Association of a functional polymorphism in the IRF5 region with systemic sclerosis in a Japanese population

Abstract

Objective

Interferon regulatory factor 5, an established susceptibility factor for systemic lupus erythematosus (SLE), plays a role in type I interferon and proinflammatory cytokine induction. A recent study showed association of a functional single-nucleotide polymorphism (SNP) in intron 1 of IRF5, rs2004640, with systemic sclerosis (SSc) in a European French population. We undertook the present study to determine whether IRF5 polymorphisms are also associated with a predisposition to SSc in Japanese.

Methods

A case–control association study was performed for rs2004640 as well as for rs10954213 and rs2280714, all of which were previously reported to be associated with SLE, in 281 SSc patients and 477 healthy controls. Patients with SSc complicated by SLE or Sjögren's syndrome were excluded. Association of the rs2280714 genotype with messenger RNA (mRNA) levels of IRF5 and adjacently located transportin 3 (TNPO3) was examined using the GENEVAR database.

Results

All 3 SNPs were significantly associated with SSc, with the rs2280714 A allele having the strongest association (allele frequency P = 0.0012, odds ratio 1.42 [95% confidence interval 1.15–1.75]). Association was preferentially observed in subsets of patients with diffuse cutaneous SSc (dcSSc) and anti–topoisomerase I antibody positivity. Conditional analysis revealed that rs2280714 could account for most of the association of these SNPs, while an additional contribution of rs2004640 was also suggested for dcSSc. The genotype of rs2280714 was strongly associated with IRF5 mRNA expression, while only marginal association was detected with TNPO3 mRNA expression.

Conclusion

Association of IRF5 with SSc was replicated in a Japanese population. Whether the causal SNP is different among populations requires further investigation.

Systemic sclerosis (SSc) is a systemic autoimmune disease characterized by tissue fibrosis of the skin and internal organs. SSc is subdivided into 2 subsets, limited cutaneous SSc (lcSSc) and diffuse cutaneous SSc (dcSSc). Epidemiologic data suggest a role of genetic factors, and association of polymorphisms with susceptibility or clinical characteristics has been reported (1).

A recent study demonstrated an interferon (IFN) signature in peripheral blood cells from SSc patients and supported the role of the type I IFN pathway in the pathogenesis of the disease (2). Polymorphisms of genes involved in IFN signaling are associated with systemic lupus erythematosus (SLE) (3). Among the IFN-related genes, IRF5 has the most well-established association with SLE (4–8). IFN regulatory factor 5 (IRF-5) plays a role in the Toll-like receptor signaling pathway, acting as a master transcription factor in the activation of genes for type I IFN as well as for proinflammatory cytokines (3).

Initially, a common haplotype of IRF5 composed of single-nucleotide polymorphisms (SNPs) rs2004640 (exon-intron border of exon 1B) and rs2280714 (5 kb downstream of the 3′-untranslated region [3′-UTR]) was found to be associated with SLE. The rs2004640 T allele creates a 5′ splice donor site, allowing the expression of IRF5 isoforms containing exon 1B, and the rs2280714 A allele is associated with overexpression of messenger RNA (mRNA) of IRF5 (4). Subsequently, the exon 6 polymorphism (which encodes an insertion/deletion of 10 amino acids) and the exon 9 SNP rs10954213 (which localizes to the 3′-UTR and disrupts the polyadenylation signal) were identified, and the risk haplotype in Caucasians was shown to contain the rs2004640 T allele, the exon 6 insertion, and the rs10954213 A allele (5).

In a previous study, we showed that IRF5 is associated with SLE in Japanese and that the allele frequencies and haplotype structures of IRF5 differed substantially between the Japanese and Caucasian populations, with population-specific SNPs appearing to play a role (7). Recently, association of IRF5 rs2004640 with SSc in a European French population was reported (9). We performed a case–control study to determine whether IRF5 is also associated with SSc in Japanese.

PATIENTS AND METHODS

Patients and controls.

