Introduction

Human leukocyte antigen (HLA) class II molecules are heterodimeric transmembrane glycoproteins that present self- and non-self-peptides to the surface of antigen-presenting cells for recognition by T cell receptors. HLA class II consists of three isotypes, including HLA-DR (encoded by HLA-DRA and -DRB1, and -DRB3, 4, and 5 in certain haplotypes), HLA-DQ (encoded by HLA-DQA1 and -DQB1) and HLA-DP (encoded by HLA-DPA1 and -DPB1). With the exception of DRA, each locus has a large number of alleles. For example, DRB1 has >1 700 alleles, DQB1 has >700 alleles and DPB1 has >500 alleles that have been registered to date according to the IMGT/HLA database.1 Polymorphic variants of HLA class II are accumulated mainly in exon 2, which encodes the α1 or β1 domain of each subunit that forms the peptide-binding groove.

Genetic associations between HLA and autoimmune diseases were reported in the early 1970s (McDevitt and Bodmer2 and references therein). Up until now, it has been confirmed that HLA have the strongest association signals, compared with any other loci, to a variety of autoimmune and inflammatory disorders. However, the mechanism that might underlie the HLA–autoimmunity associations has remained elusive for most of the autoimmune diseases, including rheumatoid arthritis (RA) and type 1 diabetes (T1D), to which the hierarchies of HLA-DR and -DQ alleles or haplotypes that are associated with susceptibility or protection have been established in Europeans and other ethnicities. The pathogenic mechanisms of RA and T1D have also been extensively analyzed in human and the murine models. In this review, we summarize what are currently known and what are remained elusive about the mechanisms of HLA associations with RA and T1D, and elucidate the potential mechanistic basis of the HLA–autoimmunity associations.

Associations between HLA class II and RA

RA is a chronic inflammatory disease of the synovial joints. Strong associations of DRB1*04:01, *04:04, *01:01 and *10:01, which carry a shared epitope (SE)3 at amino acid positions 70β to 74β (Figure 1) (QKRAA at DRB1*04:01, QRRAA at DRB1*01:01 and *04:04, and RRRAA at DRB1*10:01), with RA risk have been reported and confirmed in numerous studies. In Europeans, the association between DRB1 and RA is stronger in anti-citrullinated peptide antibody-positive RA than in anti-citrullinated peptide antibody-negative RA.4, 5 Among the SE alleles, DRB1*04:01 and *04:04 confer a stronger predisposition to RA than DRB1*01:01 and *10:01.4, 6 DRB1*04:01 homozygosity and DRB1*04:01/*04:04 heterozygosity are associated with increased risk for RA.4 The RA protective DRB1 alleles include DRB1*13:016, 7 and the alleles that carry Ile67β or Asp70β.4 In East Asian populations, DRB1*04:01 and *04:05 confer risk, whereas DRB1*13:02 and *14:05 confer protection8, 9, 10 (see Furukawa et al.11 in this issue for details of HLA association with RA).

Figure 1
figure 1

Locations of amino acid residues in DRB1 product that are associated with rheumatoid arthritis (RA). Locations of 11β, 13β and shared epitope (SE) residues in the structure of DR protein (PDB: 2seb)75 are shown.

Full size image

The associations between HLA and RA have been analyzed mainly for the DR loci. However, the strong linkage disequilibrium between DR and DQ suggests that both DR and DQ may contribute to predisposition to RA, similar to what has been observed in T1D. In Europeans, RA susceptible DRB1 alleles are found mainly in DRB1*04:01-DQA1*03-DQB1*03:01, DRB1*04:01-DQA1*03-DQB1*03:02, DRB1*04:04-DQA1*03-DQB1*03:02, DRB1*01:01-DQA1*01-DQB1*05:01 and DRB1*10:01-DQA1*01-DQB1*05:01 haplotypes. In East Asian populations, RA susceptible DR–DQ haplotypes include DRB1*04:01-DQA1*03-DQB1*03:01 and DRB1*04:05-DQA1*03-DQB1*04:01. DQA1*03-DQB1*03 and DQA1*03-DQB1*04:01 may confer susceptibility, and DQA1*01-DQB1*05 may confer mildly predisposing effect.12, 13

