Genetics of autoimmune diseases: perspectives from genome-wide association studies

Genetics of autoimmune diseases

Abstract

Genome-wide association studies (GWASs) for autoimmune diseases (ADs) have identified many risk loci and have provided insights into the etiology of each disease. Some of these loci, such as PTPN22, IL23R and STAT4, are shared among different ADs, and the combination of risk loci may determine an individual’s susceptibility for a disease. The majority of GWAS loci are expression quantitative trait loci (eQTLs), where disease-causing variants regulate expression of neighboring (or sometimes distant) genes. Because the eQTL effects are often cell type-specific, the incorporation of epigenetic data from disease-related cell types and tissues is expected to refine the identification of causal variants. The cumulative eQTL effects in multiple genes may influence the activity or fate of immune cells, which in turn may affect the function of the immune system in individuals. In this paper, I review the etiology of ADs by focusing on important immune cells (T_h1 cells, T_h17 cells and regulatory T cells), important pathways (antigen-receptor signaling and type I interferon signaling) and relevant genes identified in GWASs.

Introduction

Autoimmune diseases (ADs) include approximately 80 different disorders affecting ~3–5% of the population (1, 2). It is well accepted that a disease can be classified as autoimmune if one shows that an immune response to a self-antigen causes the disease pathology. ADs manifest a wide variability in terms of targeted tissue(s), sharing the contributions of both cellular and humoral immune responses to tissue injury. Like other complex traits, most ADs are multifactorial, where both genetic and environmental factors are involved in their etiology.

Since genome-wide association study (GWAS) has become a standard approach to the identification of susceptibility genes for complex traits, the genetic risk factors in the development of ADs have been increasingly discovered in the past decade. Furthermore, multi-ethnic meta-analysis of GWASs has been performed for some ADs such as rheumatoid arthritis (3), multiple sclerosis (4) and inflammatory bowel disease (5), where over 100 loci were identified for individual ADs. As a proportion of those loci is shared between different diseases, combinations of genetic factors may determine the phenotype (disease) of individuals. Herein, I review findings from GWASs for ADs, with special focus on important genes and pathways shared among ADs (Table 1). It should be noted that, in some of those genes, the effect of the same variant is opposite among different ADs (e.g. PTPN22 and IFIH1 genes). Moreover, the disease-associated variants are at times different among different ADs, possibly owing to the presence of multiple disease-causing variants in the locus (e.g. the IL23R–IL12RB2 region) and/or the difference in haplotype structure among different ancestral populations studied.

Table 1.

Summary of AD risk loci identified by GWAS

Genes	Relevant cells/signaling pathways	eQTL effect (risk allelea)	Missense variant effect (risk alleleb)	RA	SLE	SSc	SS	PBC	T1D	ATD	Celiac	MS	IBD	AS	PS	Behçet
PTPN22	BCR/TCR	—	R620W Gain?	●c	●				●	●			○
BLK	BCR	Down	—	○	○	○	○
STAT3	Th17	Up	—									●	●		●
STAT4	Th1/Type I IFN	Up	—	●	●	●	●	●	△		△	△	○
IL12A	Th1	Up	—			△	●	●			△	△
IL12RB2	Th1	Up	—			●		●
IL12RB2-IL23R	Th1/Th17	?	—													△
IL23R	Th17	—	R381Q Gain										●	●	●
CCR6	Th17/Treg	Up	—	●						△			△
IL2RA	Treg	Down	—	○					○			○	○
IL10	Treg	Up	—		△								△			●
CTLA4	Treg	Down (Splicing QTL)	—	○					○	○	○
FCRL3	BCR/Treg	Up	—	●						●		○
IRF5	Type I IFN	Up	—	●	●	●	●	●					●
IFIH1	Type I IFN	—	A946T Gain		●				○				●		△

Black circles indicate association identified by GWAS with up-regulated gene expression or gain-of-function in the risk allele, whereas the white circles indicate that with down-regulated expression or loss-of-function in the risk allele. The white triangles indicate GWAS association but its allelic effect on gene expression is unclear. AS, ankylosing spondylitis; ATD, autoimmune thyroid disease; BCR, B-cell receptor; IBD, inflammatory bowel disease; IFN, interferon; MS, multiple sclerosis; PBC, primary biliary cirrhosis; PS, psoriasis; RA, rheumatoid arthritis; SS, Sjögren’s syndrome; SSc, systemic sclerosis; SLE, systemic lupus erythematosus; T1D, type 1 diabetes; TCR, T-cell receptor.

