Association of the HLA–DRB1 epitope LA67, 74 with rheumatoid arthritis and citrullinated vimentin binding

Authors

Errata

This article is corrected by:

  1. Errata: Incorrect Numbers of Patients and Controls Positive for Epitope D70 Shown in Table 2 of the Article by Freed et al (Arthritis Rheum, December 2011) Volume 66, Issue 6, 1671, Article first published online: 27 May 2014

Abstract

Objective

Although rheumatoid arthritis (RA) has long been associated with an HLA–DRB1 shared epitope, a systematic search for other epitopes has never been conducted. In addition, the relationship between these epitopes and the binding of citrullinated autoantigens has not been investigated. We developed a program that can analyze HLA data for all possible epitopes of up to 5 amino acids and used this program to assess the shared epitope hypothesis in RA.

Methods

We analyzed high-resolution data from the International Histocompatibility Working Group, which included a group of 488 patients with RA and a group of 448 racially and ethnically balanced control subjects, for all combinations of up to 5 amino acids among polymorphic HLA–DRB1 positions 8–93. Statistical significance was determined by chi-square and Fisher's exact tests, with a false discovery rate correction.

Results

Three residues (V11, H13, and L67) were found to have the highest degree of association with RA susceptibility (P < 10–11), and D70 was found to correlate best with RA resistance (P = 2 × 10–11). Of >2 million epitopes examined, LA67, 74 exhibited the highest correlation with RA susceptibility (P = 2 × 10–20; odds ratio 4.07 [95% confidence interval 3.07–5.39]). HLA alleles containing the LA67, 74 epitope exhibited significantly greater binding to citrullinated vimentin65–77 than did alleles containing D70. Only 1 allele (DRB1*16:02) contained both LA67, 74 and D70; it bound citrullinated vimentin weakly and was not associated with RA.

Conclusion

The findings of these studies confirm the importance of HLA–DRB1 amino acids in pocket 4 for the binding of citrullinated autoantigens and susceptibility to RA.

Rheumatoid arthritis (RA) is a chronic disease characterized by autoimmune destruction of joints and surrounding tissues. Susceptibility to the disease has several genetic components, but the HLA–DRB1 locus is clearly the most significant (1). Several DR4 alleles, notably, DRB1*04:01, *04:04, and *04:05, are strongly associated with RA, while DRB1*04:02 is not (2–5). (Note that the alleles are designated by the newly accepted World Health Organization nomenclature [online at http://hla.alleles.org], whereby the family and alleles are separated by a colon. Thus, DRB1*04:01 corresponds to DRB1*0401 in earlier literature.) In addition, DRB1*01:01, *01:02, *10:01, and *14:02 have sometimes been associated with RA, particularly in non-Europeans (6–14). These disparate alleles have been hypothesized to contribute to RA via the presence of a “shared epitope,” a common set of amino acids at positions 70–74 of the peptide-binding groove (2).

Peptide binding to HLA–DRB1 molecules is controlled by 6 pockets, each with multiple polymorphic amino acids that create millions of potential peptide-binding epitopes. Seemingly disparate HLA alleles (e.g., DRB1*04:01 and *16:02) can share an epitope that closely related alleles (e.g., DRB1*04:01 and *04:02) do not. The concept of a shared epitope in susceptibility to RA is now widely accepted, although the exact nature of the epitope and its role in the disease has been the subject of considerable debate (15, 16). The presence of aspartic acid at position 70 (D70) alone has been reported to have a protective effect (17), and positions outside residues 70–74 have also been implicated in the disease (18, 19). However, very little is known about how these epitopes bind citrullinated autoantigens that trigger RA in the majority of patients.

