Abstract
Multiple sclerosis (MS), a common central nervous system inflammatory disease, has a major heritable component. Susceptibility is associated with the MHC class II region, especially HLA-DRB5*0101–HLA-DRB1*1501–HLA-DQA1*0102–HLA-DQB1*0602 haplotypes(hereafter DR2), which dominate genetic contribution to MS risk. Marked linkage disequilibrium (LD) among these loci makes identification of a specific locus difficult. The once-leading candidate, HLA-DRB1*15, localizes to risk, neutral, and protective haplotypes. HLA-DRB1*15 and HLA-DQB1*0602, nearly always located together on a small ancestral chromosome segment, are strongly MS-associated. One intervening allele on this haplotype, viz. HLA-DQA1*0102, shows no primary MS association. Two Canadian cohorts (n = 830 and n = 438 trios) genotyped for HLA-DRB1, HLA-DQA1 and HLA-DQB1 alleles were tested for association using TDT. To evaluate epistasis involving HLA-DRB1*15, transmissions from HLA-DRB1*15-negative parents were stratified by the presence/absence of HLA-DRB1*15 in affected offspring. All 3 alleles contribute to MS susceptibility through novel epistatic interactions. HLA-DQA1*0102 increased disease risk when combined with HLA-DRB1*1501 in trans, thereby unambiguously implicating HLA-DQ in MS susceptibility. Three-locus haplotypes demonstrated that HLA-DRB1*1501 and HLA-DQB1*0602 each influence risk. Transmissions of rare morcellated DR2 haplotypes showed no interaction with HLA-DQA1*0102. Incomplete haplotypes bearing only HLA-DRB1*1501 or HLA-DQB1*0602 did not predispose to MS. Balanced reciprocal transmission distortion can mask epistatic allelic association. These findings implicate epistasis among HLA class II alleles in human immune responses generally, provide partial explanation for intense linkage disequilibrium in the MHC, have relevance to animal models, and demonstrate key roles for DR2-specific interactions in MS susceptibility. MHC disease associations may be more generally haplotypic or diplotypic.
Keywords: genetics, MHC, linkage disequilibrium
Multiple sclerosis (MS) is an inflammatory disease of the central nervous system characterized by myelin loss, axonal pathology, and progressive neurological dysfunction (1). With a prevalence of approximately 1/1,000 in Canada, MS is the most common cause of acquired neurological disability in young adults. Although the etiology of MS remains largely unknown, it is clear that both genetic and environmental components play important roles in pathogenesis (2). It is widely believed that MS is a CD4+ TH1-mediated autoimmune disorder. Support for this view comes from a variety of studies involving both murine models (3, 4) and human genetic studies (5–8).
The major histocompatibility complex (MHC) dominates the genetic influences on MS risk (7). While the association of HLA class II alleles with MS susceptibility has been long established, the involvement of these alleles is complex. Within this region, multiple genes and several of their individual alleles have been implicated (9, 10); these alleles appear to influence MS risk through a variety of complex interactions. Considerable uncertainty remains over which of these alleles may be primarily involved (11).
The central question of whether HLA-DRB1 or HLA-DQB1 is the primary MS susceptibility gene has not been resolved due to intense linkage disequilibrium (LD). HLA-DRB1*1501 and HLA-DQB1*0602 are in tight LD on a common ancestral haplotype and in northern Europeans are almost always transmitted together. In other populations, where MS is less common, greater haplotypic diversity is frequently observed. Admixture studies have used this diversity to evaluate the effects of HLA-DRB1 and HLA-DQB1 in these populations. An early study involving a small cohort of Afro-Brazilians suggested that HLA-DQB1*0602 may be the primary locus (12). Subsequently, a larger study of African-Americans implicated HLA-DRB1*15 (13). While admixture studies have provided important insights, haplotypic differences between Africans and northern Europeans mean that the primacy of neither HLA-DRB1*15 nor HLA-DQB1*0602 has been established in northern Europeans.
Analysis of Canadian families lacking HLA-DRB1*1501 has demonstrated additional HLA-DRB1 alleles which confer susceptibility and resistance but not additional MHC loci (14). Associations at class I suggested by case-control studies to be independent of HLA class II (15) were not sustained once account was taken of linkage disequilibrium (8, 16). In addition to the MHC, modest associations with other non-MHC loci have been found for IL7R (17, 18), IL2R (19), EVI5 (20) and KIF1B (21). The latter is the largest of these with an odds ratio (OR) of 1.34 in sporadic MS and 1.74 in multiplex families (22).
The LD obstacle is compounded by complex interactions between HLA-DRB1 alleles/haplotypes. The full HLA-DRB1 genotype or diplotype (the 2 haplotypes in combination) largely determine genetic risk (7, 23). Multiple HLA-DRB1 alleles are MS-associated, some dependent on the presence/absence of HLA-DRB1*15 in trans (7). The HLA class II alleles have been implicated in MS animal models (24, 25), and epistasis between HLA class II alleles influences susceptibility and clinical course in experimental autoimmune encephalomyelitis (3, 26). Recently it has been shown that HLA class I alleles tag haplotypes differential for MS risk with HLA-DRB1*1501 present on susceptibility, neutral, and protective haplotypes (11). These data clearly show that this locus does not carry risk alone and completely changes the perspective on the HLA-DRB1 vs. HLA-DQB1 controversy; indeed, the effect of HLA-DRB1*1501 can be completely changed in polarity by adjacent variation.
In light of the complexity of these findings, systematic examination of potential epistatic interactions between HLA-DRB1, HLA-DQA1, and HLA-DQB1 was undertaken. In this, the largest family-based investigation of the HLA class II loci in MS, alleles at each of these loci were found to influence MS susceptibility through novel epistatic interactions. In particular, HLA-DQA1*0102, which instructively showed no independent association, was found to interact strongly with HLA-DRB1*15 in trans, increasing MS risk in the presence of HLA-DRB1*15 and playing a protective role in its absence.
Results
Association of HLA Class II Alleles.
