|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: LMM CEL DEA CS LSW JAT DAH. Performed the experiments: LMM CEL KD CS PMC CA. Analyzed the data: LMM CEL JC KD DEA CS BH NW. Contributed reagents/materials/analysis tools: LMM CEL JC JRO SLH AC SS TI PLDJ LSW JAT DAH. Wrote the paper: LMM CEL JC KD DEA LSW JAT DAH.
Multiple sclerosis (MS) and type 1 diabetes (T1D) are organ-specific autoimmune disorders with significant heritability, part of which is conferred by shared alleles. For decades, the Human Leukocyte Antigen (HLA) complex was the only known susceptibility locus for both T1D and MS, but loci outside the HLA complex harboring risk alleles have been discovered and fully replicated. A genome-wide association scan for MS risk genes and candidate gene association studies have previously described the IL2RA gene region as a shared autoimmune locus. In order to investigate whether autoimmunity risk at IL2RA was due to distinct or shared alleles, we performed a genetic association study of three IL2RA variants in a DNA collection of up to 9,407 healthy controls, 2,420 MS, and 6,425 T1D subjects as well as 1,303 MS parent/child trios. Here, we report “allelic heterogeneity” at the IL2RA region between MS and T1D. We observe an allele associated with susceptibility to one disease and risk to the other, an allele that confers susceptibility to both diseases, and an allele that may only confer susceptibility to T1D. In addition, we tested the levels of soluble interleukin-2 receptor (sIL-2RA) in the serum from up to 69 healthy control subjects, 285 MS, and 1,317 T1D subjects. We demonstrate that multiple variants independently correlate with sIL-2RA levels.
Multiple sclerosis (MS) and type 1 diabetes (T1D) are common, organ-specific inflammatory disorders that continue to increase in global prevalence. The processes leading to both T1D and MS are genetically determined and are thought to involve an autoimmune mechanism. After decades of research into the genetic basis of both MS and T1D, the Human Leukocyte Antigen Complex was the only known susceptibility locus for both T1D and MS. The sequencing of the human genome followed by the generation of the haplotype map, a catalogue of common genetic variation, has allowed the elucidation of allelic variants that define disease risk. Our groups have performed genome-wide association scans and candidate gene studies in both T1D and MS; the final results have identified loci outside the HLA harboring fully replicated risk alleles. Here, we show that the IL-2RA gene encoding a critical regulator of immune responses, the alpha chain of the interleukin-2 receptor, harbors variants that differentially confer risk to MS and T1D. In addition, several independent variants correlate with levels of soluble interleukin-2 receptor in the serum. This finding has critical implications for the field of complex disease genetics as it emphasizes the caution that must be taken when interpreting results for such a complex region with multiple susceptibility alleles.
Recent genome wide association (GWA) and candidate gene studies across human autoimmune disease revealed a shared genetic architecture . These include PTPN22, associated with systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), T1D, and Graves' Disease (GD) , STAT4, associated with SLE and RA , and the IL7R and KIAA0350 gene regions, which are shared between T1D and MS –. The IL2RA gene region is shared among T1D –, MS ,, GD , SLE  and RA ,. This overlap of risk loci among autoimmune diseases raises the possibilities that either (1) the same alleles, (2) non-shared, disease-specific alleles, or perhaps (3) a combination of shared and disease-specific alleles confer risk to each of the individual diseases.
In the IL-2RA gene region, a GWA study for MS risk alleles and a large-scale fine-mapping study in T1D provided compelling evidence for a shared autoimmunity locus. A GWA study for MS susceptibility genes performed by The International Multiple Sclerosis Genetics Consortium  highlighted two SNPs in the IL-2RA gene: rs12722489 (Odds Ratio (OR) for minor allele =0.80; 95% confidence interval (c.i.)=0.74–0.86, P=2.96×10−8) and rs2104286 (OR=0.84; 95% c.i.=0.79–0.90; P=2.16×10−7). These are in moderate linkage disequilibrium (LD) with each other (r2=0.62; ). The MS association at IL2RA has recently been replicated in over 600 multiplex families from Canada (rs12722489, P=0.009; OR=0.81; 95% c.i.=0.70–0.93) and 1,146 subjects with MS and 1,309 healthy controls from Australia (rs2104286, P=0.033; OR=0.86; 95% c.i.=0.75–0.99). In an extension analysis  using data from 12,360 subjects previously reported and new data from 11,019 unrelated MS subjects, 13,616 controls and 2,811 trio families (8,433 individuals) from across Europe, the association for MS risk at the two IL2RA variants became unequivocal (rs12722489, OR=0.81 (95% c.i. 0.77–0.85), P=2.24×10−15; relative risk, RR, =0.81 (95% c.i. 0.72–0.91), P=5.47×10−4; rs2104286, OR=0.80 (95% c.i. 0.77–0.84), P=2.38×10−23; RR=0.78 (95% c.i. 0.71–0.86)). Furthermore, this study demonstrated that rs2104286 is the primary association, and thus accounts for the association signal observed at rs12722489 .