The peptide-binding groove of DRB1 molecules consists of 86 amino acids arranged in 6 pockets, with >2 million possible epitopes. Karp et al (20) recently demonstrated that an epitope comprised of 8 amino acids in DRB1 pockets 4 and 7 was associated with susceptibility to systemic sclerosis. In order to undertake a thorough analysis of HLA–DRB1 epitopes in RA, we compared the frequency of 4,073 epitopes comprising 1–5 amino acids between control subjects and patients with RA. We then analyzed the ability of DRB1 alleles containing susceptibility and resistance epitopes to bind a citrullinated vimentin peptide that is known to be associated with the development of RA.

PATIENTS AND METHODS

Study subjects.

HLA data from studies on RA were downloaded from the RA component of the International Histocompatibility Working Group (IHWG) database at http://www.ncbi.nlm.nih.gov/gv/mhc/ihwg.cgi. This working group represents a consortium of international investigators who worked together to generate HLA genotyping and microsatellite data as well as clinical and demographic data for RA patients and healthy individuals derived from the same populations. Only allele-level HLA data were included in this analysis.

Of the original 2,376 subjects in this data set, we censored all subjects from centers where fewer than 80% of the typings were allele level. Similarly, 122 “matched” control subjects were deleted from the control population since they would not, by definition, represent the normal distribution of HLA alleles. However, since these data consisted of an uneven racial and ethnic distribution of subjects, which would skew the distribution of HLA alleles, we removed 36 Jewish, 23 San Bushman, 30 South African black, and 62 Thai subjects from the RA group because we did not have a suitable control population. The numbers of Spanish subjects were also unevenly distributed between the groups, so we removed the last 90 consecutive Spanish control subjects. Finally, we added to the control group 122 consecutive Caucasian subjects whose umbilical cord blood was typed at ClinImmune Labs. This resulted in a racially and ethnically balanced data set of 488 subjects and 448 controls (Table 1).

Table 1. Racial distribution of subjects in the IHWG RA study*
IHWG RA study populationRA patientsControl subjects
  • *

    IHWG = International Histocompatibility Working Group; RA = rheumatoid arthritis.

Race of subjects in unbalanced groups  
 Asian8366
 Jewish230
 Jewish Ashkenazi130
 Mexican350
 North American Amerindian10077
 North American black70
 San Bushman2355
 South African black300
 Spanish141231
 Thai620
 American Caucasian1220
 Total639429
Race of subjects in balanced groups  
 Asian8366
 Mexican3535
 North American Amerindian10077
 North American black77
 Spanish141141
 American Caucasian122122
 Total488448

Epitope analysis.

Epitope analysis was performed using the R software package version 2.6.1 (available online at http://www.r-project.org). Each possible combination of up to 5 polymorphic residues at positions 8–93 of the HLA molecule was considered a possible epitope. Amino acids outside of this range were not considered because they are unlikely to influence interactions with either the peptide or the T cell receptor. The number of individuals carrying at least 1 copy of the epitope was compared with the number of individuals not carrying the epitope in both the RA group and the control group. The distribution of these epitopes within the patient and control populations was calculated using 2 × 2 contingency tables and analyzed with either Pearson's chi-square test or Fisher's exact test, as appropriate. The resulting P value for each epitope was corrected for multiple hypothesis testing using the false discovery rate method described by Benjamini and Yekutieli (21) to control for Type I statistical errors.

Peptide binding to HLA–DRB1 alleles.

The Reveal peptide-binding assay (ProImmune) was used to measure the binding of the citrullinated vimentin peptide65–77 (SAVRLCitSSVPGVR) to one or more class II major histocompatibility complex (MHC) alleles and stabilize the MHC–peptide complex. Detection of bound peptide was based on the presence or absence of the native conformation of the HLA–peptide complex, using a relevant conformational antibody in an immunoassay. The score of the test peptide is reported quantitatively as a percentage of the signal generated by the positive control peptide. Assay performance was confirmed by including an intermediate control peptide that is known to bind with weaker affinity to the allele under investigation. In addition, experimental standard error was obtained by performing triplicate positive and negative control binding experiments.

RESULTS

HLA alleles in rheumatoid arthritis.

