Discussion
In this study, we fine mapped PsV risk within the MHC region. In addition to imputing variants of HLA genes, we evaluated risk of the HLA-like gene MICA by creating a MICA reference panel that empirically demonstrated high imputation accuracy. Our study identified multiple HLA-C∗06:02-independent risk variants of both class I and class II HLA genes for PsV susceptibility (HLA-B, HLA-A, and HLA-DQA1), but no apparent risk attributable to MICA. We also observed that risk heterogeneity between PsA and PsC could be explained by polymorphisms of a single amino acid site encoded by HLA-B, suggesting that different genetic architectures underlie the overall risk of PsV and that of its subphenotypes. To our knowledge, ours is the largest HLA fine-mapping study of PsV associations in the MHC to date and defines genetic heterogeneity between PsA and PsC subphenotypes.
HLA-C∗06:02 has the strongest association with PsV risk, as reported previously.7–12,17,18 We demonstrated that no single HLA-C amino acid polymorphism was more strongly associated than the HLA-C∗06:02 classical allele, suggesting that the haplotype sequence including HLA-C∗06:02 itself should be the origin of PsV risk. Several hypotheses might explain this. Clop et al. reported noncoding regulatory variants that are located in enhancer motifs and that are unique to the HLA-C∗06:02 haplotype.34 A combination of the polymorphisms in multiple HLA-C amino acid sites could effectively tag HLA-C∗06:02.35 We did observe a more modest independent effect at HLA-C∗12:03 in addition to the large HLA-C∗06:02 effect; Helms et al. reported that HLA-C∗06:02 and HLA-C∗12:03 share several functional domains and peptide-binding pockets of HLA-C.5 We note that HLA-C∗12:03 did not show an independent association signal after we conditioned on HLA-C∗06:02 and every classical HLA-B allele (p = 0.12), suggesting the possibility that the observed HLA-C∗12:03 association might reflect risk at other HLA-B alleles in LD. Further functional studies will be necessary for elucidating the role of HLA-C in PsV risk.
In contrast, outside of HLA-C, amino acid polymorphisms (HLA-B amino acid positions 67 and 9, HLA-A position 95, and HLA-DQα1 position 53) demonstrated stronger associations than did classical alleles HLA-B, HLA-A, and HLA-DQA1. All of these amino acid sites were located within the HLA antigen binding (Figure 5 and Figure S3). Positions 67 and 9 in HLA-B have been identified in HLA fine-mapping studies for other immune-related diseases.25–27 We note that LD structures between the amino acid positions could yield potential ambiguity in fine mapping of the causal amino acid position, and this might be clarified with larger studies.
The contribution of HLA-like genes to immune-related disease risk has long been a topic of discussion.38,39 Although our imputation of MICA alleles was highly accurate, our study did not observe independent MICA risk of PsV after we conditioned on the neighboring risk HLA genes HLA-C and HLA-B. Previous studies focusing on MICA risk did not apply robust conditioning on all classical HLA-C and HLA-B alleles and thus could have potentially reflected the associations of HLA-C and HLA-B via LD with them.19,20
Here, we were able to successfully decompose the genetic architecture of PsA and PsC to a shared component and a subphenotype-specific component. Our study demonstrates that the HLA gene associated with the risk heterogeneity between PsA and PsC (HLA-B) is distinct from the HLA gene most associated with overall PsV risk (HLA-C). Previous studies have reported that HLA-C∗06:02 has different effect sizes for PsA and PsC, naturally leading to a hypothesis that the heterogeneity is driven by the difference in HLA-C, the major risk factor.13–16,23 However, our observation is more concordant with a model where the two PsV subtypes generally share the same risk alleles, including HLA-C∗06:02, but differ at a specific locus that contributes to subtype differences.
HLA-B amino acid position 45 is the driving MHC position that modulates differential risk of PsA and PsV. The effect of Glu at HLA-B position 45 confers substantial risk of psoriatic arthritis and explains previously reported associations at HLA-B∗27. This site is located within the binding groove of HLA-B and is classified as one of the functional pockets influencing receptor cell-surface expression or antigen peptide binding or presentation.40 Being able to clinically distinguish those individuals who have isolated skin disease (PsC) from those who develop joint disease (PsA) has clinical importance. PsA often occurs in addition to psoriatic skin disease and almost always requires systemic therapy, for example, with anti-TNF or other biologic medications, for the prevention of destructive joint disease. In contrast, PsC can in many instances be managed with topical treatments alone.3 Our findings might contribute to utilizing information on HLA variants to improve diagnostic approaches for clinical subphenotypes, as suggested for other complex diseases.26 There is a possibility that the HLA-B45 amino acid residue tags other HLA-B driving risk variants at other amino acid sites, although we observed only limited LD between these sites (Figure S4). We also note that our method of evaluating OR heterogeneity (i.e., pheterogeneity) might be conservative because of shared control subjects in the PsA and PsC case-control analyses. Further studies will be required for elucidating the functional mechanisms that result in differential PsA and PsC risk.
In summary, our study fine mapped risk of multiple class I and class II HLA genes in PsV and its subphenotypes through large-scale HLA and MICA imputation. Our study should contribute to our understanding of HLA variants in the etiology of PsV.