HLA-DRB1, HLA-DQA1 and HLA-DQB1 were each assessed for association with MS using the transmission disequilibrium test (TDT) [supporting information (SI) Table S1]. HLA-DRB1*15 was positively associated with MS with an OR of 2.4 (χ12 = 88; uncorrected P = 5.3 × 10−21). Transmissions from HLA-DRB1*15-negative parents were assessed to identify additional HLA class II associations independent of HLA-DRB1*15 (7). In this analysis, 9 HLA class II alleles showed significant MS-association (Table S2).
Epistatic Interactions Between HLA-DRB1 Alleles.
Epistasis was first defined by Bateson (27) as an extension of the concept of dominance, whereby an allele at one locus may mask the effect of variation at a second locus. The term is used here broadly as an interaction between 2 alleles (in cis or trans) in which the risk associated with a particular allele depends on the presence or absence of a second allele. Because interactions between HLA-DRB1 alleles/haplotypes determine MS risk (7, 23), we assessed each of the HLA class II alleles for epistatic interactions with HLA-DRB1*15. To assess these alleles for interaction with HLA-DRB1*15, transmissions of HLA class II alleles from HLA-DRB1*15-negative parents were stratified by the presence or absence of HLA-DRB1*15 (transmitted from the other parent) in affected offspring (7). Alleles which exhibit different odds ratios in HLA-DRB1*15-positive and -negative offspring may be said to interact with HLA-DRB1*15 or another locus in strong LD with HLA-DRB1*15. Interactions of HLA-DRB1*01 and HLA-DRB1*10 with HLA-DRB1*15 observed previously (23, 28) were noted when the 2 cohorts were pooled (Table S3). HLA-DRB1*17 also showed weak nominal significance. Consistent with and extending previous studies (7, 28), HLA-DRB1*15 was strongly overtransmitted in the presence of HLA-DRB1*08 (T/NT = 22/3; χ12 = 14; P = 0.00015).
Epistasis Between HLA-DRB1*15 and HLA-DQA1*0102.
Alleles of HLA-DQA1 were examined for epistasis with HLA-DRB1*15 by similar stratification. Three alleles of HLA-DQA1 were differentially transmitted to HLA-DRB1*15-positive and HLA-DRB1*15-negative cases in the primary cohort (Table 1). Evidence for epistasis with HLA-DRB1*15 was strongest for HLA-DQA1*0102 (comparison χ12 = 7.4; P = 0.0063). With HLA-DRB1*15 present, HLA-DQA1*0102 was overtransmitted (OR = 2.1; χ12 = 3.9; P = 0.048) and in its absence, undertransmitted (OR = 0.64; χ12 = 4.0; P = 0.047). Similarly, HLA-DQA1*0201 transmission was reciprocally distorted, trending toward overtransmission to HLA-DRB1*15-positive cases and toward undertransmission in HLA-DRB1*15-negative cases. HLA-DQA1*0101 and HLA-DQB1*0501 were protective with HLA-DRB1*15 present and neutral in its absence.
Table 1.
Transmission of HLA-DQA1 and HLA-DQB1 alleles from HLA-DRB1*15-negative parents to offspring stratified by the presence or absence of HLA-DRB1*15 in the primary and secondary cohorts
| Cohort | HLA-DRB1*15-Positive Children | HLA-DRB1*15-Negative Children | Comparison | |||||
|---|---|---|---|---|---|---|---|---|
| Primary | ||||||||
| HLA-DQA1 | T | NT | OR | T | NT | OR | χ12 | P |
| *0101 | 23 | 48 | 0.48 | 60 | 62 | 0.97 | 5.2 | 0.023 |
| *0102 | 21 | 10 | 2.1 | 32 | 50 | 0.64 | 7.4 | 0.0063 |
| *0103 | 8 | 16 | 0.5 | 27 | 36 | 0.75 | 0.66 | 0.42 |
| *0201 | 51 | 37 | 1.38 | 52 | 68 | 0.76 | 4.3 | 0.037 |
| *03 | 35 | 30 | 1.2 | 66 | 61 | 1.1 | 0.061 | 0.81 |
| *04 | 19 | 6 | 3.2 | 15 | 13 | 1.2 | 2.9 | 0.089 |
| *0501 | 37 | 32 | 1.2 | 81 | 44 | 1.8 | 2.3 | 0.13 |
| *0505 | 26 | 40 | 0.65 | 53 | 53 | 1 | 1.8 | 0.17 |
| *06 | 0 | 1 | 0 | 2 | 1 | 2 | —a | 1 |
| HLA-DQB1 | ||||||||
| *0201 | 38 | 31 | 1.2 | 82 | 41 | 2. | 2.5 | 0.11 |
| *0202 | 44 | 33 | 1.3 | 41 | 50 | 0.82 | 2.4 | 0.12 |
| *0301 | 36 | 54 | 0.67 | 70 | 78 | 0.90 | 1.2 | 0.27 |
| *0302 | 24 | 15 | 1.6 | 46 | 43 | 1.1 | 1.1 | 0.30 |
| *0303 | 9 | 6 | 1.5 | 19 | 20 | 0.95 | 0.55 | 0.46 |
| *0304 | 0 | 0 | — | 1 | 0 | — | —a | 1. |
| *0305 | 0 | 0 | — | 0 | 1 | 0 | —a | 1. |
| *0402 | 19 | 6 | 3.2 | 16 | 14 | 1.1 | 3.0 | 0.082 |
| *0501 | 19 | 37 | 0.51 | 51 | 50 | 1.0 | 4.0 | 0.045 |
| *0502 | 4 | 4 | 1. | 6 | 8 | 0.75 | —a | 1. |
| *050301 | 4 | 13 | 0.31 | 9 | 15 | 0.6 | 0.90 | 0.34 |