Analysis of the original (unbalanced) allele data from the IHWG subjects described in Table 1 indicated that DRB1*04:01, *04:04, and *04:05 were associated with RA susceptibility, while DRB1*07:01, *08:01, *11:01, *13:01, and *13:02 were associated with resistance. By comparison, an allele analysis of the data using balanced controls indicated that DRB1*04:01 (P = 0.0002), *04:04 (P = 0.0002), and *04:05 (P = 0.003) were still associated with susceptibility, but only *08:01 (P = 0.031), *13:01 (P = 0.002), and *13:02 (P = 0.004) remained associated with resistance. Several other DRB1 alleles exhibited trends toward significance (P < 0.30) that ultimately proved useful in interpreting the epitope analysis. For example, DRB1*01:01 (P = 0.107) exhibited a trend toward susceptibility, while DRB1*03:02, *07:01, *08:02, *11:01, and *15:01 exhibited a trend toward resistance (P = 0.21–0.23). All but 1 of these alleles had sufficient sample size to detect statistical significance if it existed. The lack of significance therefore suggests that these alleles were not significant markers of disease susceptibility or disease resistance. The remainder of the alleles were not associated with RA susceptibility or resistance (P = 0.8–1.0).

Analysis of HLA epitopes.

Of the 86 amino acids in the DRB1 peptide-binding groove, 30 exhibited varying degrees of polymorphism, and 11 of these (positions 10, 11, 12, 13, 33, 37, 47, 67, 70, 71, and 74) were significantly associated with either susceptibility or resistance to RA, usually both (P = 10–4 to 10–15). Limiting our analysis to only these polymorphic residues, we identified 2,489,813 epitopes consisting of between 1 and 5 amino acids in the patient and control groups. However, since many of these epitopes represented combinations of significant and nonsignificant (though polymorphic) residues, we further restricted the analysis to only those positions that were significant at the single–amino acid level. Furthermore, since the side chains of amino acids at positions 10, 12, and 33 face away from the peptide-binding groove and the T cell contacts, they likely do not directly influence autoimmunity. Their association with RA susceptibility and resistance, while highly significant, is therefore most likely due to their linkage disequilibrium with other critical amino acids. These positions were not included in the epitope analysis since they would not contribute directly to either peptide binding or T cell recognition.

Positions 11 (V, L, D, P, G, or S), 13 (H, F, R, Y, G, or S), 47 (Y or F), 67 (L, F, or I), 70 (Q, R, or D) and 74 (A, E, R, Q, or L) were strongly associated with both susceptibility and resistance to RA. Amino acids V11 (P = 6 × 10–15), H13 (P = 3 × 10–14), L67 (P = 8 × 10–12), Y47 (P = 10–6), Q70 (P = 2 × 10–6), and A74 (P = 7 × 10–4) were strongly associated with susceptibility to RA, while D70 (P = 2 × 10–11), S11 (P = 7 × 10–7), I67 (P = 5 × 10–7), E71 (P = 6 × 10–6), S13 (P = 0.0001), F47 (P = 0.0001), and L74 (P = 0.0005) were strongly associated with resistance.

We then performed an epitope analysis in which combinations of up to 5 amino acids at positions 11, 13, 37, 47, 67, 70, 71, 73, and 74 were compared (Table 2). Position 73 was included, even though it was not significant, to allow comparison with the canonical QRAA70, 71, 73, 74 shared epitope. Of the 4,073 epitopes identified, LA67, 74 was most strongly associated with RA (P = 2 × 10–20, odds ratio [OR] 4.07 [95% confidence interval (95% CI) 3.07–5.39]). Extending this epitope to include other “susceptibility” amino acids (e.g., YLA47, 67, 74, VHYLA11, 13, 47, 67, 74) did not increase the statistical significance, suggesting that LA67, 74 was the critical epitope. The LA67, 74 epitope was found in several DRB1 alleles that were weakly associated with RA, including *01:01, *01:02, *10:01 and *14:02 (collective OR 1.5 [95% CI 1.16–2.06], P = 0.002) as well as several alleles (i.e., *01:04, *04:09, *04:10, and *14:06) whose association with RA has not been reported. However, the latter alleles are rare, being found in only 15 RA patients and 4 controls (collective OR 3.5 [95% CI 1.16–10.7], P = 0.02), and were therefore unlikely to have ever been associated with RA by traditional allele analyses.

