [7–9] . Stochastic fluctuations or ‘noise’ in gene expression ultimately determines which cells accumulate sufficient ComK to reach the threshold level for autoactivation [10] . In recent years it has become apparent that bimodal gene regulation processes are common and occur both in prokaryotic as well as in eukaryotic cells [11 , 12] . For example , in B . subtilis the induction of motility , expression of extracellular proteases , and sporulation are bimodal differentiation processes that use positive feedback regulation loops [13–15] . The evolutionary reason for heterogenic differentiation in isogenic cell populations is often explained as a bet-hedging strategy , since bacteria cannot predict how and when environmental conditions will change . However , there are also examples of bimodal differentiation where both cell types benefit from each other . For example , during infection , Salmonella typhimurium differentiates into a slow-growing subpopulation expressing virulence genes and a fast-growing subpopulation that is avirulent . However , the latter subpopulation is required to maintain the infection [16] . Bimodal differentiation also occurs in multicellular systems , such as the development of alternative colour vision photoreceptors in Drosophila melanogaster [17] . Because of their important role in development , bimodal regulatory feedback loops have been extensively studied and modelled . An intriguing and often debated issue is the role of expression noise in bimodal regulation . The origin of protein expression noise resides in the omnipresent stochastic fluctuations in basic biochemical processes , including transcription , translation , mRNA and protein stability . This noise leads to slight cell-to-cell variations in protein levels [18] . Expression noise is a key prerequisite for bimodal gene expression , and yet noise is an intrinsically stochastic and non-deterministic process . Here , we describe a novel genetic screen that was developed to identify possible cellular factors influencing this noise . The competence transcription factor ComK is induced in response to nutrient starvation and high cell densities . Entry into the competent state causes severe changes in the physiology of the cell , including a block in growth , cell division and DNA replication [19 , 20] . An intricate regulatory network ensures that activation of ComK is tightly controlled , and transcription from the comK promoter is regulated by five other transcription factors: Rok , AbrB , CodY , DegU and Spo0A [21–25] ( Fig 1A ) . These transcription factors are involved in the regulation of several other differentiation pathways such as sporulation and motility , and they are part of extensive and intertwined regulatory networks [26] . ComK is able to activate comK transcription without the necessity to replace the repressors CodY and Rok [27] . Binding of ComK to its own promoter is stimulated by the pleiotropic response regulator DegU [25 , 28] . Phosphorylated Spo0A also binds to the comK promoter region and transiently induces expression by antagonizing Rok [24] . However , increased concentrations of Spo0A repress comK transcription , and this master regulator imposes temporal limitations to the onset of competence [24] . Despite the presence of multiple repressors , comK is still transcribed at a basal level , and ComK is actively removed by the adaptor protein MecA , which targets it for degradation by the ClpCP protease complex [29] . This proteolytic control is alleviated by a small protein ComS that binds to MecA and prevents ComK degradation [30] . ComS synthesis depends on the production of quorum-sensing pheromones and is therefore cell-density dependent [31 , 32] . Due to the complexity of comK regulation , it seems logical that fluctuations in the different regulation pathways will result in a heterogenic development of competence . However , we have previously shown that only the autostimulation of comK expression is sufficient for bimodal expression [7] . This was illustrated by constructing a simplified ComK feedback loop , whereby the comK promoter was substituted with the promoter of the comG operon ( Fig 1B ) . This operon encodes proteins required for DNA uptake . The comG promoter is directly induced by ComK and is not controlled by any other known ComK regulator [33] . Subsequent deletion of mecA created an autostimulatory ComK loop that bypasses all known transcriptional and post-translational regulation ( Fig 1B ) . Interestingly , expression of ComK by this artificial ComK feedback loop is comparable to the bimodal ComK expression in a wild type culture ( Fig 1B ) [7] . The B . subtilis competence regulation pathway is one of the best studied and modelled natural bimodal developmental systems and is therefore a good model system to study noise in bimodal regulation . We reasoned that the simplified positive feedback loop depicted in Fig 1B provides a way to identify possible unknown cellular factors that influence noise in comK expression . If none such factor can be found then we must assume that the bimodal distribution of Fig 1B is solely determined by noise . We developed a mutagenesis screen using the artificial ComK feedback loop of Fig 1B coupled with a reporter construct to visualize its activity . Interestingly , we were able to find transposon mutants that affected the expression of this minimalistic bistable positive feedback loop . Localization of the transposon insertions revealed an unknown gene , ykyB , which influences the bimodal induction of ComK . Inactivation of this gene increases the fraction of ComK expressing cells and the gene was therefore renamed kre for ComK repressor . Kre has no homology with any other protein . Further analyses indicated that Kre influences the stability of comK mRNA . Interestingly , the activity of Kre appears to be more general and is not limited to comK , however , the expression of kre is specifically downregulated by ComK itself . Kre is only present in species that contain ComK homologues . This co-evolution raises some intriguing questions concerning the balance between benefits and fitness drawbacks of genetic competence in B . subtilis . Finally , we discuss the importance of regulated RNA degradation in ComK expression and conclude that mRNA stability requires more attention in the research of bimodal gene expression .
Previously , we have shown that an artificial autostimulatory ComK feedback loop shows a bimodal expression pattern that closely resembles the wild-type pattern of ComK expression ( Fig 1 ) [7] . Theoretical modelling has shown that a simple positive feedback loop can produce a bimodal response if there is sufficient noise in the expression of the activator and a threshold level for activation [11] . Binding of ComK to DNA is highly cooperative and presumably this non-linear reaction determines the hypersensitive response of the positive feedback loop to small fluctuations of ComK levels [34] . This leaves expression noise as an important determinant of the fraction of ComK expressing cells and therefore of the bimodal distribution . To identify possible factors that influence the bimodal outcome of this artificial feedback loop , we constructed a lacZ-gfp operon that is driven by the comG promoter . This reporter enables the screening of mutants on plate as well as by fluorescence light microscopy , which makes it possible to distinguish differences in cellular ComK levels from differences in the frequency of ComK expressing cells . The PcomG-lacZ-gfp reporter was integrated at the ectopic amyE locus and combined with the artificial ComK feedback loop resulting in strain PG401 ( PcomG-comK , ΔmecA , PcomG-lacZ-gfp ) . On nutrient agar plates containing X-gal , PG401 colonies developed a faint blue colour after 2 days of incubation . However , PG401 colonies develop a clear blue colour after overnight incubation on competence medium plates ( Fig 2 ) , indicating that medium composition still influences the artificial ComK feedback loop . This is surprising since this loop was constructed in such way that none of the known competence regulators are able to influence its activity ( Fig 1 ) . Nevertheless , the effect on nutrient agar plates could be used to our advantage since it facilitates the selection of mutants with different lacZ , thus ComK , activities . Strain PG401 was mutagenized using the mariner transposon TnYLB-1 [35] , and a library of ~30 , 000 transposons was plated on nutrient agar plates supplemented with X-gal . Colonies that were blue after one day of incubation were checked for heterogenic GFP expression by microscopy . Four independent insertions were found that mapped in the coding sequence of ykyB , a gene of unknown function . Strains with transposon insertions in ykyB grew as blue colonies of normal size on nutrient agar plates , and formed dark blue and small colonies when streaked on competence medium plates ( Fig 2B ) . Microscopic visualization revealed that inactivation of ykyB causes a strong activation of the artificial ComK feedback loop , with GFP-expressing cells appearing on nutrient agar plates , and developing with more than 4 fold higher frequency on competence medium plates . ( Figs 2 , S1A and S1B ) . To confirm that the inactivation of YkyB was responsible for this effect , a complete deletion of ykyB was constructed and introduced into the artificial ComK feedback loop strain . When the resulting strain PG539 ( PcomG-comK , ΔmecA , ΔykyB , PcomG-lacZ-gfp ) was streaked onto competence medium X-gal plates , again dark blue colonies where formed in which more than 80% of cells expressed GFP . A strong activation of the artificial ComK loop was also observed in liquid rich medium ( S1C Fig ) . Since inactivation of ykyB causes increased activation of ComK , the gene was renamed Kre for ComK repressor . To test whether inactivation of kre also influences ComK induction in wild type cells , both the kre:Tn mutation as well as the Δkre deletion were introduced into a wild-type background containing the PcomG-lacZ-gfp reporter fusion ( strains PG433 ( amyE::PcomG-lacZ-gfp , kre:Tn ) and PG488 ( amyE::PcomG-lacZ-gfp , Δkre ) ) . The resulting strains showed an approximately 3 fold increase in the number of GFP expressing cells when grown overnight on competence medium plates ( Fig 3A ) , indicating that the effect of a kre mutation is observable in wild type cells , and is not limited to strains containing the artificial ComK feedback loop . Similar results were obtained when the GFP reporter was fused to promoters of the competence genes comC , comF , addAB and nucA ( S2 Fig ) . This shows that the sensitivity for Kre is not a unique property of the comG promoter , and that Kre affects ComK activity . To test the effect of a kre deletion in liquid cultures , we made use of the sensitive luciferase reporter fusion [36] . As shown in Fig 3B , a clear induction of the PcomG-luc reporter fusion is observed when kre is deleted . ComK levels were then checked by Western blotting and , as shown in Fig 3C , a strong increase in the intensity of ComK bands was detected for the kre mutant compared to the wild type strain . Consistently , when tested under the same growth conditions , an approximately 30-fold increase in transformation frequency of a kre mutant was observed at 0 , 1 and 2 hours after the transition to stationary phase ( Fig 3D ) . The results so far suggest that kre encodes a negative regulator of ComK . To confirm this , kre was placed under control of the strong IPTG-inducible Phyper-spank promoter at the ectopic amyE