For T1D susceptibility, two associations are known to exist at IL2RA. In a large-scale fine-mapping study of over 300 SNPs in the IL2RA-RBM17 region in our T1D collection, we localized the association to T1D susceptibility to two groups of SNPs located in the 5′ region and intron 1 of IL2RA; any one or more SNPs from each group could potentially be the causal variant(s) . The minor alleles at rs41295061 and rs11594656 were found to confer protection to T1D in a case-control DNA collection of 5,312 T1D subjects and 6,855 controls (rs41295061, OR, =0.65, rs11594656, OR=0.87) and 2,612 families with T1D (rs41295061, RR=0.70, rs11594656, RR=0.89) .
The IL-2/IL-2RA(CD25) pathway plays an essential role in regulating immune responses . IL-2 is central for both expansion and apoptosis of T cells, while high concentrations of soluble IL-2RA (sIL-2RA) are found in sera from healthy subjects and are increased in subjects with autoimmune disease, inflammation and infection –. Interestingly, we have previously shown that T1D-associated variants correlate with reduced levels of sIL-2RA .
Our knowledge of the IL-2R pathway and its central role in regulating immune responses prompted us to examine whether disease susceptibility at IL2RA to T1D and MS is due to shared or distinct genetic variants. First, we demonstrate extensive allelic heterogeneity between T1D and MS, including an allelic variant that is associated with susceptibility to one autoimmune disease but protection to the other. By extending previous genotype/phenotype correlations at IL2RA, we provide insight into both common and distinct functional mechanisms. Second, we extend our findings on the correlation between sIL-2RA levels and IL2RA genotype . Using regression analyses, we show that sIL-2RA levels are determined by independent groups of SNPs, similar to what we show for disease susceptibility. Taken together, we demonstrate heterogeneity in the production of sIL-2RA in association with the genetic heterogeneity reported here. The approach described in this work will be instrumental for future investigations of complex causal mechanisms involved in human disease.
The most associated IL2RA SNP for MS susceptibility is rs2104286 located in intron 1 of IL2RA ,,,. In the MS case-control and family collections we have analyzed, rs2104286 has an OR of 0.85 (95% c.i. 0.79–0.92, P=6.27×10−7) (Table 1, Figure 1, Tables S1, S2, and S3). For T1D susceptibility, Lowe et al.  reported independent associations with two groups of indistinguishable SNPs, marked by rs41295061 (‘Group I’) and rs11594656 (‘Group II’) located in the 5′ region of the IL2RA gene. Here, we test these two SNPs for MS susceptibility. Single locus tests show no evidence of association between MS susceptibility and Group I (rs41295061; P=0.10, Table 1). We note that assuming an effect size of rs41295061 as observed for T1D susceptibility (OR in the order of 0.6), the power to detect this effect is 97% in the parent/child trios and 100% in the MS case-control collection, given a significance level of 0.05 (Table S4, S5).
Furthermore, Group II is associated with MS (rs11594656; P=7.67×10−4 (Table 1, Figure 1). Surprisingly, at rs11594656, the minor allele A is associated with protection from T1D (OR=0.87), but susceptibility to MS (OR=1.17, Table 1) in the MS case-control collection. The lack of association in the parent/child trio collection may be due to low statistical power, which is only 31% for a variant with OR=1.1 for this sample size and P<0.05 (Table S6).
The lack of MS association to Group I SNPs and the opposing effects associated with Group II SNPs indicates the presence of allelic heterogeneity between T1D and MS at Group I and Group II SNPs. In addition, we note that the MS-association observed at rs11594656 presents an independent MS-association from rs2104286 (Table S7). Taken together, rs2104286 marks an independent association from Group II SNPs (marked by rs11594656); we term this association ‘Group III’. Table S8 shows all IL2RA region SNPs in LD with rs2104286.