Table 2. HLA shared epitopes associated with rheumatoid arthritis (RA)
EpitopeRA patientsControl subjectsPcorr*Associated alleles
  • *

    P values were corrected (Pcorr) for multiple comparisons.

  • Represents the LA67, 74 epitope in the absence of D70.

Susceptibility epitopes    
 LA67, 743742002 × 10–20*01:01, *01:02, *01:04, *04:01, *04:04, *04:05, *04:08, *04:09, *04:10, *10:01, *14:02, *14:06, *16:02
 VH11, 133011562 × 10–14*04:01, *04:02, *04:03, *04:04, *04:05, *04:08, *04:09, *04:10
 QA70, 743892504 × 10–13*01:01, *01:02, *01:04, *04:01, *04:04, *04:05, *04:08, *04:09, *04:10, *14:02, *14:06, *15:01, *15:02, *15:03
 YLA47, 67, 743742007 × 10–20*01:01, *01:02, *01:04, *04:01, *04:04, *04:05, *04:08, *04:09, *04:10, *10:01, *14:02, *14:06, *16:02
 HQA13, 70, 742581047 × 10–18*0:401, *04:04, *04:05, *04:08, *04:09, *04:10
 QRAA70, 71, 73, 742951596 × 10–12*01:01, *01:02, *01:04, *04:04, *04:05, *04:08, *04:10, *14:02, *14:06
 QKAA70, 71, 73, 7496134 × 10–5*04:01, *04:09
 VHYLA11, 13, 47, 67, 7425810410–17*04:01, *04:04, *04:05, *04:08, *04:09, *04:10
Resistance epitopes    
 D70702092 × 10–11*01:03, *04:02, *07:01, *08:01, *08:02, *08:03, *08:04, *08:06, *08:07, *08:11, *11:01, *11:02, *11:03, *11:04, *12:01, *12:02, *13:01, *13:02, *13:03, *13:04, *13:05, *13:06, *13:10, *14:03, *14:16, *16:01, *16:02
 SD11, 701382272 × 10–10*08:01, *08:02, *08:03, *08:04, *08:06, *08:07, *08:11, *11:01, *11:02, *11:03, *11:04, *12:01, *12:02, *13:01, *13:02, *13:03, *13:04, *13:05, *13:06, *13:10, *14:03, *14:16
Susceptibility minus resistance epitopes    
 LA67, 74[–D70]3701943 × 10–24*01:01, *01:02, *01:04, *04:01, *04:04, *04:05, *04:08, *04:09, *04:10, *10:01, *14:02, *14:06

The epitopes at positions 67 and 74 are highly polymorphic (i.e., LA, FE, LL, LE, IL, LR, FA, IQ, FL, and IA). In contrast to the susceptibility conferred by LA67, 74, IA67, 74 was associated with resistance to RA (P = 1.7 × 10–5, OR 0.50 [95% CI 0.38–0.66]). FL67, 74 was also weakly associated with resistance, but the remaining epitopes were neutral. Several DRB1*04 alleles (i.e., *04:03, *04:06, and *04:07) contained the neutral LE67, 74 epitope instead of the LA67, 74 and were not associated with RA susceptibility or resistance (collective OR 1.13 [95% CI 0.75–1.71], P = 0.60).

