Objective. HLA association fine-mapping studies have shown the effects of missense variants in HLA-DRB1 on rheumatoid arthritis (RA) susceptibility, prognosis, and autoantibody production. However, the phenotypic effects of expression changes in HLA-DRB1 remain poorly understood. Therefore, we investigated the allele-specific expression of HLA-DRB1 and its effect on an HLA-DRβ1 structure-associated trait in RA. Methods. We quantified the allele-specific expression of each HLA-DRB1 3-field classic allele in 48 Korean RA patients with anti-citrullinated protein antibodies (ACPAs) and 319 healthy European subjects by using both RNA sequencing and HLA-DRB1 genotype data to calculate the relative expression strength of multiple HLA-DRB1 alleles (n = 14 in Koreans and n = 25 in Europeans) in each population. The known association between ACPA level and alanine at position 74 of HLA-DRβ1 in ACPA-positive RA was revisited to understand the phenotypic effect of allelespecific expression of HLA-DRB1 by modeling multivariate logistic regression with the genomic dosage or relative expression dosage of Ala-74 in 2 independent sets of 1,723 Korean RA patients with ACPA. Results. The relative expression strength was highly allele-specific, causing imbalanced allelic expression in HLA-DRB1 heterozygotes. The association between HLA-DRβ1 Ala-74 and ACPA level in RA was better explained by relative expression dosage of Ala-74 than by the genomic dosage (change in Akaike's information criterion = −6.98). Moreover, the expression variance of Ala-74 in Ala-74 heterozygotes with no genomic variance of Ala-74 was significantly associated with ACPA level (P = 2.26 × 10 −3). Conclusion. Our findings illustrate the advantage of integrating quantitative and qualitative changes in HLA-DRB1 into a single model for understanding HLA-DRB1 associations.

Background Structural variations such as copy number variations (CNVs) have a functional impact on various human traits. This study profiled genome-wide CNVs in Korean patients with rheumatoid arthritis (RA) to investigate the efficacy of treatment with TNF-α blockers. Methods A total of 357 Korean patients with RA were examined for the efficacy of TNF-α blocker treatment. Disease activity indexes were measured at baseline and 6 months after the treatment. The patients were classified as responders and non-responders based on the change in disease activity indexes according to the EULAR response criteria. CNVs in the same patients were profiled using fluorescence signal intensity data generated by a genome-wide SNP array. The association of CNVs with response to TNF-α blockers was analyzed by multivariate logistic regression accounting for genetic background and clinical factors including body mass index, gender, baseline disease activity, TNF-α blocker used, and methotrexate treatment. Results The study subjects varied in their responses to TNF-α blockers and had 286 common CNVs in autosomes. We identified that the 3.8-kb deletion at 2q14.3 in 5% of the subjects was associated with response to TNF-α blockers (1.37 × 10 − 5 ≤ P ≤ 4.07 × 10 − 4 ) at a false discovery rate threshold of 5%. The deletion in the identified CNV was significantly more frequent in the non-responders than in the responders, indicating worse response to TNF-α blockers in the deletion carriers. The 3.8-kb deletion at 2q14.3 is located in an intergenic region with the binding sites of two transcription factors, MAFF and MAFK. Conclusions This study obtained the CNV landscape of Korean patients with RA and identified the common regional deletion associated with poor response to treatment with TNF-α blockers. Electronic supplementary material The online version of this article (10.1186/s13075-019-1983-y) contains supplementary material, which is available to authorized users.