In order to explore the association of Group III SNPs to T1D susceptibility, we performed forward logistic regression analysis of the Group I, II and III SNPs in 6,425 T1D cases and 6,862 controls with complete genotyping data. The results are consistent with our previous study : Group I has the strongest association with T1D (rs41295061, P=6.43×10−25; Table 1). The first selected SNP in the regression analysis is rs41295061 and the second SNP to be added to the model including rs41295061 is rs11594656 (P=2.07×10−10, Table S9). Interestingly, Group III also shows association with T1D (rs2104686, P=1.27×10−13, Table 1). When we add rs2104286 to the model that includes both rs41295061 and rs11594656, this SNP adds to the model (P=1.30×10−5; Table S10). These data indicate that rs2104286 (marking Group III) is independently associated with T1D. At rs2104286, it is the minor allele G of rs2104286 that confers protection from both MS and T1D (Table 1, Figure 1). We note here that the major allele at all T1D-associated loci discovered so far at IL2RA encodes the susceptibility allele.
Defining the heterogeneous genetic basis at IL2RA is critical for the success of functional studies aiming to connect the risk alleles with immunophenotypes and autoimmune mechanisms controlled by this locus. In a collection of T1D plasma samples, Lowe et al.  reported a correlation between the T1D IL2RA susceptibility alleles and decreased levels of sIL-2RA. This raised the possibility of a link between T1D susceptibility and the levels of this biomarker of peripheral inflammation . Here, we investigate the correlation of sIL-2RA and the newly identified Group III SNPs, marked by rs2104286, which associates with both MS and T1D. In a replication study of up to 69 healthy control samples and 285 MS case samples we first confirm the previously observed correlation between rs11594656 with sIL-2RA levels; however, the low minor allele frequency of rs41295061 results in statistical power that was too low to detect the association with sIL-2RA in these sample collections (Table 2; Tables S11, S12). Most interestingly, however, an additional correlation between genotype and sIL-2RA level is observed at rs2104286 in our healthy control, MS and T1D collections, where the minor allele associates with decreased sIL-2RA levels. Given that the minor allele at rs2104286 associates with protection from both MS and T1D, this finding is unexpected because decreased sIL-2RA levels correlate with T1D susceptibility alleles at rs41295061 and rs11594656. This led us to investigate whether the three SNPs were marking independent associations with sIL-2RA levels, similarly to what we have observed for disease susceptibility at IL2RA. Indeed, using regression analyses in the T1D case collection, we show that the associations between sIL-2RA levels and rs11594656, rs41295061 and rs2104286 (Table 2) are independent from each other (Tables S14, S15).
Our combined genetic analyses of IL2RA variants in MS and T1D result in the discovery of a third, novel group of associated SNPs with T1D (Group III) and identifies a remarkable degree of allelic heterogeneity at this autoimmune susceptibility locus. This demonstrates the presence of (1) a T1D allele not associated with MS (rs41295061 marking Group I), (2) an allele conferring susceptibility to T1D but protection from MS (rs11594656 marking Group II) and (3) an allele shared between T1D and MS (rs2104286 marking Group III). The discovery of allelic heterogeneity between MS and T1D at IL2RA may only be a small window into the complexities that the IL2RA region harbors: GWA studies for both RA and SLE have also observed associations at IL2RA (Figure S1, Table S16); the overlap of associations among these and other diseases should be the focus of future studies. While any of the tested SNPs may be in LD with the true causal variant, the allelic heterogeneity we observe between MS and T1D provides strong evidence for the necessity of performing fine-mapping studies in each disease individually that associates with IL2RA. Another example of such allelic heterogeneity has been observed at the shared autoimmunity locus PTPN22 encoding a lymphotyrosine phosphatase. While T1D, RA and CD all show disease associations that map to the same R620W variant (rs2476601) –, it is the 620W variant that associates with risk to T1D and RA, but protection from CD . We note that R620W has not shown association with MS susceptibility in the populations analyzed thus far ,, but studies employing larger sample sizes will need to further address this variant in MS.