Consistent with the findings of previous studies (17), the epitope most strongly associated with resistance to RA was D70. Epitopes made up of 2 or more resistance residues (e.g., SD11, 70) did not show greater significance than D70 alone, which supports the theory that the presence of D70 is the major contributor to resistance (17). Only 1 allele contained both the LA67, 74 and D70 epitopes (DRB1*16:02), and it was neutral with respect to disease susceptibility. The susceptibility of DRB1*04:01, *04:04, and *04:05 appeared to be due to the presence of susceptibility residues at positions 11, 13, 47, 67, 70, and 74 and the lack of resistance residues. DRB1*04:02 was unique among the DRB1*04 alleles in that it also contained 2 amino acids, D70 (P = 2 × 10–11, OR 0.37 [95% CI 0.28–0.49]) and I67 (P = 5 × 10–7, OR 0.46 [95% CI 0.36–0.61]), that were significantly associated with resistance to RA.

Among the alleles most strongly associated with resistance to RA, DRB1*13:01 and *13:02 contained 8 resistance residues and only 1 susceptibility residue (A74). The other statistically significant resistance allele, DRB1*08:01, contained 6 resistance residues and 1 susceptibility residue (Y47). It would therefore appear that disease susceptibility and resistance may be defined by the combination of susceptibility and resistance residues, and resistant amino acids, rather than by a single epitope. Defining a shared epitope as the presence of the susceptibility epitope LA67, 74 and the absence of the resistance epitope D70 resulted in a novel epitope LA67, 74[-D70] that was even more strongly associated with RA (P = 3 × 10–24, OR 4.11 [95% CI 3.11–5.43]) than was LA67, 74 by itself.

Within the well-known shared epitope consisting of amino acid positions 70–74, D70 and E71 (P = 6 × 10–6, OR 0.38 [95% CI 0.26–0.55]) were found to be associated with resistance, but neither R70 (P = 1.0), R71 (P = 0.16), nor K71 (P = 0.27) was independently associated with susceptibility. R72 is not polymorphic, and neither A73 (P = 0.80) nor G73 (P = 0.07) was statistically significant. These observations suggest that the significance of this region is primarily restricted to positions 70 and 71. Among the 7 epitopes at these positions (QR, QK, RR, DK, QA, DR, and DE), only the QR70, 71 epitope was strongly associated with susceptibility to RA (P = 1.4 × 10–11, OR 2.6 [95% CI 2.0–3.4]). With respect to resistance, DR70, 71 (P = 3 × 10–6, OR 0.51 [95% CI 0.39–0.66]) and DE70, 71 (P = 2 × 10–6, OR 0.4 [95% CI 0.3–0.5]) were nearly equal. The fact that position 71 could be either a basic or an acidic amino acid suggests that the resistance associated with this epitope is due to the aspartic acid at position 70. The remaining epitopes (QK, RR, DK, and QA) were neutral with respect to disease susceptibility and resistance. These data suggest that, with respect to conferring susceptibility to RA, the absence of an acidic amino acid at position 70 is more important than the presence of Q, R, or K.

HLA susceptibility and resistance to RA appeared to have a strong gene-dose effect. Compared to subjects lacking LA67, 74, those who were homozygous for LA67, 74 (i.e., LA67, 74+/+) were significantly more susceptible to RA (OR 5.9 [95% CI 4.0–8.9], P < 0.0001) than were LA67, 74+/– heterozygotes (OR 1.69 [95% CI 1.13–2.5], P = 0.009). Similarly, compared to individuals lacking D70 on either allele, those who were homozygous for D70 were more resistant to RA (OR 0.20 [95% CI 0.13–0.31], P = 0.0001) than were D70+/– heterozygotes (OR 0.45 [95% CI 0.29–0.69], P = 0.0002).

Binding of citrullinated vimentin to DRB1 epitopes.

