Article Text
Abstract
Background and objectives: The Fc receptor-like 3 (FCRL3) gene −169T>C single nucleotide polymorphism (SNP) has been reported to be associated with several autoimmune diseases (AIDs) in Japanese populations. However, association results in other populations have been conflicting. Therefore, we investigated this SNP in a Scandinavian panel of AIDs.
Methods: We genotyped patients with rheumatoid arthritis (RA; n = 708), juvenile idiopathic arthritis (JIA; n = 524), systemic lupus erythaematosus (SLE; n = 166), ulcerative colitis (UC; n = 335), primary sclerosing cholangitis (PSC; n = 365), Crohn disease (CD; n = 149), a healthy control group (n = 1030) and 425 trio families with type 1 diabetes (T1D). Statistical analysis consisted of case–control and family-based association tests.
Results: RA was associated with the C allele (odds ratio (OR) = 1.16, 95% CI 1.01 to 1.33) and the CC genotype (OR = 1.30, 95% CI 1.01 to 1.67) of the FCRL3 −169T>C SNP in our material. Suggestive evidence for association was also found for JIA (CC genotype: OR = 1.30, 95% CI 0.99 to 1.70), and clinical subgroup analysis indicated that this was connected to the polyarticular subgroup. No significant association was found with SLE, UC, CD, PSC or T1D. In patients with RA, we found no significant interaction between the FCRL3 −169T>C and PTPN22 1858C>T SNPs, nor between the FCRL3 −169CC genotype and IgM-rheumatoid factor or anti-cyclic citrullinated peptide titre levels.
Conclusion: We found an association between the FCRL3 −169T>C SNP and RA, and suggestive evidence for involvement with JIA, in a Norwegian population. These findings lend support for a role for this SNP in RA across ethnically diverse populations, and warrant follow-up studies in JIA.
Statistics from Altmetric.com
Common autoimmune diseases (AIDs), including several rheumatic diseases, develop as a result of numerous predisposing genetic and triggering environmental factors.1 Genome-wide linkage scans2 and/or association studies have revealed genetic susceptibility regions shared between multiple AIDs, most notably the human leukocyte antigen (HLA) complex3 and the CTLA4 and PTPN22 genes.4
Recently, the gene encoding Fc receptor-like 3 (FCRL3) was proposed as a novel autoimmune predisposing factor.5 6 A single nucleotide polymorphism (SNP) located in the promoter region (−169T>C) of this gene was reported associated with rheumatoid arthritis (RA), systemic lupus erythaematosus (SLE), Graves disease (GD) and Hashimoto thyroiditis (HT) in a Japanese population.6 The strongest association was reported with RA assuming a recessive model (CC genotype vs CT and TT, odds ratio (OR) = 2.15, 95% CI 1.58 to 2.93). Functional studies suggested that the disease-susceptible C allele increases the binding affinity for the transcription factor nuclear factor-κB, and causes increased transcription of FCRL3 in B cells. Moreover, the C allele was significantly associated with augmented autoantibody production in patients with RA. The association with RA has been replicated in independent Japanese populations.6 7 However, results from studies in other populations and AIDs have been conflicting, where some have been positive8–12 and others negative.11 13–20 Hence, more studies in different populations and diseases are needed to clarify the role of the FCRL3 −169T>C polymorphism in AID.
For this reason, we examined the FCRL3 −169T>C SNP in a Scandinavian panel of AIDs, including RA, juvenile idiopathic arthritis (JIA), SLE, ulcerative colitis (UC), Crohn disease (CD), primary sclerosing cholangitis (PSC), and type 1 diabetes (T1D). JIA and PSC have not previously been investigated with respect to this SNP.
PATIENTS AND METHODS
Study populations
The panel of AIDs consisted of patients with JIA (n = 524), RA (n = 708), SLE (n = 166), UC (n = 335), PSC (n = 365), CD (n = 149) and 425 trio families with T1D. Blood donors recruited at the Department of Immunology and Transfusion Medicine, Ullevål University Hospital, Oslo, Norway were used as healthy controls (n = 1030).