A case–control association study was performed with 281 patients (27 men and 254 women, mean ± SD age 43.0 ± 12.8 years) and 477 controls (228 men and 249 women, mean ± SD age 34.1 ± 11.3 years) recruited at the University of Kanazawa, the University of Tokyo, and the Institute of Rheumatology, Tokyo Women's Medical University. All patients and controls were unrelated Japanese. All patients fulfilled the criteria for SSc proposed by the American College of Rheumatology (formerly, the American Rheumatism Association) (10) and were classified as having either dcSSc (n = 142) or lcSSc (n = 139) according to the classification described by LeRoy and coworkers (11). Anticentromere antibody (ACA) positivity was determined by the presence of a discrete speckled pattern on indirect immunofluorescence using HEp-2 cells and was confirmed by enzyme-linked immunosorbent assay (ELISA) using recombinant human CENP-B (Medical and Biological Laboratories, Nagoya, Japan). Anti–topoisomerase I (anti–topo I) antibody levels were determined using ELISA (Medical and Biological Laboratories), and the specificity was confirmed by immunoprecipitation. Eighty-seven patients were positive for anti–topo I antibodies, and 91 were positive for ACAs.

Because the association of IRF5 with SLE and primary Sjögren's syndrome (SS) had already been reported (3–8), patients with SSc complicated by these conditions were excluded. The sample size of this study provided detection powers of 0.95 and 0.64 for a susceptibility gene with the genotype relative risk of 1.5 and 1.3, respectively, when the risk allele frequency was 0.3 (rs2004640) and detection powers of 0.97 and 0.69 for a susceptibility gene with the genotype relative risk of 1.5 and 1.3, respectively, when the risk allele frequency was 0.5 (rs10954213 and rs2280714) (12).

This study was reviewed and approved by the research ethics committees of the University of Tsukuba, the University of Tokyo, the University of Kanazawa, and Tokyo Women's Medical University. Informed consent was provided by all donors.

Genotyping.

SNPs rs729302 A>C, rs2004640 G>T, rs3807306 C>A, rs10954213 G>A, and rs2280714 A>G were genotyped using the TaqMan genotyping system (ABI 7300; Applied Biosystems, Foster City, CA). Genotyping of rs6953165 C>G, rs2004640 G>T, rs41298401 C>G, rs10954213 G>A, and rs2280714 A>G was performed by sequencing on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems) in a proportion of samples. The intron 1 primers 5′-CACCGCAGACAGGTGGG-3′ (forward) and 5′-GGGAGGCGCTTTGGAAGT-3′ (reverse) were used for rs6953165, rs2004640, and rs41298401; the 3′-UTR primers 5′-CCCTGATTTCCCTGGTTTG-3′ (forward) and 5′-AGCCAGCCAGGTGAGTGTT-3′ (reverse) were used for rs10954213; and 5′-GCTGCAATTGGAAGAAGAGGG-3′ (forward) and 5′-TGATGTGGATTGGAAGTGGA-3′ (reverse) were used for rs2280714. The results from TaqMan genotyping were confirmed to be identical to those from direct sequencing.

Association of the rs2280714 genotype with mRNA expression levels.

Association of the rs2280714 genotype with mRNA expression levels of IRF5 and the adjacently located transportin 3 (TNPO3) was examined using the HapMap database (http://www.hapmap.org/index.html.en) and the GENEVAR database at the Wellcome Trust Sanger Institute (http://www.sanger.ac.uk/humgen/genevar/) as previously described (7). Normalized mRNA data from lymphoblastoid cell lines from 43 Japanese in Tokyo, Japan, and 45 Han Chinese in Beijing, China, in the HapMap database were obtained from the GENEVAR database. Association between the rs2280714 genotype and mRNA levels of IRF5 and TNPO3 was analyzed using simple regression analysis.

Statistical analysis.

Association analyses were conducted by chi-square test. Because this study aimed to test the specific hypothesis of whether association of IRF5 with SSc was replicated in Japanese, correction for multiple comparisons was not applied. Conditional logistic regression analysis was conducted to examine the effect of each SNP on the susceptibility to SSc after controlling for the genotype of another SNP.

Calculation of linkage disequilibrium (LD) parameters (D′ and r²) from the genotypes of 477 healthy controls was performed using Haploview version 4.0 (Broad Institute, Cambridge, MA) (http://www.broad.mit.edu/mpg/haploview/index.php). Using Haploview version 4.0, a permutation test (1 million permutations) was performed to obtain experiment-wide P values of the single SNPs (rs2004640, rs10954213, and rs2280714) and of haplotypes consisting of them.