Antibodies to citrullinated vimentin have been detected in >70% of patients with RA (22), suggesting that it is a major autoantigen. Since it was not practical to analyze all RA autoantigens for HLA epitope binding, we investigated the ability of various RA susceptibility and RA resistance DRB1 alleles to bind citrullinated vimentin65–77, a peptide that has been shown to activate T cell receptors when presented by DRB1*04:01 (23). As can be seen in Figure 1, the alleles that were associated with significant resistance to RA (DRB1*08:01 and *13:01) as well as 11 of 13 other alleles containing the D70 resistance epitope did not bind citrullinated vimentin65–77. DRB1*01:03 was the only D70-positive allele that bound citrullinated vimentin65–77. Citrullinated vimentin65–77 bound strongly to DRB1*04:01 and to the other DR4 alleles containing the LA67, 74 epitope (DRB1*04:04, *04:05, and *04:08), as well as 4 of 5 other alleles with the LA67, 74 epitope, but it did not bind to *04:02. Overall, the level of citrullinated vimentin65–77 binding to HLA–DRB1 alleles containing the LA67, 74 epitope was significantly greater than the binding to alleles containing the D70 epitope (Figure 2).

Figure 1.

Binding of citrullinated vimentin peptide65–77 (SAVRLCitSSVPGVR) to HLA–DRA1/B1 molecules. Results are expressed as a percentage of the binding observed with a positive control specific for each allele. The broken red line demarcates positive from negative binding.

Figure 2.

Citrullinated vimentin65–77 binding to HLA–DRB1 molecules according to the presence of LA67, 74 or D70 epitopes. Symbols represent individual data points; horizontal lines show the mean.

A model of DRB1*04:01 binding to citrullinated vimentin65–77 is shown in Figure 3. The LA67, 74 epitope in DRB1*04:01 likely favors the binding of the neutral citrulline residue. Despite the statistical significance of LA67, 74 in association with RA and citrullinated vimentin binding, our model suggests that Q70 and K71 would form hydrogen bonds with the peptide, and they therefore play critical roles in peptide binding (Figure 4). In contrast, the presence of D70 and E71 in *04:02 results in a more acidic pocket and would repel the weak negative citrulline dipoles and greatly reduce peptide binding.

Figure 3.

Models of DRA*01:01/B1*04:01 and DRA*01:01/B1*04:02 molecules showing the presence of the LA67, 74 and D70 epitopes. The models were created with the use of Swiss PDB Viewer version 4.0.1 and Discovery Studio Visualizer version 3.1 software, using coordinates for DRA*01:01/DRB1*04:01 (30) from the Research Collaboratory for Structural Bioinformatics Protein Data Bank. The influenza hemagglutinin peptide was replaced with citrullinated vimentin65–77, which is depicted with valine67 in the P1 pocket and citrulline70 projecting into the P4 pocket of DRB1*04:01. DRB1*04:02 was modeled from the DRB1*04:01 coordinates by modifying the residues at positions 67 (L→I), 70 (Q→D), 71 (K→E), and 86 (G→V).

Figure 4.

Model of the interaction between citrullinated vimentin65–77 and DRB1*04:01 residues 67–74. Hydrogen bonds between the citrullinated peptide and DRB1 residues Q70 and K71 were calculated with Discovery Studio Visualizer version 3.1 software and are shown by the broken black lines.

DISCUSSION

Rheumatoid arthritis has long been associated with DR4 in general and with amino acids 67–74 in particular. However, a systemic analysis of all amino acids in the DRB1 peptide-binding groove has never been undertaken. The main difficulty in undertaking such an analysis is the fact that >2 million potential DRB1 epitopes exist. When the analysis was restricted to the individual residues that were statistically significant, the number of epitopes comprising 1–5 amino acids was reduced to <5,000. The majority of these epitopes (86%) were neutral with respect to disease susceptibility, and the remaining epitopes exhibited >95% concordance with respect to alleles (data not shown).

Our findings confirm the significance of amino acids at positions 67–74 in susceptibility to RA. However, the epitope most strongly associated with RA was LA67, 74 rather than QRRAA70–74, and susceptibility correlated with the binding of citrullinated vimentin65–77. We also confirmed that resistance to RA was associated with several positions, the strongest of which was D70. The presence of D70 abrogated both the binding of citrullinated vimentin65–77 and the susceptibility to RA conferred by the LA67, 74 epitope.