^aThe risk allele is for the majority of ADs associated. ‘Down’ indicates the gene expression is down-regulated in the risk allele, whereas ‘Up’ indicates that is up-regulated.

^bThe amino acid change for the risk allele is underlined.

^cGWAS association and its eQTL effect were reviewed by using web databases such as ImmunoBase (www.immunobase.org/) and HaploReg (www.broadinstitute.org/mammals/haploreg/).

GWASs and expression quantitative trait loci studies

A GWAS usually screens 500000 to 1000000 single-nucleotide polymorphisms (SNPs) by comparing the allele/genotype frequency between subjects affected by disease and those not affected. Although a GWAS can identify disease risk loci, further analysis is required to determine the responsible genes or disease-causing variants in the loci.

Coding and regulatory variants are known to be enriched in the GWAS loci of complex diseases; such variants can affect the function of genes qualitatively and/or quantitatively. Recent in silico analysis that examined functional categories of disease-associated SNPs in 11 common diseases including 6 ADs showed that coding variants explained <10% of heritability in complex diseases, whereas DNase I hypersensitivity sites (DHSs; these indicate transcriptional activity of loci) from multiple cell types, especially enhancer DHSs and cell-type-specific DHSs, explained approximately 80% of heritability (6). Although those DHSs may be also involved in the transcription of non-coding RNAs such as lncRNA, these results may suggest that the majority of disease-causing variants in complex diseases, including ADs, may be regulatory variants that affect gene expression, which we now term expression quantitative trait loci (eQTLs). In fact, 44% of rheumatoid arthritis risk loci are detected as eQTLs in peripheral blood cells (3). This pattern is also the case for other ADs (7–9), indicating that the cumulative effects of quantitative differences in risk genes lead to the onset of AD.

Although the DNA sequence variations are shared among the cell types and tissues in an individual, the regulatory effect of variants (the eQTL effect) would be different between cell types and tissues. Therefore, to analyze the causal mechanism of complex traits, cell-type-specific eQTL analyses are needed. To this end, several eQTL studies such as the ImmVar project (10) and the GTEx project (11) are being performed.

In addition, recent human epigenome projects, such as the ENCODE project (12) and the RoadMap epigenomic project (13), provide clues that may clarify the mechanism of cell-specific eQTL effects. By combining epigenome and transcriptome data, genomic regulatory elements can be comprehensively revealed in individual cell types. A recent study proposed a fine-mapping algorithm to identify candidate causal variants by integrating the GWAS data from 21 ADs with those epigenomic and transcriptomic data (14). This study demonstrated that 60% of causal SNPs in ADs mapped to immune-cell enhancers, represented by the H3K27ac histone marker. In addition, causal variants likely occur near the binding sites of super-enhancers, which are characterized by clusters of regulatory elements bound with very high amounts of transcription factors and especially important for genes associated with determining cell identity (15, 16).

The MHC and antigen presentation

The major histocompatibility complex (MHC) locus is also known as the human leukocyte antigen (HLA) locus and is located on human chromosome 6. In the context of ADs, the MHC was the subject of intensive investigation even before the era of GWASs and is believed to play a central role in the genetic predisposition to AD. In fact, the ‘Manhattan plot’ of GWASs, which can display the results of association tests in a single figure, visually confirms that the MHC locus is the major determinant in most ADs (17). Because the classical MHC genes HLA-A/B/C and HLA-DP/DQ/DR play an essential role in antigen presentation in adaptive immune responses, it is reasonable to suggest that these genes also have critical roles in the autoimmune response in AD. The specific genes and alleles that are associated with individual diseases differ. For instance, in ankylosing spondylitis, a single allele of HLA-B27 accounts for approximately half of disease susceptibility (18), whereas multiple HLA-DRB1 alleles such as *01:01, *04:01, *04:05 and *09:01 are associated with the risk of rheumatoid arthritis (19).

