Article Text

Download PDFPDF

Extended report
The interferon regulatory factor 5 gene confers susceptibility to rheumatoid arthritis and influences its erosive phenotype
  1. Karen Dawidowicz1,2,
  2. Yannick Allanore3,
  3. Mickaël Guedj4,
  4. Céline Pierlot1,
  5. Stefano Bombardieri5,
  6. Alejandro Balsa6,
  7. René Westhovens7,
  8. Pilar Barrera8,
  9. Helena Alves9,
  10. Vitor Hugo Teixeira1,10,
  11. Elisabeth Petit-Teixeira1,
  12. Leo van de Putte8,
  13. Piet van Riel8,
  14. Bernard Prum4,
  15. Thomas Bardin11,
  16. Olivier Meyer2,
  17. François Cornélis1,11,12,
  18. Philippe Dieudé1,2,
  19. for ECRAF
  1. 1GenHotel-EA 3886, University Evry-Paris 7 Medical School, Member of the AutoCure European Consortium, Evry-Genopole cedex, France
  2. 2Rheumatology Department, Bichat Claude Bernard Hospital, Paris, France
  3. 3Paris Descartes University, Rheumatology Department A, Cochin Hospital, Paris, France
  4. 4Laboratoire Statistique et Génome, Université d'Evry, Val d'Essonne, France
  5. 5Pisa University, Pisa, Italy
  6. 6La Paz Hospital, Madrid, Spain
  7. 7KUL Leuven University, Leuven, Belgium
  8. 8Nijmegen University, Nijmegen, The Netherlands
  9. 9Porto San Joao Hospital, Porto, Portugal
  10. 10Faculty of Medicine, University of Coimbra, Coimbra, Portugal
  11. 11Rheumatology Federation, Pôle Appareil Locomoteur, Lariboisière Hospital, APHP, Paris Diderot University, Paris, France
  12. 12Clinical Genetic Unit, Pôle L I P, Lariboisière Hospital, APHP, Paris Diderot University, Paris, France
  1. Correspondence to Dr Philippe Dieudé, Service de Rhumatologie, Hôpital Bichat Claude Bernard, 46 rue Henri Huchard, 75018 Paris, France; philippe.dieude{at}


Background Increased expression of type I IFN genes, also referred to as an IFN signature, has been detected in various autoimmune diseases including rheumatoid arthritis (RA). Interferon regulatory factors, such as IRF5, coordinate type I IFN expression. Multiple IRF5 variants were suggested as autoimmunity susceptibility factors.

Objective As the linkage proof remains important to establish fully any genetic RA susceptibility factor, the authors took advantage of the largest reported European trio family resource dedicated to RA to test for linkage IRF5 and performed a genotype–phenotype analysis.

Methods 1140 European Caucasian individuals from 380 RA trio families were genotyped for IRF5 rs3757385, rs2004640 and rs10954213 single nucleotide polymorphisms (SNP).

Results Single marker analysis provided linkage evidence for each IRF5 SNP investigated. IRF5 linked to RA with two haplotypes: the CTA risk haplotype ‘R’ (transmission (T)=60.6%, p=23.1×10−5) and the AGG protective haplotype ‘P’ (T=39.6%, p=0.0015). Linkage was significantly stronger in non-erosive disease for both IRF5 R and P haplotypes (T=73.9%, p=4.20×10−5 and T=19.6%, p=3.66×10−5, respectively). Multivariate logistic regression analysis found IRF5 linked to RA independently of the rheumatoid factor status. IRF5 RR and PP haplotypic genotypes were associated with RA, restricted to the non-erosive phenotype: p=1.68×10−4, OR 4.80, 95% CI 2.06 to 11.19; p=0.003, OR 0.17, 95% CI 0.05 to 0.57, respectively.

