Complex interactions among MHC haplotypes in multiple sclerosis: susceptibility and resistance

Abstract

Mechanisms for observed associations within the major histocompatibility complex (MHC) and autoimmune diseases including multiple sclerosis (MS) remain uncertain. Genotyping of the HLA Class II DRB1 locus in 4347 individuals from 873 multiplex families with MS highlights the genetic complexity of this locus. Excess allele sharing in sibling pair families lacking DRB1*15 and DRB1*17 (58.5% sharing; P=0.012) was comparable to that seen where parents were DRB1*15 positive (62%, P=0.0006). DRB1*17 (P=0.00027) was clearly established as an MS susceptibility allele in addition to DRB1*15 (P<10⁻¹⁴). DRB1*14 showed striking under-transmission (P=0.000032) to affected offspring newly establishing this allele as a broadly acting resistance factor. Trans interactions were seen in both DRB1*15 and non-DRB1*15 bearing genotype combinations. DRB1*08 was transmitted preferentially with DRB1*15 (P=0.0114) and, in the presence of DRB1*08, the transmission of DRB1*15 was almost invariable (37 transmissions to one non-transmission). DRB1*01 was under-transmitted to offspring in the presence of DRB1*15 (P=0.019). Both DRB1*01 and DRB1*14 haplotypes carry DQA1*01-DQB1*05 alleles, suggesting a common DQ-related mechanism for the protection mediated by these haplotypes. These studies demonstrate that it is the Class II genotype that determines susceptibility and resistance to MS. By analogy with celiac disease and type I diabetes, the pattern of susceptibility strongly supports an autoimmune aetiology.

INTRODUCTION

Genetic epidemiological studies overwhelmingly indicate the influence of genes on susceptibility to multiple sclerosis (MS) (1–3). Nevertheless, there has been little progress in the identification of non-MHC susceptibility loci. For decades, the major histocompatibility complex (MHC) has been unambiguously associated with MS susceptibility (4,5). Initial associations were observed with the human leukocyte antigen (HLA) Class I region of the MHC (6,7) and later with polymorphisms in the HLA Class II region (8). Susceptibility has been fine-mapped to an extended HLA Class II haplotype DQA1*0102 DQB1*0602 DRB1*1501 DRB5*0101(9) with an estimated relative risk (RR) for this haplotype between 2 and 4 (10). Larger overall locus effects have been reported in relatively small samples using haplotype sharing as a measure (11,12). In larger samples, the same calculation gives a population RR (λ_MHC) of 1.5, implying that this locus accounts for a modest 14% of the total λ_sibs (13,14). However, prior calculations based on association with DRB1*1501 do not allow for the complex and genotypic-dependent interactions observed in other complex diseases such as coeliac disease (15), IDDM (16) and narcolepsy (17).

In addition to the association with the initially observed DRB1*1501 haplotype, there have been reports demonstrating the allelic heterogeneity at the DRB1 locus in various Southern European Ethnic groups (18–20). There is also evidence for non-Class II MHC susceptibility loci (14,21,22) and candidates have included TNF (23) and HLA-A (24). There have been sufficient data to dispel the notion that MHC-related susceptibility reflects the role of the HLA Class II allele DRB1*1501 alone.

The degree of polymorphism and linkage disequilibrium in the MHC region necessitate very large sample sizes to adequately examine allelic interactions. Here, we present a well-powered investigation into the complexity of the MHC within samples of affected relative MS pairs of siblings, cousins and aunt/uncle/niece/nephews (AUNN) and in addition, singleton ‘sporadic’ case families. We demonstrate considerable complexity at this locus, finding several features of the MHC common to disorders for which there is more direct evidence of an autoimmune pathogenesis. These data provide some explanation for previous observations and highlight additional challenges for understanding the nature of the MHC association.