locus [37] . Indeed , overexpression of Kre reduced the fraction of PcomG-gfp expressing cells approximately 5 fold , and a strong repression was observed even when the wild type kre allele was deleted ( Fig 3E ) . Overexpression of Kre also reduced the transformation efficiency ( Fig 3F ) . To confirm that the effect was due to the Kre protein , a frame-shift mutation in the start codon of kre was introduced . The resulting strain ( PG548 ) was unaffected by the addition of IPTG and showed normal transformation efficiencies ( S3 Fig ) . kre encodes a hypothetical protein of 154 amino acids with no homology to any known protein . A recent comprehensive transcriptome analysis revealed that kre is expressed as a monocistronic mRNA ( S4 Fig ) at moderate levels in different growth conditions [38] . In this analysis , slightly higher expression levels were observed in M9 medium , as well as under salt , ethanol , and heat stress , and no major difference were observed between exponential growth and stationary growth , at least in rich medium [38] . Possibly , Kre functions as a transcription factor and regulates ComK expression , although no DNA binding or any other conserved motifs are apparent from its amino acid sequence . Many transcription factors bind to the nucleoid , owing to their DNA binding property [39] . To examine whether Kre co-localizes with the nucleoid , GFP fusions to the N- and C-terminal ends of the protein were constructed ( S5 Fig ) . Overexpression of both GFP-fusions reduced the fraction of PcomG expressing cells , indicating that the fusions are at least partially functional ( S6 Fig ) . However , both fusions showed a diffuse cytoplasmic GFP signal ( S5 Fig ) , suggesting that the protein does not function as a simple DNA binding transcription factor . Recent transcriptome experiments revealed the presence of a counter transcript , S365 RNA , which overlaps with the comK gene and is transcribed from the downstream located yhxD gene [38] ( S7 Fig ) . YhxD is strongly upregulated under stress conditions such as the presence of high salt , ethanol or high temperatures , conditions that also result in some increase in kre expression [38] . Possibly , the induction of yhxD is regulated by Kre and the anti-sense S365 transcript interferes with comK expression . To test this , the PcomG-comK construct was relocated from the comK locus to the ectopic aprE locus , and the wild type comK gene was replaced with a phleomycin resistance cassette . The resulting strain ( PG461; aprE::PcomG-comK , ΔmecA , ΔcomK , amyE::PcomG-lacZ-gfp ) showed GFP expression comparable to PG401 ( S7 Fig ) . Introduction of the kre mutation into PG461 resulted in an increase in GFP expressing cells very similar to what was observed in previous experiments with strains that contain the PcomG-comK construct at the wild-type comK locus ( Figs 2 and S7 ) . Thus , the activity of Kre is not based on the induction of the S365 anti-sense transcript . So far , we have tested the effect of a kre mutation in the presence of ComK autostimulation . To further dissect at which level Kre controls the bimodal induction of ComK , we uncoupled ComK expression from its autostimulatory transcription by removing the native comK gene and by placing a copy under control of the xylose inducible Pxyl promoter at the ectopic amyE locus . To monitor the effect on ComK , the protein was N-terminally fused to GFP . The mecA gene was also deleted to prevent possible proteolytic regulation effects . Since the GFP-ComK translational fusion is partially active and binds to DNA , a clear fluorescent nucleoid signal is observed ( Fig 4A ) . Interestingly , when a kre mutation was introduced into the strain , the fluorescence signal increased significantly ( Fig 4A and 4B ) . The increase in GFP-ComK expression also resulted in a further reduction in growth rate ( Fig 4C ) . Since the kre deletion has an effect on GFP-ComK accumulation even in the absence of a comK promoter , we conclude that Kre is not directly regulating the capacity of ComK to activate its own promoter . As a negative control for the experiments of Fig 4 , the fluorescence levels in a strain that expresses GFP instead of ComK-GFP were measured . Surprisingly , it appeared that the introduction of a kre deletion in this strain also resulted in increased GFP expression levels ( Fig 5A ) . To examine whether this effect might be linked to the Pxyl promoter or to the mecA comK double mutant background that was used , the kre mutation was introduced into a wild type background strain containing an IPTG inducible Physp-gfp reporter fusion ( strain PG820 ) . As shown in Fig 5B , also this promoter produced higher levels of GFP when kre was mutated . Finally , to determine whether the Kre activity might be specific for GFP , we tested another reporter and used the β-galactosidase expressing lacZ-gfp operon . This time the reporter was driven by the Pveg promoter , which is assumed to be unregulated during logarithmic growth [40] . When the Pveg-lacZ fusion was measured in a kre mutant background ( strain PG512 ) , a modest but significant increase in β-galactosidase levels was detected ( Fig 5C ) . Another promoter , PpksA , which also seemed to be unregulated according to a recent comprehensive transcriptome study [38] , was tested as well and gave a similar increase in expression ( Fig 5C ) . These results suggest that Kre functions as a more general repressor of gene expression . To determine the genome wide expression effect of a kre mutation , a micro-array experiment was performed . The transcriptome analysis was executed with samples taken from logarithmic growing cultures in rich LB medium . These conditions repress competence development [26] , and were chosen to prevent induction of competence genes that might mask indicative gene regulation events . A table of the 68 most relevant affected genes , i . e . genes whose expression difference was more than 4-fold with an adjusted P-value <0 . 05 , is presented in Table 1 . The list comprises a mixture of metabolic genes , genes involved in iron uptake , as well as several genes with unknown activities . Two of the genes ( ssbB and dprA ) are part of the ComK regulon . However , the list of genes does not reveal a clear regulation pathway that could explain the mechanism of Kre activity . Kre reduces the expression of different unrelated genes but it is unclear whether this control occurs at the transcriptional or translational level . Therefore , we determined the levels of the veg , pksA and comK transcripts using qPCR . As shown in Fig 6A , the veg and pksA mRNA levels are higher in a kre mutant background , associated with p-values of 0 . 002 and 0 . 06 , respectively . The effect on comK mRNA is the strongest ( Fig 6A , p-value 0 . 013 ) , which is presumably a consequence of the autostimulatory transcription of this gene . Not all genes are upregulated when Kre is deleted , as is apparent from the transcriptome data ( Table 1 ) , and a qPCR experiment showed that the mRNA levels of the cell division gene ftsZ are unaffected in a kre mutant strain ( Fig 6A , p-value 0 . 88 ) . These and previous data suggest that Kre is not a general inhibitor of RNA polymerase or protein translation , but that the protein affects mRNA levels and possibly influences mRNA stability . To test this , comK mRNA levels were measured after addition of the RNA polymerase inhibitor rifampicin . In the absence of Kre , an increase in stability was detected , with the half-life increasing on average from 3 . 9 min ( SE = 0 . 4 min ) to 5 . 4 min ( SE = 0 . 6 min ) . Such increase was consistently observed in three biological replicates ( Fig 6B and S1 Table ) . Comparing these 3 independent replicate measurements at each time point using a statistical test , showed that the increase in stability was significant ( false discovery rate corrected p-value ≤ 0 . 05 , S1 Table ) . As a control , we measured the stability of ftsZ mRNA , but there was no apparent effect when Kre was absent ( Fig 6C , mRNA half-life of ~2 . 2 min in both strains , and S1 Table ) . We then measured the stability of the same transcripts upon Kre overexpression by using a strain containing an extra copy of kre under control of the strong IPTG inducible Phyperspank promoter . As shown in Fig 6D , a significant decrease in stability was detected for comK mRNA ( false discovery rate corrected p-value ≤ 0 . 05 , S1 Table ) while , again , the stability of ftsZ mRNA was unaffected ( Fig 6E ) . The half-life of comK mRNA was ~2 . 6 min in the absence , and ~1 . 3 min in the presence of IPTG , respectively ( S1 Table ) . We note that , even in the absence of inducer , the half-life of comK mRNA was shorter compared to a wild type background . This might be due to leakiness of the Phyperspank promoter , and suggests that small variations of Kre levels may be sufficient to alter comK levels . Under the same conditions , the half-life of the ftsZ transcript was ~1 . 9 min and ~2 . 1 min , respectively . Based on these data , we conclude that Kre controls the bimodal response of ComK induction by affecting the stability of comK mRNA . The effect of Kre appears modest . However , the autostimulatory feedback will amplify small variations . Kre affects the expression of many genes , yet there is a significant ‘presence-absence’ correlation between kre and comK in different bacterial genomes ( Fig 7 ) . A closer inspection of the kre promoter revealed the presence of at least 3 potential ComK dimer binding sites ( Fig 8A ) . These so-called AT-boxes are spaced by 8 nucleotides , which is the correct distance to allow for the strong binding of a ComK tetramer [34] . Thus the kre promoter contains at least two ComK binding sites , one of which overlaps with the RNA polymerase binding site ( -35 region ) . To examine whether ComK influences kre promoter activity , a Pkre-lacZ-gfp reporter fusion was cloned into a mecA deletion strain , which overproduces ComK due to the absence of the regulatory proteolytic control of ComK [41] . As shown in Fig 8B , deletion of mecA decreases the β-galactosidase activity by half , and this reduction was ComK dependent . The fact that overproduction of ComK suppresses this promoter implies a new negative feedback loop in the control of ComK expression in B . subtilis ( Fig 8C ) . To examine whether this feedback control occurs in wild type cells expressing normal levels of ComK , we measured the activity of both the kre and comG promoters in single cells . The latter promoter was used as a reporter for ComK expression . The promoter of kre was fused to GFP by means of a Campbell-type integration ( kre:Pkre-gfp ) , and the comG promoter was fused to mCherry and cloned into the amyE locus ( amyE::PcomG-mcherry ) . Fluorescence light microscopy images of cells from a competent culture showed a clear reciprocal staining in the green and red channels ( Fig 8D ) . Quantification of the fluorescent signals indicated that comG expressing cells show on average a 60% reduction in the Pkre-GFP signal ( Fig 8E ) . The heterogenic expression of Pkre-GFP disappeared in a comK mutant strain ( S8 Fig ) . Thus , the negative feedback control of kre is active in wild type cells . When kre was placed under control of the PcomG promoter , and therefore activated by ComK instead of repressed , a strong reduction in transformation efficiency was observed ( S9 Fig ) . We conclude that , even though Kre affects the expression of many genes , its activity is closely intertwined with the development of genetic competence in B . subtilis .