RESULTS

To determine whether the association of IRF5 polymorphisms with SSc could be replicated in a Japanese population, we genotyped 3 SNPs: rs2004640, rs10954213, and rs2280714. The intron 1 SNP rs2004640 has been shown to be associated with SSc in a European French population (9), SNP rs10954213 directly affects the IRF5 mRNA level by altering the polyadenylation signal, and SNP rs2280714, located 5 kb downstream of the 3′-UTR of IRF5, has also been shown to be associated with IRF5 mRNA expression (3–5). SNP rs10954213 was in absolute LD with the previously reported susceptibility SNP rs11770589 (r² = 1, D′ = 1) and also efficiently tagged the exon 6 insertion/deletion in Japanese (r² = 0.98, D′ = 0.99) (7). Deviation from Hardy-Weinberg equilibrium was not observed for any of the SNPs in the control samples.

The genotyping data are shown in Table 1. The association of rs2004640 with SSc was replicated in our Japanese patients. Similar to the European data, the association was compatible with the recessive model and was observed in dcSSc but not in lcSSc.

Table 1. Association of IRF5 region SNPs with SSc in a Japanese population*

SNP, subject subset	No. (%) with genotype†				Risk allele frequency‡	Allelic association§		Dominant association§		Recessive association§
	1/1	1/2	2/2	Total		OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
* SSc = systemic sclerosis; dcSSc = diffuse cutaneous SSc; lcSSc = limited cutaneous SSc; anti–topo I = anti–topoisomerase I; ACA = anticentromere antibody. † For the single-nucleotide polymorphism (SNP) rs2004640, 1 = the G allele and 2 = the T allele; for SNPs rs10954213 and rs2280714, 1 = the A allele and 2 = the G allele. ‡ For SNP rs2004640, T = the risk allele; for SNPs rs10954213 and rs2280714, A = the risk allele. § Odds ratios (ORs), 95% confidence intervals (95% CIs), and P values were calculated by chi-square test using 2 × 2 contingency tables.
rs2004640
SSc	123 (44)	117 (42)	41 (14)	281	0.35	1.27 (1.02–1.59)	0.032	1.23 (0.91–1.65)	0.18	1.72 (1.10–2.24)	0.018
dcSSc	61 (43)	54 (38)	27 (19)	142	0.38	1.43 (1.08–1.88)	0.012	1.27 (0.87–1.85)	0.22	2.37 (1.42–3.96)	0.00096
lcSSc	62 (45)	63 (45)	14 (10)	139	0.33	1.13 (0.85–1.51)	0.40	1.19 (0.81–1.73)	0.38	1.13 (0.60–2.13)	0.71
Anti–topo I positive	33 (38)	38 (44)	16 (18)	87	0.40	1.56 (1.12–2.18)	0.0081	1.56 (0.98–2.49)	0.061	2.27 (1.23–4.20)	0.0086
ACA positive	44 (48)	37 (41)	10 (11)	91	0.31	1.06 (0.75–1.49)	0.74	1.02 (0.65–1.60)	0.93	1.25 (0.60–2.58)	0.55
Control	233 (49)	201 (42)	43 (9)	477	0.30	–	–	–	–	–	–
rs10954213
SSc	80 (28)	131 (47)	70 (25)	281	0.52	1.30 (1.05–1.60)	0.014	1.41 (1.01–1.96)	0.042	1.38 (0.98–1.93)	0.063
dcSSc	46 (32)	60 (42)	36 (25)	142	0.54	1.39 (1.07–1.81)	0.015	1.38 (0.90–2.10)	0.14	1.66 (1.10–2.50)	0.016
lcSSc	34 (24)	71 (51)	34 (24)	139	0.50	1.21 (0.92–1.58)	0.17	1.44 (0.94–2.22)	0.094	1.12 (0.72–1.74)	0.62
Anti–topo I positive	27 (31)	35 (40)	25 (29)	87	0.51	1.27 (0.92–1.75)	0.15	1.16 (0.70–1.92)	0.56	1.56 (0.94–2.57)	0.083
ACA positive	25 (27)	43 (47)	23 (25)	91	0.51	1.26 (0.92–1.73)	0.15	1.38 (0.83–2.30)	0.21	1.31 (0.79–2.18)	0.30
Control	107 (22)	218 (46)	152 (32)	477	0.45	–	–	–	–	–	–
rs2280714
SSc	99 (35)	136 (48)	46 (16)	281	0.59	1.42 (1.15–1.75)	0.0012	1.72 (1.18–2.50)	0.0047	1.48 (1.08–2.04)	0.015
dcSSc	59 (42)	63 (44)	20 (14)	142	0.64	1.70 (1.30–2.23)	0.00013	2.05 (1.23–3.41)	0.0056	1.94 (1.32–2.85)	0.00080
lcSSc	40 (29)	73 (53)	26 (19)	139	0.55	1.18 (0.91–1.55)	0.22	1.46 (0.91–2.34)	0.12	1.10 (0.72–1.68)	0.65
Anti–topo I positive	34 (39)	41 (47)	12 (14)	87	0.63	1.62 (1.17–2.26)	0.0041	2.10 (1.12–3.95)	0.021	1.75 (1.09–2.81)	0.020
ACA positive	28 (31)	44 (49)	18 (20)	91	0.56	1.21 (0.88–1.66)	0.25	1.34 (0.77–2.34)	0.30	1.23 (0.75–2.01)	0.40
Control	128 (27)	229 (48)	120 (25)	477	0.51	–	–	–	–	–	–

