Systemic lupus erythematosus (SLE) and systemic sclerosis (SSc) share some pathophysiologic bases as evidenced by individual and familial polyautoimmunity and common susceptibility genetic factors. With regard to the latter, there has been a recent shift from the “common variant” to the “rare variant” paradigm, since rare variants of TNFAIP3 and TREX1 with large effect sizes have recently been discovered in SLE. The present study was undertaken to investigate whether rare variants of TNFAIP3 and TREX1 are also associated with SSc.
TREX1 single-nucleotide polymorphisms (SNPs) rs3135946, rs7626978, rs3135943, and rs11797 and TNFAIP3 SNPs rs9494883, rs72063345, rs5029939, rs2230926, rs117480515, and rs7749323 were genotyped in a discovery set (985 SSc patients and 1,011 controls), and replication analysis of the most relevant results was performed in a second set (622 SSc patients and 493 controls).
No association between TREX1 variants and SSc was observed. For TNFAIP3, we first demonstrated that a low-frequency variant, rs117480515, tagged the recently identified TT>A SLE dinucleotide. In the discovery sample, we observed that all tested TNFAIP3 variants were in linkage disequilibrium and were associated with SSc and various SSc subsets, including the polyautoimmune phenotype. We subsequently genotyped rs117480515 in the replication sample and found it to be associated solely with the SSc polyautoimmune subset (odds ratio 3.51 [95% confidence interval 2.28–5.41], P = 8.58 × 10−9) in the combined populations. Genotype–messenger RNA (mRNA) expression correlation analysis revealed that the TNFAIP3 rs117480515 risk allele was associated with decreased mRNA expression.
The present findings establish the TNFAIP3 locus as a susceptibility factor for the subset of SSc with a polyautoimmune phenotype. Our results support the implication of rare/low-frequency functional variants and the critical role of A20 in autoimmunity.
Evidence of a shared pathogenic basis across connective tissue diseases (CTDs) is provided by familial and individual aggregation of autoimmune diseases and by the well-known pleiotropism of autoimmune susceptibility genes (1, 2). In recent years, a number of novel and unbiased genetic associations have been revealed, notably through genome-wide association studies (GWAS) (3). The underlying rationale for GWAS is the “common disease, common variant” hypothesis, positing that common diseases, such as autoimmune diseases, are attributable in large part to allelic variants (single-nucleotide polymorphisms [SNPs]) present in >5% of the population. However, the attributable risk conferred by the susceptibility alleles identified to date explains only 15–20% of heritability (4). There is now evidence that a number of low-frequency (i.e., 1–5%) and rare (<1%) variants that have larger effect sizes and more functional consequences with regard to phenotypic effects than do common variations are involved in the genetic susceptibility to complex traits, thus resulting in a shift from the “common variant” hypothesis to the “rare variant” paradigm (5). In systemic lupus erythematosus (SLE), rare alleles with a major effect on disease susceptibility have recently been discovered (6). Among these, mutations in TREX1 and low-frequency variants in TNFAIP3 have been highlighted (7–10).
TREX1 (3p21.31), the gene for 3′ repair exonuclease 1 (Trex1), encodes the major 3′→5′ DNA exonuclease involved in a proofreading function during genome replication. There is evidence that Trex1 might be a negative regulator of the interferon-stimulatory DNA response and that it could minimize immune activation by self DNA and prevent cell-intrinsic initiation of autoimmune processes. TNFAIP3 (6q23.3), the tumor necrosis factor α–inducible protein 3 gene, encodes A20, a zinc-finger protein required for inhibition of the NF-κB signaling axis. Recently, several polymorphisms in the TNFAIP3 locus have been found to be associated with increased susceptibility to a number of autoimmune diseases, i.e., SLE, rheumatoid arthritis, Crohn's disease, psoriasis, multiple sclerosis, type 1 diabetes, and systemic sclerosis (SSc) (9, 11, 12).
SSc is a rare multisystem autoimmune disease classified as a CTD. Numerous genetic factors underlying susceptibility to SSc have been identified, the vast majority belonging to autoimmune pathways. Among these genetic factors, the TNFAIP3 SNP rs5029939, located in the second intron, was recently identified by our group as being strongly associated with SSc, with the strongest effect size reported to date in SSc outside the HLA locus (odds ratio [OR] >2) (12). However, the TNFAIP3 causal variant still remains to be identified, and functional data are needed to clarify its functional consequences. A recent genetic study in SLE identified a functional TT>A dinucleotide variant, located 41.5 kb downstream of the TNFAIP3 promoter, that could be the causal variant (10).