Stochastic fluctuations in protein expression are a key prerequisite for the bimodal activation of positive feedback regulation systems . These random fluctuations in gene regulation pathways are often compared to the ‘noise’ in electronic circuits . The way electrical noise in circuits can be dampened , so can random spikes in protein levels be dampened too . This has consequences for bimodal processes , since a decrease in the amplitude or frequency of these spikes will reduce the chance that an activator reaches the threshold level necessary for auto-activation . Peaks in stochastic protein expression can be moderated by ( i ) suppressing the chance of transcription , ( ii ) reducing the life time ( stability ) of the mRNA , ( iii ) suppressing the chance of translation , and ( iv ) reducing the life time of the protein . The first and fourth mechanism are well known control pathways in the bimodal induction of ComK: Transcription from the comK promoter is repressed by 4 different transcription factors ( Rok , AbrB , CodY and Spo0A ) , and the adaptor protein MecA stimulates degradation of ComK by the ClpCP protease complex . The identification of Kre reveals the presence of a third mechanism: control of comK mRNA stability . The complexity of ComK regulation remains puzzling , especially since bimodal expression can be obtained without the necessity of an intricate regulation network ( Fig 1B ) . However , there are two main reasons why additional regulation is required , the timing of competence development and the escape from the competence state [26] . The latter is achieved by proteolytic degradation of ComK due to the reactivation of MecA as a consequence of dwindling ComS levels late in stationary phase [29 , 42] . Proper timing of ComK expression is essential since competent cells do not grow . Therefore , this developmental process should only be induced when nutrients become limiting . This explains for example the control of comK by the metabolic regulator CodY and the transition state regulator AbrB [22 , 23] . Competence induction should also not occur when cells are sporulating , therefore the control by the key sporulation activator Spo0A [24] . However , the reasons for the regulation by Rok and Kre are not immediately apparent . What is interesting is that these proteins were acquired relatively recently in evolutionary terms ( Fig 7 ) . Possibly , the origin of Rok and Kre regulation resides in a high fitness burden of the competence state ( competent cells do not grow ) relative to fitness benefits . In fact , there is only a remote chance that a genetic competent cell will acquire genetic material from which it can immediately benefit . In this respect , it is important to realize that most wild B . subtilis isolates are poorly competent , at least under laboratory conditions , and only the domesticated and mutagenized B . subtilis 168 strain shows high levels of competence [43] . Presumably , most of the time it is better for cells to circumvent the induction of competence , and acquiring new repressors that reduce the fraction of competent cell , like Rok and Kre , might therefore be beneficial . However , in the long term , the capacity to obtain new genetic material benefits the species , and this might explain why the negative feedback regulation of kre by ComK has evolved . Related to this , it is maybe interesting to note that the expression of rok is also repressed by ComK [21] . Of course , we cannot rule out that the main function of Kre is to restrict the time cells stay in the dormant genetic competent state . Our data suggests that Kre regulates mRNA stability . The protein does not have a known RNA or nucleotide binding pocket and its activity seems to be more general and not restricted to comK transcripts . It is unclear by which mechanism Kre influences mRNA decay . Within the ComK regulation cascade there is one other pathway that is affected by RNA modification . The conserved exoribonuclease PnpA , which is involved in cellular RNA homeostasis [44] , is required for the expression of ComS [45] . The small comS gene is embedded within a very long ( ~26kb ) mRNA encoding the synthetase subunits for the lipopeptide antibiotic surfactin [31 , 32] . Why PnpA is required for the expression of ComS is not known . Interestingly , in our screen we found two mutations that repressed the artificial ComK autostimulatory loop . These mutants contained transposon insertions into pnpA and cshA . The latter gene encodes a conserved RNA helicase which is also involved in cellular RNA homeostasis [46] . Thus , it seems there are more factors influencing comK mRNA stability . In a recent study it was shown that in Halobacterium salinarum and Escherichia coli there are specific RNases that control transcriptional positive autoregulation loops involved in certain energy-related processes [47] . Clearly , regulation of mRNA life-time is an efficient and presumably common mechanism to control transcriptional positive feedback loops . Two features of the artificial bimodal ComK loop remain unexplained . Even when kre is deleted , the induction of ComK is still growth phase and medium dependent ( Figs 2 and 3 ) . One explanation for the growth phase dependent expression is that ComK expressing cells are unable to divide . However , preliminary time lapse microscopy experiments showed a strong induction in the number of ComK expressing cells after the logarithmic growth phase has ceased . A more plausible explanation is that the exponential increase in cell volume during logarithmic growth dilutes any ComK that is expressed , and only when growth slows down will ComK accumulate to levels necessary to pass the threshold level required for auto-activation [48] . Optimal induction of competence in B . subtilis occurs in minimal medium with glucose as energy source . In contrast to this , in rich Luria Broth ( LB ) medium almost no competent cells can be detected [49] . Surprisingly , activation of the artificial ComK feedback loop is still medium dependent and clearly more efficient in minimal competence medium rather than in rich medium , even when kre is absent ( Fig 2 ) . One key difference between LB and minimal competence medium is the presence of glucose . Interestingly , when glucose was added to LB medium , there was a substantial increase in cells activating the artificial bimodal ComK loop ( S2 Table ) . There is an intriguing link between glycolysis and the cellular RNA processing and degradation machinery [50 , 51] . The core of the RNA degradosome in B . subtilis exists of the essential endoribonuclease RNase Y that forms a complex with other RNases , including PnpA , and the RNA helicase CshA . Importantly , the glycolytic enzymes enolase and phosphofructokinase are also part of this large protein complex [52] . This interaction is found in many other bacterial species [53] . It is as yet unknown how these glycolytic enzymes influence the RNA degradosome activity , but it might provide a clue for the glucose dependent regulation of competence . Here , we have described a new level of regulation of a well-known bimodal developmental pathway . Further research is required to elucidate the molecular mechanism of action of Kre and to identify its functional partners . However , it is clear that the role of mRNA stability in noise control may play a more significant role than previously appreciated .
Strains and plasmids used in this study are listed in Table 2 . All the B . subtilis strains were derivatives of BSB1 , a tryptophan-prototrophic ( trp+ ) derivative of the 168 trpC2 strain [38] . B . subtilis strains were grown at 30°C or 37°C on Nutrient agar plates ( Oxoid ) , or in liquid LB or competence medium ( 78 mM K2HPO4 , 42 . 8 mM KH2PO4 , 14 . 7 mM ( NH4 ) 2SO4 , 6 . 6 mM MgSO4 , 3 . 3 mM Na3-citrate , 26 . 9 mM glucose , 95 μM tryptophan , 4 . 2 μM ferric ammonium citrate , 0 . 02% Casamino acids ) . Solid competence medium was prepared with 1 . 5% Purified Agar ( Oxoid ) and was supplemented with a mixture of 13 amino acids ( Gly , Asn , Val , Glu , Leu , Asp , Ile , Pro , Phe , Ser , Ala , Thr , Gln ) at a final concentration of 10 μg/ml each to improve growth . To relieve catabolite repression of the Pxyl promoter in competence medium , glucose was replaced by fructose [54] . For selection , nutrient agar plates were supplemented with 10 μg/ml tetracycline , 5 μg/ml chloramphenicol , 50 μg/ml spectinomycin , 5 μg/ml kanamycin , 1 μg/ml phleomycin or 0 . 5 μg/ml erythromycin together with 25 μg/ml lincomycin . Xylose and IPTG were used as inducers at concentrations of 0 . 05–0 . 1% and 50 μM-1 mM respectively . E . coli was used as cloning intermediate . Molecular cloning , PCRs and E . coli transformations were carried out using standard techniques . Oligonucleotides used in this study are listed in S3 Table . Plasmids pPG22 and pPG23 were used to construct promoter-gfp and promoter-mcherry fusions , respectively , at the amyE locus . Plasmid pMutin-GFP+ [55] contains a gfp reporter with three terminators ( t1 , t2 , t0 from the rrnB operon of E . coli ) upstream the multiple cloning site in front of gfp and a trpA terminator downstream of gfp . pMutin-GFP+ was amplified with primers PG187 and PG188 in order to remove the Pspac promoter and to introduce 5 unique restriction sites in the multiple cloning site ( AgeI , BglII , PmlI , BlnI , SacII ) . Digestion with PmlI and subsequent self-ligation resulted in plasmid pPG20 . The gfp region with terminators was amplified from pPG20 with primers PG195 and PG196 , digested with ApaI and NotI and ligated into a similarly cut amyE-integration vector pPG2 [56] , obtaining plasmid pPG22 . To construct plasmid pPG23 , the mcherry gene from plasmid pHM232 [57] , was amplified with primers PG189 and PG190 , digested with EagI and SpeI and inserted into pPG20 , obtaining pPG21 . The mcherry region with terminators was amplified from pPG21 with primers PG195 and PG196 and , after digestion , ligated into pPG2 , obtaining pPG23 . comG promoter reporters were constructed by amplifying the comG promoter region with primers PG201 and PG202 and genomic DNA of strain 168 as template . PCR fragments were digested with BglII and BlnI and ligated into digested pPG22 or pPG23 , resulting in plasmids pPG34 and pPG38 , respectively . To construct the lacZ-gfp+ operon reporter for transposon screening , the lacZ sequence was amplified from pMutin4 [58] with primers PG203 and PG204 , digested with SacII and KpnI and ligated to a similarly cut pPG34 , resulting in plasmid pPG35 . Plasmid pPG40 was derived from pPG35 by replacing the spectinomycin resistance marker with the chloramphenicol resistance cassette cat from pSG1186 [59] , which was amplified with primers PG209 and PG210 , digested with SphI and XmaI and ligated to pPG35 . To integrate a PcomG-lacZ-gfp+ reporter at the aprE genomic locus , plasmid pPG63 was constructed by amplifying the comG promoter with primers PG330 and PG202 , and ligating it to pAWC3 , a plasmid based on pAPNC213 [60] that carries the lacZ-gfp+ operon ( Gamba et al . , in preparation ) , after digestion with XbaI and BlnI . For the luciferase reporter fusion ( PcomG-luc+ ) , plasmid pPG118 was constructed by amplifying the comG promoter with primers PG418 and PG419 , subsequent digestion with HindIII and BamHI and ligation into pUC18Cm::luc [36] . Inducible GFP fusions of Kre to msfGFP ( monomeric superfolder GFP ) were made at the N-terminal or C-terminal end of the protein by cloning the coding sequence of kre into plasmids pHJS105 and pPG49 , respectively . To make an N-terminal fusion , kre was amplified with primers PG287 and PG288 , digested with BamHI and EcoRI , and ligated to pHJS105 , resulting in plasmid pPG54 . To make a C-terminal fusion , plasmid pPG49 was first constructed by amplifying mSFgfp from pHJS105 with primers PG279 and PG280 , digesting the fragment with SacII and SpeI , and ligating it to the gel-extracted backbone of pPG22 . Then , Pxyl promoter and kre coding sequences were introduced at the same time into pPG49 with a double ligation step . Pxyl was amplified with primers PG320 and PG321 from pSG1729 and digested with BglII-BlnI , while the kre gene was amplified with primers PG319 and PG282 and digested BlnI-SacII . The two fragments were ligated to a BglII-SacII cut pPG49 , obtaining plasmid pPG61 . Reporter fusions with the promoters of other competence genes were made by amplifying the promoter regions of comC , comF , addAB and nucA with primer pairs PG289-PG290 , PG291-PG292 , PG293-PG294 , PG295-PG296 , respectively . After digestion with BglII and