Abstract
Objective. Rheumatoid arthritis (RA) is an inflammatory joint disease with features of an autoimmune disease with female predominance. Candidate genes located on the X-chromosome were selected for a family trio-based association study.
Methods. A total of 1452 individuals belonging to 3 different sample sets were genotyped for 16 single-nucleotide polymorphisms (SNP) in 7 genes. The first 2 sets consisted of 100 family trios, each of French Caucasian origin, and the third of 284 additional family trios of European Caucasian origin. Subgroups were analyzed according to sex of patient and presence of anti-cyclic citrullinated peptide (anti-CCP) autoantibodies.
Results. Four SNP were associated with RA in the first sample set and were genotyped in the second set. In combined analysis of sets 1 and 2, evidence remained for association of 3 SNP in the genes UBA1, TIMP1, and IL9R. These were again genotyped in the third sample set. Two SNP were associated with RA in the joint analysis of all samples: rs6520278 (TIMP1) was associated with RA in general (p = 0.035) and rs3093457 (IL9R) with anti-CCP-positive RA patients (p = 0.037) and male RA patients (p = 0.010). A comparison of the results with data from whole-genome association studies further supports an association of RA with TIMPL The sex-specific association of rs3093457 (IL9R) was supported by the observation that men homozygous for rs3093457-CC are at a significantly higher risk to develop RA than women (risk ratio male/female = 2.98; p = 0.048).
Conclusion. We provide evidence for an association of at least 2 X-chromosomal genes with RA: TIMP1 (rs6520278) and IL9R (rs3093457).
- RHEUMATOID ARTHRITIS
- POLYMORPHISM
- GENETIC PREDISPOSITION TO DISEASE
- TISSUE INHIBITOR OF METALLOPROTEINASES
- GENETIC STUDIES
Rheumatoid arthritis (RA) is an inflammatory joint disease with features of an autoimmune disease and a prevalence of about 1% in the European Caucasian population1. There is evidence for genetic influences on RA and heritability is estimated to be 60%2. Female sex is a well known risk factor for RA. The female to male ratio ranges between 3 and 43. There may be a link between heritability and sex, as the female genome differs crucially from the male genome. The Y chromosome supplies males with several genes absent in the female4, while incomplete X-inactivation or varying inactivation patterns may lead to gene-dosage skewing in females5. X-chromosomal abnormalities were observed in immunological diseases, e.g., a significantly higher rate of acquired X-monosomy6 and significantly skewed X-inactivation7. Patients with Turner’s syndrome are known to manifest common autoimmune features8. Whole-genome linkage studies for RA suggest among others the presence of loci of interest on chromosome X9,10. Thus, X-chromosomal genes are highly relevant candidate genes to test for association with RA.
For this study we selected 7 genes, CD40LG, CD99, EIF2S3, IL9R, TIMP1, UBA1, and XIAP (Table 1). Most of these genes are involved in pathways thought to be crucial for RA etiology, and evidence for their involvement in other immunological diseases exists as well.
CD99 and IL9R are situated within pseudoautosomal regions and have a functional homologue on the Y chromosome, whereas the other genes are restricted to the X chromosome. To our knowledge none of the genes we selected, with the exception of TIMP1, has been investigated for association with RA in candidate gene studies.
CD40LG is involved in the regulation of B cell functions and the production of autoantibodies11. CD99 is described to play a role in transport regulation of MHC class I molecules12, lymphocyte adhesion13, and induced T cell death14. EIF2S3 is the γ-subunit of the eukaryotic translation initiation factor (EIF2) and is only partially affected by X-inactivation15. EIF2 is involved in stress responses and apoptosis16. Insufficient apoptosis of inflammatory cells in synovial membrane as well as increased apoptosis, especially within the synovial lining, has been demonstrated in RA17,18. IL9R is a receptor for the cytokine interleukin 9 (IL-9) expressed on many hematopoietic cells including T cells19, and it is also involved in early T cell development20. The gene product of TIMP protects extracellular matrix from degradation by inhibiting metalloproteinases (MMP)21. Secretion of MMP is required for the initial stage of angiogenesis22, contributing to pannus formation in RA23. TIMP1 (SNP rs5953060) was described to be associated with RA in a small Japanese cohort24 and has also shown association with other immunity disorders like Crohn’s disease25 and systemic sclerosis26. UBA1 (also known as UBE1) catalyzes the first step in ubiquitin conjugation to mark cellular proteins for degradation27. Involvement of UBA1 in cell-cycle regulation and apoptosis can be demonstrated and provides a functional link to RA28. XIAP is a potent inhibitor of apoptosis and is involved in regulation of lymphocyte homeostasis29.