Conclusion This study provides the ‘association and linkage proof’ establishing IRF5 as a RA susceptibility gene and the identification of a genetic factor that seems to contribute to the modulation of the erosive phenotype. Further studies are warranted to clarify the role of IRF5 in RA and its subphenotypes.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The type I IFN pathway has been postulated to play a pivotal role in autoimmune diseases.1 Indeed, an increased expression of type I IFN genes by peripheral blood mononuclear cells, also referred to as an IFN signature, has been detected in multiple autoimmune diseases including rheumatoid arthritis (RA).2,,5 The IFN regulatory factors (IRF) are major regulators of genes activated by the type I IFN. Recently, variants of the IRF5 gene (IRF5), a member of IRF, were found to be involved in the susceptibility to systemic lupus erythematosus (SLE), Sjögren syndrome (SS), systemic sclerosis (SSc) and RA genetic background.6,,9

RA, the most frequent autoimmune human disease affects approximately 1% of the adult population worldwide. RA is a multifactorial disease for which the genetic component accounts for approximately 50% of the disease susceptibility. A recent genome-wide association study revealed a marginal association between RA and the single nucleotide polymorphism (SNP) rs3807306, which tags a haplotype block containing IRF5.10 Multiple case–control association studies performed in distinct populations observed an association between IRF5 variants and RA. Several risk variants have been identified: rs3757385, rs2004640 and rs10954213. Therefore, the contribution of IRF5 in RA subphenotype susceptibility remains to be clarified.10

Multiple functional IRF5 variants were reported to contribute to susceptibility to autoimmunity: rs2004640 creates a donor splice site in intron 1, which results in expression of the alternative exon 1B; rs10954213 alters the polyadenylation site, the rare A allele leading to a shorter and more stable messenger RNA.11 An insertion deletion variant (CGGGG indel) located in the IRF5 first intron, resulting in an increased binding of the transcription factor SP1 to the risk allele, was also reported to confer susceptibility to autoimmunity.12,,14 Interestingly, the functional CGGGG indel polymorphism could be unambiguously inferred from haplotype reconstructions due to the high linkage disequilibrium (LD) with the three following SNP: rs3757385, rs2004640 and rs10954213.13 15

The increasing number of IRF5 association reports in various autoimmune diseases provides compelling evidence for an implication of this gene in susceptibility to autoimmune diseases, but the ‘gold standard’ for multifactorial diseases, that is, the linkage proof, was convincingly met only for SLE.6 As human populations show a great genetic diversity, population structure must be taken into account when conducting population-based genetic association studies becase it might lead to false positive results. To avoid this drawback, we applied the complementary transmission disequilibrium test (TDT) approach.

Therefore, our objective was to provide the linkage proof of IRF5 to RA and to perform a haplotype–phenotype correlation analysis. To that aim, we selected the three IRF5 rs3757385, rs2004640, rs10954213 SNP as they were reported to be associated with RA and the functional CGGGG indel polymorphism could be inferred from haplotype reconstruction. This study was performed taking advantage of the largest reported European family resource dedicated to RA linkage studies collected by the the European Consortium on Rheumatoid Arthritis Families (ECRAF).

Materials and methods

Patients and families

We analysed in total 1140 European Caucasian individuals from 380 trio families (one RA case and both parents). The sample consisted of the DNA from 218 trio families from French Caucasian origin and 162 trio families from west European Caucasian origin, as recorded for each of the four grandparents, who had been recruited through ECRAF. RA patients fulfilled the 1987 American College of Rheumatology (formerly the American Rheumatism Association) criteria.16 Clinical data were reviewed by three rheumatologists. All individuals provided written informed consent and ethics committees in each country approved the study. For each index, characteristics collected were sex, age at RA onset, disease duration in years, presence of bone erosions at x-ray examination, presence of rheumatoid nodules, rheumatoid factor (RF) status (determined by latex fixation, by Waaler–Rose assay or by laser nephelometry) and anti-cyclic citrullinated peptide (CCP) status determined by ELISA (Immunoscan RA, Euro-Diagnostic, Malmö, Sweden). Clinical characteristics of the RA index cases are summarised in table 1. Regarding the erosion status, an erosive disease was defined as the presence of typical erosions on both feet or wrists/hands x-rays that were collected for all RA index cases at their inclusion in the study.