Technical advances in statistical genetics have made it possible to identify the AD-relevant alleles of all classical HLA genes by imputation analysis using SNP array data. This analysis reveals the roles of individual amino-acid-encoding sequences of all classical HLA genes. In rheumatoid arthritis, the shared epitope of HLA-DRB1 at amino acid positions 70–74 has long been believed to be responsible for the disease (20), but the new amino-acid-based association analysis revealed that the amino acid at residue 11 is also crucial (21). In addition, single-amino-acid polymorphisms in HLA-B (at position 9) and HLA-DPB1 (at position 9), which are both located in peptide-binding grooves, are also associated with rheumatoid arthritis risk independently of HLA-DRB1. Furthermore, a recent study of patients with auto-antibodies for anti-citrullinated proteins identified additional association in HLA-A (at position 77) (22).

Involvement of multiple alleles in multiple MHC genes also is seen in other ADs such as type 1 diabetes (23, 24), multiple sclerosis (25) and Graves’ disease (26), implying that multiple autoantigen–MHC allele pairs are involved in the etiology of a single disease. However, because the patients have heterogeneous etiology (even if the same disease is diagnosed), whether these multiple autoantigen–MHC allele pairs are also involved in an individual patient has yet to be clarified.

TCR and BCR signaling

The strength of antigen-receptor signaling is crucial in determining the fate of clones in both T cells and B cells. In several murine models of ADs such as SKG arthritic mice (27) and NZM2410 lupus mice (28), a reduction in antigen-receptor signaling because of mutations in genes involved in antigen-receptor signaling leads to a failure to delete autoreactive cells.

In human ADs, the most important and best-studied genetic factor is PTPN22, which encodes the phosphatase Lyp (lymphoid-specific tyrosine phosphatase) (29). In human primary cells, both TCR and BCR signaling are attenuated in Lyp620W carriers, which may suggest that the phosphatase activity of Lyp620W is enhanced compared with the non-risk Lyp620R protein (30, 31). In fact, the amino acid change of 620W leads to interference with the physical association between Lyp and c-Src kinase (Csk), resulting in increased phosphatase activity of Lyp620W (31, 32). This indicates Lyp620W is a gain-of-function variant in human lymphocytes. The findings in the mouse are more complicated. Ptpn22 knock-out (KO) mice and Ptpn22-619W (a mutant analogous to the human variant) knock-in (KI) mice show similar phenotypes, including an enlarged spleen and thymus, an expanded population of memory/effector T cells and a lymphocyte hyper-responsiveness, suggesting that Ptpn22-619W is a loss-of-function variant (33). Of note, the murine variant protein exhibited reduced stability because of enhanced calpain-mediated degradation, which may be the mechanism of functional loss (33). However, whether this observation is also the case for the human protein is controversial and will require further validation (33, 34). The differences between mouse and human studies and underlying potential models for disease-causing mechanism for PTPN22 are reviewed in detail in several excellent articles elsewhere (35, 36).

Polymorphism in CSK, which physically interacts with Lyp, is also reported to be associated with the risk of systemic lupus erythematosus (SLE) (37). In combination with Lyp, Csk can modify the activation state in lymphocytes of downstream Src kinases such as Lyn in the human B cells (37). The risk allele of CSK is associated with increased expression and augments inhibitory phosphorylation of Lyn. Carriers of the risk allele exhibit increased BCR-mediated activation of mature B cells, as well as higher concentrations of plasma IgM. Moreover, the fraction of transitional B cells in these patients is increased because of an expansion of late-transitional cells in a stage targeted by selection mechanisms.

