|Home | About | Journals | Submit | Contact Us | Français|
The severity of joint destruction in rheumatoid arthritis (RA) is highly variable from patient to patient and is influenced by genetic factors. Genome-wide association studies have enormously boosted the field of the genetics of RA susceptibility, but risk loci for RA severity remain poorly defined. A recent meta-analysis of genome-wide association studies identified 6 genetic regions for susceptibility to autoantibody-positive RA: CD40, KIF5A/PIP4K2C, CDK6, CCL21, PRKCQ, and MMEL1/TNFRSF14. The purpose of this study was to investigate whether these newly described genetic regions are associated with the rate of joint destruction.
RA patients enrolled in the Leiden Early Arthritis Clinic were studied (n = 563). Yearly radiographs were scored using the Sharp/van der Heijde method (median followup 5 years; maximum followup 9 years). The rate of joint destruction between genotype groups was compared using a linear mixed model, correcting for age, sex, and treatment strategies. A total of 393 anti–citrullinated protein antibody (ACPA)–positive RA patients from the North American Rheumatoid Arthritis Consortium (NARAC) who had radiographic data available were used for the replication study.
The TT and CC/CG genotypes of 2 single-nucleotide polymorphisms, rs4810485 (CD40) and rs42041 (CDK6), respectively, were associated with a higher rate of joint destruction in ACPA-positive RA patients (P = 0.003 and P = 0.012, respectively), with rs4810485 being significant after Bonferroni correction for multiple testing. The association of the CD40 minor allele with the rate of radiographic progression was replicated in the NARAC cohort (P = 0.021).
A polymorphism in the CD40 locus is associated with the rate of joint destruction in patients with ACPA-positive RA. Our findings provide one of the first non–HLA-related genetic severity factors that has been replicated.
Rheumatoid arthritis (RA) is characterized by inflammatory arthritis and localized destruction of bone and cartilage. The severity of joint destruction is highly variable between patients and, according to the findings of twin studies, substantially influenced by genetic factors (1). Nevertheless, the precise contribution of genetic factors has yet to be determined. To date, only a small number of genetic risk factors have been identified, and apart from HLA, the association of none of these factors with RA severity has been convincingly replicated.
In contrast, the genetics of susceptibility to RA has been considerably boosted, largely due to the findings of genome-wide association studies. In addition to the HLA–DRB1 shared epitope alleles, several new susceptibility factors have been identified, and their association with RA has been independently replicated: PTPN22, TRAF1/C5, OLIG3/TNFAIP3, and STAT4. Intriguingly, for many of these genetic risk factors, the associations are confined to RA patients positive for anti–citrullinated protein antibody (ACPA). It remains unknown whether genetic factors also affect the severity of joint destruction in ACPA-positive and ACPA-negative RA in different ways. Nonetheless, compelling evidence demonstrates that ACPA-positive RA patients have a more destructive disease course than do ACPA-negative RA patients.
A recent meta-analysis of 2 genome-wide association studies identified 6 new risk loci (rs4810485 [CD40], rs1678542 [KIF5A/PIP4K2C], rs42041 [CDK6], rs2812378 [CCL21], rs4750316 [PRKCQ], and rs3890745 [MMEL1/TNFRSF14]) as susceptibility factors for autoantibody-positive RA (2). In the present study, we investigated the association between these single-nucleotide polymorphisms (SNPs) and the rate of radiologic joint destruction in RA patients and in ACPA-positive RA in particular using a large longitudinal cohort. A cohort of ACPA-positive RA patients was used for replication analysis. We found that a genetic variant in the CD40 gene is associated with the rate of joint destruction in ACPA-positive RA patients.
We studied 563 RA patients who were consecutively included in the Leiden Early Arthritis Clinic (EAC) cohort between 1993 and 2006 and who had both DNA and radiographs available. The RA patients fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 criteria for classification of RA (3). Followup visits were performed yearly. Treatment strategies changed over time and differed during the different periods of inclusion into the cohort (before 1996, 1996–1998, 1999–2001, and after 2001) (for a detailed description of the EAC cohort, see ref. 4). Anti–cyclic citrullinated peptide 2 (anti–CCP-2) antibodies were measured using stored serum samples that had been obtained at baseline (Immunoscan RA Mark 2; Euro-Diagnostica, Arnhem, The Netherlands).
For the replication analysis, we studied 393 ACPA-positive RA patients who were included in the North American Rheumatoid Arthritis Consortium (NARAC) and who had hand radiographs available. Since the radiographs had been obtained at different disease durations, the estimated radiologic progression per year was determined by dividing the total Sharp/van der Heijde score for the hands (5) by the disease duration at the time the radiograph was obtained.
