Rheumatoid arthritis (RA) and systemic lupus erythaematosus (SLE) are autoimmune/inflammatory disorders with environmental and genetic components. Familial and linkage studies have provided strong evidence of multiple genetic factors for RA and SLE.1 2 These studies have also shown familial aggregation of RA, SLE and other autoimmune diseases, suggesting the presence of common genetic factors that confer predisposition to these diseases. In fact, variation in the human leukocyte antigen gene (HLA) region is a crucial determinant of susceptibility to many autoimmune diseases including RA and SLE.3^–6 Of the non-HLA genes, PTPN22 and STAT4 have been repeatedly associated with predisposition to RA and SLE.7^–9

Recent studies have demonstrated the interferon regulatory factor 5 gene (IRF5), a member of the IRF family of transcription factors, is an important genetic risk factor for SLE in Caucasian and Asian subjects.10^–17 A common haplotype composed of a single nucleotide polymorphism (SNP) at the intron–exon border of exon 1B (rs2004640) and an SNP in the 3′ flanking region (rs2280714) is associated with genetic predisposition to SLE.11 Both polymorphisms show functional relevance. The rs2004640 T allele creates a 5′ donor splice site in an alternative exon 1 of IRF5, allowing expression of several unique IRF5 isoforms, while the rs2280714 T allele has been identified as a cis-acting determinant of increased baseline expression level of IRF5.11 18 19 Subsequent studies have identified other polymorphisms in IRF5: a CGGGG insertion–deletion polymorphism (four or three repeats of CGGGG) (in the promoter region), rs3807306 (in intron 1), rs2070197(in the 3′ untranslated region (UTR)), rs10954213 (in the 3′UTR), rs10488631 (in the 3′ flanking region) and a 30-base pair (bp) in-frame insertion–deletion variant of exon 6, as polymorphisms that could be implicated in susceptibility to SLE.13 15^–17 20 Above all, recent studies have indicated the CGGGG indel as a strong candidate for causing SLE as well as inflammatory bowel diseases (IBDs).20 21 This indel has not only increased disease risk but also displayed disease-causing potential; the longer allele of CGGGG repeats contains an additional binding site of the transcription factor SP1 and affects IRF5 gene expression.

Previous studies using RA cohorts of Caucasian subjects showed that IRF5 functional polymorphisms (rs2004640 and rs2280714) did not increase genetic susceptibility to RA, unlike SLE.22 23 However, a recent association study in Caucasian subjects has indicated that a haplotype in the promoter region of IRF5 is significantly associated with susceptibility to RA in patients negative for anti-cyclic citrullinated peptide antibody (anti-CCP antibody).24 Until recently, the association between RA susceptibility and IRF5 polymorphisms remained to be clarified in Asian populations.

In the present work we investigated whether an association exists between RA susceptibility and IRF5 variants, using two independent case–control cohorts of Japanese subjects.

PATIENTS AND METHODS

Subjects

The study subjects comprised 2 sets of Japanese patients with RA and unrelated controls, with the first set comprising 830 patients and 658 unrelated controls, and the second set comprising 1112 patients and 940 unrelated controls. The first set of subjects was the same as used in our previous association studies of RA, PADI4, SLC22A4 and FCRL3.25^–27 To validate the case–control association test, we evaluated the impact of population stratification on the case–control study (the first set), and detected no significant evidence of population stratification. Details have been described previously.27 All cases met the revised American College of Rheumatology (ACR) criteria for RA.28 Our subjects were all Japanese and were recruited via several medical institutes in Japan. All subjects provided informed consent to participate in the study, as approved by the ethical committee of the SNP Research Center, RIKEN.