| *0601 | — | — | — | — | — | — | — | — |
| *0602 | 1 | 1 | 1. | 2 | 1 | 2. | —a | 1. |
| *0603 | 8 | 16 | 0.5 | 23 | 36 | 0.64 | 0.23 | 0.63 |
| *060401 | 14 | 8 | 1.75 | 17 | 24 | 0.71 | 2.8 | 0.093 |
| *0605 | 6 | 2 | 3. | 9 | 11 | 0.82 | —a | 0.22 |
| Secondary | ||||||||
| HLA-DQA1 | ||||||||
| *0101 | 12 | 23 | 0.52 | 24 | 25 | 0.96 | ||
| *0102 | 15 | 8 | 1.9 | 17 | 30 | 0.57 | 5.3 | 0.022 |
| *0103 | 5 | 8 | 0.62 | 14 | 20 | 0.7 | ||
| *0201 | 25 | 26 | 0.96 | 27 | 26 | 1.0 | ||
| *03 | 29 | 26 | 1.1 | 36 | 38 | 0.95 | ||
| *04 | 4 | 10 | 0.4 | 6 | 8 | 0.75 | ||
| *0501 | 20 | 19 | 1.0 | 36 | 28 | 1.3 | ||
| *0505 | 30 | 19 | 1.6 | 51 | 33 | 1.5 | ||
| *06 | 0 | 1 | 0 | 0 | 1 | 0 | ||
| HLA-DQB1 | ||||||||
| *0201 | 22 | 20 | 1.1 | 36 | 24 | 1.5 | ||
| *0202 | 27 | 25 | 1.1 | 27 | 22 | 1.2 | ||
| *0203 | 1 | 0 | — | 0 | 0 | — | ||
| *0301 | 35 | 37 | 0.95 | 48 | 44 | 1.1 | ||
| *0302 | 26 | 16 | 1.6 | 30 | 23 | 1.3 | ||
| *0303 | 8 | 7 | 1.1 | 5 | 8 | 0.62 | ||
| *0402 | 7 | 10 | 0.7 | 6 | 11 | 0.55 | ||
| *0501 | 15 | 22 | 0.68 | 30 | 26 | 1.2 | ||
| *0502 | 7 | 4 | 1.8 | 4 | 4 | 1. | ||
| *050301 | 0 | 3 | 0 | 0 | 4 | 0 | ||
| *0602 | 1 | 0 | — | 0 | 1 | 0 | ||
| *0603 | 6 | 10 | 0.6 | 14 | 22 | 0.64 | ||
| *060401 | 9 | 11 | 0.82 | 5 | 19 | 0.26 | ||
| *0605 | 1 | 0 | — | 4 | 1 | 4. | ||
aFisher's exact test (two-tailed) was used if any of the expected cell counts were <5.
In a secondary cohort, we posed the specific question generated in the first cohort. We replicated the proposed HLA-DRB1*15–HLA-DQA1*0102 epistasis (comparison χ12 = 5.3; P = 0.022; Table 1). Reciprocal transmission distortion of HLA-DQA1*0102 was proportional to that seen in the primary cohort when stratified by the presence/absence of HLA-DRB1*15.
The magnitude of the HLA-DRB1*15–HLA-DQA1*0102 interaction was estimated by pooling the 2 cohorts. HLA-DQA1*0102 showed no overall transmission distortion (T/NT = 85/98; χ12 = 0.92; P = 0.34) but significant overtransmission from HLA-DRB1*15-negative parents to HLA-DRB1*15-positive (OR = 2.0; χ12 = 6.0; P = 0.014) and significant undertransmission to HLA-DRB1*15-negative offspring (OR = 0.61; χ12 = 7.4; P = 0.0063). This reciprocal distortion (comparison χ12 = 13) remained significant after robust permutation correction for multiple testing (corrected P = 0.0080). Transmissions of HLA class II alleles from parents negative for both HLA-DRB1*15 and HLA-DQA1*0102 showed no transmission distortion (data not shown).
We next asked if these results were independent of HLA-DRB1*15 interactions, and no influence of these results was detected. HLA-DQA1*0102 from parents negative for alleles previously confirmed to interact specifically with HLA-DRB1*15 (7, 28) (HLA-DRB1*01, HLA-DRB1*08, and HLA-DRB1*10) was still significantly overtransmitted to HLA-DRB1*15-positive (OR = 2.1; χ12 = 5.8; P = 0.016) and significantly undertransmitted to HLA-DRB1*15-negative offspring (OR = 0.59; χ12 = 6.44; P = 0.011). Comparison of the 2 strata revealed epistasis (comparison χ12 = 12), which remained significant after permutation correction for multiple comparisons (corrected P = 0.0095), demonstrating that HLA-DRB1*15–HLA-DQA1*0102 epistasis is independent of potential HLA-DRB1 allelic interactions.
Inversely, transmissions from parents to offspring for the rare HLA-DQA1*0102-negative but HLA-DRB1*15-positive haplotypes stratified by the presence of HLA-DQA1*0102 in affected offspring (received in trans) were as predicted by the epistatic interaction observed above in pooled data: HLA-DRB1*15 trends toward overtransmission (T/NT = 8/3) with HLA-DQA1*0102 present and toward undertransmission (T/NT = 1/5) in the absence of HLA-DQA1*0102. Comparison of these strata reveals a marginally significant interaction (Fisher's exact test; comparison P = 0.049). More clearly, in larger numbers HLA-DQA1*0102 transmission to 338 unaffected siblings stratified by the presence/absence of HLA-DRB1*15 was neutral for both HLA-DRB1*15-positive (T/NT = 8/7) and HLA-DRB1*15-negative (T/NT = 19/18) individuals.
Demonstration that epistasis between HLA-DRB1*15 and HLA-DQA1*0102 is independent of HLA-DRB1 allelic interactions very strongly implicates an independent role for HLA-DQ in susceptibility.