Table 1

Characteristics of rheumatoid arthritis index cases

Each IRF5 SNP (rs3757385, rs2004640, rs10954213) and the resulting observed haplotypes were tested for linkage in the global sample and in subsets according to RF status and the presence or absence of erosive disease status. Testing for the influence of the HLA-DRB1*SE and the anti-CCP status was performed in 200 and 239 RA trio families, respectively, when these data were available. The control haplotype derived from untransmitted parental chromosomes provided a control population perfectly matched for the population of origin.17


Blood samples were collected for DNA extraction and genotyping. IRF5 rs3757385, rs2004640 and rs10954213 SNP were genotyped by the Taqman SNP genotyping assay-allelic discrimination method (Applied Biosystem, Foster City, California, USA) and using the KASpar genotyping system (Kbioscience, The average genotype completeness was 98% for the three SNP investigated in both RA index cases and their parents. The reliability of the genotypes was established by re-genotyping 30% of both samples with 100% consistency. We assessed the quality of the genotype data by testing for Hardy–Weinberg equilibrium in the control samples, using Fisher's exact test (p>0.05) and using the program PedCheck to identify possible non-Mendelian inheritance of alleles in the trio families.18

Statistical analysis

The linkage analysis relied on the TDT, which compares, for a given allele, the transmission of that allele from heterozygous parents to RA patients, with the transmission expected from Mendel's first law (ie, 50%).19 TDT analyses were performed using GENEHUNTER v2.0 β.20 Pairwise LD of the three IRF5 SNP (IRF5 rs3757385, rs2004640 and rs10954213 SNP) was obtained from family-based association tests.21 Bonferroni's correction was applied for TDT univariate analysis (eight phenotypic subsets defined by erosive, RF, ACPA and HLA-DRB1*SE status). To test for linkage evidence of IRF5 to RA restricted to a particular subset we perform a multivariate logistic regression analysis adjusted for the following covariates (erosion and RF status). Similar to a case–control study, genotype relative risk compares the SNP genotype/haplotype distribution in RA cases and in controls (each control being perfectly matched to a RA index case, as controls were defined by reconstructed genotypes/haplotypes derived from untransmitted parental chromosomes).17 Association analysis was performed using the χ2 test of 2-by-2 contingency tables of haplotype frequencies. The corresponding OR were assessed using a standard logistic regression analysis.


Overall, 1140 European Caucasian individuals from 380 RA trio families were analysed. No significant deviation from the Hardy–Weinberg equilibrium was observed in the controls (p=0.99 for IRF5 rs375785, p=0.42 for IRF5 rs2004640 and p=0.70 for IRF5 rs10954213) and no Mendelian inheritance errors were observed for any of the SNP included in the study.

IRF5 linkage study in the global sample

We observed linkage evidence with, compared with the Mendel's expectation of 50%, an overtransmission of IRF5 rs3757385 C, rs2004640 T and rs10954213 A risk alleles, in the global RA sample: T (transmission testing Mendel's law) T=58.4%, p=0.0022, T=57.9%, p=0.0020 and T=56.5%, p=0.013, respectively (table 2). Following those results and according to the pairwise LD of the three IRF5 SNP obtained from family-based association tests (table 3),21 we tested for linkage the IRF5 rs3757385 C–rs2004640 T–rs10954213 A (CTA) risk haplotype, taking advantage of trio families, which provide the ideal tool to test genuine haplotypes. In agreement with the previously reported LD analysis, IRF5 rs3757385 and IRF5 rs2004640 were found to be in LD (D′=0.97, r2=0.54). IRF5 rs10954213 was also in LD with both IRF5 rs3757385 and IRF5 rs2004640 (D′=0.90, r2=0.70 and D′=0.88, r2=0.52, respectively; table 3).

Table 2

RA linkage study for the IRF5 rs3757385 C, rs2004640 T and rs10954213 A alleles according to bone erosion and RF status of RA index cases

Table 3

Linkage disequilibrium of IRF5 rs3757385, rs2004640, rs10954213

IRF5 linked to RA with two significant haplotypes out of eight: a strong overtransmission of the IRF5 CTA risk haplotype (R) was observed (T=60.6%, p=23.1×10−5). Conversely, an undertransmission of its mirror haplotype (ie, the IRF5 AGG protective haplotype (P)) was detected (T=39.6%, p=0.0015; table 4).