Genotyping

To cover the IRF5 gene and neighbouring sequences (chromosome 7 position; 128 347 997–128 384 962 bp), we identified 10 tag SNPs that captured the majority of the common SNP variation (ie, that present on more than 5% of chromosomes), using an r² threshold of 0.8 or greater. For this selection, we analysed the HapMap Phase II data of Japanese and Han Chinese using Haploview software, V. 4.0 (http://www.broad.mit.edu/haploview/haploview). We added another three SNPs (rs3757385, rs2004640 and rs10488631) and the CGGGG indel polymorphism for comparison to previous association studies in RA and SLE.11 12 16 20 22^–24 We ultimately selected 13 SNPs and the CGGGG indel for genotyping (fig 1A). We genotyped SNPs using TaqMan assays in accordance with the manufacturer’s instructions. For all SNPs but the rs2004640, predesigned TaqMan SNP genotyping assays were used (Applied Biosystems, Foster City, California, USA) (see Supplementary material). For the rs2004640 polymorphism, custom-made primers and probes were provided by Applied Biosystems. Fluorescence was detected using an ABI Prism 7900 HT Sequence Detection System (Applied Biosystems). The CGGGG indel was genotyped as previously described; after the indel sequences were amplified with a fluorescent PCR primer, the amplified fragments were analysed using an ABI 3700 sequencer (Applied Biosystems) (see Supplementary material).21 The fragment analysis was performed using the GeneMapper v.4.0 software (Applied Biosystems). Genotyping calls were made on >98% of samples, for all polymorphisms genotyped. All the SNPs and the CGGGG indel were in Hardy–Weinberg equilibrium in control subjects according to χ² statistics (p>0.01).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Interferon regulatory factor 5 gene (IRF5) structure and pairwise linkage disequilibrium (LD) patterns for polymorphisms. (A) Exons and polymorphisms involved in the present study. The exons are represented by numbered black boxes. The untranslated alternative 5′-exons are represented by grey boxes (exon 1A–C) and the 3′-untranslated region (UTR) is shown as a white box. (B) Pairwise LD patterns for the 13 polymorphisms studied (except for rs10488631). The diagram shows pairwise LD values as quantified using r². Stronger LD is depicted graphically by denser shading. The haplotype block is outlined by a solid line.

Genotyping of HLA-DRB1

Genotyping of HLA-DRB1 was performed for 830 patients with RA (subjects in the first set). HLA-DRB1 alleles were genotyped using the HLA-DRB BigDye Terminator Sequencing typing kit (Applied Biosystems) according to the manufacturer’s instructions. Details have been described previously.4 The shared epitope (SE) alleles were HLA-DRB1*0101, *0401, *0404, *0405,*0410, *1001, *1402 and *1406.

Statistical analysis

We used χ² contingency table tests to evaluate the significance of differences in genotype and allele frequencies in the case–control subjects. The genotype and allele frequencies for cases and controls were used to calculate the odds ratio (OR) and the 95% CI using the Woolf method. For the combined analysis, we used the method of Mantel–Haenszel. We used the Bonferroni method for a multiple testing correction in the combined analysis. We calculated the statistical power of association using R software (http://www.r-project.org). We performed conditional logistic regression analysis to evaluate the effect of each polymorphism conditional on the remaining, using the WHAP software program (http://pngu.mgh.harvard.edu/purcell//whap). We estimated haplotype frequencies and calculated pairwise linkage disequilibrium (LD) indices between pairs of polymorphisms (r²) using Haploview V. 4.0. We defined haplotype blocks using the ‘solid spine of linkage disequilibrium’ definition available in the Haploview program. Permutation tests were also performed for the case–control haplotype analyses using Haploview v4.0 (with 100 000 iterations).

