HLA-DRB1*11 and variants of the MHC class II locus are strong risk factors for systemic juvenile idiopathic arthritis

Abstract

Systemic juvenile idiopathic arthritis (sJIA) is an often severe, potentially life-threatening childhood inflammatory disease, the pathophysiology of which is poorly understood. To determine whether genetic variation within the MHC locus on chromosome 6 influences sJIA susceptibility, we performed an association study of 982 children with sJIA and 8,010 healthy control subjects from nine countries. Using meta-analysis of directly observed and imputed SNP genotypes and imputed classic HLA types, we identified the MHC locus as a bona fide susceptibility locus with effects on sJIA risk that transcended geographically defined strata. The strongest sJIA-associated SNP, rs151043342 [P = 2.8 × 10(-17), odds ratio (OR) 2.6 (2.1, 3.3)], was part of a cluster of 482 sJIA-associated SNPs that spanned a 400-kb region and included the class II HLA region. Conditional analysis controlling for the effect of rs151043342 found that rs12722051 independently influenced sJIA risk [P = 1.0 × 10(-5), OR 0.7 (0.6, 0.8)]. Meta-analysis of imputed classic HLA-type associations in six study populations of Western European ancestry revealed that HLA-DRB1*11 and its defining amino acid residue, glutamate 58, were strongly associated with sJIA [P = 2.7 × 10(-16), OR 2.3 (1.9, 2.8)], as was the HLA-DRB1*11-HLA-DQA1*05-HLA-DQB1*03 haplotype [6.4 × 10(-17), OR 2.3 (1.9, 2.9)]. By examining the MHC locus in the largest collection of sJIA patients assembled to date, this study solidifies the relationship between the class II HLA region and sJIA, implicating adaptive immune molecules in the pathogenesis of sJIA.

Keywords: Still’s disease; autoinflammation; human leukocyte antigen; systemic juvenile idiopathic arthritis.

Fig. S1.

Multidimensional scaling plots of directly genotyped sJIA cases and healthy control subjects with multiancestry HapMap3 reference populations. MDS plots demonstrate the ancestral distribution of cases and controls from each stratum, as determined with the set of directly genotyped SNPs from each stratum after performing LD pruning for pairs with EM r² < 0.4 and removing the 24 regions of long-range LD. Excluded individuals in each plot are marked with a red boxed X.

Fig. S2.

PCA plots from nine sJIA case-control collections. PC plots demonstrate the ancestral distribution of each sJIA case-control stratum, as determined in the set of directly genotyped SNPs from each stratum after performing LD pruning for pairs with EM r² < 0.4 and removing the 24 regions of long-range LD.

Fig. S3.

P-P plots from additive association testing of directly genotyped SNPs from nine sJIA case-control collections. P-P plots were generated and genomic control inflation factors (λ_GC) were calculated for the set of directly genotyped SNPs from each stratum after performing LD pruning for pairs with EM r² < 0.4 and removing the 24 regions of long-range LD.

Fig. 1.

The MHC is an sJIA susceptibility locus. (A) Dominant-model association results from fixed-effect meta-analysis of MHC-region SNPs in nine independent sJIA case-control populations are depicted by blue circles. The results of association testing conditioned on the effect of the top sJIA-associated MHC-region SNP (rs151043342, blue star) are shown as orange circles, and the top SNP from the conditional analysis (rs12722051) is marked by an orange star. (B–D) Forest plots depict the magnitude of the effects of rs151043342 (B), rs12722051 (C), and HLA-DRB1*11 (D) on sJIA risk in the individual study populations and by meta-analysis. (E) Association results from fixed-effect meta-analysis of best-guess MHC-region SNPs (squares) and classic HLA types (triangles) from six Western European study populations are depicted by red shapes. Association results conditioned on the effect of the top SNP from A (rs151043342, red star) are shown as green shapes. Association results conditioned on the effect of the study’s top risk factor (HLA-DRB1*11, red triangle) are shown as orange shapes. *rs12722051 was not imputed with high quality in this population.

Fig. S4.

Examination of MHC region SNPs with evidence of heterogeneity in meta-analysis. Regional association plots demonstrates the results of fixed effect meta-analysis (circles) and random-effects meta-analysis (squares) of dominant model associations for the 398 MHC region SNPs for which a high probability of heterogeneity (I² > 0.7) was identified.

Fig. 2.

Multiple HLA-DRB1*11 alleles are associated with sJIA. Forest plots depict the odds ratios of the sJIA-associated HLA-DRB1*11 family members, HLA-DRB1*11:01 (A) and HLA-DRB1*11:04 (B). Ribbon models of an HLA-DR molecule (C and D) demonstrate the two defining features of HLA-DRB1*11 molecules: a glutamate residue at position 58 (white) on the exterior surface of the molecule with its side-chain pointing away from the peptide-binding groove, and the combination of peptide-binding groove residues that is unique to HLA-DRB1*11 (shown in blue). The models were created from PDB ID code 3LQZ. *Allele frequency < 0.01%.

Fig. 3.

Extended, HLA-DRB1*11-containing haplotypes are associated with sJIA. Forest plots depict the magnitude of the effects of the HLA-DRB1*11:01 (A) and HLA-DRB1*11:04 (B) containing forms of the HLA-DRB1*11-HLA-DQA1*05-HLA-DQB1*03 haplotype on sJIA risk. To define haplotypic boundaries (C), we independently analyzed haplotypes of HLA types with the top 400 population-specific, sJIA-associated SNPs in the United Kingdom (blue circles) and the United States (red circles) populations. This revealed a 397-kb sJIA-associated haplotype in the United Kingdom population [P = 2.6 × 10⁻¹⁰; OR 3.4 (2.3, 5.0)] and a 243-kb disease-associated haplotype in the United States population [P = 1.8 × 10⁻⁵; OR 1.8 (1.3, 2.4)] that was fully contained within the United Kingdom haplotype. Comparison of the alleles of the 113 markers in common between the two haplotypes (open boxes) revealed that the United States and United Kingdom haplotypes were identical.