Rheumatoid arthritis is one of the most common autoimmune diseases in Western countries, with an estimated prevalence of 1% in the Spanish population. Recent advances in its genetic basis have been fuelled by an ongoing Japanese genomewide study, which has described several causative polymorphisms in the past few years.^1–3 However, some of these outstanding findings have been difficult to replicate in European populations, as has recently been shown by our group and by others with regard to the gene peptidylarginine deiminase 4 (PADI4),^4–6 the haematopoietic isoform of the citrullinating enzyme. SLC22A4, an organic cation transporter gene, is another example of lack of association in Caucasian populations of a gene previously shown by the Japanese group^7,8 to be associated with the disease. The causes underlying such discrepancies may vary, but arguably they include both distinct environmental inputs and different genetic structure. The purely genetic reasons for a failure to replicate the association of a polymorphism described as aetiological are many. Firstly, the polymorphism studied may be found in another population in such a low frequency that a significant difference may be difficult to achieve. Secondly, the supposedly aetiological polymorphism may not be really aetiological, but instead it may act as a marker of a nearby causative polymorphism. The associations between different polymorphisms in the same or nearby genes (linkage disequilibrium) may vary widely among populations. Finally, a true aetiological polymorphism may act only on a defined genetic background, and therefore discloses itself only in specific populations. A similar situation has been reported repeatedly in mice, where a defined inactivating mutation may yield different phenotypic outcomes, depending on genetic background.⁹

A recent genetic report from the Yamamoto group³ has described the association of four single-nucleotide polymorphisms (SNPs) located in the Fc receptor-like 3 (FCRL3) gene with an increased susceptibility to rheumatoid arthritis and several other autoimmune diseases, thereby raising the possibility of FCRL3 being a general autoimmune susceptibility factor. FCRL3, also termed FCRh3, is one of a family of six related genes located on chromosome 1q21–23, sharing homologies with other classic Fc receptor genes located in the same cluster. These previously known classic genes code for receptors for the constant portion (Fc) of the immunoglobulin molecules. The Fc receptors have relevant roles in diverse aspects of immunobiology.¹⁰ This fact, along with their polymorphic nature, led long ago to their consideration as natural candidates to be susceptibility genes in several autoimmune inflammatory conditions. The region 1q21–23 has been reported to be linked to systemic lupus erythematosus in familial linkage studies with several affected members, but it has not been detected in a recent study on rheumatoid arthritis (interestingly, in the Japanese population) using microsatellites,¹¹ or in previous genomewide scans in populations of European ancestry. Kochi et al³ examined the region with high-density SNP mapping and found a peak of association not in the classic genes, but on FCRL3, a family member expressed on some subsets of B cells. Although they were cautious enough not to discard fully other genes located in the vicinity, the main candidate they propose is clearly FCRL3 itself. They suggest that the −169C susceptibility allele binds the transcription factor nuclear factor κB (NFκB) more strongly than the non-susceptible −169T allele, thereby enhancing the transcription rate of the FCRL3 gene.

Although some recent research in the Spanish population has been undertaken to ascertain the role of the classic Fc receptor genes in rheumatoid arthritis,¹² we have tried to replicate directly the Japanese findings, as their functional studies convincingly showed that the polymorphism may be mechanistically relevant to the pathological process. We also looked for possible genetic interactions between the polymorphisms in the FCRL3 promoter and the functional polymorphism in the promoter of NFKB1, coding for the p105 subunit of NFκB. This p105 subunit is proteolytically activated on cell stimulation, yielding the p50 subunit, a component (in combination with c-Rel) of the active transcription factor binding to the −169 FCRL3 promoter polymorphism.³

METHODS

Our study included 734 white Spanish patients with rheumatoid arthritis, consecutively recruited in two hospitals in Madrid (n = 489) and one hospital in Granada (n = 245). The diagnosis was established on the basis of the 1987 American College of Rheumatologists (formerly ARA) criteria.¹³ Table 1 shows their main clinical characteristics. Most of these patients have been included in previous studies from our groups. The written consent of the participants was obtained according to the Declaration of Helsinki. The control group included 736 people, 489 from Madrid and 247 from Granada: white Spanish blood donors and healthy laboratory staff and students. Although the blood donors were not specifically asked about the presence of rheumatoid arthritis, the 1% frequency of this disease in the Spanish population precludes this fact from being a major concern. The ethics committees of the participating hospitals (Clinico San Carlos, La Paz, Virgen de las Nieves, Spain) approved this study.

The FCRL3 polymorphisms were analysed using the C_1741825_10 (FCRL3 −169, rs7528684, in Madrid only), C_1741826_10 (FCRL3 −110, rs11264799) and C_2741972_10 (FCRL3 intron 3, rs1537947, in Granada only) Assays-on-Demand of Applied Biosystems (Foster City, California, USA), according to the conditions recommended by the manufacturer. In all, 94 samples were genotyped for both the −169 SNP and the polymorphism in the intron 3 of the FCRL3 gene. As described previously (see supplementary table in the Japanese study³), the concordance was almost total (>97%; FCRL3 −169T is equivalent to FCRL3_6G, and FCRL3 −169C is equivalent to FCRL3_6A), and it allowed us to consider them as quasi-equivalent markers. They are both referred to as −169SNP hereafter, for simplicity. Alleles were discriminated in an ABI 7900 Sequence Detector (Applied Biosystems). The −94ins/delATTG NFκB1 promoter polymorphism was analysed as explained previously.¹⁴

The presence of the shared epitope (defined as DRB1*0101, DRB1*0102, DRB1*0401, DRB1*0404, DRB1*0405, DRB1*0408 or DRB1*1001 alleles) was ascertained as described elsewhere.^15,16 The shared epitope is a common stretch of amino acids present in human leucocyte antigen (HLA)-DRB1 alleles associated with the disease, and it is the strongest rheumatoid arthritis susceptibility factor identified.