RESULTS

Association analyses of individual polymorphisms in IRF5

We first performed allele frequency comparisons between patients with RA and control subjects. All polymorphisms genotyped were polymorphic in our population except for rs10488631 in which the disease-risk allele was rarely observed (frequency<0.01). A significant association of the rs729302 A allele with RA susceptibility was replicated in both sets (first set, OR 1.21, 95% CI 1.03 to 1.43, p = 0.021; second set, OR 1.22, 95% CI 1.06 to 1.40, p = 0.005) (table 1). The combined analysis also revealed a statistically significant association of this allele with RA susceptibility (OR 1.22, 95% CI 1.09 to 1.35, p<0.001), and retained the statistical significance even when corrected for multiple hypothesis testing (corrected p = 0.004, by the Bonferroni method). No genotyped polymorphisms were significantly associated with RA susceptibility in both sets (including rs3757385 and rs3807306, which are implicated in genetic predisposition to RA in the study of RA in Caucasians), but rs729302 was significantly associated.24 In our analysis of RA in Japanese patients, no statistically significant difference between cases and controls was observed for allele frequency of rs2004640, of which the T allele confers genetic susceptibility to SLE.10^–17 The statistical power to detect association in rs2004640 exceeded 0.99 at a significance level of α = 0.05, if we presumed the OR of the allele for RA was at similar level with that for SLE previously reported (OR 1.47, 95% CI 1.36 to 1.60 in Caucasians or OR 1.32, 95% CI 1.14 to 1.54 in Koreans).11 12 If the presumed OR of rs2004640 for RA was as low as 1.13, the power still remained 0.80. We found moderate associations of rs6968225, the CGGGG indel and rs2280714 with RA susceptibility in the second set (rs6968225: OR 1.16, 95% CI 1.02 to 1.31, p = 0.024; CGGGG indel: OR 1.21, 95% CI 1.00 to 1.46, p = 0.044; rs2280714: OR 1.19, 95% CI 1.05 to 1.35, p = 0.006), and weak associations of rs4728142, the CGGGG indel, rs3807306, rs13242262 and rs2280714 in combined analyses (rs4728142: OR 1.16, 95% CI 1.01 to 1.35, p = 0.042; CGGGG indel: OR 1.20, 95% CI 1.04 to 1.39, p = 0.015; rs3807306: OR 1.13, 95% CI 1.01 to 1.26, p = 0.031; rs13242262: OR 1.12, 95% CI 1.02 to 1.23, p = 0.019; rs2280714: OR 1.12, 95% CI 1.02 to 1.23, p = 0.023). To distinguish between independent effects of these polymorphisms and those due to LDs, we also analysed the effects of each polymorphism conditional on the most strongly-associated SNP, rs729302, using conditional logistic regression analysis. This analysis showed that none of these polymorphisms retained significant association when evaluated conditionally on rs729302 (rs6968225, p = 0.97; rs47218142, p = 0.21; CGGGG indel, p = 0.14; rs3807306, p = 0.46; rs13242262, p = 0.52; rs2280714, p = 0.33 in pooled samples). This result implies that genetic associations observed in these polymorphisms would be secondary owing to modestly strong LD with rs729302 (fig 1B).

In order to investigate the genetic mode of effect of IRF5 on RA susceptibility, we studied the genotype frequencies for rs729302 (see Supplementary material). We found that the OR for the AA+AC vs CC genotype (OR 1.67, 95% CI 1.29 to 2.17, p<0.001 in the combined analysis) was higher than that for the AA vs AC+CC genotypes (OR 1.18, 95% CI 1.03 to 1.36, p = 0.014 in the combined analysis), suggesting an autosomal dominant pattern of inheritance. This mode of association was replicated in both sets.

Association analysis of IRF5 haplotypes

We analysed the haplotype structure of the polymorphisms genotyped in IRF5, and compared their frequencies between patients and controls to detect any combinational effects of these polymorphisms on RA susceptibility (table 2). A haplotype block composed of all polymorphisms but rs960633 was defined (fig 1B). Then we estimated haplotype frequencies in the block. Among the seven haplotypes with a frequency over 2% in control subjects, the frequency of haplotype 2 was significantly lower in cases compared with controls (OR 0.73, 95% CI 0.64 to 0.84, p<0.001), which indicates that this haplotype plays a protective role against disease susceptibility. Next, we focused on rs729302, the CGGGG indel, rs2004600 and rs2280714 in the haplotype block, considering their functional relevance and significant association with disease susceptibility: rs729302 was significantly associated with RA susceptibility in this study; the CGGGG insertion has been shown to create an additional binding site for SP1, which can lead to increased expression of IRF5, and was significantly associated with SLE as well as IBDs; rs2004640 is relevant to alternative splicing and is significantly associated with SLE susceptibility;10–17 29 rs2280714 has been identified as a cis-acting determinant of IRF5 expression.18 19 In an analysis of four-marker haplotypes of these polymorphisms, we observed a significant association of haplotype 8 with protection against RA (OR 0.77, 95% CI 0.69 to 0.86, p<0.001), in which the protective C allele of rs729302 seemed to have a major role. No obvious combinational effect on RA susceptibility was detected in this four-marker analysis.

Analysis of the association between IRF5 variants and susceptibility to RA, stratified by the SE

We tested for an epistatic interaction between IRF5 and HLA-DRB1. We stratified 830 patients with RA according to the presence of the SE, and examined for the association between rs729302 and disease susceptibility (table 3). When stratified according to SE, the rs729302 A allele conferred increased risk for RA in patients that were SE negative relative to patients carrying the SE (SE negative: OR 1.50, 95% CI 1.17 to 1.92, p = 0.001; SE positive: OR 1.11, 95% CI 0.93 to 1.33, p = 0.24).

DISCUSSION

IRF5 is a member of the IRF family of transcription factors and is involved downstream of the TLR-MyD88 signalling pathway.30 IRF5 induces the production of not only type I interferons (IFNs) but also proinflammatory cytokines, such as interleukin (IL)6, IL12 and tumour necrosis factor (TNF)α, which play important roles in the pathology of RA. Therefore, IRF5 is a potent candidate gene that could indicate a genetic predisposition to RA.