BlnI the promoter regions were ligated into pPG40 , resulting in plasmids pPG55 , pPG56 , pPG57 and pPG58 respectively . For overexpression of Kre , kre was cloned behind the strong IPTG inducible hyperspank promoter ( Physp ) . The kre coding sequence was amplified with primers PG299 and PG300 , digested with SalI and SphI and ligated into pDR111 [37] , resulting in plasmid pPG59 . Plasmid pPG59*fs is a variant of pPG59 with a mutation in the ATG start codon of the kre coding sequence , which becomes ATAG . This plasmid was obtained by amplifying plasmid pPG59 with oligonucleotides PG332-PG333 , which carry the desired mutation . To create a Pkre-lacZ-gfp reporter fusion , the promoter region of kre was amplified with primers PG322 and PG323 and , after digestion with AgeI and BlnI , ligated into pPG40 from which the PcomG promoter region had been removed by extraction of the cut plasmid form agarose gel , resulting in plasmid pPG62 . Plasmid pPG66 was made to create a Pkre-gfp promoter fusion at the kre locus by means of homologous Campbell-type integration . The plasmid was constructed by amplifying the promoter region of kre and its ribosome binding site with primers PG334 and PG336 , digested with KpnI and PstI and ligation into pSG1164 [61] . To create a kre gene under expression of PcomG , plasmid pPG34 was digested with BlnI and SpeI and gel extracted to remove the gfp fragment . The plasmid backbone was then ligated with the kre gene that was amplified with PG319 and PG438 from genomic template DNA , resulting in plasmid pPG126 . lacZ-reporter fusions with veg and pksA promoters were made as follow . The promoter region of veg , was amplified using primers PG317 and PG318 and , after digestion with BglII and BglI , ligated into pPG40 from which the PcomG promoter fragment was removed by extraction of the cut plasmid form agarose gel . The resulting plasmid was labelled pPG60 . The promoter region of pksA was amplified with primers PG509 and PG510 , digested with BglII and BglI and ligated into pPG60 from which the Pveg promoter fragment was removed by extraction of the cut plasmid form agarose gel . The resulting plasmid was named pPG136 . To construct strain PG368 ( mecA::tet ) , 2 . 5 kb regions upstream and downstream of mecA were amplified with primer pairs PG223-PG224 and PG225-PG226 , and subsequently digested with BlnI or XhoI respectively . A tetracycline resistance cassette was amplified from pBEST309 [62] with primer pairs PG221-PG222 , digested with BlnI and XhoI and ligated to the digested upstream and downstream amplified fragments . Competent B . subtilis cells were transformed directly with the ligation products and mutants were verified with PCR . To construct strain PG447 ( comK::phleo ) , regions upstream and downstream of comK were amplified with primer pairs PG211-PG212 and PG213-PG214 , and subsequently digested with NcoI or BamHI , respectively . A phleomycin resistance cassette was amplified from plasmid pIC22 [63] with primers PG215 and PG216 , and digested with the corresponding restriction enzymes prior to ligation . Mutants were verified by PCR and by checking the loss of transformability . To construct strain PG455 ( aprE::spc PcomG-comK ) , the PcomG-comK region , including the spc resistance cassette , was amplified from chromosomal DNA of strain PG401 with primers PG269 and PG270 . Next , 2 kb fragments comprising the 5’ or 3’ half of the aprE gene were amplified with primer pairs PG271-PG272 and PG273-PG274 , respectively . The three fragments were digested with BamHI , ligated and transformed to strain BSB1 . Integration was verified with PCR and by sequencing . To construct strain PG479 ( kre::erm ) , regions upstream and downstream of kre were amplified with primer pairs PG306-PG307 and PG308-PG309 , and subsequently digested with BamHI and NcoI , respectively . An erythromycin resistance cassette was amplified from pMutin4 [58] with primer pair PG312-PG313 , and digested with the corresponding enzymes prior to ligation . Random transposon mutagenesis of strain PG401 ( PcomG-comK , ΔmecA , PcomG-lacZ-gfp ) was carried out using the mariner transposable element TnYLB-1 [35] . PG401 is not transformable , therefore plasmid pMarB was introduced by protoplast transformation using standard protocols . Individual colonies carrying the transposon plasmid were picked and grown in LB at 30°C for 6 h . Aliquots were frozen and stored at -80°C . Serial dilutions of each culture were plated on Nutrient agar plates containing kanamycin or erythromycin and incubated at 50°C overnight to induce transposition . The following day , the clone with the highest ratio of kanR/ermR colonies , indicative of efficient transposition [35] , was chosen for further experiments . An aliquote of the selected clone was diluted and plated on Nutrient agar plates , and incubated at 50°C to construct a library of approximately 45 , 000 transposon colonies . The colonies were then scraped off the plates , aliquoted and frozen . About 30 , 000 clones of the library were plated on Nutrient agar plates supplemented with 160 μg/ml X-gal and incubated at 37°C for 24 hours . Colonies showing an intense blue colour were reisolated , checked for integration of the transposon ( kanR ) and loss of the plasmid ( ermS ) , and inspected by fluorescence microscopy to assess the frequency of GFP-expressing cells . Two rounds of backcrosses were performed: First , chromosomal DNA of the selected mutant strains was transformed into strain PG389 ( PcomG-lacZ-gfp ) . The resulting strains were transformed with chromosomal DNA of PG401 so to introduce simultaneously the PcomG-comK and ΔmecA mutations and reconstitute the artificial ComK feedback loop . Chromosomal DNA from colonies that still showed an increase frequency of competent cells on nutrient agar was then re-introduced into PG401 by SPP1 phage transduction [64] . Transposon insertions were located by arbitrary PCR followed by sequencing . Cells were mounted on microscope slides coated with a thin layer of 1 . 2% agarose . Images were acquired with a Zeiss Axiovert 200M or a Nikon T1 microscope coupled to a Sony Cool-Snap HQ cooled CCD camera ( Roper Scientific ) , and using Metamorph imaging software ( Universal Imaging ) . Images were analysed and prepared for publication with ImageJ [65] . Initially ( S2 Table and Figs 3 , 8 , S1 , S2 , and S6 ) , GFP intensities of individual cells were measured manually with ImageJ [65] , and subtracted by background GFP intensity levels , measured for each image . In later experiments ( Figs 4 and 5 ) , an in house developed ImageJ plugin ( NucTracer ( Syvertsson and Hamoen ) ) was used to semi-automatically determine cellular GFP levels . NucTracer , which uses nucleoids as region of interest ( ROI ) to measure GFP intensities , was employed to determine the GFP-ComK signals in Fig 4B . NucTracer was also used in Fig 5A and 5B , but here nucleoids were outlined with DAPI staining . To this end , the Pxyl-gfp and Physp-gfp reporter containing cells were grown in LB at 37°C in the presence or absence of 0 . 1% xylose or 50 μM IPTG , respectively . When cultures reached O . D . 600 of ~0 . 2 , aliquots were concentrated 4 times in PBS supplemented with 2 μg/ml DAPI and transferred onto microscope slides . To determine the fraction of ComK expressing cells using the PcomG-GFP reporter , a threshold value , generally 100 or 200 A . U . , was used to separate cells in expressing and non-expressing categories . This threshold value is well above ( ~ 3 to 5 times ) the fluorescent level of wild type ( non GFP-expressing ) cells . Overnight cultures in competence medium were diluted 20 fold in fresh medium and grown at 37°C until OD600 0 . 1 , then diluted 10 fold and 150 μl distributed into a black 96-well plate . Beetle Luciferin ( Potassium salt , Promega ) was added to a final concentration of 1 . 5 mg/ml ( 4 . 7 mM ) , and the cultures were incubated at 37°C in a FluoStar Optima plate reader ( BMG-LabTech ) . Relative luminescence units ( R . L . U . ) and OD600 were measured with 10 min time intervals . Transformation of competent B . subtilis cells was performed using a two-step starvation procedure [4 , 49] . Briefly , overnight cultures were diluted 10 fold in 10 ml competence medium and incubated at 37°C under vigorous shaking . After 3 hours of growth , an equal volume of prewarmed “starvation medium” ( competence medium lacking tryptophan , Cas aminoacids and ferric ammonium citrate ) was added and incubation was continued for another 2 hours , prior to DNA addition . DNA was added to 400 μl aliquots , and incubation was prolonged for 1 hour at 37°C prior to plating onto selective nutrient agar plates . Transformation frequency was determined by transforming competent cultures with genomic DNA carrying an antibiotic resistance marker . To test the transformation frequency of a kre mutant compared to the wild type strain BSB1 , exponentially growing cultures were diluted to OD600 ~ 0 . 01 in warm competence medium and grown at 37°C . The optical density of the cultures was measured at regular intervals . At the time of transition to stationary phase ( T0 ) , as well as 1 and 2 hours afterwards , DNA was added to 400 μl aliquots to a final concentration of 2 . 5 μg/ml , and incubation was prolonged for 1 hour at 37°C . Serial dilutions were plated on selective and unselective LB plates respectively . Transformation frequencies were calculated as 100 x ( transformants/ml / CFU/ml ) . Relative transformation frequencies were normalized to the frequency of wild type strain . To test the transformation frequency upon kre overexpression , overnight cultures were diluted 10 fold in the presence or in the absence of 1 mM IPTG and the protocol used for routine transformations was followed as described in the previous section . DNA was added at a final concentration of 2 μg/ml . Exponentially growing cultures were diluted to OD600 ~ 0 . 01 in warm competence medium and incubated at 37°C . Optical density was measured at regular intervals and 1 ml samples were collected , spun down and flash frozen in liquid nitrogen at the time of transition to stationary phase ( T0 ) and 1 , 2 , 4 hours after that time point . Incubation was prolonged overnight and one last sample ( Ton ) was collected the following morning . Cell pellets were resuspended in 100 μl of lysis buffer ( 100 mM Tris-Cl pH 7 . 5 , 2 mM EDTA , supplemented with Roche Complete mini protease inhibitor ) containing 10 μg/ml lysozyme , incubated 10 min at 37°C and then sonicated . Cell debris were removed by centrifugation . Relative protein concentrations were estimated with a Bio-Rad protein assay and equal amount of proteins were loaded on NuPAGE 4–12% Bis-Tris gradient gels which were run in MES buffer ( Life Technologies ) . Proteins were transferred onto a Hybond-P PVDF membrane ( GE Healthcare ) by using a wet procedure and western blotting was performed according to standard methods . A 1:5 , 000 dilution of rabbit polyclonal anti-ComK serum was used . Anti-rabbit horseradish peroxidase-linked antiserum ( Sigma ) was used as secondary antibody at a dilution of 1:10 , 000 . Protein bands were detected using an ImageQuant LAS 4000 mini digital imaging system ( GE Healthcare ) . Overnight cultures grown at 37°C in fructose-based competence medium were washed in 0 . 2 μM filtered starvation medium , stained with the red-fluorescent membrane dye FM5-95 , diluted 300 fold in filtered starvation medium and directly analyzed on a CyFlow Space flow cytometer ( Partec ) . Cell particles were selected based on the red-fluorescent signal . For each sample , 200 , 000 cells were analyzed and GFP signals were collected . Data were captured using FlowMax software ( Quantum Analysis GmbH ) and further analyzed using Cyflogic software ( http://www . cyflogic . com ) , which was also used for graph preparation . β-galactosidase assays were performed in exponentially growing cultures as described by Daniel et al . [66] and the units of enzymatic activity calculated as described by Miller [67] . To analyse the differences in transcriptome expression between wild-type B . subtilis ( strain 168 ) and the kre mutant ( PG479 ) , microarray analyses were performed using an 8x15k Custom Agilent microarray . The NCBI annotation BSU41030 B . subtilis subsp . subtilis str . 168 , complete genome , 2006-05-02 GenBank , containing information for 4105 transcripts , was used to design three probes per transcript . To isolate RNA , cell pellets were flash frozen in liquid nitrogen immediately after harvesting and stored at -80°C . Frozen pellets were grounded and subjected to RNA extraction as described previously [68] , yielding RIN values of ≥ 9 . 