Our aim was to investigate genetic variants of selected X-chromosomal genes in a candidate gene association study based on a family-trio approach in a European Caucasian population.
MATERIALS AND METHODS
Three sets of family trios, RA patient (i.e., the affected individual) and both parents, were genotyped. Detailed characteristics of the first 2 and parts of the third set have been described30. Briefly, the first 2 sets consisted of 100 family trios of French Caucasian origin. The third set consisted of 284 additional European Caucasian families, from France, Germany, Italy, Portugal, Spain, The Netherlands, and Belgium. All affected individuals fulfilled the American College of Rheumatology 1987 revised criteria for RA31. In addition the status of anti-cyclic citrullinated peptide autoantibodies (anti-CCP, also known as ACPA) was available for French and German RA patients (CCP-positive, n = 226; CCP-negative, n = 73). In our multistage approach all SNP were genotyped in the first sample set (“exploration set”). Markers with a significant association with RA (uncorrected p < 0.05) were then genotyped in the second sample set (“replication set”). When evidence increased in favor of an association, i.e., the p value of the association decreased in the combined analysis of both sets, markers were genotyped again in the third sample set (the multinational European replication set).
Genomic DNA was purified from fresh peripheral blood leukocytes or from Epstein-Barr virus-transfected cell lines using standard methods.
SNP were chosen based on their position in the gene, depending on gene length and validation status. Information from public databases (PupaSNP, UCSC Genome Browser, Ensembl) was used to aid in SNP selection. Selected SNP are listed in Table 1.
Genotyping was carried out using the genoSNIP technique (Bruker Daltonics, Billerica, MA, USA)32. Polymerase chain reaction primers were designed using MuPlex Vs 2.2. SBE-primer design was carried out using PrimExtend, an in-house software tool based on CalcDalton33. Primer sequences are shown in Table 2.
Samples of the third set were genotyped by applying a TaqMan 5’ allelic discrimination assay (Applied Biosystems, Foster City, CA, USA) following the manufacturer’s protocols.
For quality control purposes Mendelian laws of inheritance and Hardy-Weinberg equilibrium (HWE) in nontransmitted controls had to be fulfilled (p > 0.01). HWE analysis for nonpseudoautosomal genes was carried out in healthy female controls (mothers) only. All genotyping results fulfilled the quality control criteria. Genotype call-rate was more than 95%.
Statistical analysis
HWE was investigated using a chi-square test with 1 degree of freedom. Linkage and association analyses were performed using the transmission disequilibrium test (TDT)34 and the genotype relative risk (GRR) test35. The TDT compares the transmission of SNP alleles from heterozygous parents to affected offspring with a transmission ratio of 50% as expected by Mendel’s law. The GRR test compares differences in genotype distribution between RA cases and “virtual controls” reconstructed from nontransmitted parental alleles. Haploview 4.1 software was used for gene-wide haplotype analysis36. Tests were also done in sample sets stratified for sex or anti-CCP status of RA patients. We used a 2-tailed test of interaction to assess significance of differences between subgroups37.
For nonpseudoautosomal genes XTDT was applied as implemented in Haploview 4.138. As proposed39, males were treated like homozygous females for comparing allele frequencies by allele counting. Additionally, for GRR tests (Lathrop tests) only maternal reconstructed control genotypes and genotypes from corresponding affected female children were included.
RESULTS
In the first set, consisting of 100 French Caucasian family trios, 3 genes, IL9R, TIMP1, and UBA1, showed evidence for association. SNP with evidence for association were again genotyped in the second French Caucasian family trio set (100 additional trios). A combined analysis of set 1 and set 2 revealed a decreased p value for 3 markers. These SNP, rs4239963 (UBA1), rs6520278 (TIMP1), and rs3093457 (IL9R), were genotyped in the third European Caucasian sample set (284 additional trios). These data are summarized in Tables 3, 4, and 5. Details of family trio-based association analysis for all markers are shown in Tables 6, 7, and 8.