Given that there is no known physical interaction between HLA-DQA1 and HLA-DRB1, HLA-DQB1 was targeted because it forms a functional dimer with HLA-DQA1. Problematically, HLA-DQB1*0602 is not prevalent on non-DR2 haplotypes, whereas HLA-DQA1*0102 is highly so. Three-locus class II haplotypes were assessed for transmission distortion (Table 2). Expectedly, HLA-DRB1*15–HLA-DQA1*0102–HLA-DQB1*0602 haplotypes were significantly overtransmitted (OR = 2.6; χ12 = 120; P = 1.2 × 10−27). Uncommon to rare incomplete haplotypes lacking any one or 2 of these loci (Table 2) were neutrally transmitted (T/NT = 146/153; χ12 = 0.16; P = 0.69), suggesting that susceptibility is attenuated or absent on incomplete haplotypes. In particular, those lacking either HLA-DRB1*15 or HLA-DQB1*0602 were neutral (T/NT = 114/129 and T/NT = 139/145, respectively).
Table 2.
Transmission of HLA-DRB1*15—HLA-DQA1*0102—HLA-DQB1*0602 and morcellated haplotypes lacking HLA-DRB1*15, HLA-DQA1*0102 or HLA-DQB1*0602
| Complete vs. incomplete haplotypes | T | NT | OR |
|---|---|---|---|
| HLA-DRB1*15—HLA-DQA1*0102—HLA-DQB1*0602 | 431 | 165 | 2.6 |
| HLA-DRB1*15-positive, HLA-DQB1*0602-negativea | 14 | 13 | 1.1 |
| HLA-DRB1*15—HLA-DQA1*04—HLA-DQB1*0402 | 1 | 1 | 1 |
| HLA-DRB1*15—HLA-DQA1*06—HLA-DQB1*0301 | 1 | 2 | 0.5 |
| HLA-DRB1*15—HLA-DQA1*0102—HLA-DQB1*0303 | 0 | 2 | 0 |
| HLA-DRB1*15—HLA-DQA1*0102—HLA-DQB1*0501 | 2 | 2 | 1 |
| HLA-DRB1*15—HLA-DQA1*0102—HLA-DQB1*0502 | 2 | 2 | 1 |
| HLA-DRB1*15—HLA-DQA1*0102—HLA-DQB1*0603 | 18 | 8 | 2.2 |
| HLA-DRB1*15—HLA-DQA1*0102—HLA-DQB1*05031 | 1 | 1 | 1 |
| HLA-DRB1*15—HLA-DQA1*0103—HLA-DQB1*0601 | 3 | 3 | 1 |
| HLA-DRB1*15—HLA-DQA1*0505—HLA-DQB1*0301 | 4 | 0 | 4 |
| HLA-DRB1*15—HLA-DQA1*0505—HLA-DQB1*0603 | 0 | 1 | 0 |
| HLA-DRB1*15-negative, HLA-DQB1*0602-positive | 7 | 6 | 1.2 |
| HLA-DRB1*08—HLA-DQA1*03—HLA-DQB1*0602 | 1 | 0 | — |
| HLA-DRB1*08—HLA-DQA1*04—HLA-DQB1*0602 | 1 | 1 | 1 |
| HLA-DRB1*08—HLA-DQA1*0102—HLA-DQB1*0602 | 0 | 1 | 0 |
| HLA-DRB1*12—HLA-DQA1*0102—HLA-DQB1*0602 | 2 | 0 | — |
| HLA-DRB1*13—HLA-DQA1*0102—HLA-DQB1*0602 | 0 | 2 | 0 |
| HLA-DRB1*13—HLA-DQA1*0103—HLA-DQB1*0602 | 2 | 1 | 2 |
| HLA-DRB1*17—HLA-DQA1*0102—HLA-DQB1*0602 | 1 | 1 | 1 |
| HLA-DRB1*15-positive, HLA-DQB1*0602-positive | 0 | 2 | 0 |
| HLA-DRB1*15—HLA-DQA1*0101—HLA-DQB1*0602 | 0 | 2 | 0 |
| HLA-DRB1*15- and HLA-DQB1*0602-negative | 107 | 123 | 0.87 |
aThe total for HLA-DRB1*15-positive, HLA-DQB1*0602-negative haplotypes excludes HLA-DQB1*0603 already implicated in MS susceptibility, and supported here.
Among haplotypes simultaneously HLA-DRB1*15-positive and HLA-DQB1*0602-negative, it became apparent that the single haplotype accounting for overtransmission contained the closely-related HLA-DQB1*0603, reportedly overtransmitted in MS (29) and similarly protective against type I diabetes (30). Among HLA-DRB1*15-positive, HLA-DQB1*0602/*0603-negative haplotypes, transmission was neutral (T/NT = 14/13). Similarly, among very rare HLA-DRB1*15-negative, HLA-DQB1*0602-positive haplotypes, transmission was also neutral (T/NT = 7/6). Further complexity is suggested by comparison of the transmission of morcellated haplotypes that contain or do not contain HLA-DQA1*0102 (Table 2). Comparison of transmission of pooled haplotypes negative for one of either HLA-DRB1*15 or HLA-DQB1*0602 with transmission for intact DR2 haplotypes was marginally significant (χ12 = 4.3; P = 0.038) and comparison of transmissions of the complete DR2 haplotype to the entire pool of incomplete haplotypes was also significant (χ12 = 46; P = 9.2 × 10−12).
Furthermore, in MS offspring selected for having HLA-DRB1*X–HLA-DQA1*0102–HLA-DQB1*Z haplotypes (where *X and *Z represent any non-DR2 allele), we saw overtransmission of archetypical DR2 3-locus haplotypes from the other parent (T/NT = 35/7) which differed from that of HLA-DRB1*15–HLA-DQA1*Y–HLA-DQB1*Z (T/NT = 6/6), (comparison P = 0.027) implying trans interactions of HLA-DQA1*0102 specific to DR2 haplotypes. This argues for interaction between DQA1*0102 and DQB1*0602.