6 . Labeling was performed by reverse transcription using random octamers , incorporating Cy3 for the test samples and Cy5 for the common reference , as described [69] . The common reference was a pool of equal amounts of total RNA taken from all test samples . Hybridization , washing , and scanning was performed as described in the Two-Color Microarray-Based Gene Expression Analysis manual ( Version 6 . 6 , Agilent Technologies ) . Briefly , hybridization mixtures were made by combining 300 ng test ( Cy3 ) and 300 ng common reference ( Cy5 ) material and were subsequently hybridized to the Agilent SurePrint Custom 8x15k microarrays G2509F ( Agilent Technologies ) . Two biological replicates were used for strain 168 , while three biological replicates were used for strain PG479 . The raw and normalized data from all arrays were subjected to various quality control checks [68] . Normalized expression values were calculated by using the robust multi-array average ( RMA ) algorithm [70] , collecting and summarizing the intensity values of probes associated with a specific BSU locus tag . Differences in gene expression between wild-type and the kre mutant strain ( PG479 ) were statistically analysed using the Limma package in R 2 . 14 . 1 ( http://cran . r-project . org/ ) . Empirical Bayes test statistics were used for calculating P-values [71] , and for calculating false discovery rate corrected P-values [72] . Gene expression data and array design have been deposited at the public repository Gene Expression Omnibus , accession number GSE61757 . Cultures were grown in LB at 37°C and , at O . D . 600 ~0 . 25 , 5 ml volumes were spun at 6 , 000 rpm for 4 min and flash frozen in liquid nitrogen . Samples were processed with FastRNA Pro kit ( MP Biomedicals ) , cell disruption was achieved by shaking samples 4 times per 20 seconds at 6 , 000 rpm in a Precellys24 Tissue homogenizer ( Bertin technologies ) . RNA was further purified with Qiagen RNeasy kit . Total RNA ( 0 . 2 μg ) was retro-transcribed using Multiscribe reverse transcriptase and a High-Capacity cDNA reverse transcription kit ( Applied Biosystems ) . cDNA samples were diluted 1:24 and 6 μl was added to 10 μl GoTaq qPCR Master Mix ( Promega ) and 2 μl of each primer stock ( final concentration of 0 . 5 μM for each primer ) . qPCR was performed on a Rotor-Gene Q cycler ( QIAGEN ) with 40 cycles of 5 s at 95°C and 10 s at 60°C . Cycle threshold ( CT ) values were obtained according to the software instructions . Relative quantification was performed with the 2-ΔΔCT method [73] . pfkA mRNA levels were used as normalizer in Fig 6A . Changes in expression given are the average of 3 biological replicates , and the differences were statistically tested using an ANOVA model with coefficients for strain and replicate batch [74] . Oligonucleotides pairs used for qPCR were PG475-PG476 ( pfkA ) , PG456-PG474 ( veg ) , PG466-PG486 ( pksA ) , PG495-PG496 ( ftsZ ) , PG489-PG490 ( comK ) and PG471-PG472 ( kre ) , and their sequences are listed in S3 Table . Strains PG500 ( amyE::Pveg-lacZ-gfp ) , PG512 ( amyE::Pveg-lacZ-gfp , Δkre ) and PG474 ( amyE::Physp-kre ) were grown in LB at 37°C . At O . D . 600 of ~0 . 2 , T0 samples were collected ( 1 ml ) and rifampicin added to a final concentration of 150 μg/ml . Samples were taken at time intervals ( minutes ) after rifampicin addition , and immediately stabilized by mixing them with equal volumes of RNAlater solution ( Ambion ) . RNA was isolated and quantified using qPCR . Abundance of comK and ftsZ transcripts relative to the T0 sample was calculated with the 2ΔCT equation , and average values and standard deviations were calculated from 3 biological replicates . mRNA half-lives were determined from an exponential fit to a plot of relative mRNA abundance versus time . The logit transformed relative mRNA abundances were subjected to an ANOVA , to test for differences at each time point . The p-values were corrected for false discoveries using Benjamini-Hochberg correction . Calculations were carried out using Microsoft Excel and R statistical software ( http://cran . r-project . org/ ) .
|
Gene expression can be highly heterogeneous in clonal cell populations . An extreme type of heterogeneity is the so-called bistable or bimodal expression , whereby a cell can differentiate into two alternative expression states , and consequently a population will be composed of cells that are ‘ON’ and cells that are ‘OFF’ . Stochastic fluctuations of protein levels , also referred to as noise , provide the necessary source of heterogeneity that must be amplified by autostimulatory feedback regulation to obtain the bimodal response . A classical model of bistable differentiation is the development of genetic competence in Bacillus subtilis . Noise in expression of the transcription factor ComK ultimately determines the fraction of cells that enter the competent state . Due to its intrinsic random nature , noise is difficult to investigate . We adapted an artificial autostimulatory loop that bypasses all known ComK regulators , to screen for possible factors that affect noise in the bimodal regulation of ComK . This led to the discovery of Kre , a novel factor that controls the bimodal expression of ComK . Kre appears to affect the stability of comK mRNA . Interestingly , ComK itself represses the expression of kre , adding a new double negative feedback loop to the intricate ComK regulation circuit . Our data emphasize that mRNA stability is an important factor in bimodal regulation .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[] |
2015
|
A Novel Feedback Loop That Controls Bimodal Expression of Genetic Competence
|
Mitochondrial translation , essential for synthesis of the electron transport chain complexes in the mitochondria , is governed by nuclear encoded genes . Polymorphisms within these genes are increasingly being implicated in disease and may also trigger adverse drug reactions . Statins , a class of HMG-CoA reductase inhibitors used to treat hypercholesterolemia , are among the most widely prescribed drugs in the world . However , a significant proportion of users suffer side effects of varying severity that commonly affect skeletal muscle . The mitochondria are one of the molecular targets of statins , and these drugs have been known to uncover otherwise silent mitochondrial mutations . Based on yeast genetic studies , we identify the mitochondrial translation factor MEF2 as a mediator of atorvastatin toxicity . The human ortholog of MEF2 is the Elongation Factor Gene ( EF-G ) 2 , which has previously been shown to play a specific role in mitochondrial ribosome recycling . Using small interfering RNA ( siRNA ) silencing of expression in human cell lines , we demonstrate that the EF-G2mt gene is required for cell growth on galactose medium , signifying an essential role for this gene in aerobic respiration . Furthermore , EF-G2mt silenced cell lines have increased susceptibility to cell death in the presence of atorvastatin . Using yeast as a model , conserved amino acid variants , which arise from non-synonymous single nucleotide polymorphisms ( SNPs ) in the EF-G2mt gene , were generated in the yeast MEF2 gene . Although these mutations do not produce an obvious growth phenotype , three mutations reveal an atorvastatin-sensitive phenotype and further analysis uncovers a decreased respiratory capacity . These findings constitute the first reported phenotype associated with SNPs in the EF-G2mt gene and implicate the human EF-G2mt gene as a pharmacogenetic candidate gene for statin toxicity in humans .
The primary function of the mitochondria is the aerobic production of ATP , a process that is reliant on a series of protein complexes that comprise the electron transport chain . Several components of the electron transport chain are encoded in the mitochondrial genome , the translation of which is governed largely by nuclear encoded genes . Increasingly , mutations within these genes are being implicated with respiratory deficiency , an underlying factor in a number of diseases , including myopathies and liver failure [1] , [2] , [3] , [4] . For example , pathogenic mutations in the human mitochondrial elongation factor genes , EF-G1mt and EF-Tu ( mt ) , have been implicated with severe lactic acidosis and encephalopathy [1] , [2] , [3] . Recently a mutation in a novel gene , believed to be a member of the class of mitochondrial peptide release factors , was identified in patients exhibiting symptoms of Leigh syndrome [4] . In addition to disease , there is also emerging evidence that respiratory deficiencies are responsible for adverse drug reactions . Consequently , treatment with certain drugs have uncovered otherwise silent mitochondrial mutations [5] , [6] . The group of cholesterol-lowering drugs , statins , are one example . The primary target of statins is 3-hydroxy-3-methylglutaryl-coenzyme A ( HMG-CoA ) reductase , the rate limiting enzyme of the sterol synthesis pathway , but increasingly , studies are reporting signs of statin-induced mitochondrial dysfunction [7] , [8] . This is believed to be a factor in the myopathic side-effects of statins . Approximately 0 . 1 to 0 . 5 percent of statin users experience severe myopathic symptoms ( defined as serum creatine kinase levels more than 10 times the upper limit of normal ) and many more suffer milder musculoskeletal pain [9] , [10] . Frequently such patients present symptoms that are similar to those of patients with mitochondrial myopathies [11] . To date , there have been several case studies reporting the presence of a subclinical MELAS ( mitochondrial encephalopathy , lactic acidosis and stroke-like episodes ) mutation within the mitochondrial DNA ( mtDNA ) of patients who have developed severe myopathic symptoms following statin medication [12] , [13] , [14] . It is expected that existing weakness in mitochondrial function can be exacerbated upon exposure to statin , leading to the uncovering of previously asymptomatic mutations in mitochondrial genes . The yeast Saccharomyces cerevisiae has been the model of choice for studies of mitochondrial function . In addition to mitochondrial similarities with human cells , the ability of yeast to survive in the absence of mtDNA , the simplicity with which both nuclear and mtDNA can be manipulated and the extensive number of tools and resources available specifically for yeast research has greatly contributed to an understanding of potentially pathogenic mutations [15] , [16] , [17] . Statins were first isolated as secondary metabolites from fungi , the presumption being that the strong antifungal properties of statins provide an ecological advantage for the producer over other fungi , similar to that of antibiotics . We and others have demonstrated that upon exposure to statin , yeast , as well as having reduced cell viability , also display evidence of mitochondrial dysfunction [18] , [19] , [20] . In this study , we identify a nuclear gene encoding a mitochondrial translation factor as a modulator of atorvastatin toxicity in yeast ( MEF2 ) and human cell lines ( EF-G2mt ) . The eukaryotic mitochondrial protein synthesis system consists of four phases; initiation , elongation , termination and ribosome recycling , each carefully orchestrated by a series of nuclear encoded proteins [21] , [22] . The human EF-G2mt gene , originally named a mitochondrial elongation factor based on sequence homology with bacterial EF-G , has since been shown to function as a ribosome recycling factor [23] , [24] . EF-G2mt is believed to interact with the already known ribosome recycling factor ( RRF1 ) to promote dissociation of the ribosomal subunits following termination of translation [23] . In bacteria , the dual role of translocation and ribosome recycling are shared by a single EF-G protein [25] . Eukaryotic cells harbour two EF-G proteins in their mitochondria and it appears that these have distinct functions , the EF-G1mt protein for translocation and the EF-G2mt protein for ribosome recycling [23] , [26] . The human EF-G2mt protein is conserved across the majority of eukaryotic species [23] . With its yeast Mef2p counterpart , the human EF-G2mt protein shares greater than 32 percent homology and four of the five protein domains . We use the atorvastatin-sensitive phenotype of the yeast MEF2 gene to uncover naturally occurring human variants of EF-G2mt that have respiratory deficient phenotypes . These findings have ramifications for patient drug response and possibly also for disease .