RESULTS

The DRB1*15 association

Affected subjects from each type of multiplex and singleton family (see Materials and Methods) showed a positive association with DRB1*15. In the total sample, the DRB1*15 allele was transmitted 634 times and not transmitted 252 times; χ²=164.70, P<10⁻¹⁴ (Table 1). DRB1*15 was present in 55.6% (990/1781) of the MS patients and the allele frequency was 33%. In normal controls, the allele frequency of DRB1*15 was 12.6%, and 23% (41/178) of the controls were DRB1*15 positive. The RR for the DRB1*15 allele in MS patients was 2.41 (CI 1.84–3.17). The transmission of DRB1*15 to the unaffected offspring was also tested in these families. There was no significant transmission distortion. Eighty DRB1*15 alleles were transmitted and 93 DRB1*15 alleles were not transmitted, χ²=0.98, P=0.32. A difference in the transmission of DRB1*15 was not significant between the multiplex (OR=2.5) and the singleton families (OR=2.8), although the number of singleton families was relatively small (n=73).

Evidence for other allelic associations at the DRB1 locus

The over-transmission of DRB1*15 confounds the identification of other susceptibility and protective alleles. Therefore, to assess allelic associations other than DRB1*15, transmissions from non-DRB1*15 parents (i.e. X/Y—where X and Y are alleles other than DRB1*15) were examined and are presented in Table 2. This was assessed in matings where at least one parent was X/Y; the other parent could carry zero, one or two copies of DRB1*15 or could be of ‘unknown’ genotype. The most prominent associations in the combined sample are with the DRB1*17 allele (P=0.0030) and with the protective DRB1*14 allele (P=0.000055). A conservative Bonferroni correction for the 13 comparisons was applied and both DRB1*17 (P_c=0.039) and DRB1*14 (P_c=0.00072) associations remained significant. Alleles 11 and 12 were also under-transmitted and allele 8 was over-transmitted in the combined sample, but failed statistical significance upon stringent correction for multiple testing.

To test whether the transmissions from DRB1*15 negative parents were consistent with other families, we counted transmission of alleles from DRB1*15 heterozygous parents. The known association of DRB1*15 was subtracted from the total, in order to control for the known under-transmission of the other non-DRB1*15 alleles (Table 3). No transmissions were formally significant; however, the odds ratios of DRB1*14 (OR=0.46) and DRB1*17 (OR=1.2) were comparable with the odds ratios observed in Table 2. The same is observed for DRB1*11 and DRB1*12, both are again under-transmitted and consistent with the odds ratios in Table 2.

An interaction with DRB1*15

Transmission to MS offspring from a non-DRB1*15 bearing parent i.e. X/Y was also assessed, stratified by the presence or absence of DRB1*15. For example, in a mating of an X/Y parent and an X/15 parent, the child could be DRB1*15 positive or negative based upon transmission from the X/15 bearing parent. We then assessed the transmission of alleles from the non-DRB1*15 X/Y parent in both scenarios (Table 4). If there were no interactions between DRB1*15 and other DRB1 alleles, the transmission ratios should be the same for DRB1*15 positive and negative offspring.

The previously associated alleles DRB1*14 and DRB1*17 were found to be consistently under- and over-represented, respectively, in both DRB1*15 positive and negative children. For a number of other alleles including DRB1*09, *11, *12, *13 and *16, transmission was similar in 15 positive and negative children. For others, this was less clearly the case. Several alleles were dichotomized by stratification into DRB1*15 and non-DRB1*15 positive offspring, for example, DRB1*01. This allele was under-transmitted to affected DRB1*15 positive offspring with an odds ratio of 0.57 (P=0.011), however, when DRB1*15 was not present in the affected offspring, there was no distortion of transmission, suggesting that DRB1*01 is protective only in the presence of DRB1*15 (Table 4).

DRB1*08 showed the opposite trend as it was preferentially transmitted to DRB1*15 positive offspring with an odds ratio of 2.38 (P=0.0067), but was not over-transmitted when DRB1*15 was absent in the affected offspring (OR=1.15); this observation suggests that DRB1*08 is predisposing only in the presence of DRB1*15.

DRB1*10 also showed differences between the DRB1*15 defined groups (P=0.022) in that it was under-transmitted to the DRB1*15 positive offspring with an odds ratio of 0.17, whereas it was over-transmitted to the DRB1*15 negative children with an odds ratio of 2.25.