DRB1 positions 67 and 74 have been reported to play variable roles in peptide binding and T cell recognition. Doherty et al (24) demonstrated that an I67→F67 substitution inhibited T cell proliferation to a viral peptide without affecting peptide binding. Similarly, Fu et al (25) demonstrated that an A74→E74 substitution inhibited T cell responses to hemagglutinin peptide307–319 without inhibiting peptide binding. In contrast, Jurcevic et al (26) demonstrated that positions 67 and 71 contributed equally and synergistically to human immunodeficiency virus peptide binding to DRB1*01:03. More recently, Menconi et al (27) reported that DRB1 residues 67 and 74 are strongly associated with the development of autoimmune polyglandular syndrome. Thus, while LA67, 74 may not directly interact with the citrulline residue, it may create the appropriate pocket in which citrullinated vimentin can bind to some DRB1 alleles.

As illustrated in a model of DRB1*04:01 binding to citrullinated vimentin65–77 (Figure 4), the carbonyl group of the deiminated arginine side chain in peptide position 70 hydrogen bonds with the amides of Q70 and K71. DRB1*04:04, *04:05, and *04:08 have a K→R substitution at position 70, which appears to allow both peptide binding and disease susceptibility. However, the presence of the acidic D70 within the LA67, 74 epitope (e.g., DRB1*16:02) would produce a repulsive force. The only D70-positive allele to bind citrullinated vimentin was DRB1*01:03, which is not associated with RA. Similarly, DRB1*03:01 and *15:01 also bound citrullinated vimentin and were not associated with RA susceptibility. It is possible that these alleles bind the citrullinated peptide in a conformation that does not trigger T cell activation, but this hypothesis is yet to be tested.

It should be noted that the peptide-binding assay used in this study employs monoclonal antibodies that recognize HLA–DRB1 molecules in conformations that are associated with the ability to activate T cells. However, the interaction between a T cell receptor and its cognate HLA allele is highly specific, and binding in this assay cannot be unequivocally related to activation of T cells in response to citrullinated vimentin. For example, citrullinated vimentin bound strongly to DRB1*15:01 in this assay, but not to *15:02 or *15:03, which differ from *15:01 by only a single amino acid. Although DRB1*15:01 has recently been associated with the development of anti–citrullinated fibrinogen antibodies (28), this allele has not been unequivocally linked to RA. Finally, we only tested a single vimentin peptide, and many other citrullinated vimentin peptides, as well as citrullinated forms of other protein peptides, are likely to be important in RA (29, 30). Despite these caveats, the immunoassay demonstrated remarkable concurrence between citrullinated peptide binding and susceptibility to RA, both at the epitope level and the allele level. We predict that this approach will allow detection of more obscure HLA epitopes in other immunologic diseases.

When all the subjects with LA67, 74 epitopes were censored from the data and the remaining 114 patients and 248 controls were reanalyzed, the epitope analysis only identified epitopes associated with the DRB1*09:01 allele (OR 3.4 [95% CI 1.7–6.7], P = 0.0004, Pcorrected = 0.02). DR9 has been shown to be associated with RA in Chileans (31). Of the 23 RA patients with DR9, 15 were Korean, 3 were Amerindian, 3 were Caucasian, and 2 were Spanish. Although DRB1*09:01 does not appear to bind citrullinated vimentin65–77, it has an epitope in pocket 4 (FLERRRAE67–74) that may favor the binding of other citrullinated peptides. Thus, a more thorough analysis of HLA epitopes will likely provide a better understanding of both genetic susceptibility and target antigens in this and other autoimmune diseases.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Freed had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Freed, Schuyler, Aubrey.

Acquisition of data. Freed, Schuyler, Aubrey.

Analysis and interpretation of data. Freed, Schuyler, Aubrey.

Ancillary