HLA loci heterozygosity modulates genetic risk in idiopathic inflammatory myopathies

Abstract

Objectives: Idiopathic inflammatory myopathies (IIMs) are rare autoimmune disorders. Genetic association studies have highlighted the role of human leukocyte antigen (HLA) polymorphisms in IIM. We aimed to characterise the nonadditive effects (dominance and interaction) of HLA alleles on IIM risk.

Methods: This study included a total of 3206 IIM cases and 11,697 controls of European ancestry. HLA alleles were imputed using a multiancestry HLA reference panel. Logistic regressions were conducted to estimate the nonadditive effects of HLA alleles. Clinical subgroup analysis, calculation of phenotypic variance explained, and stepwise conditional analyses were conducted to further characterise these effects.

Results: We identified significant nonadditive effects in 5 HLA genes, particularly in the core alleles of ancestral haplotype 8.1 (8.1 AH), including HLA-B*08:01 (P = 3.93 × 10^-13), HLA-C*07:01 (P = 3.14 × 10^-8), HLA-DQA1*05:01 (P = 3.03 × 10^-9), HLA-DQB1*02:01 (P = 3.53 × 10^-23), and HLA-DRB1*03:01 (P = 8.47 × 10^-21). Notable risk difference between heterozygotes and homozygotes was observed in IIM, such as HLA-DRB1*03:01 (homozygote odds ratio [OR], 2.17; heterozygote OR, 3.13). In the interaction model, HLA-DQA1 and HLA-DRB1 showed specific significant allelic interactions. The nonadditive effect model explained a larger proportion of phenotypic variance than the model with additive effects alone. Conditional analysis indicated the independent nonadditive effect of HLA-DRB1*03:01 in 8.1 AH and amino acid residue Arg-74 in HLA-DRB1.

Conclusions: This study identified significant nonadditive effects within the HLA region of IIM. A genetic risk model including nonadditive effects could provide more accurate individual risk estimates. These findings highlight a complex role of HLA heterozygosity in the development of IIM and support further research into HLA nonadditive effects with clinical relevance.

Figure 1.. HLA and non-HLA Variant Risk Patterns in IIM.

(A) Comparison of odds ratios (OR) for common high-resolution HLA allele heterozygotes and homozygotes in the ImmunoChip cohort. The allele non-carriers serve as the reference group, with a risk set to 1. The solid line represents the expected relationship under the additive assumption. Error bars indicate 95% confidence intervals. (B) Comparison of odds ratios for non-HLA variants in ImmunoChip dataset (P-value < 5×10⁻⁶ in previous meta-analysis). The OR calculation and visualization used here are the same as those of HLA alleles. The SNP non-carriers serve as the reference group, with a risk set to 1. (C) De Finetti diagrams of common high-resolution HLA alleles in ImmunoChip cohort cases and controls. Dots represent observed frequencies of homozygotes and heterozygotes, while the solid line shows the expected proportion of heterozygotes under Hardy-Weinberg equilibrium. (D) De Finetti diagrams of non-HLA variants in ImmunoChip dataset (P-value < 5×10⁻⁶ in previous meta-analysis).

Figure 2.. Non-additive interaction effect of HLA-DQA1 and HLA-DRB1 combinations in IIM.

(A, B) Logistic regression model with an interaction term in the HLA-DQA1 and HLA-DRB1 regions. Common alleles and others (rare alleles aggregated) were included in the model. The additive odds ratio (OR) of HLA alleles represents their individual risk without interaction with others. Interaction ORs and their 95% confidence intervals are presented in the upper triangle; corresponding P values are shown in the lower triangle (red color indicates significant interaction). The HLA-DQA1*05:01 and HLA-DRB1*03:01 were used as the reference allele, with a risk set to 1.

Figure 3.. Additive and non-additive effects of the 8.1 ancestral haplotype.

(A) Linkage structure of core alleles within the 8.1 AH. (B) P-values and effect sizes for the additive and non-additive effect tests for the 8.1 AH proxy (HLA-B*08:01-HLA-DRB1*03:01). The additive and non-additive bars correspond to the model fit improvement test after including the additive and non-additive term of HLA-B*08:01-HLA-DRB1*03:01. The odds ratios (ORs) and 95% confidence intervals (CIs) for the additive model, heterozygotes, and homozygotes are reported relative to non-carriers of the haplotype. (C) Conditional analyses of the 8.1 AH alleles using both additive and non-additive effects. P.additive and P.non-additive indicate the significance level of the alleles at each round of conditioning. Conditioning alleles were selected based on the model fit improvement P-value after including both additive and non-additive terms. The dotted lines represent the Bonferroni-corrected significance threshold (0.05). ORs and CIs for the additive model, heterozygotes, and homozygotes at each step are shown in the third column. The three rows of the plot show allele significance and effect with no allele conditioned, conditioned on HLA-DRB1*03:01, and conditioned on both HLA-DRB1*03:01 and HLA-B*08:01.

Figure 4.. Stepwise conditional analyses of amino acid residues in HLA-DRB1*03:01.

(A) Unconditioned. (B) Conditioned on 74: R. The conditioning residues were selected based on their combined additive and non-additive effects. The “additive effect” and “non-additive effect” panel displays the significance levels of amino acids in each round of conditioning. The dotted lines indicate the Bonferroni-corrected significance threshold (0.05). The top amino acids with the strongest significance are labelled. Colours represent the LD r² between the marked residues and the residue with the strongest significance.