Recently, another interesting mechanism of BCR-mediated autoimmunity was suggested by the analysis of the BLK gene, which is associated with rheumatoid arthritis, SLE and other autoimmune conditions (38, 39). BLK encodes B lymphocyte kinase, a member of the Src family of kinases, and a risk haplotype is associated with reduced expression in B cells. Blk phosphorylates tyrosine residues in the immunoreceptor tyrosine-based activation motifs (ITAMs) of the BCR (40). As expected, the phosphorylation of downstream molecules such as PLCγ2 and SHP-2 was shown to be decreased in the carriers harboring the BLK risk haplotype. However, after BCR cross-linking, B cells from these individuals can be hyperactivated, exhibiting enhanced phosphorylation of these molecules, as well as up-regulation of CD86; the latter may relate to enhanced potential for T-cell activation by B cells. In addition, the number of isotype-switched memory B cells was increased in the risk-allele carriers (38). These lines of evidence imply that thresholds for BCR signaling are lowered with a reduced function of BLK, leading to autoimmunity.

The observations in PTPN22, CSK and BLK suggest that the increased susceptibility to AD reflects modulation of antigen-receptor signaling resulting from either enhancement or reduction of the signal at multiple maturation and activation points in lymphocytes. Therefore, careful interpretation is needed to understand the mechanism of T- and B-cell abnormalities caused by these genetic variants.

T_h1 and T_h17 cells

CD4⁺ T cells are critical players in host defense, as well as in autoimmunity. In the context of autoimmune conditions, T_h1 and T_h17 cells, characterized by the production (respectively) of IFN-γ and IL-17, are thought to be the main pathogenic players, although their roles in individual diseases may be different. The importance of genetic factors regulating T_h1 and T_h17 cells in AD is evident from the fact that GWAS SNPs of ADs are highly enriched for CD4⁺ T-cell super-enhancers (41).

Among the cytokine and cytokine-receptor genes that regulate the differentiation of helper T cells, the missense variant (R381Q) in IL23R, encoding the receptor molecule for IL-23, may have been the best studied of AD-associated alleles. This variant has been associated with multiple ADs, including inflammatory bowel disease (42), psoriasis (43) and ankylosing spondylitis (44). Reduced phosphorylation of signal transducer and activator of transcription 3 (STAT3) upon stimulation with IL-23 and decreased production of IL-17 were observed in human CD4⁺ T cells carrying the R381Q protective allele (45). This result suggests that reduced responsiveness in T_h17 cells protects against those ADs. Interestingly, this missense variant is rare in Asian populations, and an independent association signal in the intergenic region of IL23R–IL12RB2 was observed in patients with Behçet’s disease (46, 47).

The IL12RB2 gene, which encodes a subunit of the IL-12 receptor, is essential in the differentiation of T_h1 cells, and thus is another candidate gene in this locus. Although the gene responsible for Behçet’s disease has yet to be determined, the intergenic variant is inferred to have cis-eQTL effects on both genes (48). Moreover, the IL12RB2 locus showed an association signal independent from that of the IL23R locus in primary biliary cirrhosis and systemic sclerosis in European populations (49, 50). Association with the IL12A locus was also reported for these two diseases (49, 51), suggesting that the IL-12-driven T_h1 response may be essential in those ADs.

Another example of differential roles of T_h1 and T_h17 cells in AD is suggested by the associations observed between ADs and the STAT genes STAT3 and STAT4. STAT3, which transduces signals from cytokines such as IL-6 and IL-23, plays an important role in the differentiation of T_h17 cells. The STAT3 locus exhibits associations with multiple ADs, including multiple sclerosis (52), inflammatory bowel diseases (5, 53) and psoriasis (54). Although the disease-causing mechanism for this locus remains unclear, a potential cis-eQTL effect was observed in this locus, with up-regulated gene expression in the risk allele (11, 55).