Genotype frequencies (3×2 contingency tables), carriage rates (2×2 tables) and allelic frequencies (2×2 tables) were compared with the χ² statistics and p<0.05 was considered to be significant. The p values were calculated with a standard free software package (Epi Info V.2000, CDC, Atlanta, Georgia, USA). Statistical power (indicated below each table as the minimum detectable relative risk (RR) value for p = 0.05 and a statistical power of 80%) was calculated with the online calculator at http://calculators.stat.ucla.edu/powercalc/binomial/case-control/index.php.

RESULTS

SNPs located in the FCRL3 region (the ones identified as primary sources of the association signal described in the Japanese study³) were examined in the Spanish population. The Madrid and Granada populations were very similar in terms of their genotypic distributions. We found no evidence of phenotypic, genotypic or allelic association with rheumatoid arthritis (tables 2 and 3), even when results from both populations were pooled. Shared-epitope stratification did not improve the significance. We could not confirm the association described in the Japanese population.

Table 4 shows the allelic distribution of the −94ins/delATTG NFKB1 promoter polymorphism found in the promoter region of the NFKB1 gene. Data from Granada have been published previously¹⁴ and they are included here for completeness, with only minor modifications. No differences were found when patients and controls from each separate location were compared. When controls from Madrid were compared with controls from Granada, we found no differences, showing that the two populations are homogeneous. Similarly, when patients with rheumatoid arthritis from Madrid were compared with those from Granada, we found no differences. This prompted us to pool both populations, to improve statistical power, but again no differences were found between patients with rheumatoid arthritis and controls. No differences were apparent when patients were stratified according to the number of shared epitope alleles carried (data not shown).

As the functional interaction between NFκB1 and FCRL3 is proved, we decided to investigate whether any genotype combination showed an altered pattern when patients were compared with controls. Table 5 shows the compound genotypes observed in patients with rheumatoid arthritis and controls. Among patients heterozygotic for the −94ins/delATTG NFκB1 promoter polymorphism, there is a skewed distribution of the −169 polymorphism (p = 0.001, corrected by three different NFκB1 genotypes, p_c = 0.003). Similarly to the Japanese population, the less common allele was over-represented in patients with rheumatoid arthritis when compared with controls.

DISCUSSION

We report that the association of the promoter polymorphism at position −169 on the FCRL3 gene depends on the NFκB1 genotype, thereby suggesting that a genetic and functional link between both genes is relevant in onset of rheumatoid arthritis. This is to our knowledge the first attempt to replicate the FCRL3 association reported by the Yamamoto group.³ In contrast with that report, our results show no association of FCRL3 polymorphisms when considered alone.

The previous Japanese report did not include the polymorphism on the NFκB1 gene in its analysis. However, their results were highly relevant per se. Our finding of an interaction between both genes seems to reconcile their results with the negative results we obtained when FCRL3 alone is considered. If the polymorphisms are indeed aetiological this lack of association is otherwise problematic to explain, as the allelic frequencies in our population are by no means low enough to cause problems related to statistical power.

A genetic cooperation between distinct loci is not new in the genetics of rheumatoid arthritis. The relevance of rheumatoid arthritis susceptibility of mutations affecting an intronic RUNX1 binding site on the SLC22A4 gene (an organic cation transporter) and of mutations in the RUNX1 gene itself has been recently described.^1,17 A few years ago, we also described the protection against rheumatoid arthritis afforded by major histocompatibility complex class I chain-related gene A alleles in the presence of the shared epitope at the HLA-DRB1 locus.¹⁵

A functional hypothesis explaining our results is not straightforward. The ins/del polymorphism in the NFκB1 promoter has been related to transcriptional activity of the gene encoding this transcription factor,¹⁸ fundamental in orchestrating immune and inflammatory reactions. But this differential activity is probably not the sole cause behind the results we have presented, because a specific association of FCRL3 genotypes in NFκB1 heterozygotes (as opposed to homozygotes or carriers of any allele) is difficult to interpret. There have been a few reports in the literature of the effect of heterozygous genotypes on susceptibility to autoimmune diseases. An IL4 polymorphism has been described as negatively associated with multiple sclerosis in heterozygosis.¹⁹ The authors suggest that perhaps the association could result from linkage disequilibrium with distant markers. This situation is reminiscent of the one found at the HLA-DR locus in coeliac disease. The association of DR7-DR5 heterozygotes with the disease stems from the fact that DR7 is strongly linked in our population to DQB1*02, and DR5 is almost invariably linked to DQA1*05. These two alleles, located in distinct but nearby genes, are the primary susceptibility factors in coeliac disease, and DR is acting only as a linkage disequilibrium marker. A speculative scenario would be that a specific genetic configuration at the NFκB1 locus confers the ability to become a susceptibility factor to variations found in FCRL3. And perhaps what is different between our population and the Japanese population is the distribution of genetic configurations in the NFκB1 locus.

In summary, our results show complex genetic interactions between different genes to determine the final outcome: susceptibility to rheumatoid arthritis. Further studies are necessary to expand this knowledge and to delineate the genetic susceptibility pathways operating in distinct populations.