We hypothesized that the transmission of DRB1*15 from the DRB1*15 heterozygous parents would also be skewed when the children are stratified based upon the presence of DRB1*01 and DRB1*08. Moreover, any transmission differences should expectedly be in the same direction as that observed in Table 4. This is a test of the RR of the allele in the presence of DRB1*15; e.g. for DRB1*01, it is a test of DRB1*15/1 versus DRB1*15/X, where X is not DRB1*01. Transmissions were counted from heterozygous DRB1*15 positive parents to those children who are positive or negative for DRB1*01 or *08 inherited from the other parent (Table 5).

The findings observed in Table 5 are largely consistent with Table 4; the odds ratio of DRB1*15 transmission from a DRB1*01 negative parent to a DRB1*01 positive child (OR=1.6) is less than the odds ratio for a DRB1*01 negative child (OR=2.5). Although not significant (P=0.08), the trend is ‘as expected’ from the original observation (Table 4). The same findings are observed with DRB1*08 stratified children, DRB1*15 was transmitted with an odds ratio of 2.3 to the non-DRB1*08 bearing children, but when DRB1*08 was present, DRB1*15 was transmitted almost invariably (37 transmissions to one non-transmission).

Evidence for non-DRB1*15 bearing haplotypes linked to MS

We have previously shown increased sharing at the HLA DRB1 locus in MS sibling pairs having both parents DRB1*15 negative (13). For Table 6, allele sharing at DRB1 was assessed in sibling pairs from the total non-HLA DRB1*15-bearing parents. For example, sharing from a DRB1*X/Y mother was included in the analysis even when the father was DRB1*15/15 or DRB1*15/X. This is different from our previous analysis (13) when only families with both parents negative for DRB1*15 were counted.

The haplotype sharing from the DRB1*X/Y parents from the total sibling pair sample was 58.0% (χ²=7.68; P=0.0056). As we have shown that DRB1*17 is also positively associated with MS in the Canadian sample, we repeated the analysis with parents who were DRB1*15 and DRB1*17 negative (Table 7). In families with both parents negative for the two susceptibility alleles (n=30), the sharing was 70.5% (χ²=10.25; P=0.0013). Sharing from all parents lacking DRB1*17 and DRB1*15 was 58.5% (χ²=6.31; P=0.012).

As a comparison, the sharing of DRB1 alleles was also assessed among sibling pairs with parents heterozygous DRB1*15/X. The sharing observed from these heterozygous parents was 62.5% (χ²=16.1; P=0.0006) and not significantly different than the sharing observed when parents lacked DRB1*15 or DRB1*17.

DISCUSSION

The mechanisms by which HLA Class II associations impart risk for most diseases are yet to be defined. However, it is widely believed that the primary association for these putative autoimmune diseases is with alleles at these loci. In the case of MS, the HLA Class II DQA1*0102-DQB1*0602-DRB1*1501-DRB5*0101 haplotypic association provides evidence in support of an autoimmune reaction against myelin-related antigens presented to T-cells in the restricting context of HLA DRB1*1501.

Despite circumstantial evidence in support of this hypothesis, there have been a number of observations suggesting that this interpretation is facile. The data in this study suggest a more complex picture characterized not only by locus and allelic heterogeneity, but also by evidence of interactions that modify the RR to develop MS.

DRB1*15

The well-documented association with HLA DRB1*15 was observed in this sample of 873 Canadian MS families (Table 1). However, the effect of DRB1*15 is modest with an RR of 2.4, with only 55% of MS patients bearing this susceptibility allele (25,26).

Other allelic associations

If DRB1*15 was itself the sole susceptibility allele in the MHC, we should observe no increased sharing in sibling pairs where both parents are DRB1*15 negative. However, we observed 62.7% sharing at the DRB1 locus (13). Our conclusion, based upon the previous finding, was that there are haplotypes or loci other than DRB1*15 contributing to MS risk. We have extended this linkage evidence in the absence of the DRB1*15 bearing haplotypes by examining linkage in sibling pairs with any parent DRB1*15 negative (Table 6). The results show an increased sharing of 58% among the sibling pairs, from all DRB1*15 negative parents, with an mlod=1.7; χ²=7.68; P=0.0056. This is similar to the 62% sharing seen in sibling pairs from a DRB1*15/X parent.