Patients with RA and SLE were classified according to the American College of Rheumatology (ACR) criteria.21 22 Patients with RA from the Norwegian part of the European Research on Incapacitating Disease and Social Support (EURIDISS) cohort (n = 215) had clinical data available, and among these patients 47.4% and 60.5% were positive (levels above 25 U/ml) for RF or anti-cyclic citrullinated peptide (CCP), respectively. Analyses of RF and anti-CCP have been described elsewhere.23 Patients and clinical subgroups of JIA (npolyarthritis = 127; noligoarthritis = 215; nenthesis-related arthritis = 72; npsoriatic arthritis = 43; nundifferentiated arthritis = 38; nsystemic arthritis = 29) were classified according to the International League of Associations for Rheumatology (ILAR) criteria.24 Stratification for IgM-rheumatoid factor (RF) status in patients with JIA was not performed, as only 24 patients (5%) were positive.
Patients with inflammatory bowel disease (IBD) were recruited through the Inflammatory Bowel in South-Eastern Norway (IBSEN) study.25 26 Diagnosis and classification (UC or CD) were based on standard endoscopic and histological criteria.27 Patients with PSC were recruited on admission to the Medical Department, Rikshospitalet HF, Oslo, Norway (n = 232) and Division of Gastroenterology and Hepatology, Huddinge University Hospital, Stockholm, Sweden (n = 133). Diagnosis of PSC was based on accepted criteria and typical findings on endoscopic retrograde cholangiography.28 79.5% of the patients with PSC had a concurrent diagnose of IBD (nUC = 236; nCD = 32; nindeterminate colitis = 22).
T1D families were recruited as described elsewhere29 and patients diagnosed before the age of 15 years according to the EURODIAB criteria.30
All patients, families and controls were of Norwegian origin, except for the PSC material, which also included Swedish patients (n = 133). The study was approved by the Regional Committees for Research Ethics.
Statistical power
A priori statistical power was calculated using the Genetic Power Calculator,31 based on the association with RA reported by Kochi et al (recessive model, allele C frequency = 0.345, type I error rate 0.05).6 In our data sets we obtained a power ranging from 76% to >99% given an OR of 2.15 for the CC genotype (CD 76%, SLE 80%, T1D 91%, UC 96%, PSC 97%, JIA and RA >99%), or ranging from 30% to 74% given the lower CI limit of OR = 1.58.
Genotyping
Genotyping was performed on whole genome amplified (WGA) DNA (GenomiPhi DNA Amplification Kit, GE Healthcare, Piscataway, New Jersey, USA) except for the control, SLE and parts of the RA materials, which were genotyped on genomic DNA. WGA DNA has been thoroughly validated for SNP genotyping.32 Allelic discrimination of the FCRL3 −169T>C (rs7528684) and PTPN22 1858C>T (rs2476601) SNPs were carried out using TaqMan SNP Genotyping Assays (ID: C___1741825_1 from Applied Biosystems, Foster City, California, USA, and as described in Viken et al,33 respectively).
Accuracy of genotyping
The parents (n = 798) in the T1D data set showed significant deviation from Hardy–Weinberg equilibrium (HWE), with a decrease in observed heterozygosity compared to the expected frequency (p = 0.0014). Although such deviation could indicate genotyping errors,34 only two Mendelian inconsistencies were observed (identified using PedCheck).35 Moreover, no significant deviation from HWE was observed in our case–control materials (p>0.05), allelic discrimination by clustering of genotypes did not reveal any anomalies and call rates were >95% for all populations. Genotyping was therefore assumed to be accurate.
Statistical analysis
HWE was tested using the exact test (with the founders only option for the family material) implemented in the program Pedstats.36
Association analyses in the case–control data sets were performed in EpiCalc 2000 V.1.02 (http://www.brixtonhealth.com/epicalc.html), using the Pearson χ2 test, or Fisher exact test when any of the cell counts were low (<5). Association tests in the family data were performed using the transmission–disequilibrium test (TDT) in the TDTPHASE application of the UNPHASED V.3.0 program package,37 after removal of families (n = 2) with Mendelian inconsistencies. The TDT statistic is not affected by the observed deviation from HWE in the family data, as it measures differences in transmission and non-transmission of alleles from heterozygous parents to affected offspring, which is independent of population structure.38
Statistical interaction between the FCRL3 −169T>C and PTPN22 1858C>T polymorphisms was tested using genotype stratification and the Breslow–Day test in SPSS V.14.0 (SPSS, Chicago, Illinois, USA), and the epistasis test in PLINK v0.99p.39 RF and anti-CCP titre levels were tested with SPSS, using the Mann–Whitney U test on a recessive genetic model. Genotypes were also stratified according to RF or anti-CCP positivity and tested using Pearson χ2 test in EpiCalc 200.