The 6 recently identified risk loci (2) were genotyped in the 563 RA patients from the Leiden EAC cohort using allele-specific kinetic polymerase chain reaction analysis, as previously described (6). The data were manually curated by one of us (FASK), who had no knowledge of the clinical characteristics of the study patient, before statistical analysis. There was a 98% genotyping success rate; previous analyses suggested a genotyping accuracy of >99%. For the MMEL1/TNFRSF14 locus, a perfect proxy for rs3890745 (as reported in ref. 2) was used (rs6684865; r2 = 1).
In the NARAC cohort, genotyping was performed using the Illumina HapMap500 BeadChip (Illumina, San Diego, CA), as described previously (7). SNP rs4810485 was not typed in the whole-genome study, but a perfect proxy for this variant was genotyped (rs1569723; r2 = 1). For CDK6, neither rs42041 nor a perfect proxy was genotyped; therefore, the data on rs42041 were imputed as described elsewhere (2).
In the EAC cohort, radiographs of the hands and feet, which had been obtained over consecutive years, were scored according to the Sharp/van der Heijde method (5). To encompass a reliable sample size, radiographic followup data were restricted to a maximum of 9 years (median 5 years). All radiographs were scored by a single experienced scorer (MPMvdL) who was blinded with respect to the clinical and genetic data. A total of 499 radiographs were rescored (149 baseline radiographs and 350 radiographs obtained during followup from 60 randomly selected RA patients). The intraclass correlation coefficient (ICC) was 0.91 for all radiographs, 0.84 for baseline radiographs, and 0.97 for the radiographic progression rate.
In the NARAC cohort, the radiographs were scored by a single reader (JvN), who was blinded to the clinical and genetic data. A total of 25% of the radiographs were rescored, and the ICC was 0.99.
Analyses were performed using SPSS version 16.0 software (SPSS, Chicago, IL). Since the radiographic data were not normally distributed, the raw data for the Sharp/van der Heijde scores are presented as medians and were log-transformed in preparation for analysis. In the EAC cohort, a linear model for longitudinal data was used to compare progression rates between groups. Age, sex, and inclusion period (a proxy for treatment strategy), as well as their interactions with time, were entered in the model to correct for possible confounding effects (see below). Since 6 SNPs were evaluated, Bonferroni correction for multiple testing was applied; the P value for significance was set at P < 0.008. Only the SNPs that were clearly related to the progression rate in the EAC cohort were analyzed in the replication cohort.
In the NARAC cohort, the estimated radiologic progression per year was compared using the Kruskal-Wallis test. No corrections were made for age, sex, or treatment in this cohort.
To take advantage of the prospective character of the data from the EAC cohort, which consisted of repeated measurements, and to avoid multiple testing by performing statistical tests at each time point, a linear model for the longitudinal data was used, with the log-transformed Sharp score as the response variable, to compare the radiologic progression rates between genotype groups. Different correlation structures between the repeated measurements were explored, and based on Akaike’s information criterion, an autoregressive correlation structure with heterogeneous variances was chosen. Due to the study design (an inception cohort), not all patients achieved a similar duration of followup. The model takes missing observations into account, assuming that the observation is missing at random.
Differences in progression rates between the different genotypes were tested by considering the significance of the interaction between genotype and time, with time as a linear covariate. Age, sex, and inclusion period (before 1996, 1996–1998, 1999–2001, and after 2001), as well as their interactions with time, were entered in the model to correct for possible confounding effects. To prevent overfitting of the data, no corrections were applied for other variables. Inclusion period is a proxy for treatment modalities, because treatment strategies improved over time, and an influence of the treatment strategy on the progression of radiographic joint damage was observed previously, as well as in the present study. The following treatment strategies were applied in the subsequent inclusion periods. Patients included between 1993 and 1995 were treated initially with analgesics and subsequently with chloroquine or sulfasalazine if they had persistent active disease (delayed treatment). From 1996 to 1998, RA patients were promptly treated with either chloroquine or sulfasalazine (early treatment) (4). From 1998 to 2002, patients were promptly treated with either sulfasalazine or methotrexate (early treatment), and patients included during 2002 or later were promptly treated with either sulfasalazine or methotrexate, combined with treatment adjustments based on the disease activity (early and disease activity–based treatment).