While the UBA1 SNP rs4239963 showed significant association with RA in the first 2 sample sets, it was not found to be associated with RA in the combined analysis of all 3 sets, although the trend was the same as in sets 1 and 2, with the minor allele (C) being undertransmitted (Table 3).
In contrast, SNP rs6520278 (TIMP1) was found to be significantly associated with RA in general, which is indicated by significant p values in the combined analysis of all 3 sets (p = 0.035; Table 4). The TDT showed the minor allele T was undertransmitted. Association of rs6520278 in families with male offspring could not be replicated in the European replication set. Additionally, the test of interaction revealed no significant difference between female and male subgroups for the SNP, as effect sizes (GRR minor vs major genotype) of the 2 subgroups did not differ significantly (p for interaction = 0.071). Another SNP, rs6520277 of TIMP1, also showed significant p value in families with male offspring in set 1, but this result could not be replicated in the second sample set.
SNP rs3093457 of IL9R was found to be significantly associated with RA in 2 subgroups in the combined analysis: families with anti-CCP-positive patients (p = 0.037) and families with male patients (p = 0.010), while an association of rs3093457 was only marginally significant in all family trios (p = 0.056; Table 5) and was not significant in families with female RA patients. In both subgroups the association was due to an increase of the homozygous minor genotype rs3093457-CC in RA cases. We also performed an interaction test to identify specific effects concerning anti-CCP status and/or sex. Comparing the GRR of rs3093457-CC for male and female subgroups revealed that the SNP affected males significantly more than females (p = 0.048). GRR in families with male offspring was about 3 times greater in the combined analysis of all sample sets (ratio of male/female GRR 2.98, 95% CI 1.01–8.79; Table 9). No significant difference between effect sizes was observed for anti-CCP-positive and negative subgroups.
DISCUSSION
We investigated SNP in 7 X-chromosomal genes for association with RA and were able to detect evidence for association for markers of 2 genes, TIMP1 and IL9R. SNP rs6520278 of TIMP1 showed a significant association in the combined analysis of all 3 sets (n = 484 family trios), with the minor T-allele being undertransmitted in RA patients (affected children), indicating a protective effect for this allele.
SNP rs6520278 was measured directly in at least 3 whole-genome association studies (WGAS) [the Spanish Upstream Regulatory Region study40; the British Wellcome Trust Case-Control Consortium (WTCCC) study39; the North American Rheumatoid Arthritis Consortium and Swedish Epidemiological Investigation of Rheumatoid Arthritis41 studies], but p values were not significant. This might be due to disease heterogeneity or, if the analyzed variant is not a causative variant, to differences in the linkage disequilibrium of the various sample groups. On the other hand, we found several markers in the WGAS in proximity (± 200 kb, as proposed40) to TIMP1 associated with RA (Table 10) at the single-marker level.
UBA1 and TIMP1 are both situated on the same chromosomal band (Xp11.23) and about 370 kb apart. We could not confirm an association of the UBA1 gene with RA in the analysis of all 3 sample sets. However, in WGAS several SNP near the gene showed significant p values as well (Table 10). Given the proximity of UBA1 and TIMP1, these data might indicate the presence of causative variants in this chromosomal region.
Linkage disequilibrium (correlation of alleles of 2 polymorphisms in a given population) was examined between SNP associated with RA in our study and SNP in proximity that are also associated with RA in WGAS. Because SNP data for rs4239963 (UBA1) were not available from HapMap (release 23) and the IL9R region was not covered by the cited WGAS, only TIMP1 could be investigated. The SNP rs760580 correlated with rs6520278 of TIMP1 as shown by high D’ (0.545) and r2 (0.222). Moreover, SNP rs760580 was associated with RA in the WTCCC study at the single-marker level (p = 0.044) and showed a protective effect of the minor allele, as did rs6520278.
TIMP1 SNP rs5953060 was described to be associated with RA in a small Japanese cohort (p = 0.02)42. While we could not replicate this association (p = 0.228; Table 6), we found linkage disequilibrium between rs5953060 and rs65020278 (D’ = 1, r2 = 0.607). Therefore it appears possible that rs5953060 in the Japanese study reflects association of the same unknown causative locus in the TIMP1 region as did rs6520278 in our study due to different linkage disequilibrium among populations.