Interestingly, the other 2 SNPs, especially rs2280714, showed a stronger association with SSc. With respect to the autoantibody profile, rs2004640 and rs2280714 exhibited a significant association in the anti–topo I antibody–positive subset of patients, but none of the SNPs showed an association in the ACA-positive subset of patients.

To exclude the possibility that other SNPs might exhibit stronger associations, we additionally genotyped SNPs that have previously been implicated in autoimmune diseases, rs729302 A>C, rs6953165 C>G, rs41298401 C>G, and rs3807306 C>A, in 106 patients and 290 controls. None showed significant association (data not shown).

Because LD was present among SNPs rs2004640, rs10954213, and rs2280714, the contribution of each SNP was examined using conditional logistic regression analysis. In the group of all patients with SSc, conditioning by rs2280714 eliminated the significant association of other SNPs, while the association of rs2280714 remained significant after conditioning by rs10954213 or rs2004640 (Table 2). These results indicated that rs2280714 accounts for the genetic effect of the IRF5 region in SSc. When the same analysis was used for the subgroup of patients with dcSSc, the results were essentially the same, except that the association of rs2004640 remained marginally significant after conditioning by rs2280714 (Table 2).

Table 2. Logistic regression analysis of the 3 SNPs examined*

Group, SNP	Model	r^2†		P‡	P§
		With rs2004640	With rs2280714		Adjusted for rs2280714	Adjusted for rs10954213	Adjusted for rs2004640
* NA = not applicable (see Table 1 for other definitions). † Pairwise linkage disequilibrium as measured by r^2 in the controls. ‡ P values for each SNP under the recessive, additive, or dominant model that provided the best fit by logistic regression analysis. § P values adjusted for each SNP under the indicated model.
All SSc patients
rs2004640	Additive	NA	0.26	0.035	0.58	0.18	NA
rs10954213	Additive	0.14	0.77	0.018	0.34	NA	0.089
rs2280714	Additive	0.26	NA	0.0014	NA	0.018	0.014
All dcSSc patients
rs2004640	Recessive	NA	0.26	0.00093	0.033	0.0058	NA
rs10954213	Additive	0.14	0.77	0.020	0.051	NA	0.15
rs2280714	Additive	0.26	NA	0.00018	NA	0.00043	0.0058

A permutation test demonstrated that the most significant association was observed when the rs2280714 A allele was used as a single marker (permutated P = 0.0076 for total SSc, permutated P = 0.0016 for dcSSc) rather than when the haplotype containing each susceptibility allele of the 3 examined SNPs was considered (permutated P = 0.097 for total SSc, permutated P = 0.21 for dcSSc). This further supported the significance of rs2280714 in the Japanese population.

SNP rs2280714 is located in the intergenic region between IRF5 and TNPO3. To gain insight into the functional significance of rs2280714, we investigated whether this SNP was associated with the expression of IRF5 and TNPO3 using the mRNA expression profiling data of the B cell lines from HapMap subjects. As shown in Figure 1, the IRF5 mRNA level was strongly correlated with the number of A alleles (r² = 0.447, P = 1.02 × 10⁻¹², slope = 0.506), while the rs2280714 genotype was only marginally associated with TNPO3 mRNA (r² = 0.056, P = 0.025, slope = −0.071).

