Rheumatoid arthritis (RA) is a heterogeneous disease characterised by a chronic, fluctuating, peripheral, symmetrical and erosive polyarthritis. It has been reported throughout the world, with a prevalence rate of approximately 1% in most populations 1. Persistent synovial inflammation leads to progressive joint destruction, which in turn produces deformity and significant disability. In addition, RA is a systemic disease and some patients develop subcutaneous rheumatoid nodules, secondary Sjögren's syndrome, episcleritis and scleritis, interstitial lung disease, pericardial involvement, systemic vasculitis and Felty's syndrome. These extra-articular manifestations of RA appear to be rare in the absence of rheumatoid factor (RF), and IgG RF titres correlate with articular disease severity and with the extra-articular manifestations of RA 2. In addition to RF, many other IgG autoantibodies are found in RA, most notably anti-cyclic citrullinated peptide 3 and anti-type II collagen antibodies 4.

The Fcγ receptors (FcγRs), which bind these IgG autoantibodies and IgG-containing immune complexes, have been shown to play important roles in the initiation and regulation of many immunological and inflammatory processes. In humans, there are three classes of FcγRs (I-III) encoded by eight genes, which produce at least 15 different membrane-bound and soluble isoforms that vary in their cellular distribution and affinity for different IgG isotypes. This molecular and expression diversity restricts specific biological properties to certain cell types. Activating (FcγRIIa, FcγRIIIa and FcγRIIIb) and inhibitory (FcγRIIb) FcγRs are frequently coexpressed on the same cell, thus providing a means for regulating signalling thresholds 56. Furthermore, the absolute level of receptor expression is modulated by proinflammatory and anti-inflammatory cytokines 7. Activating functions include uptake and clearance of immune complexes (complement dependent and independent mechanisms), activation of phagocytes (trigger the oxidative burst, cytotoxic granule and cytokine release), antigen presentation and antibody-dependent cellular cytotoxicity 6. Conversely, FcγRIIb contains an inhibitory motif in the cytoplasmic tail and abrogates cellular activation. FcγRIIb may also play a role in maintaining peripheral B cell tolerance and prevention of autoimmunity 8. Single and multiple FcγR knockout mouse models have demonstrated that the balance between activating and inhibitory FcγRs influences the development of both immune complex-mediated and collagen-induced arthritis 910.

Polymorphic variants that increase the expression or affinity of these IgG receptors or that enhance their ability to bind specific IgG isotypes may therefore play an important role in determining the severity and persistence of inflammation to IgG (auto)antibodies and immune complexes in RA. Our previous studies have supported an association between FCGR3A and RA. The higher affinity FCGR3A-158V allele was associated with an increased susceptibility to RA, with homozygotes demonstrating a 1.5-fold to twofold increased risk for RA and a twofold to fourfold increase in nodules 1112. In keeping with several other RA susceptibility genes, this association has not been replicated in all populations 131415161718. There was a trend toward an increased frequency of the FCGR3A-158V allele in Norwegian and Dutch RA patients 1618, with a skewing toward the FCGR3A-158F allele in Spanish, Japanese and Indian populations 13141517. Many reasons for the lack of reproducibility of association studies have been proposed and include differences in the design, power and accuracy of genotyping strategies in the various studies. The apparently conflicting results may also be a consequence of a genuine difference in the genetic and environmental susceptibility factors according to the precise ethnic group or disease phenotype under investigation. Nonreplication may thus indicate a real biological difference between populations. Alternatively, if FCGR3A is in linkage disequilibrium with the true RA-susceptibility locus then, because the extent of linkage disequilibrium varies between different populations, the association may only exist in certain populations and contribute to nonreplication of findings 1219.

FCGR3A lies in a 200 kilobase FCGR gene cluster on 1q22-23 in close molecular proximity to FCGR2A, FCGR3B and FCGR2B 20. Functional single nucleotide polymorphisms (SNPs) in the latter genes have been investigated as genetic susceptibility and severity factors in multiple infectious and autoimmune diseases 56. We therefore undertook additional genotyping in our original RA cohorts and examined SNPs in FCGR2A, FCGR2B and FCGR3B; examined the extent of linkage disequilibrium at this locus; and analyzed FCGR haplotypes for association with disease in order to investigate the possibility that there are other RA susceptibility variants at this locus. We demonstrate an increased association with a FCGR3A–FCGR3B haplotype, which suggests that other polymorphic variants within FCGR3A or FCGR3B, or in linkage disequilibrium with this haplotype may additionally contribute to disease pathogenesis.

