In the present study, we conducted the Japanese GWAS for SSc comprising a total of 114,027 subjects consisting of 1,428 cases and 112,599 controls, resulting in the largest Asian GWAS for SSc ever, and identified three novel significant loci. One of the SNPs, rs6697139 located at the intergenic region of FcγR family genes as well as its complete LD SNP, rs10917688, had a strong effect size (OR ~ 2.0) and were found to be plausible causal SNPs by fine-mapping. Considering the much higher MAF but with no significant association in the European dataset, this Japanese-specific loci was worth further investigation. Notably, rs10917688 is positioned within a cCRE and likely to be a part of binding motifs of IRF8, a key TF for the developmental trajectory of B cells (34) as well as dendritic cells, macrophages, and NK cells (35). One of the enhancer-related histone marks, H3K4me1, was identified in B cells and heritability of active histone marks was enriched in B cells both in European and Japanese populations, suggesting IRF8-FCGR/FCRL axis in B cells might be a novel pathological mechanism in SSc.
FcγRs are expressed on the surface of both innate and adaptive immune cells and confer immune modulatory responses by binding the Fc portion of immunoglobulin G (IgG). Based on their binding affinity, FcγRs are divided into low-affinity and high-affinity FcγRs and there are five low-affinity FcγRs, FcγRIIa, FcγRIIb, FcγRIIc, FcγRIIIa, and FcγRIIIb, each of which is encoded by FCGR2A, FCGR2B, FCGR2C, FCGR3A, and FCGR3B, respectively (36). Due to the close proximities of FcγR gene locations and segmental duplications (SD), loci with two or more highly similar and duplicated regions, it is difficult to determine which gene is critically affected by a causal SNP for SSc development. SD loci are enriched for immune genes and often show copy number variations (CNVs) (36). Indeed, CNVs in the FCGR region have been suggested to be associated with multiple AIDs including RA, SLE, celiac disease, and inflammatory bowel disease (36). However, the association of CNVs with SSc has never been reported, and thus the effect of the SNPs rather than CNVs is likely to be associated with SSc. In addition, the association was not significant despite higher MAFs in European populations, suggesting the association of these loci is specific to Japanese, which may be independent of CNV.
FCRLA encodes a protein similar to the FcγR and is selectively expressed in peripheral and germinal center B cells, but not in terminally differentiated plasma cells. Thus, it has been considered to be involved in B-cell development through consensus Immunoreceptor Tyrosine-based Activation Motifs (ITAMs) and Immunoreceptor Tyrosine-based Inhibitory Motifs (ITIMs) (37). The ligands had long been unknown, but it has recently been found that they bind IgA, IgM, and IgG inside endoplasmic reticulum (ER) and hence they are now supposed to work as ER chaperones (38). Although they are mainly retained in ER, one of the isoforms has a longer signal peptide that allows for secretion outside cells, implying its role as a secreted protein (39). Despite these advances in understanding, FCRLA is still a gene of unknown function both for physiological and pathological conditions.
FcγRIIa, CD32a, is expressed in monocytes, neutrophils, and eosinophils, and mediates phagocytosis of opsonized antigens or immune complexes. FcγRIIa also has ITAMs and ITIMs and thus can be functionally divergent. A well-known functionally relevant SNP, rs1801274, is a nonsynonymous mutation, which alters arginine (R) to histidine (H) at amino acid position 131 of the extracellular domain and confers binding to IgG2 and IgG3 with higher affinity than RR receptors (40). The SNP has been implicated for the susceptibility to SLE, Kawasaki disease, cystic fibrosis, and several infectious diseases including invasive pneumococcal or meningococcal disease, severe malaria, dengue fever, respiratory syncytial virus, and SARS-CoV (40).
FcγRIIb, CD32b, is a solely inhibitory receptor among FcγRs expressed mainly on B cells but is also expressed on dendritic cells, macrophages, and mast cells. It inhibits phagocytosis of immune complexes and antibody production by B cells (41). FCGR2B has been implicated for its association with multiple AIDs including RA (42), SLE (43, 44), type I diabetes (45), and IgG4-related disease (46). Although the genetic association of FCGR2B with SSc has never been reported, one small study observed higher levels of anti-FcγRIIB/C antibodies in sera of Japanese dcSSc patients compared to those of lcSSc or non-SSc controls (47).
Despite limited sample numbers and genotype information, our eQTL analysis using the Japanese eQTL dataset of six WBC subpopulations (26) showed a trend of decreased FCGR2A and FCGR2B expression in relevant cell types, B cells, NK cells, and Monocytes. Although we were not able to assess an eQTL effect on FCRLA due to a lack of genotype information in the database, the gene is also located in close proximity to rs6697139 and rs10917688. Thus, further eQTL studies with a sufficient number of samples are highly warranted to measure the effect of the variant, rs10917688, on any of the gene(s) in this region.