Given the evidence of a shared genetic background between SSc and SLE and the emerging rare variants paradigm, we investigated the association of low-frequency and rare variants of the TREX1 and TNFAIP3 gene regions, identified as susceptibility factors for SLE, with SSc and disease subsets. In addition, we performed genotype–messenger RNA (mRNA) expression correlation analyses to assess their putative functional implications.
PATIENTS AND METHODS
We performed a case–control association study, using a French Caucasian population consisting of 841 SSc patients and 997 healthy unrelated ethnically matched controls as previously described (13) for the genotyping of TREX1 variants, and an extended French cohort consisting of 1,043 SSc patients and 1,095 matched controls for the genotyping of TNFAIP3 variants. The replication set was an Italian cohort comprising 668 SSc patients and 501 controls. We included only individuals of European ancestry (defined as all 4 grandparents being of European Caucasian ancestry). Characteristics of the SSc patients in both cohorts are shown in Table 1. In addition, both French and Italian SSc patients were investigated by local physicians for the co-occurrence of autoimmune diseases known to be associated with CTDs, which were identified in 268 of them (15.7%). These autoimmune diseases were SLE, rheumatoid arthritis, Sjögren's syndrome, autoimmune thyroid disease, and primary biliary cirrhosis, and were classified in accordance with standard international criteria. All patients were screened for antinuclear antibodies and their putative antibody specificity, and for other antibodies if clinically relevant.
Table 1. Characteristics of the patients with SSc included in the discovery and replication sets*
Except where indicated otherwise, values are the percent. SSc = systemic sclerosis; lcSSc = limited cutaneous SSc; dcSSc = diffuse cutaneous SSc; anti–topo I = anti–topoisomerase I antibody; ACA = anticentromere antibody; CT = computed tomography.
Number of patients after exclusion of some patients from analysis for quality control purposes (see Patients and Methods).
Age, mean ± SD years
57.5 ± 13.9
57.6 ± 13.9
Disease duration, mean ± SD years
10.9 ± 8.2
12.1 ± 8.6
Fibrosing alveolitis seen on CT scan
At least 1 associated autoimmune disease
Primary biliary cirrhosis
Genotyping and quantification of gene expression in peripheral blood mononuclear cells (PBMCs).
We selected 4 TREX1 SNPs (rs3135946, rs7626978, rs3135943, and rs11797) and 6 low-frequency TNFAIP3 SNPs (rs9494883, rs72063345, rs5029939, rs2230926, rs117480515, and rs7749323) because of their reported associations with SLE (6–11) and/or because of their possible functional implications (10, 11). Genotyping was performed as previously described (13). In addition, direct sequencing of the region downstream of TNFAIP3 was performed in a random sample of 143 French SSc patients and 131 controls, to determine whether the polymorphic dinucleotide identified in SLE (rs117480515) was present (10). TNFAIP3 gene expression by PBMCs was measured by TaqMan quantitative real-time polymerase chain reaction. (Details on the process for direct sequencing and for measurement of gene expression in PBMCs are available from the corresponding author upon request.)
Statistical analyses were conducted as previously described (13) and included power calculation and haplotype analysis. Subjects lacking genotyping data for at least one of the TREX1 and TNFAIP3 SNPs (n = 96 and n = 178, respectively) were excluded from the analyses. In addition, SSc patients with concomitant autoimmune diseases known to be associated with the TREX1 and TNFAIP3 loci (n = 18) were excluded prior to analysis in order to avoid bias due to a possible excess of risk alleles attributable to these patients, as previously described (13). The combined data on the 2 populations were analyzed by calculating the pooled ORs using a Cochran-Mantel-Haenszel test for stratified analysis. Ninety-five percent confidence intervals (95% CIs) were calculated. The Breslow-Day method was applied to calculate the homogeneity of ORs between the 2 cohorts.
The average genotyping call rate for the TREX1 and TNFAIP3 variants was >98% in both the SSc and the control samples. The 6 TNFAIP3 variants and 3 of the TREX1 variants (rs1197, rs3135946, and rs7626978) were in Hardy-Weinberg equilibrium in the control subjects in the discovery population. Allelic frequencies were found to be in good agreement with those reported in the European Caucasian population.
No association was found between any of the 4 TREX1 variants and SSc overall or specific subphenotypes of SSc (data available from the corresponding author upon request). The TREX1 polymorphism rs3135943 described in SLE patients was not found in the French SSc and control populations, supporting the notion that it relates to a genetic mutation.