Potential mechanism of RA susceptibility and protection

It has been thought that DRB1 alleles that carry SE confer disease susceptibility through the selective presentation of self-peptides14 or the mechanism that involves alteration in the peripheral T cell repertoire (see, for example, Auger et al.15and Roudier16). It was later found that DRB1*04:01 protein interacted with citrullinated peptides with higher affinity than with non-citrullinated peptides.17 A structural study confirmed that DRB1*04:01 and *04:04 proteins presented citrullinated vimentin and aggrecan peptides via interactions between Lys71β or Arg71β and citrulline at the P4-binding pocket.18 These findings indicate that the SE alleles exert pathogenic effects through the presentation of citrullinated peptides, which are recognized as non-self by T cells. The critical roles for Lys71β and Arg71β in the presentation of citrullinated peptides are consistent with the findings that DRB1 alleles that carry Glu71β, such as DRB1*04:02, *13:01 and *13:02, are not associated with RA risk. In addition to SE, variants at amino acids 11β and 13β in DRB1 (Figure 1) also predispose strongly to RA.619 Although these two residues could affect binding preferences for peptides, self-peptide(s) that might explain the predispositions of 11β and 13β to RA have not currently been identified.

The mechanism of protective associations between DR–DQ haplotypes and RA are not well characterized. A recent study revealed that the citrullinated DERAA motif, which was found in the protective DRB1 allele products, including DRB1*13 protein, and in vinculin, can be presented by the DQ proteins that were encoded on the RA susceptible DR–DQ haplotypes.20 The protective effect of DRB1*13 against RA was explained by the cross-reactivity of self-reactive T cells to the citrullinated DERAA motif in vinculin and DRB1*13 protein, and the absence of these self-reactive T cells in the DR4/DR13 heterozygotes.20

Self-peptides that are derived from other potential self-antigens, including type II collagen and nuclear ribonucleoprotein A2 can also be presented to the susceptible DR or DQ allele products and are involved in the pathogenesis of RA.21, 22, 23, 24, 25 The binding affinity of DR4 with class II-associated invariant chain peptide (CLIP) may also affect RA risk.26 The presentation of misfolded immunoglobulin G heavy chain by DR protein is also shown to contribute to the susceptibility to RA.27

To summarize, studies of RA revealed that the association patterns of the most predisposing and the most protective DRB1 alleles can be explained, at least in part, by variants at 71β that mediate the presentation of citrullinated peptides.18 The proposed mechanism of protection against RA also involves the allele-specific presentation of self-peptides.20 Therefore, the selective presentation of self-peptides on the susceptible and protective HLA allele products might be one of the key mechanisms of DR–DQ haplotype associations with RA (Figure 2a).

Figure 2
figure 2

Hypothetical mechanisms for human leukocyte antigen (HLA)–autoimmunity associations.59 (a) HLA associations with autoimmune diseases may be explained by the selective presentation of disease-relevant self-peptides by the disease susceptible HLA allele products (grey). The disease-relevant peptides (black) and irrelevant peptides (white) are shown. (b) The HLA/major histocompatibility complex (MHC) stability model. This model proposes that intrinsically unstable HLA proteins (grey), which form unstable HLA-peptide complex through the presentation of diverse self-peptides, confer a risk for autoimmune diseases.

Full size image

Associations between HLA class II and T1D

T1D is caused by the autoimmune-mediated destruction of insulin-producing beta cells in the pancreas. DR and DQ are the strongest susceptibility loci for T1D. In European descendants, the highest risk is conferred by DR3-DQA1*05-DQB1*02 and DR4-DQA1*03-DQB1*03:02 haplotypes, and the highest protection is conferred by DR15-DQA1*01:02-DQB1*06:02.28, 29, 30 Heterozygosity of DR3-DQA1*05-DQB1*02/DR4-DQA1*03-DQB1*03:02 confers the strongest genotypic risk for T1D.28, 29, 30 These susceptible haplotypes are partly shared in a variety of other ethnic groups.29 In the Japanese population, in which these susceptible haplotypes are infrequent, susceptibility to T1D is conferred by DR9-DQA1*03-DQB1*03:03 and DR4-DQA1*03-DQB1*04:01 haplotypes.31 The strong association between non-Asp57β in DQB1 and T1D risk is found in Europeans32, 33 but not in the Japanese population, in which the T1D susceptible DQB1 alleles carry Asp57β. DR15-DQA1*01:02-DQB1*06:02 haplotype confers protection against T1D in both Europeans and Japanese populations.29, 31

It has been established that both DR and DQ loci confer a predisposing and/or protective effect to T1D in a manner independent of or dependent on the allele at the other locus. One of the examples is DR4-DQA1*03-DQB1*03:02 haplotype, which can be subgrouped into the susceptible haplotypes that contain DRB1*04:01 or *04:05, or the neutral-to-protective haplotypes that contain DRB1*04:03 or *04:06.28, 29, 30, 31 Another example is the haplotype that contains DRB1*07, the association of which can vary from susceptible to protective, depending on DQA1 and DQB1 alleles.28, 30, 34