To test the hypothesis that allelic heterogeneity may be responsible for this increased sharing, we examined whether other alleles were transmitted preferentially to affected offspring from non-DRB1*15 bearing parents. A number of other alleles were observed to increase and decrease risk of MS. Allele DRB1*17 is significantly associated with MS upon correction for multiple comparisons and is in keeping with previous case–control studies of Northern European MS patients (27,28). This finding is also consistent with findings from non-Northern European patients. Mexican and Sardinian MS patients both have an association with DRB1*17 (18,29). Consistent with this hypothesis is the observation that the Sardinian MS population is also associated with DRB1*15, but it was not originally identified due to the relatively low frequency of DRB1*15 alleles in the population (14). In this Canadian sample, 72% (1276/1781) of patients assessed were positive for DRB1*15 and/or DRB1*17 compared with 48% (85/178) of controls.

In contrast to the HLA*DRB1*15 and HLA DRB1*17 alleles, HLA DRB1*14 was identified as a strongly protective allele with an odds ratio of 0.31. This is as large a relative effect as the odds ratio for DRB1*15 (OR=2.5) for susceptibility and this has not been reported previously. DRB1*11 and DRB1*12 also showed a trend to significance with odds ratios of 0.75 and 0.55, respectively. This trend was consistent with the transmissions of DRB1*11 and DRB1*12 from DRB1*15 heterozygous parents (Table 3). Some alleles appeared to have a ‘neutral’ effect on risk (DRB1*13, *16 and *07); however, there may exist subgroups within the sample that may be associated with MS because some of the effects only emerge upon stratification.

HLA interactions with the DRB1*15-bearing haplotype

The transmission of DRB1 alleles from non-DRB1*15 parents stratified by the presence or absence of the DRB1*15 in the offspring was largely consistent; alleles DRB1*04, *09,*11, *12, *13 and *16 were observed to be equally transmitted between the two groups (Table 4). The same was true for DRB1*14 and DRB1*17, which were consistently under- and over-transmitted between the two groups, respectively.

Another positive trend for association was with the relatively rare allele DRB1*8 (Table 2). However, when the findings were restricted to patients who were DRB1*15 positive, the odds ratio for DRB1*8 increased from 1.64 to 2.39, P=0.0067 (Table 4). The non-DRB1*15 bearing MS offspring showed no transmission distortion of DRB1*8 and the odds ratio was 1.15.

DRB1*01 also showed distinct dimorphic patterns of transmission between the two groups. In the case of DRB1*01, the non-DRB1*15 bearing parents under-transmitted DRB1*01 to DRB1*15 positive offspring (Table 4). DRB1*10 seems to show a similar pattern in that DRB1*10 was over-transmitted to DRB1*15 negative children (OR=2.25) and under-transmitted to DRB1*15 positive children (OR=0.17).

To confirm or refute these findings for haplotypic interactions, we tested the transmission from DRB1*15 positive parents (versus DRB1*15 negative of Table 4) to offspring, in other words, the reciprocal of the previous analysis. The results were remarkably similar (Table 5) with DRB1*15 being relatively under-transmitted to DRB1*01 positive offspring when compared with DRB1*01 negative offspring (χ²=3.01; P=0.08). DRB1*15 was also preferentially transmitted to DRB1*08 positive offspring when compared with non-DRB1*08 bearing offspring (χ²=13.21; P=0.0002). Although DRB1*15 also showed similar findings when stratified by the presence or absence of DRB1*10, the sample size was exceedingly small with only three transmissions of DRB1*15 counted to the DRB1*10 positive offspring.

DRB1*01 has previously been implicated as a protective allele but of low magnitude; our odds ratio was 0.86 (30–34). Importantly, the three negatively associated DRB1 alleles (DRB1*01, *14 and *10) have in common the extended DQ haplotype DQA1*01-DQB1*05, suggesting that resistance may be encoded by the DQ genes versus the DRB1 alleles. This DQ haplotype is also protective in narcolepsy and DRB1*14 has been shown to be a protective haplotype in IDDM.