Table 4

RA linkage study for both IRF5CTA and AGG haplotypes according to bone erosion and RF status of RA index cases

IRF5 Linkage study conditional on RF and bone erosion status: haplotype–phenotype correlation

Univariate analysis

We observed strong linkage evidence of the IRF5 R haplotype to both RA RF negative (T=69.9%, p=27.9×10−5, padj=0.0022). An excess of transmission was also detected in RF-positive families; however, not reaching statistical significance after correction for multiple testing (padj=0.288). The T of the IRF5 P haplotype was decreased in RF-negative RA (T=32.7%, p=0.0087, padj=0.0696). No linkage evidence was found in RF-positive families (table 4).

The T of both IRF5 R and P haplotypes was not influenced by the anti-CCP status of RA index cases (data not shown, available upon request from the corresponding author).

When linkage analysis was performed according to the erosive status of RA index cases, strong linkage evidence of the IRF5 R haplotype was detected to non-erosive RA (T=73.9%, p=4.20×10−5, padj=3.36×10−4). No linkage evidence was detected to erosive RA (table 4). A strong undertransmission of the IRF5 P haplotype was only detected in non-erosive RA (T=19.6%, p=3.66×10−5, padj=2.93×10−4; table 4).

Multivariate logistic regression analysis

As the erosive status could be influenced by the RF status and to establish fully the contribution of IRF5 to the non-erosive and RF-negative phenotypes we next performed a multivariate logistic regression analysis adjusted for the following covariates (RF and erosive status). The undertransmission of the IRF5 P haplotype was found to be restricted to the non-erosive subset independently of the RF status: erosive status adjusted for RF status p=0.038 and RF status adjusted for erosive status, p=0.611. Regarding the IRF5 R haplotype we observed a trend for an excess of transmission also restricted to non-erosive RA; however, not reaching statistical significance: erosive status adjusted for RF status p=0.068 and RF.

IRF5 association study in the global sample

Having established linkage evidence for both IRF5 R and P haplotypes, we obtained from those data an unbiased estimation of their association with RA. Concordant with the linkage evidence we detected an association between both IRF5 R and P haplotypes and RA (tables 5 and 6).

Table 5

RA association study for the IRF5 CTA haplotype according to bone erosion and RF status of RA index cases

Table 6

RA association study for the IRF5 AGG haplotype according to bone erosion and RF status of RA index cases

IRF5 association study conditional on bone erosion and RF status

Following our linkage study results, we selected both non-erosive RA and RF-negative subsamples for association study. There was no practical interest in describing the association analysis in the remaining subsamples, as no other independent factor influencing linkage was detected by multivariate analysis.

In good agreement, both IRF5 R and P haplotypes were found to be associated with non-erosive RA and RF-negative RA (table 5). A strong association between the homozygous RR haplotypic genotype with both non-erosive and RF-negative RA populations was observed (OR 4.80, 95% CI 2.06 to 11.19, p=1.68×10−4 and OR 3.64, 95% CI 1.62 to 8.18, p=0.0014; table 5). The homozygous PP haplotypic genotype was associated with non-erosive RA (OR 0.17, 95% CI 0.05 to 0.57, p=0.003). This protective effect was also detected in the RF-negative RA population (OR 0.11, 95% CI 0.03 to 0.45, p=0.0010; table 6).


Our study is the first to provide the linkage proof, attesting IRF5 as a RA susceptibility gene. Among the eight IRF5 haplotypes defined by the following SNP: rs3757385, rs2004640 and rs10954213, we have identified two IRF5 haplotypes linked to and associated with RA: the IRF5 CTA risk haplotype (R), which confers susceptibility to RA and its mirrored IRF5 AGG protective haplotype (P), with a protective effect. The observed IRF5 allele and haplotype frequencies were in good agreement with those previously reported in the European Caucasian population, supporting the accuracy of the genotyping.6 8 12 13 15 22,,25

In addition, the haplotype–phenotype correlation analysis revealed that IRF5 influences both the erosive and the RF status of RA, as both linkage and association proofs were restricted to the non-erosive and RF-negative RA subsets. Following our results it could be hypothesised that some shared genetic factors, such as IRF5, could strongly influence the RA phenotype leading to phenotypic similarity with known IRF5-associated arthritis (ie, SLE and SS being non-erosive and frequently RF negative). In our study, the mean disease duration (9.7±8.5 years) of RA index cases strengthens the putative role of IRF5 on the RA non-erosive phenotype as it is well established that erosions usually appear within the first 2 years of RA.26,,28 However, to clarify the role played by IRF5 in the RA erosive phenotype, investigation into inception cohorts is mandatory.