The major autoantigens involved in the pathogenesis of T1D are insulin, glutamic acid decarboxylase, zinc transporter 8 and islet antigen-2.35, 36, 37 The HLA-peptide binding studies revealed that the self-peptides derived from these autoantigens bound promiscuously to the susceptible, neutral and protective DR and DQ allele products,38, 39, 40 which may reflect the facts that the peptide-binding spectrum of DR and DQ allele products partially overlaps across alleles.41, 42 It has remained unknown whether the HLA associations with T1D can be explained by the selective presentation of certain pancreatic self-peptides on the susceptible HLA allele products. Potential contributions of other functional variations among the alleles, such as the promoter activity43 and the dependency of HLA proteins to invariant chain and HLA-DM,44, 45, 46 to T1D risk have also been reported.

The association between non-Asp57β in DQB1 and T1D in Europeans, and the presence of non-Asp57β in I-Ag7 of the non-obese diabetogenic (NOD) mice suggested that non-Asp57β had a critical role in the pathogenesis of T1D. Studies of the structure of DQA1*03-DQB1*03:02 and I-Ag7 products revealed that non-Asp57β in DQB1 and I-Ag7 facilitated the accommodation of acidic residue at the P9-binding pocket, thereby allowing for the presentation of certain pancreatic self-peptides, such as insulin B9-23, which carried acidic residue at the p9.47, 48, 49 It was later found in both human and the NOD mice that insulin and other pancreatic self-peptides that were presented weakly on the susceptible HLA/major histocompatibility complex allele products were the targets of self-reactive T cells.50, 51, 52, 53, 54, 55, 56 These studies established a notion that the T1D pathogenesis was mediated through the formation of unstable HLA/major histocompatibility complex-peptide complex.57 The presence of non-Asp57β does not appear to be a prerequisite for the accommodation of self-peptides in a weak binding register. Non-Asp57β is present in the neutral haplotype DQA1*02-DQB1*02 in Europeans and is absent in the T1D susceptible DQB1 alleles in the Japanese population. These findings suggest that non-Asp57β in DQB1 may not be an essential component in the shared mechanism of T1D across ethnicities.

Associations between DQ protein instability and T1D

One of the additional factors that might contribute to the associations between DR–DQ haplotypes and T1D is the protein instability of DQ. It was reported in the 1990s that DQA1*05-DQB1*02 and DQA1*03-DQB1*03:02 haplotypes that predisposed to T1D in Europeans generated SDS unstable DQ protein, whereas the protective haplotype DQA1*01:02-DQB1*06:02 generated SDS stable DQ protein.58 However, SDS stability can be affected by both the stability of HLA protein and affinity of the bound peptides. We therefore validated, through a cell-surface HLA expression assay, whether the stability of DQ protein might differ intrinsically among the allele products. We confirmed a steep allelic hierarchy in the stability of DQ haplotype products.59 Consistent with previous study,58 the T1D susceptible DQ haplotypes in Europeans generated intrinsically unstable DQ proteins. Unstable DQ proteins were also generated by DQA1*05 and DQB1*03:02 allele products that can be formed in DQA1*05-DQB1*02/DQA1*03-DQB1*03:02 heterozygotes. DQA1*03-DQB1*03:03 and DQA1*03-DQB1*04:01, which were associated with T1D risk in the Japanese population, also generated unstable DQ proteins. The protective haplotype DQA1*01:02-DQB1*06:02 generated highly stable protein (Figure 3).59 When all of the major DQ haplotypes were analyzed, the protein stability of DQ was associated inversely with T1D risk, indicating that the protein instability of DQ might contribute to T1D risk irrespective of ethnicity.59

Figure 3
figure 3

Allelic hierarchy of DQ protein stability. Protein stability of DQ proteins that are encoded by major DQ haplotypes in European and Japanese populations are displayed in order of increasing protein stability (x axis) with the estimated DQ protein stability plotted on the y axis.59 The type 1 diabetes (T1D) susceptible haplotypes (black, large diamond), T1D protective haplotypes (white, large diamond) and haplotypes not associated with T1D (white, small diamond) are identified in the Swedish76 and Japanese populations.31 The following abbreviations are used for each DQ haplotype; DQA1*01:02-DQB1*06:02 (DQ0602), DQA1*02-DQB1*03:03 (DQ9.2), DQA1*01:03-DQB1*06:03 (DQ0603), DQA1*01:03-DQB1*06:01 (DQ0601), DQA1*05-DQB1*03:01 (DQ7.5), DQA1*03-DQB1*03:03 (DQ9.3), DQA1*03-DQB1*03:02 (DQ8.3), DQA1*05-DQB1*02 (DQ2.5), DQA1*03-DQB1*04:01 (DQ4.3), DQA1*05/DQB1*03:02 (trans combination) (DQ8.5) and DQA1*01-DQB1*05:03 (DQ0503).