As this is a replication study, no correction for testing of multiple diseases was performed. However, in the subgroup analysis of JIA, we applied Bonferroni correction for the number of subgroups tested.
RESULTS
The distribution of genotypes and allele frequencies among the AID case–control materials, and results of the TDT in the T1D families are presented in table 1 and table 2, respectively.
We found a significant association between RA and the C allele and CC genotype of the FCRL3 −169T>C SNP (C allele: OR = 1.16, 95% CI 1.01 to 1.33; CC genotype: OR = 1.30, 95% CI 1.01 to 1.67). A tendency for association was also observed for JIA at the genotype level, although results were not significant (OR = 1.30, 95% CI 0.99 to 1.70). However, as JIA is a clinically heterogeneous disease, we stratified for different subgroups (table 3). We found a significant association with polyarticular JIA at the allele and genotype level (C allele: OR = 1.30, 95% CI 1.00 to 1.70; CC genotype: OR = 1.67, 95% CI 1.08 to 2.60), but not with the other subgroups. However, these results were not significant after Bonferroni correction for the number of subgroups tested (p = 0.29 for the C allele and p = 0.12 for the CC genotype). No significant disease associations were detected for SLE, UC, CD, PSC or T1D (tables 1 and 2).
Concerning RF or anti-CCP, neither positivity (OR = 1.36, 95% CI 0.65 to 2.88 and OR = 1.99, 95% CI 0.88 to 4.54 for RF+ and anti-CCP+, respectively) nor titre levels (median (SD) for RF: CC: 34 (107); CT+TT: 21 (93); p = 0.36; for anti-CCP: CC: 73 (103); CT+TT: 54 (103); p = 0.41) were significantly associated with the FCRL3 −169CC risk genotype in our material (n = 206).
A recent report indicated an interaction between the FCRL3 −169T>C SNP and another RA-associated variant, the PTPN22 1858C>T SNP.8 Association between the PTPN22 1858C>T SNP and our patients with RA has been reported earlier,23 33 and was also seen when comparing the patients with the control material used in this study (TT/TC/CC for patients with RA 2/28/69%; for controls 1/19/80%; genotype TT+CT vs CC: OR = 1.73, 95% CI 1.40 to 2.14). When stratifying FCRL3 −169T>C genotypes on PTPN22 1858C>T risk or non-risk genotypes, and vice versa, the ORs for either polymorphism show a tendency towards an increased risk in the presence of risk genotypes at the other locus (table 4). However, the ORs in the risk and non-risk groups did not show significant heterogeneity (p = 0.65 by Breslow–Day test), and an epistasis test using logistic regression was negative (p = 0.095).
DISCUSSION
In this paper, we found an association between the C allele and CC genotype and RA, thus supporting the findings in Japanese, Canadian and Dutch RA populations.6–9 In addition, we found suggestive, novel evidence for involvement with JIA at the allele and genotype level, in particular for the polyarticular subgroup. Interestingly, this is also the JIA subgroup with the most clinical characteristics in common with adult RA.40 However, especially for the enthesitis-related, psoriatic, systemic and undifferentiated JIA subgroups, stratification resulted in very small study populations, and the lack of significant findings in these subgroups is therefore far from conclusive. Hence, it is difficult to say whether the observed result reflects a relationship with adult RA, whether clinical features specific for JIA are involved, or if this is a chance finding. These results should therefore be followed up in other, preferably larger and clinically well characterised, JIA populations.
The evidence for association between this SNP and RA has been conflicting, with negative results in some of the recent replication attempts in Caucasian (Spanish, North-American and UK) and Korean RA populations.13–16 However, given the propensity for first reports to overestimate the effect of a genetic association,41 the OR is likely to be substantially lower than reported in the initial study by Kochi and colleagues.6 This is also indicated in three recent meta-analyses, giving joint ORs between 1.03 and 1.36.5 9 42 A risk factor of such size would mean that most, if not all, of the studies conducted thus far have been under-powered.
In addition there may be more subtle effects, such as environmental or genetic heterogeneity between the study populations, also among those defined as Caucasian. Related to this is the possibility that the FCRL3 −169T>C SNP does not constitute the real aetiological locus, but instead acts as a proxy. For instance, in the original SNP screen performed by Kochi et al, the peak of association with RA included three more SNPs in the FCRL3 gene (rs11264799, rs945635 and rs3761959), which were in strong linkage disequilibrium (LD) with the −169T>C SNP.6 The association with rs945635 (exon 2) has been replicated in a Dutch RA study,9 and rs3761959 (intron 3) was reported associated with Graves disease in two recent studies.12 43 Therefore, it is possible that imperfect and possibly population specific LD between one of these SNPs, or a hitherto unidentified primary locus, and the FCRL3 −169T>C SNP contributes to the conflicting results.