Direct sequencing of the region downstream of TNFAIP3 revealed the presence of a single-base deletion at chromosome 6 position 138,271,732, followed by a T>A variation in 12 SSc patients and 7 control individuals (allelic frequency 4.23% and 2.69%, respectively; P = 0.33), all in a heterozygous state. The rs117480515 minor A allele that refers to the T>A variant at chromosome position 138,272,733 was detected in 100% of individuals carrying the T deletion. Conversely, the rs117480515 T allele was detected in all individuals who did not carry the deletion. We therefore considered that the rs117480515 A allele was a perfect proxy for this deletion, and that this result could be extended to our large genotyping cohort using rs117480515 as a tag SNP for the T deletion and the dinucleotide.
The case–control association study demonstrated an association of all TNFAIP3 SNPs investigated with SSc (Table 2). When subphenotypes were considered and after Bonferroni correction for multiple testing, the most significant differences in allelic frequencies and ORs were observed for the subsets of SSc patients with diffuse cutaneous disease (dcSSc), with fibrosing alveolitis, and with a polyautoimmune phenotype (Table 2). The genotypes obtained for rs117480515 were in 100% agreement with those obtained by direct sequencing in individuals randomly selected to have both methods performed. Taking into account the observed association signal for TNFAIP3, we continued the investigation solely for this gene, and TREX1 was not genotyped in the replication sample.
Table 2. Allelic distributions of the TNFAIP3 SNPs in SSc patients and controls from the French discovery sample and results of the Cochran-Mantel-Haenszel test for the combined European Caucasian populations*
TNFAIP3 haplotype analysis and linkage disequilibrium (LD) relationship.
Two common TNFAIP3 haplotypes (frequency >1% in controls) were predicted from our study sample. Only the G-C-G-G-A-A haplotype, consisting of the minor alleles of each SNP, was found to be associated with SSc (4.46% in patients versus 1.82% in controls; P = 2.63 × 10−6). The other haplotype consisted of the major alleles of each SNP. Pairwise LD was calculated by analysis of the correlation (r2) in the control population. Consistent with findings in the haplotype analysis, all SNPs demonstrated strong LD in the French Caucasian healthy control subjects (details available from the corresponding author upon request). Given that the TNFAIP3 risk haplotype could be tagged by one single SNP, we subsequently only investigated SNP rs117480515 in the replication sample.
The SNP rs117480515 was in Hardy-Weinberg equilibrium in the Italian control population. The average call rate for rs117480515 was >98% in both the SSc patients and the control subjects. The allelic association with SSc was replicated only in patients with the polyautoimmune phenotype (minor allele frequency 7.0%, versus 2.4% in controls; OR 3.02 [95% CI 1.53–5.94], P = 0.002). Associations shown in the discovery cohort were not replicated for other SSc subtypes, including dcSSc (minor allele frequency 3.8%; OR 1.58 [95% CI 0.78–3.20], P = 0.139) (other data available from the corresponding author upon request).
Despite the different geographic origins of the 2 data sets, a test of combinability performed by the Breslow-Day method identified no significant differences between them. Pooled OR analysis with the Cochran-Mantel-Haenszel test showed that the rs117480515 A allele was strongly associated with susceptibility to SSc with associated autoimmune disease(s) (pooled OR 3.51, [95% CI 2.28–5.41], χ2 = 36.6) (Table 2). Furthermore, the combined analysis of each separate autoimmune disease subgroup revealed significant associations for all 3 of the most frequently reported autoimmune diseases, i.e., Sjögren's syndrome (OR 3.66 [95% CI 1.97–6.78], P = 1.07 × 10−5), autoimmune thyroid disease (OR 4.09 [95% CI 2.17–7.71], P = 2.03 × 10−6), and primary biliary cirrhosis (OR 5.07 [95% CI 1.76–14.6], P = 0.00083). Within-cohort analyses showed an association between the TNFAIP3 rs117480515 A allele and SSc with other autoimmune disease (n = 250) as compared to SSc with no other documented autoimmune disease (n = 1,099) (OR 1.98 [95% CI 1.31–2.99], P = 0.001) (data not shown).
Significant but weaker associations were observed in the dcSSc subset and the SSc with fibrosing alveolitis subset. The within-cohort comparison between the dcSSc subset (n = 443) and the limited cutaneous SSc subset (n = 158) was not significant (OR 1.39 [95% CI 0.95–2.04], P = 0.09).
Functional consequences of the TNFAIP3 rs117480515 polymorphism with regard to TNFAIP3 mRNA expression.
We then investigated the potential influence of the risk haplotype, tagged by SNP rs117480515, on TNFAIP3 mRNA expression by PBMCs from 33 control subjects (none with the rs117480515 AA genotype, 2 with the TA genotype, and 31 with the TT genotype) and 38 SSc patients (none with the rs117480515 AA genotype, 3 with the TA genotype, and 35 with the TT genotype). TNFAIP3 mRNA expression levels were significantly lower in individuals who carried the TNFAIP3 rs117480515 A risk allele than in those who did not (P = 0.02) (Figure 1). No difference in TNFAIP3 expression between patients and controls was observed (data not shown).