The interactions observed between DRB1 haplotypes suggest that the gene products are interacting to increase or decrease MS risk. The mechanism is unlikely to be operating via antigen presentation by the HLA DRβ/HLA DRα dimer complex, as the HLA DRA1 gene is relatively invariant. If the DRB1*15 restricted DRB1*01/*08/*10 interactions are true, this would lend further evidence implicating the DQ genes in MS susceptibility. Both the HLA DQA and HLA DQB genes are highly polymorphic, and they are in tight disequilibrium. From a functional perspective, the DQA and DQB alleles interact to form dimers for T-cell presentation. It is plausible that the formation of the HLA DQ α and HLA DQ β heterodimers may provide a molecular mechanism for the haplotypes to interact in trans to increase the MS risk, as has been suggested in IDDM and coeliac disease.

In IDDM, the DRB1*03/*04 heterozygotes have a greater RR when compared with DRB1*03 or *04 homozygotes. This is thought to be the result of trans-encoded HLA DQ genes encoding highly susceptible DQ α1β1 dimers (16). A similar mechanism is observed in coeliac disease with patients carrying the DQA1*05-DQB1*02 alleles in trans and also in cis, and in narcolepsy, the DQB1*0602 shows increased risk when DQB1*0601, DQB1*0501 or DQA1*01 are present in trans (15). These similarities support the evidence that MS is an autoimmune disorder with strong parallels to IDDM. Although DRB1*1501 is a susceptibility allele in MS and imparts resistance in IDDM, other alleles, as shown in this study, behave similarly in the two conditions. We have shown here that the alleles DRB1*01, *11 and particularly DRB1*14 unexpectedly impart resistance in MS as has been reported for IDDM. With these similarities, it might be expected that MS may share other susceptibility loci with other autoimmune conditions. However, PTPN22 was shown to be associated with IDDM, rheumatoid arthritis, systemic lupus erythematous and Graves disease (35) was not associated to MS (36).

Further insight suggesting HLA DQ involvement in MS was observed in a study of Afro-Brazilian MS patients which showed a positive association for the DQ alleles in the absence of the DRB1*1501 allele (37). Moreover, a recent Sardinian study of over 400 MS patients demonstrated that both DRB1 and DQB1 act independently to increase MS risk (14). In contrast, among African-American MS patients, the DRB1 locus appears to be the susceptible locus rather than the DQ loci (38). If both DR and DQ independently and in combination determine risk, population associations at these loci could favour one or the other depending on allele frequencies and additional haplotypic features. The DQ locus was not genotyped for this investigation, though it would be beneficial for future resolution of the involvement of these loci in MS susceptibility.

Non-Class II MHC susceptibility loci

Although the alleles implicated in MS susceptibility may represent a functional hierarchy related to antigen presentation, it cannot be excluded that they are simply in functional or physical disequilibrium with as yet unidentified loci. The literature contains many reports in support of a non-Class II susceptibility locus (14,21–24). In favour of non-DRB1 susceptibility loci, the observation that haplotype sharing where both parents were DRB1*15 negative is very similar to when parents are positive (Table 6). Furthermore, when we examined sharing from parental haplotypes lacking both DRB1*15 and DRB1*17, the finding was still increased at 58.5% (Table 7). This suggests that not only allelic heterogeneity, but also locus heterogeneity exists at the MHC locus. Recent scans spanning the MHC support this hypothesis (14,21,22).

Complexity and MHC

These results serve to highlight the genetic complexity of the MHC in MS, a putative autoimmune disorder. Although it is possible that the identification of specific susceptibility genes may eliminate much of this, there remain many unresolved features of this region to date which are central to its role in disease. We observed multiple alleles, protective alleles and the effects of allelic interactions. The observations of MHC-linked complexity will be of importance to MS and other autoimmune diseases, where there is also increasing evidence of MHC heterogeneity (39,40). The MHC is a logical target for a high-density single nucleotide polymorphism map in disease (41) and we have completed a high-density map of this region in more than 2000 affected individuals (submitted for publication). The results localize MHC-related susceptibility to the MHC Class II region alone.