An alternative explanation was proposed by Newman et al, after reporting an interaction between the FCRL3 −169T>C SNP and another RA risk variant, the PTPN22 1858C>T SNP.8 The interaction was indicated by an apparent strengthening of the association between RA and either SNP in the absence of risk alleles at the other locus. Because the allele frequencies of the PTPN22 1858C>T SNP is known to vary widely among populations,44 the effect of an interaction would also be population dependent, possibly confounding the effect of the FCRL3 −169T>C SNP. However, neither we nor the authors of three recent reports were able to find statistically significant evidence for such an interaction.9 15 42 In our material, the ORs after stratification actually show the opposite tendency to that reported by Newman et al, ie, an increased risk in the presence of risk alleles at the other locus. Moreover, the original report by Newman et al may be questioned: firstly, no explicit test of genetic interaction was performed. Secondly, when testing the original data (stratified in a similar fashion as in table 4 in this paper), we found only a marginally significant difference between the ORs in high and low risk strata (p = 0.054 by Breslow–Day test). Therefore, if any joint effect of these two SNPs is present, it is small.
A similar approach was taken in two recent Spanish studies of patients with RA and UC, respectively, where evidence of association was presented with RA when stratified for genotypes of a promoter polymorphism in the NF-κB1 gene (-94ins/delATTG),13 and with UC when stratified for the presence of the HLA-DRB1*0103 allele.20 However, none of these studies showed significant associations with the FCRL3 –169T>C SNP alone, and the modes of genetic interaction were unclear. Still, interactions with other aetiological loci as a source of heterogeneity represent an interesting proposition, which should be explored further.
Whether the FCRL3 −169C allele also influences the autoantibody production in patients with RA, as suggested by Kochi et al, remains an open question. Although we found no significant association between the CC risk genotype and either presence or levels of RF or anti-CCP, the number of patients in our material that had clinical data available for such analyses was relatively small. As the tendency points in the same direction as that seen by Kochi et al, the lack of significant findings may therefore be due to insufficient power. However, so far, none of the replication studies showing positive associations with RA have been able to confirm the correlation with autoantibodies reported by Kochi et al.7–9
No significant association was found with the other AIDs investigated in our panel, including PSC, which have not been investigated for the FCRL3 −169T>C SNP previously. For SLE, the study by Kochi et al remains the only positive study so far,6 with negative results in Korean and Spanish populations,16 17 although a positive tendency was observed in the Spanish study. T1D, UC and CD have shown concurring negative results in other, Caucasian populations.11 18–20 By contrast, positive associations have been seen in GD,6 12 HT,6 and recently also in autoimmune Addison disease and multiple sclerosis (although with the opposite allele (T)).10 11 Moreover, limited statistical power could contribute to lack of associations in our AID panel, given a risk impact less than that assumed a priori. Even so, the accumulated evidence so far does not support the proposition of the FCRL3 −169T>C SNP as a factor common to all AIDs.
In conclusion, our study supports the association with RA in Caucasians and suggests an involvement of the FCRL3 −169T>C SNP also in JIA, in particular with the polyarticular subgroup. However, the impact of this polymorphism appears to be smaller than implicated by the original Japanese RA study, which could explain the conflicting results seen in other studies.
Acknowledgments
The authors wish to thank Vibke Lilleby and Inge M Gilboe for collecting the SLE samples, and Ulrika Broomé and Erik Schrumpf for collecting the Swedish and Norwegian PSC samples, respectively. We also thank the Department of Immunology and Transfusion Medicine at Ullevål University Hospital for making the control material available, and Beate Skinningsrud and Kristina Gervin for graciously contributing FCRL3 −169T>C and PTPN22 1858C>T control genotypes. Marita Olsson was of great assistance with statistical questions.
REFERENCES
Footnotes
Funding: This study was supported by the Juvenile Diabetes Foundation International (1-2004-793), the Norwegian Diabetes Association, Rikshospitalet HF and the Southern and Eastern Norway Regional Health Authorities.
Competing interests: None.
Ethics approval: The study was approved by the Regional Committees for Research Ethics.