In the present study, rs729302 in the IRF5 promoter region was significantly associated with RA susceptibility in two independent sets of samples, and the association was still significant when corrected for multiple testing. Previous studies in Caucasian patients with SLE have also indicated the contribution of rs729302 to disease susceptibility,11 16 although the functional relevance of rs729302 remains to be clarified. Because rs729302 in itself may not affect the baseline expression level of IRF5,15 this polymorphism might influence the regulation of IRF5 gene expression not at baseline, but under conditions of cell activation. In addition, there remains the possibility that some unknown polymorphisms close to rs729302 are affecting IRF5 gene expression, and that the genetic association observed for rs729302 is secondary owing to strong LD.

The CGGGG indel in the IRF5 promoter region had only moderately significant association with RA predisposition in our study. Considering its genetic and molecular potential, however, this polymorphism would be one of the candidates for causing RA; the longer allele of CGGGG repeats has not only conferred increased risk of SLE and IBD, but also increased expression of IRF5 through creating an additional SP1 binding site.20 21 Together with the genetic association observed in rs729302, our results show that polymorphisms in the IRF5 promoter region would contribute to susceptibility to RA and imply that altered expression of IRF5 may raise the disease risk. Further analyses will be needed to clarify the precise role of the promoter polymorphisms in IRF5.

Although rs2004640 is convincingly associated with susceptibility to SLE, whether this polymorphism confers risk for RA has been controversial.22^–24 Our result is consistent with the results found in two previous studies of RA in Caucasian populations. The absence of association in rs2004640 in the present study was less likely due to a statistical type II error, as described above. This difference between RA and SLE may be due to the different aetiological roles of IRF5 splicing variants in these diseases, whose expression is regulated by rs2004640. Multiple spliced isoforms of IRF5 reportedly differ with respect to regulation of expression, subcellular localisation and function.29 This implies that diversity in the function and expression of IRF5 splicing variants in various cell types (eg, lymphocytes, macrophages and dendritic cells) may change the pattern of cytokine expression and determine the characteristic pathophysiology of RA and SLE.

Previous studies have indicated that the presence of the SE-encoding alleles increased the risk of progression to severe arthritis.31 32 The SE-encoding alleles are associated with production of anti-CCP antibody, while a non-SE allele, HLA-DRB1*03, is significantly associated with an absence of anti-CCP antibody in Caucasian patients with RA.33^–35 In Caucasians, IRF5 variants are also significantly associated with disease susceptibility in patients that are anti-CCP antibody negative.24 Given the correlation between the SE and anti-CCP antibody status, our result that the IRF5 polymorphism conferred increased disease risk in patients that were SE negative is similar to the findings in Caucasian patients. In summary, the IRF5 polymorphism indeed has a major effect on predisposition to RA in patients that are anti-CCP antibody negative, but less effect in patients that are anti-CCP antibody positive.

Findings from clinical and animal studies have demonstrated that T cell-mediated antigen-specific responses and cytokine networks are the two major branches of disease pathogenesis.36^–38 The presence of anti-CCP antibodies in patients that were SE positive indicates T cell response driven by citrullinated peptides. In patients with RA that were SE negative, most of whom lack anti-CCP antibodies in their sera, the T cell response to citrullinated autoantigens is not prominent, or is even absent. This means that the proinflammatory cytokine cascade in these patients is not mainly triggered by T cell response to citrullinated proteins, but by other unknown factors. TNFα polymorphisms in the conserved A1;B8;DRB1*03 haplotype (known as the ancestral 8.1 haplotype) are thought to enhance the expression of TNFα and thus may be a candidate for triggering the proinflammatory cytokine cascade.35 39 Considering the crucial role of IRF5 in regulating the expression of proinflammatory cytokines, the IRF5 polymorphism would be another genetic factor in patients that were SE negative that would enhance cytokine production and disease progression.

In conclusion, the present study indicates that the polymorphism in the IRF5 promoter region is a genetic factor that confers predisposition to RA. There is much to learn about the role of IRF5 in immunity and autoimmune disorders. Our future interest is in elucidating the regulation of IRF5 expression and the precise function of IRF5 variants in the immune system, particularly their response to type I IFNs and influence on induction of type I IFNs and proinflammatory cytokines, which will determine the complex aetiology of autoimmune diseases.