Full size image

One of the polymorphic variants that regulated the protein stability of DQ was 57β; Asp57β stabilized DQ protein through the interactions with peptide and/or Arg79α. The variants at 47α in DQA1, which were located outside of the peptide-binding groove (Figure 4), also regulated the intrinsic stability of DQ protein. Associations between the destabilizing variants at these sites, such as non-Asp57β in DQB1 and Gln47α in DQA1, and T1D risk59 indicated that non-Asp57β in DQB1 may confer T1D risk through destabilizing the DQ proteins.

Figure 4
figure 4

Locations of amino acid residues in DQA1 and DQB1 products that are associated with type 1 diabetes (T1D). Locations of 47α and 57β in the structure of DQ protein (PDB: 1uvq)77 are shown.

Full size image

Based on the inverse association between the DQ protein stability and T1D risk, and the strong association signals detected at the protein destabilizing variants at 57β and 47α, we proposed that the intrinsic instability of HLA protein may be one of the important functional components that conferred T1D risk (Figure 2b).59 This hypothesis is consistent with the established concept of T1D pathogenesis, and indicates that an intrinsically unstable HLA may increase a risk for T1D through facilitating the formation of unstable HLA–self-peptide complex, which may permit the thymic escape of self-reactive T cells.

The mechanisms of protection against T1D of DQA1*01:02-DQB1*06:02 and DQA1*02-DQB1*03:03 haplotypes are not known. As the mechanisms of protection, the ‘affinity model’ and the ‘determinant capture model’ have been proposed.60, 61, 62

Mechanisms of DR–DQ haplotype associations with autoimmune diseases

Mechanism of HLA-associated autoimmune diseases has generally been studied through the identification of self-peptides that are presented selectively to the susceptible HLA allele products. Similar to the findings in RA, the association between DRB1*04:06-DQA1*03-DQB1*03:02 haplotype and insulin autoimmune syndrome63 has also been explained by the presentation of the disease-relevant self-peptide, the reduced form of insulin, to DRB1*04:06 protein, but not to DQB1*03:02 and the non-risk allele DRB1*04:05 products.64

Regarding the mechanism of T1D, pancreatic self-peptides that might explain the association of HLA have not currently been identified. Studies of multiple sclerosis also documented the promiscuous binding patterns and variable affinity levels in the interactions of self-peptides with the susceptible and non-susceptible HLA allele products.65, 66, 67 These findings may indicate a possibility that a variety of low-affinity self-peptides, most of which have not been identified readily in the HLA-peptide binding studies, are implicated in the pathogenesis of certain autoimmune diseases, including T1D.

Genetic association studies of other autoimmune and inflammatory diseases have revealed that the T1D susceptible haplotypes DR3-DQA1*05-DQB1*02 and DR4-DQA1*03-DQB1*03:02 in Europeans and DR9-DQA1*03-DQB1*03:03 and DR4-DQA1*03-DQB1*04:01 in the Japanese population predispose to multiple autoimmune diseases, including autoimmune polyglandular syndrome type II,68 celiac disease69, 70 and antineutrophil cytoplasmic antibody-associated vasculitis.71 The T1D protective haplotype DR15-DQA1*01:02-DQB1*06:02 confers protection against autoimmune polyglandular syndrome type II and III68, 72 and selective immunoglobulin A deficiency.73, 74 The associations of T1D susceptible/protective DR–DQ haplotypes with a variety of autoimmune diseases indicate that the mechanism of autoimmune susceptibility and/or protection may partly be shared among certain autoimmune diseases. As distinct sets of self-peptides are involved in individual diseases, it would not be possible to explain the shared associations by the peptide-binding spectrum of DR and DQ proteins. The HLA/major histocompatibility complex stability model or other molecular mechanism(s), which could operate irrespective of the presented self-peptides, may underlie these associations.

Concluding remarks

The studies of RA and T1D suggest that the two distinct mechanistic basis, which might involve allelic variations in the peptide-binding preferences and the protein stability of HLA, might explain the associations of HLA with autoimmune diseases. These two mechanisms may constitute a part of the whole mechanism of HLA–autoimmunity associations, which can also be affected by allelic variations in the expression levels and the dependency to accessory molecules. In addition to the functional diversity at each locus, their interplay between the loci may also contribute to the complex pattern of DR–DQ haplotype associations with autoimmunity.