Our analysis of how disease susceptibility correlates with sIL-2RA levels suggest discordance between sIL-2RA level and disease susceptibility and calls for studies addressing causality of sIL-2RA in autoimmune disease. It is plausible that the three independent genetic associations marked by Group I to III SNPs present independent biological pathways that contribute to disease susceptibility. These pathways may involve transcriptional regulation of IL2RA, levels of surface expression of IL-2RA, in addition to serum sIL-2RA levels. In light of multiple, independent associations present at IL2RA, the genotype/phenotype correlations observed here and previously  may require extension to haplotype/phenotype correlations in sample sizes an order of magnitude greater than are currently available. Nevertheless, these data represent a comparative study between MS/T1D susceptibility and production of sIL-2RA and show that multiple variants contribute independently not only to disease susceptibility but also to an individual's sIL-2RA level.
All case and control subjects were of self-reported white ethnicity and were enrolled under study protocols approved by the Institutional Review Board of each institution that contributed.
MS and T1D cases: Trio families and MS cases were collected as described in our recent investigation of patients with MS . Subjects with MS all meet McDonald criteria for MS. T1D subjects were recruited as part of the Juvenile Diabetes Research Foundation/Wellcome Trust Diabetes and Inflammation Laboratory's British case collection (Genetic Resource Investigating Diabetes) , which is a joint project between the University of Cambridge Department of Pediatrics and the Department of Medical Genetics at the Cambridge Institute for Medical Research. Most cases were <16 years of age at the time of collection. All were under age 17 years at diagnosis, resided in Great Britain, and were of European descent (self-reported).
Healthy Control Subjects: Healthy adult control subjects were recruited through the Brigham and Women's Hospital and the University of California at San Francisco, as previously described . They consisted of unrelated individuals who were self-reported as being of non-Hispanic white origin and having no history of chronic inflammatory disease. In addition, we included data from 1,679 control individuals collected throughout the USA as part of a GWAS of bipolar disorders sponsored by the NIMH (http://zork.wustl.edu/nimh). The GB control subjects were obtained from two collections, with 5,239 obtained from the British 1958 Birth Cohort, all born during one week in 1958 (National Child Development Study) and the remaining 1,445 controls selected from the UK Blood Services (UKBS) control collection . All GB control subjects were of white ethnicity.
SNPs were genotyped using the iPLEX Sequenom MassARRAY platform, TaqMan (Applied Biosystems), or MIP technology (Affymetrix) in accordance with the manufacturer's instructions. We analyzed only SNPs with high quality data (>95% genotype call rate, Hardy-Weinberg equilibrium in controls or unaffected parent P-value>0.001). MS collections were genotyped at the Broad Institute: rs41295061 and rs11594656 were genotyped using iPLEX Sequenom MassARRAY platform. The previously published data for the MS cases and healthy controls from the USA as well as the MS cases from GB were obtained from MIP technology . GB healthy controls and GB T1D case were genotyped for rs2104286 using TaqMan genotyping at the Diabetes and Inflammation Laboratory. The previously published T1D data for rs41295061 and rs11594656 were obtained from TaqMan and MIP technology .
ELISA measurement of sIL-2RA was performed according to the manufacturer's recommendations (BD Biosciences). Serum samples were diluted 120 using PBS supplemented with 10% FBS. Microtiter plates were read using a Biorad Benchmark microplate reader. T1D plasma samples, healthy control and MS subject serum samples were stored at −80°C prior to analyses. A log10 transformation of total sIL-2RA concentration was used to provide a Normally distributed outcome. For T1D plasma samples, the analysis was adjusted for independently associated covariates, namely, age, duration of T1D and plasma storage duration. The healthy control subject population consisted of 60.3% females, 29.7% males, with an average age of 43 (range=20–68) and an average sample storage duration of 2.1 years (range=1.27–3.15). The MS subject population consisted of 74.2% females, 25.8% males, with an average age of 43 (range=18–73) and an average sample storage duration of 2.4 (range=1.1–3.3).
All statistical analyses were performed in either the Stata or R statistical systems. Single locus tests, logistic regression analyses, 2-d.f. locus-based tests were performed as described in . Briefly, logistic regression analyses for the GB case-control collection were adjusted for 12 broad geographical regions within GB to minimize any confounding due to variation in allele frequencies across the country . A multiplicative allelic effects model was assumed as it was not significantly different from the full genotype model for any of the SNPs (except for rs2104286 in the USA case-control collection, for which a full model was chosen as it was significantly different from the multiplicative model; P=6.57×10−3). SNPs were modeled as a numerical indicator variable coded 0, 1 or 2, representing the number of occurrences of the minor allele. In the forward logistic regression analysis, we start by assessing the evidence against the most significant SNP being alone sufficient to model the association . No specific mode of inheritance for the most associated SNP (A>a) or any additional SNP with significant independent effects of disease susceptibility was assumed, so genotype risks of A/A and A/a were modeled relative to the a/a genotype. Combined P values for the USA and GB case-control were stratified by population. Measures LD, D' r2 were calculated using the Haploview package . Power calculations were performed using the method described in .