In light of the emerging evidence that mitochondria are important in dictating statin toxicity , which in turn can reveal underlying respiratory defects that have important health implications , experiments were designed to discover mitochondrial lesions that affect statin sensitivity . Two published fitness profiling experiments in yeast have observed statin sensitivity in hundreds of heterozygous deletion mutants following statin exposure for 20 generations of growth [27] , [28] . Using the Gene Ontology term finder available on the Saccharomyces genome database website ( www . yeastgenome . org ) , we discovered approximately 14–17% of genes conferring statin-sensitivity are associated with the mitochondria . One of the most sensitive of these mitochondrial associated genes was the MEF2 gene , encoding a mitochondrial translation factor believed to have a role in ribosome recycling [23] , [27] . The growth of deletion mutants in a competition style assay is a very sensitive method of detecting differences in growth rate . However , these assays are prone to a higher incidence of false positive and non-replicable results [29] . To confirm the MEF2 phenotype , a number of yeast deletion mutants that ranked as the most statin sensitive in the fitness profiling experiments were compared [27] , [28] . Both heterozygous and haploid mutants were tested and cell viability was assessed after five days exposure to 110 µM atorvastatin . The concentration of 110 µM atorvastatin is approximate to those concentrations used in the original genome-wide fitness profiling screens ( 62 . 5 µM and 125 µM atorvastatin ) [27] . Furthermore , during a previous investigation of the effects of different atorvastatin concentrations in yeast , we have shown that 110 µM atorvastatin does not inhibit cell growth but is sufficient to cause a significant decrease in intracellular ergosterol ( approximately 85% ) , accompanied by loss of cell viability after prolonged exposure ( 5 days ) [18] . Of the heterozygous mutants tested , only the hmg1Δ/HMG 1 strain was confirmed to be sensitive to atorvastatin ( Figure 1A ) . However , of the haploid mutants tested , three displayed a statin hypersensitive phenotype ( Figure 1B ) . The mef2Δ mutant emerged as the strain which exhibited the greatest loss of cell viability in the presence of atorvastatin , with an almost 20-fold reduction in cell viability compared with the wild-type . The hmg1Δ strain displayed a 5-fold reduction in cell viability and disruption of the HTZ1 gene , encoding a histone protein , resulted in a 2-fold loss of cell viability ( Figure 1B ) . Yeast mutants defective in mitochondrial translation undergo rapid loss of mtDNA and we have previously shown this to be the case for the mef2Δ mutant [30] . Consequently , mef2Δ is ρ0 ( completely devoid of mtDNA ) . To determine whether atorvastatin-sensitivity is the consequence of abrogation of the MEF2 gene or simply from the absence of mtDNA , mef2Δ statin sensitivity was compared with that of ethidium bromide generated cytoplasmic ρ0 mutants . In order to ensure that there were no secondary site mutations in the mef2Δ deletion mutant that originated from the S . cerevisiae gene deletion collection , new mef2Δ haploid strains were created . Results show that although the ρ0 strains were more sensitive to atorvastatin than the respiratory positive parent , they did not display the same degree of sensitivity as mef2Δ ( Table 1 ) . Therefore , mutation of the MEF2 gene is a critical determinant of statin sensitivity through mtDNA dependent and independent functions . The human EF-G2mt gene , ortholog of the yeast MEF2 gene , encodes a recently characterised mitochondrial ribosome recycling factor [23] , but to date , no functional analysis has been performed for this gene . To determine whether the EF-G2mt gene is essential for human cell function and to ascertain whether depletion influences statin toxicity , an siRNA pool comprising of four individual EF-G2mt targeted siRNAs was used to silence EF-G2mt expression in the human rhabdomyosarcoma ( RD ) cell line . The RD cell line has previously been established as a skeletal muscle model for mitochondrial disorders and has also been used in studies of statin toxicity [31] , [32] , [33] . At 72 hours post-transfection , greater than 80 percent silencing was consistently achieved and cells remained viable . Cells were then re-transfected at this time point to enable continued depletion of EF-G2mt activity . This strategy had previously uncovered an essential role for cell viability for the first discovered ribosomal recycling factor gene , RRF1 [34] . However , at 72 hours post-re-transfection ( six days after the initial transfection ) , RD cells remained viable even though EF-G2mt mRNA concentration had decreased by 99 . 9 percent . Additionally , there was no decrease in mtDNA levels upon EF-G2mt silencing as analysed using quantitative PCR ( Figure S1 ) [35] . Although gross inhibition of mitochondrial translation in human cell lines results in loss of cell viability [34] , a more subtle mitochondrial phenotype may be masked by the phenomenon of the Crabtree effect whereby many human cell lines , when grown in the presence of glucose , derive their energy almost solely by fermentative means [36] . To circumvent this effect and force cells to rely on mitochondrial respiration as their primary energy source , galactose was used to replace glucose as the carbon source [37] . Silencing of EF-G2mt led to a marked decline in the growth of RD cells at four days post-transfection , which was maintained for a period of seven days ( Figure 2 ) . This growth defect on galactose medium signals an impairment in oxidative phosphorylation ( OXPHOS ) . Based on the findings in yeast , it was predicted that decreased EF-G2mt activity would also enhance the effects of atorvastatin toxicity . EF-G2mt silenced RD cells were subjected to various concentrations of atorvastatin in medium containing either galactose or glucose as the carbon source for a period of 48 hours . In glucose medium , there was no difference in statin sensitivity between the EF-G2mt silenced cells and those transfected with the non-targeting control . However , based on IC50 values ( defined here as a 50% loss of viability at 48 hours ) in galactose medium , EF-G2mt silenced cells were over 20 percent more sensitive to atorvastatin than cells transfected with the non-targeting siRNA pool ( Table 2 ) . These results confirm a role for the human EF-G2mt gene in cell resistance to atorvastatin in a human skeletal muscle cell line . Notably , a similar increase in sensitivity ( 17% ) was observed using the human hepatic HepG2 cell line ( a model for statin-induced liver toxicity ) , although it should be noted that HepG2 cells are approximately 10 times more statin resistant than RD cells and this elevation in atorvastatin sensitivity was not statistically significant in these experiments ( Table 2 ) . A global alignment of the amino acid sequence of the human EF-G2mt protein ( Isoform I , AAH15712 . 1 ) with the yeast Mef2 protein ( CAA59392 ) reveals 32 . 1% amino acid sequence identity ( Figure 3 and Figure S2 ) . At the commencement of this study , there were nine published non-synonymous Single Nucleotide Polymorphisms ( SNPs ) in the human EF-G2mt gene , of which five were either conserved or semi-conserved in the yeast MEF2 gene . Three of these variants , EF-G2mtI627T , EF-G2mtE594G and EF-G2mtK334R , are considered rare , with a heterozygosity frequency below one percent . One of the variants , EF-G2mtR744G , has a heterozygosity frequency of three percent and the EF-G2mtR774Q allele has a heterozygosity frequency greater than 20 percent . These five SNPs were selected for functional analysis using the yeast MEF2 gene as a model . For each of the five selected human EF-G2mt variants , single nucleotide base pair substitutions were constructed directly into the chromosomal copy of the yeast MEF2 gene to replace the codon specific to the wild-type amino acid residue with a codon that corresponds to the amino acid present in the human EF-G2mt protein variants . The Mef2p variants constructed were mef2K769Q , mef2R740G , mef2I616T , mef2D578G and mef2K308R which correspond to EF-G2mtR774Q , EF-G2mtR744G , EF-G2mtI627T , EF-G2mtE594G and EF-G2mtK334R respectively . To assess for respiratory competence , mef2 mutants were grown on medium containing the non-fermentable carbon source glycerol . After 72 hours , all five mutants were proficient in the production of colonies on both glucose and glycerol medium . The number and size of colonies produced by the mutant strains on glycerol medium was equal to that of the wild-type , indicating that all mutants are respiratory competent ( Figure 4A ) . Moreover , based on measurements of cell growth in both glucose and glycerol liquid medium , there were no growth defects exhibited by any of the mef2 mutants ( Table S1 ) . Mitochondrial DNA stability was measured periodically for up to 32 generations . The frequency of cells which spontaneously lose mtDNA amongst populations of each mef2 mutant remained equal to that of the wild-type ( approximately 2 to 3% ) , verifying that mtDNA is stable over successive generations . The five mef2 mutants were then assayed for atorvastatin sensitivity . Following five days of exposure to 110 µM atorvastatin , three of the mef2 mutants exhibited a statin hypersensitive phenotype . Viability of the mef2I616T , mef2D578G and mef2K308R mutants was reduced to 20 . 3 , 22 . 8 and 24 . 2 percent respectively ( Figure 4B ) . The two other mef2 mutants , mef2K769Q and mef2R740G , did not exhibit a statin sensitive phenotype . The unmasking of a phenotype for the mef2K308R , mef2D578G and mef2I616T variants by atorvastatin is a strong indicator that these mutations have an effect on Mef2p function . A similar effect conferred by these alleles in the human EF-G2mt protein could have vital consequences for statin users . Although no obvious growth phenotype was observed , the statin-sensitive phenotype of three of these mutants indicates a subtle defect in mitochondrial function . Staining of mef2 mutant cells with the nucleic acid staining dye 4′ , 6-diamidino-2-phenylindol ( DAPI ) confirmed the presence of mtDNA nucleoids in cells of all five mutants and quantification of mtDNA copy number using quantitative PCR ( qPCR ) [38] showed that mtDNA levels were the same as that of the wild-type ( Figure S3 ) . It therefore appears that the EF-G2mt equivalent mutations do not destabilise Mef2p function so as to compromise mtDNA stability . To investigate the possibility of a respiratory phenotype , oxygen consumption for the three statin-sensitive mef2 mutant cultures was measured ( in the absence of atorvastatin ) using a non-invasive oxoluminescent device [35] . All three mutants , mef2K308R , mef2D578G and mef2I616T , exhibited a significantly reduced respiration capacity , approximately one third lower than that of the wild-type ( Figure 5 ) . The two mef2 mutants that did not display an atorvastatin-sensitive phenotype had respiration rates much closer to that of the wild-type . Together , these results demonstrate a respiratory phenotype conferred by at least three of the EF-G2mt equivalent mef2 variants and this correlates with the inability of these mef2 mutants to tolerate atorvastatin toxicity . Interestingly , the partially respiratory deficient mef2 mutants exhibit a greater statin sensitivity than the ethidium bromide generated ρ0 strains , which are completely devoid of respiratory function ( Table 1 ) . This , in support of the mef2Δ results , indicates the role of a non-respiratory function of MEF2 in the statin response . In addition to a lack of mtDNA , we have previously shown , by staining cells with MitoTracker Red CMXRos , that the mef2Δ strain has a reduced mitochondrial membrane potential ( ΔΨ ) and that mitochondria appear fewer , with a tendency to aggregate [30] . The MitoTracker Red CMXRos probe enters the mitochondrial matrix dependent on ΔΨ . This same method was used to test ΔΨ and mitochondrial morphology in the mef2 mutants . Cells were visualised using a laser scanning confocal microscope and , in contrast to the mef2Δ strain , all five mef2 mutants displayed mitochondria that stain brightly and are arranged in a tubular network , typical of mitochondria in wild-type yeast . Staining of the three partially respiratory deficient mutants is shown in Figure 6 . These results indicate that the mef2 mutations do not disrupt the function of Mef2p in maintaining ΔΨ and so does not explain the enhanced statin sensitivity of these mutants . It is known that statin toxicity can cause loss of ΔΨ [39] . Therefore one possibility is that atorvastatin acts synergistically with a mutated Mef2p to exacerbate loss of ΔΨ and compromise cell viability . However , other mechanisms , such as modulation of the mitochondrial retrograde response , cannot be discounted [40] . The crystal structure of the human EF-G2mt protein has not yet been elucidated . Therefore , to gain insight into the molecular effects of the five chosen amino acid variations , a computational model of the EF-G2mt protein was constructed using the SWISS-MODEL server [41] . The model is based on the experimentally determined crystal structure of the Thermus . thermophilus EF-G protein [42] which shares 39 percent identity with the human EF-G2mt protein and four of the five EF-G2mt protein domains . As the N-terminal and C-terminal regions of the EF-G2mt protein share particularly low homology with the template , the initial 65 and final 10 amino acid residues could not be accurately modelled . For this reason , the location of amino acid variant EF-G2mtR774Q , positioned very close to the C-terminus , was omitted ( Figure 7A ) . Stereochemical quality of the model was assessed by generating a Ramachandran plot using PROCHECK . Eighty seven percent of residues fall within the most favoured regions of the plot , indicating ideal stereochemistry . Using the in silico model , we can make some hypothetical predictions about the function of the protein variants . The human EF-G2mtK334R variant , which corresponds to the respiratory compromised yeast mef2K308R variant , is located on a coil close to the surface of the GTP-binding domain ( domain I ) that is highly conserved in EF-G2mt homologs from all major eukaryotic species . The presence of GTP is essential for ribosome dissociation at the termination of translation , and subsequent hydrolysis of GTP is then required in order to release the EF-G2mt protein from the ribosome [23] . The EF-G2mtE594G and EF-G2mtI627T variants , equivalent to the respiratory compromised mef2D578G and mef2I616T mutants respectively , are both located in domain IV and occur close to the EF-G2mt protein surface . The EF-G2mtE594G variant is expected to diminish the largely negative electrostatic surface potential of this domain , thereby interfering in the protein's interaction with both the mitochondrial ribosome and Rrf1 [23] , [25] . The EF-G2mtI627T variant is located on an alpha helix whose structural integrity is disturbed by the threonine hydroxyl group . The final two EF-G2mt protein variants , in which the corresponding mef2 mutants did not exhibit a phenotype , are located in domain V , the C-terminal region that accelerates ( but is not essential for ) the ribosome recycling action of the EF-G2mt protein [23] . In the last year , data from the 1000 Genomes project has expanded the number of known polymorphisms in the human genome and a further 11 non-synonymous SNPs have been discovered within the human EF-G2mt gene ( dbSNP , National Centre for Biotechnology Information ( NCBI ) ) [43] . Of these , one is fully conserved in the S . cerevisiae Mef2 protein and another five are semi-conserved ( Figure 3 and Figure S2 ) . The location of five of these variants is shown on the EF-G2mt protein model in Figure 7B . Although experimental confirmation is essential , based on the findings above it is hypothesised that the domain IV ( EF-G2mtK621N and EF-G2mtF609Y ) and domain I ( EF-G2mtV165G ) variants will affect protein activity . Therefore , these naturally occurring EF-G2mt variants may have respiratory deficient consequences .