MATERIALS AND METHODS

MS patients and their ascertainment

The families were ascertained as part of an ongoing Canadian Collaborative Project on the Genetic Susceptibility to MS (CCPGSMS), for which the methodology has been previously described (42). The material consisted of those with sibling pairs (n=442), cousin pairs (n=184), AUNN pairs (n=174) and singleton families (n=73) for a total of 873 families and 1781 affected offspring (Table 8). Families were counted once; for example, if a family included an affected aunt plus two affected nephews (i.e. a sibling pair), the family was counted once as an AUNN relative pair family. The families were of Canadian and Caucasian descent. One-hundred and seventy-eight unaffected and unrelated individuals were also genotyped to serve as a normal control sample. They were also of Canadian and Caucasian descent.

Genotyping

The genotyping of the HLA DRB1 gene was performed using a low-resolution allele-specific PCR amplification method (43). Twenty-four reactions amplified allelotypes corresponding to alleles DRB1*1, 4, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17 and 18 as well as amplicons for the DRB3, DRB4 and DRB5 genes. Allele frequencies were estimated from an additional 178 unrelated and unaffected individuals of Caucasian descent.

Four-thousand three-hundred and forty-seven individuals from the 873 families were genotyped at DRB1 locus.

Statistical analysis

Transmission disequilibrium test (TDT) was performed using the sib_tdt program of ASPEX 2.3 analysis package available at (ftp://lahmed.stanford.edu/pub/aspex). The TDT counts the number of times an allele is transmitted to affected offspring from heterozygous parents (44). Parents who were not genotyped were reconstructed whenever possible from the unaffected siblings only. In the case when one parent was unknown, the transmissions were counted only in those instances where both the genotyped parent and the affected offspring were heterozygous for different alleles (e.g. a 7, 15 parent and 7,15 child were not counted), in order to avoid directional bias. The 184 cousin pair families and the 174 AUNN pair families were separated into independent nuclear families for the TDT analysis. Statistical significance was determined empirically with Monte Carlo simulation. The allele labels of heterozygous parents were inverted randomly with probability of 50%, and the TDT χ² statistics was calculated for each permutation step. The proportion of simulated values that exceed the observed value is the derived empirical significance level (45).

Non-parametric linkage analyses were performed with the sib_ibd program of Aspex 2.3 package.

The odds ratios and the chi-squared values for the combined data were calculated by Mantel–Haenszel analysis (46).

MEMBERS OF THE CANADIAN COLLABORATIVE STUDY GROUP

Vancouver, British Columbia: D. Paty, MD, J. Oger, MD, V. Devonshire, MD, S. Hashimoto, MD, J. Hooge, MD, L. Kastrukoff, MD, T. Traboulsee, MD; London, Ontario: G. Rice, MD; M. Kremenchutzky; Calgary, Alberta: L. Metz, MD; Edmonton, Alberta: S. Warren, PhD; Saskatoon, Saskatchewan: W. Hader, MD; Toronto, Ontario: P. O'Connor, MD, T. Grey, MD, M. Hohol MD; Ottawa, Ontario: M. Freedman, MD; Kingston, Ontario: R. Paulseth, MD; Montreal, Quebec: Y. Lapierre, MD (Montreal Neurological Institute), P. Duquette, MD (Hopital Notre Dame); Quebec City, Quebec: J-P. Bouchard, MD; Halifax, Nova Scotia: T. Murray, MD, V. Bhan, MD, C. Maxner, MD; St. John's, Newfoundland: W. Pryse-Phillips, MD, M. Stefanelli, MD.

ACKNOWLEDGEMENTS

We would especially like to thank Holly Armstrong, Beverly Scott, Dr Jan Hillert, Dr Arturs Ligers, Dr Jamie Steckley, Katie Morrison and Angie Shaw for their help with genotyping and data organization. Informed consent was obtained from all subjects and the experiments performed for this investigation comply with current guidelines and ethics. Funding for the Canadian Collaborative Project on Genetic Susceptibility to MS is by the MS Society of Canada Scientific Research Foundation. The MS Society of Canada has funded studentships to D.A.D. and C.J.W. A.D.S. is a Michael Smith Distinguished Scholar.

Conflict of Interest statement. A. Sadovnick is a member of the Speaker's Bureau for Berlex, Biogen, Teva and Serono. No other conflicts of interest were declared.

References