Comparison of IL2RA variants genotyped in T1D, MS, RA and SLE. Minor allele associations with disease are shown. Odds ratios and 95% confidence intervals of association results are shown from the current study and previously published studies ,,.
(0.11 MB DOC)
Single-locus test P values for rs2104286, rs11594656 and rs41295061 in 2,115 MS cases and 6,902 healthy controls with complete genotype information (analysis stratified by population). MAF, minor allele frequency. OR, odds ratio.
(0.03 MB DOC)
Single-locus test P values for rs2104286, rs11594656 and rs41295061 in 1,183 MS cases and 582 healthy controls from the USA with complete genotype information. MAF, minor allele frequency. OR, odds ratio.
(0.03 MB DOC)
Single-locus test P values rs2104286, rs11594656 and rs41295061 in 932 MS cases and 6,320 healthy controls from GB with complete genotype information. MAF, minor allele frequency. OR, odds ratio.
(0.03 MB DOC)
Power calculations to detect variants with odds ratios ranging from 1.1 to 1.4 and a minor allele frequency (MAF) of 0.10 using 1,250 parent/child trios. MAF, minor allele frequency. OR, odds ratio.
(0.03 MB DOC)
Power calculations to detect the effect of variants with odds ratios ranging from 1.1 to 1.3 and a minor allele frequency (MAF) of 0.10 using 2,400 MS cases and 9,100 healthy controls. MAF, minor allele frequency. OR, odds ratio.
(0.03 MB DOC)
Power calculations to detect the effect of variants with odds ratios (OR) ranging from 1.1 to 1.4 and a minor allele frequency (MAF) of 0.25 using 1,250 parent/child trios.
(0.03 MB DOC)
Regression analysis (a) adding rs11594656 and rs41295061 to rs2104286 and reverse regression analysis (b) adding rs2104286 to rs11594656 and rs41295061 for 2,115 MS cases and 6,902 controls with complete genotype information (analysis stratified by population). 1 Results for a model assuming multiplicative effects and 2 for a model assuming genotype effects (full model) are shown. OR, odds ratio; Pdiff=P value for tests between multiplicative and full models.
(0.05 MB DOC)
LD between rs2104286 and all common variants located between 5.904,506 bp and 6.403,954 bp on chromosome 10. Positions are given using NCBI build 36. rs and in-house (DIL) SNP numbers are shown. LD measures are calculated based on 32 CEPH individuals; genotyping data are obtained from HapMap (www.hapmap.org) and .
(0.12 MB DOC)
Regression analysis adding rs2104286 and rs11594656 to rs41295061 and regression analysis adding rs41295061 to rs2104286 and rs11594656 in 6,425 T1D cases and 6,862 controls. 1 Results for a model assuming multiplicative effects and 2 for a model assuming genotype effects (full model) are shown. OR, odds ratio; Pdiff=P value for tests between multiplicative and full models.
(0.05 MB DOC)
Regression analysis (a) adding rs2104286 to rs41295061 and rs11594656 and reverse regression analysis (b) adding rs41295061 and rs11584656 to rs2104286 in 6,425 T1D cases and 6,862 controls. 1 Results for a model assuming multiplicative effects and 2 for a model assuming genotype effects (full model) are shown. OR, odds ratio; Pdiff=P value for tests between multiplicative and full models.
(0.06 MB DOC)
Power calculations to detect the effect of variants with a minor allele frequency of 0.10 using 70 subjects. Power calculations were performed using the method described in .
(0.04 MB DOC)
Power calculations to detect the effect of variants with a minor allele frequency of 0.10 using 280 subjects. Power calculations were performed using the method described in .