Interferon-regulatory factor 8 (IRF8), also known as interferon consensus sequence binding protein (ICSBP), is a TF exclusively expressed in hematopoietic lymphoid and myeloid cells. IRF8 deficiency in humans was previously documented and the subjects suffered from severe immunodeficiency due to depletion or impaired functions of dendritic cell subsets, monocytes, and NK cells (48). Sequence variants near IRF8 have repeatedly been identified as risk factors for various AIDs including SSc (12) according to the previous GWASs. IRF8 has also been implicated in early B-cell development, Igk rearrangement, germinal center formation, and plasma cell generation (49). Several mouse experiments showed that IRF8 worked together with other TFs, such as PU.1, IRF4, IKAROS, or E2A, to modulate lineage specification, commitment, and differentiation in B cells (50, 51). An intronic variant, rs8057456, was one of GWAS significant variants for serum immunoglobulin levels in the GWAS of Scandinavian populations (52). Together these studies strongly suggest that genetic variations of IRF8 in B cells can modulate susceptibility to autoimmunity including SSc. The present study indicated that IRF8 may bind to a motif containing FCGR/FCRL variant, rs10917688, within a cCRE, which has never been reported so far. Furthermore, our genotype combination analysis (Fig. 4B) revealed that IRF8-binding to this motif is indispensable for the risk effect of rs10917688, supporting an interactive effect of IRF8 and the FCGR/FCRL variant. Although we were not able to identify eGene(s) for rs10917688 and thus the result has not been fully convinced, further independent studies for both east Asians and other populations and validation experiments such as reporter assays or single nucleotide editing approaches will validate our findings as well as clarify more precise molecular mechanisms.
As can be observed in other AIDs, our study revealed that SSc has also polygenic architecture. Intriguingly lcSSc or ACA-positive SSc tended to more fit the PRS than systemic types of SSc, implying different genetic backgrounds and hence pathological mechanisms between these distinctive phenotypes. The PRS constructed from GWAS SNPs moderately fit the predictive model indicating a potential utility of PRS for a risk assessment of SSc in clinical settings. Furthermore, prioritizing the top 5% SNPs annotated according to IRF8-binding in RAMOS cells improved the predictive performance showing the better trans-ancestral portability of PRS with the use of IMPACT-annotated SNPs.
The trans-ethnic GWAS meta-analysis identified a total of 30 GWAS loci, most of which were derived from those identified in European populations. Three of these loci including EOMES, ESR1, and SLC12A5 have never been reported. Eomesodermin encoded by EOMES and T-bet are TFs belonging to the T-box family and known to differently regulate CD8+ T cell differentiation and function as well as exhaustion. (53–55). Considering the association of EOMES variation with multiple AIDs such as RA (56), multiple sclerosis (57), and ankylosing spondylitis (58), as well as immune-related traits including lymphocyte count and granulocyte count (59), variations of EOMES may be a trigger of autoimmunity shared among multiple AIDs including SSc. A nuclear hormone receptor, estrogen receptor 1, encoded by ESR1 is involved in various gene expression which affects cellular proliferation and differentiation in given tissues. ESR1 binds to nuclear factor-κB (NF-κB) and these two molecules mutually trans-repress each other to regulate cellular response including cytokine production (60). Of note, rs230534 in the NFKB1 region was also one of the lead SNPs of our meta-analysis (Table 2). Since females are ~ five times more affected by SSc than males, the association of ESR1 with SSc is quite reasonable, but on the other hand, it was surprising that previous studies have never identified this locus; it might be due to the relatively weak association and effect sizes. Potassium-chloride transporter member 5 (KCC2) encoded by SLC12A5 is a brain-specific chloride potassium symporter and is well-characterized in neuronal cells for its function of maintaining intracellular chloride concentrations (61). However, the functional relation of these loci to immune regulations or tissue fibrosis has never been known thus far and is remained to be investigated.
As we expected, most of the susceptibility genes including novel risk loci identified in the present study overlapped with those associated with other AIDs, especially SLE. Consequently, there was a significant genetic correlation between SSc and SLE, and to a lesser extent RA despite the relatively small number of cases in the present study. Our study did not answer several questions, such as how identified and unidentified susceptibility variants, HLAs, interaction with other genes (62), or environmental factors contribute to the difference in disease phenotypes among AIDs, and this question is also still open to be explored.
As is often the case of current GWASs, we were not able to include SNPs with very low allele frequencies and other types of genetic variations such as structural variants not tagged by SNPs, which are the main limitations of the present study and remained for future studies with more sample sizes and better imputation utilizing a WGS-based reference panel. Nevertheless, our largest-scale Asian GWAS provides valuable insights into the complex genetic architecture of SSc as well as the prioritization of targets for future functional studies.