Findings of the present association study provide strong evidence of a positive association between TNFAIP3 low-frequency variants and the subset of SSc with concomitant autoimmune disorders in European Caucasian populations. This result was obtained in analyses of large cohorts and included a replication step confirming the strong association observed in the subset of SSc with polyautoimmunity. Within-cohort analyses further supported this association. It is noteworthy that the association with the TNFAIP3 rs117480515 A allele was consistently observed when each associated autoimmune phenotype (Sjögren's syndrome, autoimmune thyroid disease, and primary biliary cirrhosis) was analyzed individually, further demonstrating that the TNFAIP3 gene confers susceptibility to the polyautoimmune trait in general. This is of particular interest since the risk haplotype may therefore be considered a specific marker of polyautoimmunity.
We investigated a TT>A functional dinucleotide polymorphism in the downstream region of TNFAIP3 (10). The presence of the T deletion composing this risk variant was identified recently by resequencing of the TNFAIP3 locus in SLE patients, and was reported to be associated with reduced A20 protein expression (10). However, whether this deletion is present in patients with other autoimmune diseases and in healthy individuals was still unknown. Our results shed new light on this dinucleotide variant, since we found that the deletion 1) was present in 2.69% of control subjects, confirming that it is a low-frequency polymorphism and not a mutation, and 2) could be tagged by the contiguous SNP rs117480515. Furthermore, our haplotype analysis showed that the risk haplotype could be tagged by the rs117480515 A allele. When we addressed the question of the potential functional impact of this deletion on modulating TNFAIP3 mRNA expression, we found that carriers of the TNFAIP3 rs117480515 A risk allele displayed decreased expression of TNFAIP3 mRNA compared to noncarriers, consistent with the results observed in SLE (10). However, this first attempt to investigate the functional effect of the risk haplotype was limited by the small sample size and will need to be confirmed using a larger sample.
Recently reported converging data provide evidence of a central role of A20 in autoimmune pathways and suggest that defects in A20 expression might be involved in the development of autoimmune conditions. In SSc, gene expression profile analysis of endothelial cells derived from progenitor cells revealed a trend toward reduced TNFAIP3 mRNA expression in SSc patients compared to healthy controls (14). In addition, reduced A20 protein expression was observed in skin tissue of SSc patients compared to controls. Furthermore, studies using conditional A20-knockout mouse models have shown that A20 deficiency in dendritic cells results in systemic B and T cell autoimmunity resembling an SLE phenotype, while selective A20 deficiency in myeloid cells triggers an erosive polyarthritis phenotype (15).
The results obtained in this case–control association study demonstrate that the investigated polymorphisms of the TREX1 gene do not contribute to susceptibility to SSc or its subphenotypes. This lack of association strengthens the notion that while some genetic loci may confer susceptibility to several autoimmune phenotypes and support the concept of shared autoimmunity, other genes may be restricted to specific diseases. However, due to the low frequency of the TREX1 variants investigated, our study was statistically underpowered to allow us to draw firm conclusions regarding the lack of association observed for TREX1. Studies of larger populations of cases and controls are thus needed to confirm this result. Furthermore, although we found similar allelic frequencies of the TNFAIP3 SNP rs117480515 and used the Breslow-Day test to rule out genetic heterogeneity, the use of ancestry-informative markers would have further confirmed the absence of heterogeneity between French and Italian cohorts.
Our study establishes the TNFAIP3 rs117480515 variant as a susceptibility factor in the subset of SSc with associated autoimmune disease, mostly represented by Sjögren's syndrome, autoimmune thyroiditis, and primary biliary cirrhosis. Moreover, the current results, obtained through the analysis of enriched phenotypes, also provide further evidence supporting the implication of rare variants in susceptibility to complex diseases. These findings should encourage investigators to perform more in-depth phenotyping of patients with autoimmune disease and to take into account polyautoimmunity in statistical analyses. Indeed, genetic factors such as PTPN22, IRF5, STAT4, and now TNFAIP3 not only confer common genetic susceptibility to multiple autoimmune diseases, but also show potential as possible markers of individual and familial polyautoimmunity (1). Taken together, our data support the emerging role of A20 in various autoimmune diseases as highlighted in recent in vitro and in vivo studies (14, 15). This molecule is progressively being identified as a crucial player in shared autoimmunity, and as a potential new therapeutic target that will hopefully lead to common treatment for different autoimmune diseases.
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. Allanore 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. Koumakis, Dieudé, Cuomo, Matucci-Cerinic, Diot, Amoura, Sibilia, Carpentier, Guiducci, Meyer, Kahan, Boileau, Allanore.