Heterogeneity among HLA-DRB1*15-bearing haplotypes prompted examination of other alleles/haplotypes. As the most “neutral” control (T/NT = 150/151 in a previous study of a northern European-derived population; ref. 7), HLA-DRB1*13 was selected. Unexpectedly, this neutrality similarly masks several balanced reciprocal transmissions to affected, much as is seen for HLA-DQA1*0102 (Table 3). HLA-DRB1*13–HLA-DQA1*0505–HLA-DQB1*0301 showed marked overtransmission (T/NT = 76/41;χ12 = 10; P = 0.0012), while HLA-DRB1*13–HLA-DQA1*0103–HLA-DQB1*0603 was undertransmitted (T/NT = 52/107; χ12 = 19; P = 1.3 × 10−5) and HLA-DRB1*13–HLA-DQA1*0102–HLA-DQB1*06041 trended toward undertransmission (T/NT = 46/64; χ12 = 2.9; P = 0.086).
Table 3.
Common 3-locus HLA Class II haplotypes in MS: transmission heterogeneity
| Haplotype | T | NT | OR | χ12 | P |
|---|---|---|---|---|---|
| HLA-DRB1*13—HLA-DQA1*0505—HLA-DQB1*0301 | 76 | 41 | 1.9 | 10 | 0.0012 |
| HLA-DRB1*13—HLA-DQA1*0102—HLA-DQB1*06041 | 46 | 64 | 0.72 | 2.9 | 0.086 |
| HLA-DRB1*13—HLA-DQA1*0103—HLA-DQB1*0603 | 52 | 107 | 0.49 | 19 | 1.3 × 10−5 |
| HLA-DRB1*04—HLA-DQA1*03—HLA-DQB1*0301 | 53 | 103 | 0.51 | 16 | 6.2 × 10−5 |
| HLA-DRB1*04—HLA-DQA1*03—HLA-DQB1*0302 | 150 | 127 | 1.2 | 1.9 | 0.17 |
These results find independent confirmation in Israel. There, MS is not HLA-DRB1*1501- but HLA-DRB1*13-associated, the same HLA-DRB1*13–HLA-DQA1*0505–HLA-DQB1*0301 haplotype coherently unmasked here (30). Similarly, of the 2 common HLA-DRB1*04-bearing haplotypes, only HLA-DRB1*04–HLA-DQA1*03–HLA-DQB1*0301 was undertransmitted; HLA-DRB1*04–HLA-DQA1*03–HLA-DQB1*0302 showed no evidence for transmission distortion (Table 3). These haplotypes differ only at HLA-DQB1, but the difference in transmission was statistically significant (χ12 = 16; P = 5.3 × 10−5), illustrating the clear advantages of MHC haplotypes to dissect associations.
Epistasis involving HLA-DQA1 and the haplotype interactions involving HLA-DQB1 implicate HLA-DQ in MS susceptibility. To examine whether HLA-DRB1 exerts effects independent of HLA-DQ or is secondarily associated, haplotypes sharing HLA-DQA1*0505 and HLA-DQB1*0301 alleles were assessed for transmission distortion (Table 4). The HLA-DRB1*11-bearing haplotype was significantly undertransmitted (OR = 0.68; χ12 = 9.8; P = 0.0017), whereas that bearing HLA-DRB1*13 was overtransmitted (OR = 1.9; χ12 = 10; P = 0.0012). These results cohere with the incomplete haplotype data, implying roles for HLA-DRB1 independent of HLA-DQ loci.
Table 4.
Haplotypes sharing HLA-DQA1*0505 and HLA-DQB1*0301 in MS: the effect of varying HLA-DRB1 on transmission
| Haplotype | T | NT | OR | χ12 | P |
|---|---|---|---|---|---|
| HLA-DRB1*01—HLA-DQA1*0505—HLA-DQB1*0301 | 11 | 8 | 1.4 | 0.47 | 0.49 |
| HLA-DRB1*04—HLA-DQA1*0505—HLA-DQB1*0301 | 0 | 1 | 0 | —a | 1 |
| HLA-DRB1*11—HLA-DQA1*0505—HLA-DQB1*0301 | 107 | 158 | 0.68 | 9.8 | 0.0017 |
| HLA-DRB1*12—HLA-DQA1*0505—HLA-DQB1*0301 | 16 | 27 | 0.59 | 2.8 | 0.093 |
| HLA-DRB1*13—HLA-DQA1*0505—HLA-DQB1*0301 | 76 | 41 | 1.9 | 10 | 0.0012 |
| HLA-DRB1*14—HLA-DQA1*0505—HLA-DQB1*0301 | 1 | 1 | 1 | —a | 1 |
| HLA-DRB1*15—HLA-DQA1*0505—HLA-DQB1*0301 | 4 | 0 | — | —a | 0.12 |
| HLA-DRB1*16—HLA-DQA1*0505—HLA-DQB1*0301 | 1 | 0 | — | —a | 1 |
| HLA-DRB1*17—HLA-DQA1*0505—HLA-DQB1*0301 | 0 | 2 | 0 | —a | 0.5 |
| HLA-DRB1*18—HLA-DQA1*0505—HLA-DQB1*0301 | 1 | 1 | 1 | —a | 1 |
aBinomial test (two-sided) was used if T + NT <10.
Discussion
Searches for the primary MHC association in multiple sclerosis have inconclusively implicated HLA-DRB1*1501 or HLA-DQB1*0602. The central problem is one common to other HLA-associated disorders: more generally the dilemma is how to identify a disease allele (say, a relatively common variant) in a region of tight linkage disequilibrium when one allele almost completely predicts the adjacent alleles. In the case of MS, the relevant candidate alleles are common in both MS patients and unaffected family members.
Emphasis on HLA-DRB1 and HLA-DQB1 has shifted focus away from other functional loci in the HLA class II region, including HLA-DQA1*0102. Understandably this locus had been discounted since HLA-DQA1*0102 has not been shown to have a primary association with MS.