An association with the STAT4 locus has also been reported for multiple ADs, including rheumatoid arthritis, SLE (56), systemic sclerosis (57) and Sjögren’s syndrome (58). Expression of STAT4 was increased in cells harboring the risk allele, suggesting that these ADs result from a gain of STAT4 activity (59). STAT4 is involved in the differentiation of T_h1 cells, and thus increased expression of STAT4 may enhance T_h1 cell activity. Interestingly, although the STAT4 locus is also associated with inflammatory bowel diseases (5), the eQTL effect of the risk allele is in the opposite direction when compared with that of other ADs (Table 1). Therefore, both up-regulation of STAT3 and down-regulation of STAT4 are involved in the disease pathogenesis, suggesting the role of T-helper subsets may be mutually exclusive. STAT4 also is involved in type I interferon signaling, as will be discussed below.

Regulatory T cells

The role of regulatory T cells (T_reg cells, usually defined as CD4⁺CD25⁺FOXP3⁺ T cells) in murine autoimmunity is well established; in the mouse, T_reg-cell deficiency abrogates self-tolerance and causes AD (60). However, the roles of T_reg cells in complex human ADs remain unknown, as do the identities of genetic factors that influence the activity of T_reg cells. Analyses of the risk loci for rheumatoid arthritis showed that H3K4me3 (a promoter- and enhancer-specific modification associated with active transcription) is particularly enriched in primary T_reg cells compared with the levels in another 33 cell types (61). Another study showed that causal variants of some ADs such as celiac disease, primary biliary cirrhosis and SLE are enriched in acetylated cis-regulatory elements of T_reg cells (14). These observations suggest that a substantial proportion of AD risk variants are involved in transcriptional regulation of genes that affect T_reg-cell function.

The interpretation of these findings is complicated by the fact that the majority of genes implicated in T_reg-cell function is also expressed in other CD4⁺ T-cell subsets. For instance, the CCR6 gene, for which a regulatory variant was associated with multiple ADs (including rheumatoid arthritis), is expressed in both T_h17 cells and T_reg cells (62). Although increased CCR6 expression was observed in cells harboring the risk allele, the cell types in which this variant drives the disease pathology are not known.

A recent eQTL analysis demonstrated that some genes (including CTLA4, IL2RA, IL2RB, FCRL3 and RTKN2) identified via GWASs of ADs were up-regulated in T_reg cells when levels were compared with those in conventional CD4⁺ T cells (63). Among the candidate loci, FCRL3 and RTKN2 are of potential interest, because their functional roles in human T_reg cells are not well established. FCRL3, which exhibits association with rheumatoid arthritis and Graves’ disease (64), was previously shown to have inhibitory potential in B cells (65). More recently, FCRL3 was shown to be expressed in a subset of T_reg cells in humans that have decreased capacity to suppress the proliferation of effector T cells (compared with FCRL3 ^– T_reg cells) (66).

RTKN2, another risk gene for rheumatoid arthritis (67), encodes a Rho-GTPase effector protein. Although the major long transcript is expressed in multiple tissues, recent CAGE (cap analysis gene expression)-seq analysis of human T_reg cells demonstrated that a shorter transcript is selectively expressed in T_reg cells (68). Both of these isoforms have potential to enhance NF-κB signaling (Y. Kochi and K. Myouzen, unpublished data), but the role of the short isoform in human T_reg cells is unknown. This shorter transcript may impair the regulatory function of T_reg cells or, alternatively, the shorter version may affect the plasticity of T_reg cells for development into effector cells such as T_h1-like or T_h17-like cells.