(0.04 MB DOC)
Covariates associated with sIL-2RA concentrations in T1D analysis. We adjusted for the covariates year of birth, duration of disease and duration of storage of the plasma sample prior to processing, as these were all independently associated with log10-transformed sIL-2RA concentrations. Covariates were selected using forward and then reverse regression. The following covariates: gender, broad geographical region and the age and month when the plasma sample was collected all had P-values>0.05 when added to the selected covariates. Year of birth, duration of disease and the duration of storage of the plasma sample prior to processing were independently associated (P>0.05) and together account for 10.3% of the total variation with log10-transformed sIL-2RA concentration with the direction and magnitude shown in the table below. *%CV is the variation accountable by each covariate.
(0.04 MB DOC)
Regression analysis adding rs2104286 and rs41295061 to rs11594656 and the reverse regression analysis adding rs11594656 to rs2104286 and rs41295061 in complete data for 1,167 T1D cases using log10-transformed sIL-2RA concentrations. 1 Results for a model assuming multiplicative effects and 2 for a model assuming genotype effects (full model) are shown. Pdiff=P value for tests between multiplicative and full models.
(0.06 MB DOC)
Regression analysis (a) adding rs2104286 to rs11594656 and rs41295061 and reverse regression analysis (b) adding rs11594656 to rs2104286 and rs41295061 and adding rs41295061 to rs2104286 and rs11594656 in complete data for 1,167 T1D cases using log10-transformed sIL-2RA concentrations.
(0.07 MB DOC)
r2 values for the four SNPs associated with T1D, MS or RA and the four SNPs studied in a GWA study for SLE susceptibility loci ,,. r2 values are based on a maximum of 32 CEPH individuals. Allele frequencies are shown in the top diagonal line.
(0.05 MB DOC)
We gratefully acknowledge the participation of all the patients and control subjects. We acknowledge use of the DNA from the British 1958 Birth Cohort collection, funded by the Medical Research Council and Wellcome Trust. We thank The Avon Longitudinal Study of Parents and Children laboratory in Bristol and the British 1958 Birth Cohort team, including S. Ring, R. Jones, M. Pembrey, W. McArdle, D. Strachan and P. Burton for preparing and providing the control DNA samples. Thanks to the National Institutes of Mental Health for generously allowing the use of their genotype data.
The International Multiple Sclerosis Genetics Consortium
Mark J. Daly, Broad Institute of Harvard University and Massachusetts Institute of Technology, Cambridge, MA, and Massachusetts General Hospital, Harvard Medical School, Boston; Paul I. W. de Bakker, Broad Institute of Harvard University and Massachusetts Institute of Technology, Cambridge, MA and Harvard Partners Center for Genetics and Genomics, Boston, MA; Stacey B. Gabriel, Broad Institute of Harvard University and Massachusetts Institute of Technology, Cambridge, MA; Daniel B. Mirel, Broad Institute of Harvard University and Massachusetts Institute of Technology, Cambridge, MA; Adrian J. Ivinson, Harvard Center for Neurodegeneration and Repair, Harvard Medical School, Boston, MA Margaret A. Pericak-Vance, Duke University Medical Center, Durham, NC and University of Miami School of Medicine, Miami, FL; Simon G. Gregory, Duke University Medical Center, Durham, NC; John D. Rioux, Broad Institute of Harvard University and Massachusetts Institute of Technology, Cambridge, MA and Universite de Montreal, Montreal Heart Institute, Montreal, Canada; Jacob L. McCauley, Center for Human Genetics Research, Vanderbilt University Medical Center, Nashville, TN; Jonathan L. Haines, Center for Human Genetics Research, Vanderbilt University Medical Center, Nashville, TN; Lisa F. Barcellos, University of California at Berkeley, Berkeley, CA and University of California at San Francisco, San Francisco, CA.
The authors have declared that no competing interests exist.
This work was funded by the Juvenile Diabetes Research Foundation (JDRF) International, the Wellcome Trust, the National Multiple Sclerosis Society (NMSS), the National Institute for Health Research Cambridge Biomedical Research Centre, and the National Institutes of Health (P01 AI39571). The Cambridge Institute for Medical Research is the recipient of a Wellcome Trust Strategic Award (079895). Further supported by a JDRF Postdoctoral Fellowship (LMM), a grant by the American Cancer Society (DEA), an NIH grant (to The IMSGC, NS049477), a Harry Weaver Neuroscience Scholar award by the NMSS (PLD), and a Jacob Javits Merit Award (NS2427, to DAH) from the National Institute of Neurological Disorders and Stroke. LSW is a Wellcome Trust Principal Research Fellow.