The presence of HLA-DQA1*0102 on HLA-DRB1*15-negative haplotypes crucially enabled investigation for epistasis. Trans interactions were demonstrated by a modification of the TDT, where transmissions of HLA-DQA1*0102 from HLA-DRB1*1501-negative parents were shown to depend on the HLA-DRB1*1501 status of the offspring (Fig. 1). HLA-DQA1*0102 was overtransmitted if HLA-DRB1*1501 was the homologous allele present and undertransmitted if HLA-DRB1*1501 was absent. Interactions in cis were demonstrated by examination of 3 locus haplotypes (Fig. 2). Observed cis interactions between HLA-DRB1*15 and HLA-DQA1*0102 and between HLA-DRB1*15 and HLA-DQB1*0602 may reflect either functional relationships between these alleles or additional haplotype-specific effects. The availability of a large sample of MS families allowed this effect to be separated from previously identified interactions at HLA-DRB1. HLA-DQA1 cis and trans effects, independent of HLA-DRB1, strongly implicated HLA-DQ in the MS immune response, perhaps through formation of cis or trans HLA-DQA1*0102–HLA-DQB1*0602 encoded dimers. However, the morcellated haplotypes studied here appear to implicate the intact HLA-DRB1–HLA-DQA1–HLA-DQB1 haplotype at a minimum. Intriguingly, DR2 haplotypes exhibit the strongest LD among HLA class II haplotypes (3, 31), implying functional requirements under selection (3).
Fig. 1.
Trans Epistasis between HLA-DRB1*1501 and HLADQA1*0102. (A) HLA-DQA1*0102 (blue) is overtransmitted (OR = 2.1) when HLADRB1*1501 (red) is also present in affected offspring. (B) When HLADRB1*1501 is absent, HLA-DQA1*0102 is undertransmitted (OR = 0.64).
Fig. 2.
Transmission of morcellated DR2 haplotypes reveals cis epistasis in the MHC class II region. (A) The classical MS susceptibility haplotype bearing HLA-DRB1*1501, HLA-DQA1*0102, and HLA-DQB1*0602 confers susceptibility to MS with an OR of 2.6. (B) Haplotypes bearing HLA-DRB1*1501 in the absence of HLA-DQB1*0602 confer no susceptibility (OR = 1.1). (C) Haplotypes bearing HLA-DQB1*0602 in the absence of HLA-DRB1*1501 are also neutral (OR = 1.2). (D) HLA-DQA1*0102-bearing haplotypes which lack HLA-DRB1*1501 and HLA-DQB1*0602 are protective (OR = 0.87).
Epistasis among alleles at different HLA class II loci which exhibit reciprocal transmission distortion warrants consideration in future association studies. Concentration of epistatic alleles on the same susceptibility haplotype supports complex functional relationships between HLA-DR and HLA-DQ loci determining MS susceptibility. Genetic interactions between alleles at HLA-DRB1, HLA-DQA1, or HLA-DQB1 have been reported in narcolepsy (32) and diabetes (33) and functional epistasis between HLA-DRB1 and HLA-DRB5 has been demonstrated in murine models of MS (3). In addition to the allelic interactions reported here, the differential susceptibility associated with different HLA-DRB1*15-bearing haplotypes (11) may imply structural and epigenetic influences. The latter may explain the reciprocal distortion seen for HLA-DQA1.
These findings may be relevant to other aspects of the immune response (34) and to finding genes for complex traits. We have previously shown how false positives can result from LD (16). Here we show how balanced reciprocal transmission distortion masks association and envision similar limitations in detecting epigenetic effects which have now been shown to be important in MS at the same loci (35) and may be responsible for this phenomenon. Reciprocal distortion suggests that pairings of HLA-DQA1 molecules with HLA-DQB1 are selected, not necessarily independent of HLA-DR–HLA-DQ epistasis. The variable effect of HLA-DRB1 on DQ pairings may have analogies for HLA-DQA1*0102–HLA-DQB1*0602 haplotypes, explaining populations in which the HLA-DQB1*0602 association exceeds that for HLA-DRB1*1501.
Epistatic interactions between HLA-DRB1, HLA-DQA1, and HLA-DQB1 appear to play defining roles in MS susceptibility. Analysis stringently restricted to parents negative for any HLA-DRB1 allele having nominally significant interactions showed that the epistatic effect of HLA-DQA1*0102 was undiminished, demonstrating independence from HLA-DRB1. Previously described “dose effects” of HLA-DRB1*1501 may reflect, at least in part, epistasis between HLA-DR and HLA-DQ loci.
Strategic accounting for allelic associations was essential since haplotypic transmission probability is influenced by alleles on a parent's other haplotype. Large samples with available parents enabled the obligate haplotype construction. Assumptions made (36) in measuring individual contributions of loci using sib-pair haplotype sharing may not apply in complex circumstances. Haplotypic associations in MS, different by population, suggest additional yet unidentified haplotype-specific features and 2 recent papers indicate that there may be additional haplotypic-specific influences within the MHC (35, 37), as improbable as this may have seemed a priori. However, the results presented here illuminate how MS risk is influenced by the DR-DQ region, strongly implying that allelic variations in all 3 genes in the interval localized by dense SNP mapping (8) are operative. This paradigm may be more generally applicable to human immune and autoimmune responses. It is possible that MHC-disease associations will more generally be haplotypic rather than allelic in nature. Furthermore, the recent observation that vitamin D may influence HLA-DRB1*1501 expression via a vitamin D response element which is largely specific to HLA-DRB1*1501 haplotypes shows that gene-environment interactions may act within the MHC (37), potentially adding a fourth element to an archetypic risk haplotype in MS.
Materials and Methods
Subjects.
To minimize the number of false positive test results, a 2-stage replication design was chosen in advance. The primary cohort consists of 2,024 individuals (830 trios) derived from 394 Canadian multiplex MS families. The secondary cohort similarly consists of 982 individuals (438 trios) derived from 295 families. Both cohorts were ascertained through the Canadian Collaborative Project on Genetic Susceptibility to MS (CCPGSMS). Ascertainment methodology has been described in detail previously (38). Informed consent was obtained from all subjects and the experiments performed for this investigation comply with current guidelines and ethics.