By exploiting the genetic tractability of yeast , complemented by siRNA silencing studies in human cell lines , we have identified a mitochondrial translation factor as a mediator of atorvastatin toxicity and also made fundamental discoveries about the function of human variants within the EF-G2mt gene . The human EF-G2mt gene , ortholog of the yeast MEF2 gene , was originally identified as a mitochondrial elongation factor gene . However , a recently published comprehensive analysis of the EF-G2mt protein has shown that it functions as a ribosome recycling factor , interacting with the first discovered ribosome recycling factor protein , Rrf1 , to dissociate the ribosomal subunits at the termination of translation [23] . In yeast , deletion of the MEF2 gene results in loss of mtDNA [30] , a circumstance which would be lethal in higher eukaryotic cells without the supplementation of uridine and pyruvate [44] . Nevertheless , siRNA silencing experiments show that in contrast to its RRF1 counterpart , knockdown of human EF-G2mt expression does not compromise cell viability in glucose medium . Furthermore , silencing of EF-G2mt expression does not deplete cellular mtDNA content . It is possible that some compensatory mechanism enables ribosome recycling to continue to a sufficient degree to maintain cell viability in fermentative cell lines . When galactose is used as the carbon source instead of glucose , the ATP produced via glycolysis is insufficient for cell energy requirements; therefore , there is a greater reliance on the production of ATP through the oxidative metabolism of glutamine . This more closely resembles the metabolic activity of cells in a human physiological system [37] , [45] . By using galactose to force reliance on the mitochondria for cell energy production , it was shown that reduced EF-G2mt activity does indeed compromise cell proliferation in respiring cells . It is consequently expected that the EF-G2mt gene is essential for cell function in a human system . Furthermore , in support of the notion that abnormalities in mitochondrial function sensitise cells to statin toxicity , the growth defect of EF-G2mt silenced cells in galactose medium was even further exacerbated upon exposure to atorvastatin . For cells in which mitochondrial function is challenged , exposure to mitochondrial toxicants , such as statins , places additional stress on mitochondrial function and this has the potential to trigger pathogenicity . Indeed we have shown that ρ0 strains , lacking respiratory capacity , are more sensitive to atorvastatin than those with a functioning mitochondrial genome in both yeast and human cells . Studies have shown that statins exert their mitochondrial toxicity effects by inhibiting function of the electron transport chain but there is also evidence of non-respiratory mitochondrial consequences [46] , [47] , [48] . These include a loss of mitochondrial membrane potential , aberrant mitochondrial morphology and apoptosis [39] . These effects , in combination with the absence of respiratory function may explain the hypersensitivity of ρ0 cells to statin . The hypersensitivity of respiratory deficient cells to statins may have clinical ramifications for patients that have variations within mitochondrial functioning genes . Statins have been known to aggravate clinically silent disease associated mutations resulting in myopathies . In fact , mutations ( which in many cases were asymptomatic ) for three common myopathic diseases; carnitine palmitoyltransferase II deficiency , McArdle disease and myoadenylate deaminase deficiency ( AMPD deficiency ) , are thought to be the underlying determinants responsible for statin-induced myopathy in up to 10 percent of patients showing adverse effects [5] . There are also reports of the statin-induced triggering of MELAS syndrome in patients whose MELAS mutations were clinically silent [13] , [14] . As MELAS syndrome arises from mutations in mtDNA , it strongly implicates mitochondrial dysfunction in susceptibility to statin toxicity . This is further supported by the identification of two commonly occurring SNPs in the human COQ2 gene that are associated with an increased risk of statin intolerance [49] . The COQ2 gene is required for the synthesis of coenzyme Q10 , an essential component of the mitochondrial electron transport chain [49] . Despite the presence of a number of SNPs within the EF-G2mt gene , the function of these variants for cell fitness had never been investigated . In this study , five EF-G2mt equivalent SNPs were selected for functional analysis in the yeast MEF2 gene . Unlike the mef2Δ deletant , all mef2 mutants grew proficiently on both glucose and the non-fermentable carbon source glycerol . Furthermore , the mutations had no effect on mtDNA stability , ΔΨ or mitochondrial morphology . However , exposure of the mutants to toxic concentrations of atorvastatin has uncovered a phenotype for three of the mef2 mutants; mef2K308R , mef2D578G and mef2I616T , equivalent to the EF-G2mtK334R , EF-G2mtE594G and EF-G2mtI627T alleles respectively . In silico protein homology modelling reveals these three mutations are located in either the GTP binding domain ( domain I ) of the EF-G2mt protein or an external domain ( domain IV ) necessary for ribosomal interaction . Based on the observations that statins exacerbate clinically silent disease associated mutations , it was predicted that the three mutations compromise mitochondrial function in a subtle yet potentially significant way . The subsequent observation that oxygen consumption was significantly reduced in the statin-sensitive mef2 mutants confirmed this hypothesis and demonstrates a sub-optimal mitochondrial function for the three EF-G2mt equivalent mef2 mutants . This sub-optimal mitochondrial function is expected to contribute to the atorvastatin-sensitive phenotype of the three mef2 mutants . However , comparison of the statin sensitive phenotype of the mef2K308R , mef2D578G , mef2I616T and mef2Δ mutants with cytoplasmic ρ0 strains indicates that statin sensitivity is not fully explained by the reduced respiratory capacity of these mutants and further studies are required to completely elucidate the precise mechanism . This study constitutes the first report of a phenotype associated with EF-G2mt , demonstrating an essential role for aerobic respiration in human cell lines and an importance for cell tolerance to atorvastatin . Atorvastatin constituted the focus of this study and is the highest selling and also one of the more potent of the statins [50] . Although the various statins have been shown to differ in their cellular toxicity effects , all of them have been implicated with mitochondrial dysfunction [48] , [50] , [51] , [52] , [53] . It would therefore be expected that a mitochondrial mediator of atorvastatin toxicity may also mediate cell response to the other statins . In support of this , preliminary experiments confirm that the atorvastatin sensitive mef2 mutants also exhibit sensitivity to lovastatin . With an estimated 38 million people around the world undertaking statin treatment [9] , the identification of novel biomarkers for statin toxicity has the potential to personalise therapy for millions of individuals . To date , only a handful of genes have been associated with statin toxicity [9] , but the finding of the EF-G2mt gene as a potential pharmacogenetic candidate has strengthened the association between existing mitochondrial dysfunction and statin hypersensitivity . Importantly , the discovery of naturally occurring human polymorphisms within the EF-G2mt gene that affect respiratory function indicates that these variants , either alone or in combination with other polymorphisms , have significant pathogenic consequences . This opens avenues for further clinical investigations into a possible association between EF-G2mt variants and disease .