Figure 1. Association of the rs2280714 genotype with levels of mRNA for IRF5 and TNPO3. The rs2280714 genotype was strongly associated with IRF5 mRNA, but only marginally associated with TNPO3 mRNA. Simple regression analyses were performed based on the mRNA expression profiling data in the B cell lines from HapMap subjects (obtained from the GENEVAR database).

DISCUSSION

In this study, we replicated the association of IRF5 with SSc in a Japanese population. In addition to the originally reported rs2004640 T allele (9), we also detected an association of the rs2280714 A allele and the rs10954213 A allele with SSc. SNP rs2280714 appeared to account for most of the genetic effects of this region, but an additional effect was suggested for the rs2004640 T allele in the subset of patients with dcSSc. Similar to findings in the European study, the association of IRF5 was observed in dcSSc but not in lcSSc. Furthermore, the association was observed in the anti–topo I antibody–positive subset of patients, but not in the ACA–positive subset of patients. The risk allele of rs2280714 was associated with a higher IRF5 mRNA level in an allele number–dependent manner. Taken together with its functional relevance and association with susceptibility to SLE, rheumatoid arthritis, SS, and inflammatory bowel disease in multiple populations (3–8), IRF5 is postulated to be a common genetic factor for autoimmunity.

An “IFN signature” has been shown in the expression profiling of SSc (2). Our findings not only support the role of type I IFN in the pathogenesis of SSc, but also suggest that IRF-5 overexpression may play a causal role in a proportion of patients. Of interest, treatment with type I IFN has been suggested to trigger the development of SSc in patients with multiple sclerosis (13). If such a hypothesis is proven, type I IFN should be considered as a potential therapeutic target for SSc.

Unexpectedly, the primary role was detected for rs2280714 (located 5 kb downstream of the 3′-UTR of IRF5) rather than for rs2004640 (which creates the intron 1 splice donor site and is associated with alternative splicing of exon 1B) or for rs10954213 (which directly disrupts the polyadenylation signal of IRF5). Because investigators in the European study only reported the data for rs2004640 (9) and because the haplotype structure of the IRF5 region differs considerably between Caucasians and Japanese (7), it remains unclear at this point whether rs2280714 also shows the strongest association in Caucasians. Interestingly, a previous study demonstrated that the association of rs2280714 remained significant after conditioning by rs10954213, suggesting that the effect of rs2280714 cannot be fully explained by LD with rs10954213 (5).

SNP rs2280714 is located only 223 bp downstream of the 3′-UTR of TNPO3. TNPO3 imports multiple proteins into the nucleus, including serine/arginine-rich protein (one of the substrates of topo I), which regulates the splicing of mRNA (14). The LD block encompassing rs2280714 contains the entire TNPO3 gene; thus, it is theoretically possible that the molecular mechanism of association with SSc may involve TNPO3. However, although the TNPO3 mRNA level was marginally correlated with the rs2280714 genotype, the slope of the regression line and the r² value were much lower than those for the IRF5 mRNA level. Thus, it is reasonable to assume that the molecular mechanism is largely mediated by up-regulation of IRF5 mRNA.

The molecular mechanism by which IRF5 confers risk of SSc requires further study. Investigators in a recent study reported that anti–topo I–containing sera induced IFNα from normal peripheral blood mononuclear cells, implicating a role of IFNα in vascular damage and lung fibrosis (15). Our observations support this hypothesis, because overexpression of IRF-5 is supposed to promote such a pathway. The additional effect of rs2004640 in dcSSc might suggest a role of exon 1B–containing isoforms in the fibrotic phenotype. Such a hypothesis needs to be tested in the future.

Although the healthy controls were younger than the patients, this difference should not affect the results, because the risk that any of the controls would develop SSc later in life is extremely low due to the rarity of the disease. Furthermore, if such potential misclassification of future patients in the control group is taken into account, our current analysis should be interpreted as a conservative one.

In conclusion, our findings replicated the association of IRF5 with SSc in a Japanese population, which supported a role of type I IFN in pathogenesis of SSc. Whether the causal SNP is different among populations requires further investigation.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Tsuchiya had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ito, Tsuchiya.

Acquisition of data. Ito, Kawaguchi, Kawasaki, Hasegawa, Kawamoto, Fujimoto, Takehara, Sato, Hara.

Analysis and interpretation of data. Ito, Kawasaki, Ohashi, Hikami, Tsuchiya.