The baseline characteristics of the RA patients in the EAC and NARAC cohorts are shown in Table 1. In the EAC cohort, the minor allele frequencies were 0.242, 0.340, 0.267, 0.366, 0.204, and 0.307 for rs4810485, rs1678542, rs42041, rs2812378, rs4750316, and rs6684865, respectively. These frequencies are consistent with previous results (2).
The raw data for the Sharp/van der Heijde scores in EAC cohort patients with each of the 3 genotypes at each SNP are depicted in Figure 1. To study the influence of the SNPs on the rate of joint destruction, a linear mixed model analysis was performed for each SNP. For rs4810485 (CD40), the GG and GT genotypes showed comparable radiographic scores; therefore, the genotype data were combined, and a carriership analysis was performed. Similarly, the CC and CG genotypes of rs42041 (CDK6) were pooled. In the total group of RA patients, an association was observed for rs42041 (CDK6) (P = 0.033). For the other SNPs, no significant association with radiologic progression over time was detected (P = 0.268, P = 0.369, P = 0.679, P = 0.583, and P = 0.451 for rs4810485, rs1678542, rs2812378, rs4750316, and rs6684865, respectively).
Because the genetic regions studied have thus far been observed to be susceptibility factors only for autoantibody-positive RA patients, the analyses were repeated in the ACPA-positive subgroup from the EAC cohort. In these analyses, 2 polymorphisms, rs4810485 (CD40) and rs42041 (CDK6), were shown to affect the rate of joint destruction (Figure 2). For rs4810485, the G allele was associated with a lower progression rate (GG/GT versus TT; P = 0.003). Back transforming the regression coefficient of the genotype in the model to the original scale yielded a 1.12 times greater increase in the Sharp score per year (95% confidence interval [95% CI] 1.04–1.21) for carriage of the risk genotype. For rs42041, the C allele was associated with a higher rate of joint destruction (CC/CG versus GG; P = 0.012). For carriership of the C allele, we observed a 1.09 times greater increase in the Sharp score per year (95% CI 1.02–1.16). After correction for multiple testing, only the association with rs4810485 was statistically significant. The interaction between inclusion period and time was significant in all 6 analyses (P < 0.001), demonstrating the effect of inclusion period on the radiologic progression rate. Neither sex nor age was independently associated with progression.
In an effort to replicate the findings, the effect of CD40 and CDK6 on radiologic progression was analyzed in 393 ACPA-positive RA patients from the NARAC cohort. Using a perfect proxy for rs4810485, the genotype associated with severity in the EAC cohort also revealed a higher estimated radiologic progression rate per year in the NARAC cohort: 3.40 Sharp units/year (n = 23) versus 2.83 and 1.83 Sharp units/year (n = 122 and n = 248, respectively; P = 0.021). Using imputed data for rs42041, no significant differences between the 3 genotypes were observed (2.76, 2.38, and 2.07 Sharp units/year; n = 32, n = 163, and n = 188, respectively [P = 0.327]). The total number of patients available for analysis of rs42041 was 383; genotyping data were missing in 10 patients.
Although several clinical and serologic risk factors for RA severity are known, thus far, the interindividual variance in joint destruction is insufficiently explained and genetic factors are scarcely investigated. A better comprehension of the factors that mediate joint damage in RA may lead to the development of targeted therapies or may contribute to the prediction of disease outcome in individual RA patients. Most recently, 6 new loci were reported to predispose to autoantibody-positive RA (2). Although susceptibility factors do not necessarily affect disease progression, we investigated whether these 6 SNPs were also risk factors for a severe course of RA, as measured by the rate of joint damage. The present data suggest that 2 SNPs, rs4810485 (CD40) and rs42041 (CDK6), influence the rate of joint destruction in ACPA-positive RA. Of these, only rs4810485 was significantly associated after correction for multiple testing, and the association was replicated in an independent cohort of ACPA-positive RA patients. Thus, CD40 is the first non–HLA-related genetic risk factor for RA severity that has been independently replicated.
A recent study (2) identified a common variant at the CD40 locus (the minor T allele) as being protective against the development of RA. Surprisingly, in our study, the minor T allele associated with a higher rate of joint destruction in 2 cohorts. This finding is counterintuitive if one assumes that genetic variants associating with susceptibility also associate with severity. Although our findings were observed in 2 independent cohorts, and were thus replicated, a type I error cannot be ruled out. The disease-associated (common) allele marks a haplotype of CD40 that contains a polymorphism in the upstream Kozak sequence that results in increased surface expression on B cells (8). To our knowledge, the effect of this haplotype on CD40 surface expression in synovial fibroblasts has not been directly studied. However, CD40 expression is increased on synoviocytes in RA, and triggering of CD40 in synovial fibroblasts is associated with the production of proinflammatory cytokines and osteoclastogenesis (9,10). It is likely that the biologic pathways underlying susceptibility and severity are distinct with respect to the triggering of CD40. This would provide an explanation for the finding that the minor T allele has a protective effect in susceptibility studies but is associated with a more severe disease course. Clearly, it is essential to perform further studies on the mechanisms by which CD40 polymorphisms associate with erosive outcome in RA.