Atorvastatin calcium was purchased from 7 Chemicals ( India ) . Stock solutions were prepared by dissolving atorvastatin in methanol at a concentration of 20 mg/mL and solutions were stored at −20°C . 5-fluoroorotic acid and geneticin ( G418 ) were purchased from Sigma . The haploid wild-type S . cerevisiae strains used were of the background BY4741 ( MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 ) and BY4742 ( MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0 ) . The diploid strain was BY4743 ( MATa/α his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 LYS2/lys2Δ0 MET15/met15Δ0 ura3Δ0/ura3Δ0 ) [54] . The mef2Δ/MEF2 , mef2Δ and the cytoplasmic ρ0 strains were previously described [30] . Prior to experiments , yeast strains were cultured in liquid YEPD ( 1% yeast extract , 2% peptone , 2% dextrose ) medium for 24 hours , subcultured and then grown to exponential phase in Synthetic Complete ( SC ) medium ( 0 . 67% Difco yeast nitrogen base without amino acids , 2% dextrose , 0 . 79 g L−1 amino acid supplement ( Sunrise Science Products , Australia ) ) . Cells were incubated at 30°C with shaking . Triplicate exponential phase cultures grown in SC medium were diluted to an optical density ( OD595 nm ) of 0 . 2 and 1 . 25 mL of this culture was added to 3 . 75 mL SC medium with the appropriate concentration of atorvastatin . After a maximum of five days growth , cells were diluted and plated onto solid YEPD ( YEPD plus 2% agar ) medium and viability counts were performed 48 hours later . HepG2 cells and RD Cells are from the American Type Culture Collection ( ATCC ) . Cells were cultured at 37°C and 5% CO2 in Dulbecco's Modified Eagle's Medium ( DMEM ) medium ( Gibco ) containing 4 . 5 g/L glucose and 10% fetal bovine serum ( FBS ) ( Bovogen Biologicals , Australia ) . For experiments in which galactose was used as the carbon source , cells were grown in glucose free DMEM with 10% fetal bovine serum and 4 . 5 g/L galactose . To generate the ρ0 cell lines , used as respiratory deficient controls , HepG2 and RD cells were cultured for eight weeks in DMEM medium containing 100 ng/mL ethidium bromide and supplemented with 10% FBS , 50 µg/mL uridine and 100 µg/mL sodium pyruvate [31] , [44] . Following ethidium bromide treatment , ρ0 cells were maintained in medium supplemented with uridine and sodium pyruvate . Depletion of mtDNA was confirmed using qPCR . The ON-TARGETplus SMARTpool , comprising of four siRNAs targeting the EF-G2MT transcript ( NM_170691 ) was purchased from Dharmacon ( cat . # L-017534-01-0005 , Dharmacon , Thermo Fisher Scientific , Lafayette , CO . ) . The corresponding ON-TARGETplus non-targeting SMARTpool ( cat . #D-001810-10-05 ) and DharmaFECT transfection agents were also purchased from Dharmacon . DharmaFECT agent 2 ( 0 . 2 µL/100 µL ) and DharmaFECT agent 4 ( 0 . 4 µL/100 µL ) were used to transfect RD cells and HepG2 cells respectively . All siRNAs were used at a final concentration of 25 nM . To transfect cells , equal volumes of the siRNA and transfection agent were mixed , allowed to incubate for 30 minutes , and 100 µL of the solution was added to 400 µL DMEM medium containing 10% FBS . This solution was added to the attached cells which had been washed in PBS . Gene expression silencing was assessed at 72 hours post-transfection by qPCR . For experiments in which cells were re-transfected , the above procedure was performed again at 72 hours subsequent to the initial transfection . At 24 hours post-transfection , approximately 1×104 cells/well of each culture were seeded into eight wells ( one for each day of the eight-day proliferation assay ) of a 96-well plate and allowed to attach overnight . Each day , the number of viable cells in one of the eight wells was assessed using the CellTiter-Glo assay ( Promega ) as described above . Cell medium was changed every two days . Approximately 5×104 cells/well were seeded into wells of a 96-well plate . For siRNA transfected cells , seeding occurred at 24 hours post-transfection . Following overnight incubation to enable attachment , media was changed to DMEM containing 10% FBS plus the appropriate atorvastatin concentration . Atorvastatin concentrations ranged from 0 to 128 µM for RD cells and 0 to 1024 µM for HepG2 cells . Cell survival after 48 hours in the presence of atorvastatin was estimated using the CellTiter-Glo luminescent cell viability assay ( Promega ) , which measures intracellular ATP concentration . The assay was performed according to the manufacturer's instructions and luminescence was quantified using a Tecan Genios microplate reader . IC50 values were calculated from dose-response curves that were generated using least-squares linear regression . Codon modifications used to alter yeast Mef2p amino acid residues to those of the corresponding EF-G2mt variants are A2305C , A2219G , T1848C , A1734G and A924G for the yeast mef2K769Q , mef2R740G , mef2I616T , mef2D578G and mef2K308R respectively . The GenBank accession number for the MEF2 sequence used was NC_001142 . 9 . The single base pair substitutions were created in the yeast MEF2 gene according to the double-strand break mediated delitto perfetto method [55] . The pGSKU plasmid containing the CORE-I-SceI cassette was kindly provided by Francesca Storici ( Georgia Tech , Atlanta , GA ) . The CORE-I-SceI cassette was PCR amplified with chimeric primers ( Table S2 ) that contain 50 bp homologous to the site of insertion and 20 bp for amplification of the cassette . Cassette amplification was performed in 50 µL PCR reactions ( 25 µL 2× Phusion Flash PCR Master mix ( Finnzymes ) , 0 . 5 µM each primer and approximately 1–10 ng purified pGSKU plasmid ) using a PTC-200 thermal cycler ( MJ Research ) ( Initial denaturation , 98°C for 10 seconds; denature , 98°C for 1 second; anneal/extend , 72°C for 75 seconds; repeat denature and anneal/extend for 30 cycles; final extension , 72°C for 1 min ) . Cells were transformed with 10 µL of the concentrated PCR product as previously described [55] . Transformants were selected by plating onto synthetic medium lacking uracil ( 0 . 67% Difco yeast nitrogen base without amino acids , 2% dextrose , 0 . 79 g L−1 uracil dropout amino acid supplement ( Sunrise Science Products , Australia ) and 2% agar ) and after 24 hours , G418 resistance was checked by replica plating onto solid YEPD containing 200 µg/mL G418 ( Sigma ) . Colony PCR was performed to confirm accurate integration of the cassette . Cells containing the integrated CORE-I-SceI cassette were then transformed with 0 . 5 nM of each strand of a pair of complementary 80 bp oligonucleotides ( Table S2 ) containing the desired substitution and possessing 40 bp on either side of the oligo which is targeted to the regions adjacent to the integrated marker ( GeneWorks , Adelaide ) . To induce the I-SceI mediated double strand break at the recombination site , prior to transformation cells were cultured in 50 mL of synthetic complete medium containing 2% galactose instead of glucose and incubated at 30°C with shaking for a period of six to eight hours . Following transformation , cells were plated onto solid YEPD medium and after 24 hours , replica plated onto synthetic complete medium containing 1 g/L 5-fluoroorotic acid and 60 mg/L uracil to check for loss of the URA3 marker . Loss of the CORE-I-SceI cassette was confirmed using colony PCR ( primers listed in Table S2 ) and the resulting PCR product was sequenced to ensure the desired mutation was present . Due to complete disruption of the MEF2 gene during insertion of the cassette , all resulting mef2 mutants were lacking mtDNA . Therefore , to reintroduce mtDNA , mutant strains were mated with the wild-type of opposite mating type ( the BY4741 strain ( MATa ) ) . Diploid strains were then sporulated and dissected , resulting in 2∶2 segregation of wild-type and mef2 mutant colonies , all of which were ρ+ . Sequencing was used to identify the colonies which possessed the desired mutation . Cell respiration was measured by monitoring dissolved oxygen levels in 30 mL of exponential phase yeast cells at an optical density of 0 . 2 in YEPD medium . Cells were cultured in 50 mL Erlenmeyer flasks sealed with a rubber bung to minimise the exchange of oxygen with the external environment . Each flask was equipped with a PreSens PSt3 oxygen sensitive spot ( NomaCorc ) and the percentage of dissolved oxygen in the medium was measured using an oxoluminescent device , the NOMASense oxygen analyser system ( NomaCorc ) [35] , every 15 minutes until percent oxygen reached below 10% for the wild-type strain . MitoTracker Red CMXRos ( Molecular Probes ) was added directly to a 500 µL volume of exponential phase culture in YEPD to a final concentration of 250 nM ( 1× ) or 2500 nM ( 10× ) . Cells were incubated at 30°C for 30 minutes , washed with fresh medium and resuspended in YEPD medium . Cells were mounted onto a glass slide and viewed immediately using a laser scanning confocal microscope ( Zeiss LSM 710 , Carl Zeiss Microimaging , Germany ) controlled using the ZEN 2010 software ( Carl Zeiss Microimaging , Germany ) . The excitation line used was 543 nm and the laser power was set at 2% . Cells were viewed using 630× optical magnification and 3× digital magnification . All samples were analysed using the same settings . The protein model of the human EF-G2mt protein was based on the crystal structure of the Thermus thermophilus EF-G protein ( Protein Data Bank 2bm0 ) [42] . This shares 39% residue identity . Selection of the T . thermophilus template and homology modelling was carried out using the SWISS-MODEL server in ‘project mode’ to enable alignment inspection prior to modelling [41] . The completed model was then submitted to the SWISS-MODEL suite of quality check programs which tests for model quality and stereochemistry using algorithms such as ANOLEA [56] and PROCHECK [57] . The model was visualised using the Visual Molecular Dynamics program ( VMD ) , version 1 . 8 . 7 ( University of Illinois ) and this software was also used for the assignment of protein secondary structure and the assessment of electrostatic potential . All sequences , both yeast and human , were obtained from the Ensembl database . There are three known human EF-G2mt protein isoforms but isoform I ( AAH15712 . 1 ) was used throughout this study . Human EF-G2mt SNPs were identified using the dbSNP database available on the National Center for Biotechnology Information ( NCBI ) website ( www . ncbi . nlm . nih . gov/SNP/index . html ) . The Lalign program available on the Swiss EMBnet server ( www . ch . embnet . org/software/lalign_form . html ) was used to generate global alignments of protein sequences , based on the BLOSUM50 matrix [58] . Graphical representations were constructed using the BioEdit version 7 . 0 . 5 sequence alignment editor .
|
The mitochondria are responsible for producing the cell's energy . Energy production is the result of carefully orchestrated interactions between proteins encoded by the mitochondrial DNA and by nuclear DNA . Sequence variations in genes encoding these proteins have been shown to cause disease and adverse drug reactions in patients . The cholesterol-lowering drugs statins are one class of drugs that interfere with mitochondrial function . Statins are one of the most prescribed drugs in the western world , but many users suffer side effects , commonly muscle pain . In severe cases this can lead to muscle breakdown and liver failure . In this study , we discover that disruption of a mitochondrial translation gene , EF-G2mt , impedes respiration and increases cell death when exposed to statin . Using the simple unicellular organism yeast as a model , the activity of naturally occurring human EF-G2mt variants is tested . Three of these variants render yeast cells more sensitive to statin . Patients who possess these EF-G2mt variations may be more susceptible to statin side effects . Importantly , the test for statin sensitivity also led to the discovery of mutants that have a reduced energy production capacity . The decreased ability to produce energy is linked to a number of diseases , including myopathies and liver failure .
|
[
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] |
[
"biotechnology",
"cellular",
"structures",
"subcellular",
"organelles",
"rna",
"interference",
"microbiology",
"model",
"organisms",
"mitochondrial",
"myopathy",
"genetics",
"and",
"genomics",
"mitochondrial",
"diseases",
"gene",
"expression",
"biology",
"molecular",
"biology",
"cell",
"biology",
"genetics",
"yeast",
"and",
"fungal",
"models",
"saccharomyces",
"cerevisiae",
"human",
"genetics",
"molecular",
"cell",
"biology",
"genetics",
"of",
"disease",
"pharmacogenomics"
] |
2012
|
Polymorphisms in the Mitochondrial Ribosome Recycling Factor EF-G2mt/MEF2 Compromise Cell Respiratory Function and Increase Atorvastatin Toxicity
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.