HLA Typing.
Each sample was genotyped for HLA-DRB1 with either a low- (7, 39) or a high-resolution allele-specific PCR amplification (8). Before analysis, all HLA-DRB1 genotypes were converted to 2-digit precision. HLA-DQA1 and HLA-DQB1 were genotyped using the same protocol. Fifteen allele-specific PCR reactions were used to type HLA-DQA1 and 23 were used to characterize HLA-DQB1. For each Mendelian error, the entire family was regenotyped for each locus. Where Mendelian errors remained, the family was eliminated from subsequent analyses.
Statistical Methods.
Individuals for whom consistent genotypes could not be obtained for each of the loci under study were removed from the study. Missing parental genotypes were reconstructed from the genotypes of unaffected offspring where possible. Hardy-Weinberg equilibrium was assessed at each locus, with none showing significant deviation among founder genotypes.
HLA class II alleles were assessed for association with the TDT (40). Tests for interaction between alleles were performed on the raw transmission counts as described previously (7). To correct for multiple testing, permutation tests were performed (106 permutations). Within each family, the transmission status of each allele was permuted randomly in a fashion that preserved the IBD status of affected offspring and the haplotypic relationships between alleles.
Three-locus haplotypes of HLA class II alleles were constructed using an E-M algorithm (8). Only haplotypes which could be inferred unambiguously were included in these analyses.
Supplementary Material
Acknowledgments.
The authors thank R.E. Phillips and C. Bangham for helpful comments on the manuscript. This research was done with the help of core facilities at the Wellcome Trust Centre for Human Genetics and support from the United Kingdom MS Society. The Canadian Collaborative Study on Genetic Susceptibility to MS was instrumental in the work leading up to these investigations. S.V.R. is supported by a Medical Research Council studentship. B.M.H was supported by the Oxford department of clinical neurology. M.R.L. was supported by the Clarendon Fund and a summer studentship from the University of Toronto.
Footnotes
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/cgi/content/full/0812664106/DCSupplemental.
References
- 1.Paty DW, Ebers GC. Multiple sclerosis. Philadelphia: F.A. Davis; 1998. [Google Scholar]
- 2.Ebers GC, Sadovnick AD, Risch NJ. A genetic basis for familial aggregation in multiple sclerosis. Nature. 1995;377:150–151. doi: 10.1038/377150a0. [DOI] [PubMed] [Google Scholar]
- 3.Gregersen JW, et al. Functional epistasis on a common MHC haplotype associated with multiple sclerosis. Nature. 2006;443:574–577. doi: 10.1038/nature05133. [DOI] [PubMed] [Google Scholar]
- 4.Zamvil SS, Steinman L. The T lymphocyte in experimental allergic encephalomyelitis. Annu Rev Immunol. 1990;8:579–621. doi: 10.1146/annurev.iy.08.040190.003051. [DOI] [PubMed] [Google Scholar]
- 5.Ebers GC, et al. A full genome search in multiple sclerosis. Nat Genet. 1996;13:472–476. doi: 10.1038/ng0896-472. [DOI] [PubMed] [Google Scholar]
- 6.Dyment DA, Ebers GC, Sadovnick AD. Genetics of multiple sclerosis. Lancet Neurol. 2004;3:104–110. doi: 10.1016/s1474-4422(03)00663-x. [DOI] [PubMed] [Google Scholar]
- 7.Dyment DA, et al. Complex interactions among MHC haplotypes in multiple sclerosis: Susceptibility and resistance. Hum Mol Genet. 2005;14:2019–2026. doi: 10.1093/hmg/ddi206. [DOI] [PubMed] [Google Scholar]
- 8.Lincoln MR, et al. A predominant role for the HLA class II region in the association of the MHC region with multiple sclerosis. Nat Genet. 2005;37:1108–1112. doi: 10.1038/ng1647. [DOI] [PubMed] [Google Scholar]
- 9.Marrosu MG, et al. Dissection of the HLA association with multiple sclerosis in the founder isolated population of Sardinia. Hum Mol Genet. 2001;10:2907–2916. doi: 10.1093/hmg/10.25.2907. [DOI] [PubMed] [Google Scholar]
- 10.Rubio JP, et al. Genetic dissection of the human leukocyte antigen region by use of haplotypes of tasmanians with multiple sclerosis. Am J Hum Genet. 2002;70:1125–1137. doi: 10.1086/339932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Chao MJ, et al. HLA class 1 alleles tag HLA-DRB1*1501 haplotypes for differential risk in multiple sclerosis susceptibility. Proc Natl Acad Sci USA. 2008;105:13069–13074. doi: 10.1073/pnas.0801042105. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Caballero A, et al. DQB1* 0602 confers genetic susceptibility to multiple sclerosis in Afro-Brazilians. Tissue Antigens. 1999;54:524–526. doi: 10.1034/j.1399-0039.1999.540511.x. [DOI] [PubMed] [Google Scholar]
- 13.Oksenberg JR, et al. Mapping multiple sclerosis susceptibility to the HLA-DR Locus in African Americans. Am J Hum Genet. 2004;74:160–167. doi: 10.1086/380997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ligers A, et al. Evidence of linkage with HLA-DR in DRB1*15-negative families with multiple sclerosis. Am J Hum Genet. 2001;69:900–903. doi: 10.1086/323480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Fogdell-Hahn A, Ligers A, Gronning M, Hillert J, Olerup O. Multiple sclerosis: A modifying influence of HLA class I genes in an HLA class II associated autoimmune disease. Tissue Antigens. 2000;55:140–148. doi: 10.1034/j.1399-0039.2000.550205.x. [DOI] [PubMed] [Google Scholar]