We did not find a significant sex-specific effect of rs6520278, although another TIMP1 SNP investigated in our study, rs6520277, did hint at sex-specific effects of the gene. This SNP was significantly associated with RA in families with male children in the first set and in the combined analysis of the first and second sets. However, the small number of informative families of male RA patients did not allow for final conclusions. Further investigations are required to clarify possible sex-specific effects of TIMP1.
TIMP1 could influence the etiology of RA in several ways. It inhibits MMP43,44 and subsequently prevents the degradation of cartilage22. The inhibition of MMP also may inhibit angiogenesis required for pannus formation23,45. A genetic association of TIMP1 with RA therefore supports the hypothesis that modified angiogenesis might play an important role in the etiology of RA due to altered regulation of MMP via their interactions with TIMP1.
Synovial endothelial cells of patients with RA secrete decreased levels of TIMP146. Levels of TIMP1 expression are affected by X-chromosomal inactivation47,48, but TIMP1 partially escapes X-chromosomal gene silencing49. TIMP1 variants may also lead to differences in the level of expression, e.g., SNP might be involved in incomplete gene silencing or in other regulatory mechanisms. It remains to be seen whether allele-specific effects contribute to differences in TIMP1 expression.
Another SNP associated with RA in our study was rs3093457 in the IL9R gene. SNP near IL9R were not investigated in any of the WGAS, thus our findings are the only data available for this gene and this region. The homozygous minor genotype CC was marginally increased in all cases (p = 0.056) and was significantly increased in the anti-CCP-positive subgroup (p = 0.037) and in male RA patients (p = 0.01). The interaction test result further supports the sex-specificity of the association with males, who are 3 times more affected by this genotype than females. Sex-specific effects for IL9R have been described for bipolar disorder as well as childhood wheezing, an asthma characteristic, with associations limited to males50,51. The observed association of the X-chromosomal IL9R with RA would therefore provide further evidence for sex-specific disease mechanisms in RA.
There are several possibilities for IL9R involvement in the etiology of RA. Different IL9R splice variants affect the influence of IL-9, because they differ in IL-9-binding abilities52. Expression of IL-9 was shown to be correlated with inflammation events and infiltration of lymphocytes in allergic diseases53. The STAT pathway is the main signaling pathway of IL-9/IL-9R54, and its role in RA is discussed55. IL9R is also involved in early T cell development20, which is relevant for RA, as the balance between autoreactive T cells and regulatory T cells is essential for immune tolerance.
We provide evidence suggesting association of 2 X-chromosomal genes, TIMP1 and IL9R, with RA. As in other studies of RA39, the effects of the observed associations were modest. This might be a reason why only nominal significance was achieved. However, our multistage approach analyzing and combining multiple study cohorts allowed testing for such modest genetic effects56. It is necessary to verify the associations we observed in additional larger cohorts.
While our findings might not explain the female predominance in RA, they point out that different disease mechanisms might exist in females and males. To elucidate the genetic background of complex diseases such as RA it might be beneficial to consider sex-specific effects, e.g., using sex-stratified sample subsets for association studies.
Acknowledgments
We thank RA family members and their rheumatologists for participation. We thank Carlos Vaz, Antonio Vaz-Lopes, and Manuela Fernandes for their medical assistance. We also thank Rita Rzepka and Annette Feldmeyer for expert technical assistance. We acknowledge as well the Association Française des Polyarthritiques, Association Rhumatisme et Travail, Societe Francaise de Rhumatologie, Association Polyarctique, Groupe Taitbout, Genopole; and the European Union for AutoCure. The family data were collected by ECRAF (European Consortium on Rheumatoid Arthritis Families) members Paola Migliorini, Alejandro Balsa, René Westhovens, Pilar Barrera, Helena Alves, Carlos Vaz, Manuela Fernandes, Dora Pascuale-Salcedo, Stefano Bombardieri, Jan Dequeker, Timothy R. Radstake, Piet van Riel, Leo van de Putte, Antonio Lopes-Vaz, and François Cornélis.
Footnotes
-
Supported by grant 7692/1187 from the Sächsische Aufbaubank-Förderbank, grant 4212/04-04 from the European Fund for Regional Development (EFRE), the German Federal Ministry for Education and Research (Hochschul- und Wissenschaftsprogramm; “Kompetenznetz Rheuma” 01GI9949 to IM), by the Rosa-Luxemburg-Stiftung, and by the Foundation for Science and Technology, Portugal (grant SFRH/BD/23304/2005).
- Accepted for publication May 12, 2009.