These findings illustrate that IRF5, an autoimmune susceptibility gene shared by various connective tissue diseases, may contribute to the typical heterogeneity of a given disease such as RA. Of note, our results and those previously reported in SSc, indicate that IRF5, a genetic factor shared by distinct autoimmune diseases, could strongly influence the phenotype.15 25 This suggests that information from common risk polymorphisms could improve disease prediction and may be useful for risk stratification of a given disease. Nonetheless, it should be noted that the individual co-occurrence of known IRF5 associated diseases (SS, SSc, SLE, multiple sclerosis, Wegener's disease, inflammatory bowel diseases) in RA index cases was not available (respective diagnosis criteria are not systematically assessed). Therefore, this point should be taking into account for future studies.

Of most interest, IRF5 haplotypes related to those identified in this study (ie, CTA risk and the AGG protective haplotypes), were also reported to be associated with SLE, SS, Wegener's disease and more recently SSc.8 11 15 29

A limitation of this familial study is that it provides reliable association estimates only for the western European Caucasian population. In addition, if the ‘control’ frequencies can be considered reliable for the western European Caucasian population, the RA index case frequencies are representative only for early-onset RA, taking into account the requisites for both parents' participation in trio families. The involvement of IRF5 in RA heterogeneity has to be refined, and a model integrating IRF5 with other RA susceptibility genes is warranted.

Identification of new RA susceptibility genes may contribute to the better understanding of RA pathogenesis providing rational grounds for a subclassification based on both genetic background and phenotype.30 Nonetheless, the biological consequences of the observed IRF5 risk haplotype remains to be investigated. Taking into account that the CGGGG indel is unambiguously inferred from both IRF5 R and P haplotypes, it could be hypothesised that the indel polymorphism plays a pivotal role in RA genetic susceptibility.13 However, a re-sequencing of IRF5 of individuals carrying the RR genotype is required before concluding that the CGGGG indel is the definite and exclusive IRF5 RA causal variant.

In conclusion, this study provides both ‘association and linkage proof’ for IRF5 as a RA susceptibility gene. We also provide evidence for an involvement of IRF5 in RA heterogeneity, notably the non-erosive and RF-negative phenotypes. Elucidating how both IRF5 R and P haplotypes modulate the type I IFN pathway leads to an improved understanding of RA pathophysiology. Our results provide new insight into the pathogenesis of RA suggesting a pivotal role of type I IFN, at least in non-erosive and RF-negative RA.


The authors are grateful to the RA family members and their rheumatologists for participation.



  • Funding This work was supported by Association Française des Polyarthritiques, Société Française de Rhumatologie, Association Rhumatisme et Travail, Foundation for Science and Technology, Portugal (grant SFRH/BD/23304/2005), Association Polyarctique, Groupe Taitbout, Académie Nationale de Médecine, Association de Recherche sur la Polyarthrite, Genopole, Conseil Régional Ile de France, Fondation pour la Recherche Médicale, Université Evry-Val d'Essonne and unrestricted institutional support from Wyeth, Schering-Plough, Pfizer and Amgen. The European Consortium on Rheumatoid Arthritis Families (ECRAF) was initiated with funding from the European Commission (BIOMED2) by: T Bardin, D Charron, F Cornélis (coordinator), S Fauré, D Kuntz, M Martinez, J F Prudhomme, J Weissenbach (France); R Westhovens, J Dequeker (Belgium); A Balsa, D Pascuale-Salcedo (Spain); M Spyropoulou, C Stavropoulos (Greece); P Migliorini, S Bombardieri (Italy); P Barrera, L Van de Putte (The Netherlands); H Alves, A Lopez-Vaz (Portugal).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval This study was conducted with the approval of the CPPRB Kremlin Bicêtre Hospital, AP-HP.

  • Provenance and peer review Not commissioned; externally peer reviewed.