A second SNP, rs42041, tended to associate with the rate of joint damage in RA in the EAC cohort. Lack of replication in the NARAC cohort indicates that the observed association with the progression rate in the EAC cohort cannot be interpreted. Nonetheless, it will be interesting to see the results of other studies analyzing CDK6 and RA severity. Thus, at present, of the 2 SNPs that tended to show an association with the rate of joint destruction, only the genetic variant in CD40 was statistically significant after correction for multiple testing and was replicated in a second cohort; this variant in CD40 is therefore identified as a severity factor for RA.
The other 4 SNPs we examined in the loci encoding for KIF5A/PIP4K2C, CCL21, PRKCQ, and MMEL1/TNFRSF14 were not observed to associate with the severity of joint destruction. Therefore, these polymorphisms appear to be genetic risk factors that are primarily associated with RA susceptibility. Indeed, all of these SNPs were recently replicated as being true susceptible loci in RA patients of European ancestry (11).
The prospective nature of the data from the EAC cohort strengthens the impact of the findings because higher radiologic scores for risk genotypes were present at subsequent time points; as such, the present data set is advantageous in comparison to studies that assessed cross-sectional radiologic data. The fact that a large number of patients with a followup of as long as 9 years were included for analysis is clearly an advantage, but also has a limitation. Inherent in the design of an inception cohort is the fact that not all patients had achieved maximum followup, so the number of missing data that the mixed model had to take into account increased with longer followup. Small numbers of radiographs available at the latest time points are also the most likely explanation for the observed “bump” at the 8-year time point for genotypes GG, CC, and TT of SNPs rs2812378, rs4750316, and rs6684865, respectively (Figure 1).
Evaluation of the effect of genetic factors on the rate of joint destruction during the disease course inevitably implies that other factors that affect the disease course should be taken into consideration as well. Analyses of all 6 SNPs revealed that the inclusion period, a proxy for treatment strategy, was significantly associated with the rate of joint damage, which is consistent with previous results from the EAC cohort (12). The analyses of CD40 and CDK6 showed that these SNPs were associated with joint damage, independently of the treatment strategy. Nevertheless, corrections for treatment strategy were made at the group level and, thus, were an approximation of the real effect of treatment on the rate of joint destruction in individual RA patients.
In conclusion, a polymorphism in the CD40 locus showed a significant association with the rate of joint destruction in ACPA-positive RA patients, a finding that was replicated in an independent cohort. Although further studies are needed to identify the causal variant, the data presented provide a foundation for further investigations of the role of CD40 in joint destruction of RA.
Dr. Gregersen’s work was supported by the NIH (National Institute of Arthritis and Musculoskeletal and Skin Diseases/National Institute of Allergy and Infectious Diseases grants N01-AR-2-2263 and R01-AR-44422). Dr. Toes’ work was supported by the Dutch Arthritis Foundation, the European Commission Seventh Framework Programme (projects Masterswitch and AutoCure), and the Centre for Medical Systems Biology. Dr. van der Helm-van Mil’s work was supported by The Netherlands Organization for Health Research and Development and by the Dutch Arthritis Foundation.
Dr. Begovich owns stock in Celera and stock options in Roche. Dr. Gregersen has received consulting fees, speaking fees, and/or honoraria from Roche (less than $10,000). Ms Chang and Mr. Catanese own stock or stock options in Celera.
AUTHOR CONTRIBUTIONSAll 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. van der Linden 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. Van der Linden, Kurreeman, van Nies, Gregersen, Huizinga, Toes, van der Helm-van Mil.
Acquisition of data. Van der Linden, Kern, Raychaudhuri, Begovich, Catanese, Gregersen, Huizinga, Toes, van der Helm-van Mil.
Analysis and interpretation of data. Van der Linden, Feitsma, le Cessie, Olsson, Raychaudhuri, Chang, Kurreeman, van Nies, van der Heijde, Gregersen, Huizinga, Toes, van der Helm-van Mil.