In addition to the classical T_reg cells (defined as CD4⁺CD25⁺FOXP3⁺ T cells), various subsets of T_reg cells have been proposed in both physiological and pathological conditions. Among these subpopulations, the CD4⁺CD25^–LAG3⁺ T-cell subset (designated LAG3⁺ T_reg cells) has regulatory potential to suppress antibody production by B cells in both humans and mice (69, 70). LAG3⁺ T_reg cells also are characterized by the production of large amounts of IL-10 and TGF-β3; the latter factor is important in the regulation of antibody production (69). The transcriptional factor Egr2 is essential in the regulation of LAG3⁺ T_reg cells in the mouse (70). In addition, lymphocyte-specific Egr2 defects lead to a lupus-like autoimmune phenotype with no impact on the development of Foxp3⁺ T_reg cells (71). In humans, significant association signals were observed in the EGR2 locus in the GWASs of rheumatoid arthritis (72) and inflammatory bowel diseases (5). Additionally, regulatory polymorphisms in EGR2 have been reported to be associated with SLE (73). Notably, decreased frequencies of LAG3⁺ T_reg cells were observed in SLE patients (69). These data imply that abnormality in LAG3⁺ T_reg cells is involved in these ADs. Although IL-10 is a hallmark cytokine for LAG3⁺ T_reg cells, an association with the IL10 locus is restricted to SLE, inflammatory bowel diseases and Behçet’s disease, suggesting that a pathological role for LAG3⁺ T_reg cells may be restricted to these autoimmune conditions.

Type I interferons

For ADs involving the adaptive immune system, an immune response to a self-antigen is considered a principal disease mechanism. Nonetheless, multiple lines of evidence from disease models of AD suggest that the innate immune system also plays an important role in such conditions. Among the players of the innate immune system, type I interferons, which are activated by pattern-recognition receptors such as Toll-like receptors and which elicit antiviral and immunomodulatory responses, are implicated in the pathogenesis of some ADs such as SLE and myositis (74). The importance of type I interferon signaling in SLE is evident from the expression profile of peripheral blood cells; patients with SLE exhibit a signature up-regulation of type I interferon genes.

In addition, GWASs performed for SLE have identified multiple loci associated with type I interferon signaling pathways, including TLR7, IRAK1, IRF5, TYK2, IRF7 and IFIH1 (75). Some of these genes also are shared with other ADs. For instance, IRF5 is associated with rheumatoid arthritis, inflammatory bowel diseases and primary biliary cirrhosis; IFIH1 is associated with type 1 diabetes, psoriasis and inflammatory diseases.

STAT4, as mentioned above, is also activated by type I interferons, and STAT4 is essential for the production of IFN-γ in T cells and natural killer cells (76). Higher IFN-α activity was observed in SLE patients with the IRF5 risk haplotype, whereas increased sensitivity to IFN-α signaling was associated with the STAT4 risk allele (77). As observed for IRF5, the risk allele of IRF7 is associated with elevated serum levels of IFN-α in SLE patients (78).

These lines of evidence indicate that the combination of genetic variants determines the level of the type I interferon activity in an individual, which in turn affects susceptibility to disease. Interestingly, the PTPN22 risk variant Lyp620W, which has been associated with SLE, as well as with many other ADs, has been shown to be associated with reduced TLR7-induced type I interferon signaling in SLE patients (79), suggesting that genetic heterogeneity exists in the disease.

Conclusions and future directions

As seen in this review, GWASs have identified many genetic factors for ADs and have contributed to characterization of the pathogenesis of individual diseases. Utilization of these findings in clinical practice is still under way, but holds promise for future success. Notably, genetic etiology (association with IL23R-encoding and IL23A-encoding loci) and response to drug treatment (anti-IL-12/IL-23 and anti-IL-17 therapy) have been linked in the case of psoriasis (80). However, before such clinical applications can be achieved in other ADs, many issues will have to be addressed.

The causal mechanisms suggested by GWASs remain to be dissected. Missing heritability that cannot be identified by GWAS will need to be explored and may reflect the occurrence of rare variants. Innovative statistical approaches using known genetic and environmental factors are needed to better predict an individual’s pathological condition and prognosis for precision medicine. I am optimistic about these applications, as I have, just in the last decade, witnessed scientific progress beyond my imagination.

Conflict of interest statement: The author declared no conflicts of interest.