- 16.Chao MJ, et al. Transmission of class I/II multi-locus MHC haplotypes and multiple sclerosis susceptibility: Accounting for linkage disequilibrium. Hum Mol Genet. 2007;16:1951–1958. doi: 10.1093/hmg/ddm142. [DOI] [PubMed] [Google Scholar]
- 17.Lundmark F, et al. Variation in interleukin 7 receptor α chain (IL7R) influences risk of multiple sclerosis. Nat Genet. 2007;39:1108–1113. doi: 10.1038/ng2106. [DOI] [PubMed] [Google Scholar]
- 18.Peltonen L. Old suspects found guilty—the first genome profile of multiple sclerosis. N Engl J Med. 2007;357:927–292. doi: 10.1056/NEJMe078147. [DOI] [PubMed] [Google Scholar]
- 19.Hafler DA, et al. Risk alleles for multiple sclerosis identified by a genomewide study. N Engl J Med. 2007;357:851–862. doi: 10.1056/NEJMoa073493. [DOI] [PubMed] [Google Scholar]
- 20.Hoppenbrouwers IA, et al. EVI5 is a risk gene for multiple sclerosis. Genes Immun. 2008;9:334–337. doi: 10.1038/gene.2008.22. [DOI] [PubMed] [Google Scholar]
- 21.Aulchenko YS, et al. Genetic variation in the KIF1B locus influences susceptibility to multiple sclerosis. Nat Genet. 2008;40:1402–1403. doi: 10.1038/ng.251. [DOI] [PubMed] [Google Scholar]
- 22.D'Netto, et al. Risk alleles for multiple sclerosis in multiplex families. Neurology. 2009 doi: 10.1212/WNL.0b013e3181a92c25. in press. [DOI] [PubMed] [Google Scholar]
- 23.Barcellos LF, et al. Heterogeneity at the HLA-DRB1 locus and risk for multiple sclerosis. Hum Mol Genet. 2006;15:2813–2824. doi: 10.1093/hmg/ddl223. [DOI] [PubMed] [Google Scholar]
- 24.Madsen LS, et al. A humanized model for multiple sclerosis using HLA- DR2 and a human T-cell receptor. Nat Genet. 1999;23:343–347. doi: 10.1038/15525. [DOI] [PubMed] [Google Scholar]
- 25.Mustafa M, et al. Protective influences on experimental autoimmune encephalomyelitis by MHC class I and class II alleles. J Immunol. 1994;153:3337–3344. [PubMed] [Google Scholar]
- 26.Khare M, Mangalam A, Rodriguez M, David CS. HLA DR and DQ interaction in myelin oligodendrocyte glycoprotein-induced experimental autoimmune encephalomyelitis in HLA class II transgenic mice. J Neuroimmunol. 2005;169:1–12. doi: 10.1016/j.jneuroim.2005.07.023. [DOI] [PubMed] [Google Scholar]
- 27.Bateson W. Facts limiting the theory of heredity. Science. 1907;26:649–660. doi: 10.1126/science.26.672.649. [DOI] [PubMed] [Google Scholar]
- 28.Ramagopalan SV, et al. The Inheritance of Resistance Alleles in Multiple Sclerosis. PLoS Genet. 2007;3:1607–1613. doi: 10.1371/journal.pgen.0030150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Stewart GJ, Teutsch SM, Castle M, Heard RN, Bennetts BH. HLA-DR,-DQA1 and -DQB1 associations in Australian multiple sclerosis patients. Eur J Immunogenet. 1997;24:81–92. doi: 10.1046/j.1365-2370.1997.00252.x. [DOI] [PubMed] [Google Scholar]
- 30.Kwon OJ, et al. HLA class II susceptibility to multiple sclerosis among Ashkenazi and non-Ashkenazi Jews. Arch Neurol (Chicago) 1999;56:555–560. doi: 10.1001/archneur.56.5.555. [DOI] [PubMed] [Google Scholar]
- 31.Ahmad T, et al. Haplotype-specific linkage disequilibrium patterns define the genetic topography of the human MHC. Hum Mol Genet. 2003;12:647–656. [PubMed] [Google Scholar]
- 32.Mignot E, et al. Complex HLA-DR and-DQ interactions confer risk of narcolepsy-cataplexy in three ethnic groups. Am J Hum Genet. 2001;68:686–699. doi: 10.1086/318799. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Koeleman BP, et al. Genotype effects and epistasis in type 1 diabetes and HLA-DQ trans dimer associations with disease. Genes Immun. 2004;5:381–388. doi: 10.1038/sj.gene.6364106. [DOI] [PubMed] [Google Scholar]
- 34.Owens T, Wekerle H, Antel J. Genetic models for CNS inflammation. Nat Med. 2001;7:161–166. doi: 10.1038/84603. [DOI] [PubMed] [Google Scholar]
- 35.Chao MJ, et al. Epigenetics in multiple sclerosis susceptibility: Difference in transgenerational risk localizes to the major histocompatibility complex. Hum Mol Genet. 2009;18:261–266. doi: 10.1093/hmg/ddn353. [DOI] [PubMed] [Google Scholar]
- 36.Risch N. Assessing the role of HLA-linked and unlinked determinants of disease. Am J Hum Genet. 1987;40:1–14. [PMC free article] [PubMed] [Google Scholar]
- 37.Ramagopalan SV, et al. Expression of the multiple sclerosis associated allele HLA-DRB1*1501 is influenced by vitamin D. PLoS Genet. 2009 doi: 10.1371/journal.pgen.1000369. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Sadovnick AD, Risch NJ, Ebers GC. Canadian collaborative project on genetic susceptibility of MS—Phase II: Rationale and method. Can J Neurol Sci. 1998;25:216–221. doi: 10.1017/s0317167100034041. [DOI] [PubMed] [Google Scholar]
- 39.Bunce M. PCR-sequence-specific primer typing of HLA Class I and Class II alleles. Methods Mol Biol. 2003;210:143–171. doi: 10.1385/1-59259-291-0:143. [DOI] [PubMed] [Google Scholar]
- 40.Spielman RS, McGinnis RE, Ewens WJ. Transmission test for linkage disequilibrium: The insulin gene region and insulin-dependent diabetes mellitus (IDDM) Am J Hum Genet. 1993;52:506–516. [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.