Haplotype analyses have started to assume an increased importance in genetic studies of human disease because they can be more informative in their ability to identify unique chromosomal segments that are likely to harbour disease predisposing genes. They may also provide additional evidence for the presence of further unidentified polymorphic variants that are the true disease-susceptibility variants 36. We have demonstrated an increased level of association between FCGR3A–FCGR3B haplotypes and RA compared with FCGR3A alone. The effects were stronger in the subset of RA patients with nodules. The two-locus haplotype showing the strongest association with RA susceptibility in each group was the FCGR3A–FCGR3B 158V-NA2 haplotype. UK Caucasian individuals who were homozygous for this haplotype were estimated to be at threefold risk for disease compared with all others (OR 3.18, 95% CI 1.13–8.92; P = 0.03), and this analysis does not depend on inference of uncertain phase because individuals that are homozygous for a haplotype are unambiguously identified by their genotype. The relative importance of the FCGR3A-158V and FCGR3B-NA2 polymorphic variants were assessed further using stepwise regression analyses 34. These analyses showed that, although only FCGR3A has been shown to be associated with RA 11, the model including both loci provided an improved fit for RA susceptibility, but not necessarily so for the development of nodules. In the North Indian/Pakistani population there was a 6% increase in the FCGR3A–FCGR3B 158V-NA2 haplotype, but this failed to reach statistical significance.

Several methods have been proposed for the analysis of such data in the absence of family data and the consequent presence of phase uncertainty. The HTR method we have chosen to use has various advantages. The method uses all the data, including that from individuals with uncertain phase. Because the data are analyzed in a regression framework, the usual regression diagnostics are available, other factors can be included in the model and tests for interaction can be performed. A potential disadvantage of the method is that weights used in the regression analysis are based on estimated haplotype frequencies, and the uncertainty inherent in the estimates is ignored in the model. This leads to an anti-conservative test and could lead to false-positive results. However, Stram and coworkers 37 evaluated the method in comparison with other more sophisticated approaches and found it to perform well in most situations.

Haplotype frequencies were estimated using the expectation-maximization algorithm, which has been shown to perform well, even in the presence of some departure from Hardy–Weinberg equilibrium 38. Errors due to sampling are generally of much greater concern than inaccuracies due to the estimation process.

We acknowledge that none of our results are highly statistically significant when a consideration is made for multiple tests. However, the consistent pattern of results, taken in conjunction with prior findings in relation to these genes and their known biological functions in humans and mice, gives additional credence to them. Replication of these findings in other populations will ultimately be required.

The FcγRs play important roles in the initiation and propagation of many different immunological and inflammatory processes. Consequently, they may act as susceptibility factors for RA through a variety of mechanisms. FCGR3A was the most significant gene in this study, and we have previously discussed the role that this receptor may play in RA pathogenesis 1112. In humans, FcγRIIIa is expressed on natural killer cells, macrophages, γδ T cells, a subset of monocytes and cultured mast cells 56. Higher levels of FcγRII and FcγRIII expression have been demonstrated in synovial biopsy specimens from RA patients compared with control individuals 39. Similarly, an increase in the expression level and proportion of circulating FcγRIIIa-positive monocytes has been observed in RA and may correlate with disease activity 4041. In addition, in vitro derived macrophages from RA patients expressed more FcγRII and FcγRIIIa, and released higher levels of tumour necrosis factor-α and matrix degrading enzymes in response to heat-aggregated IgG 39 compared with controls. These findings are supportive of our own work whereby the higher affinity genetic variant of FCGR3A may sensitize FcγR-bearing cells to IgG-containing immune complexes. FcγRIIIa may also play an important role delivering (auto)antigens, and activation and maturation signals to dendritic cells 42. This may provide an explanation for the over tenfold increased risk for RA in SE-positive Caucasian individuals homozygous for the FCGR3A–FCGR3B 158V-NA2 haplotype, which has been a consistent finding by a number of groups 1112131518.

FcγRIIIb is selectively expressed on neutrophils and eosinophils, and has a low affinity for IgG. It is linked to the membrane by a glycosylphosphatidylinositol anchor and does not appear to associate with the known transmembrane adapter molecules 56. However, FcγRIIIb appears to interact with FcγRIIa in the phagocytosis of immune complexes and subsequent cellular activation, with signalling being mediated through the ITAM (immunoreceptor tyrosine-based activation motif) of FcγRIIa 4344. FCGR3B has two common polymorphic forms, namely NA1 and NA2, which differ in five nucleotides that produce four amino acid differences. This alters the number of glycosylation sites, and neutrophils from individuals homozygous for the FCGR3B-NA2 allele have been found consistently to exhibit lower levels of phagocytosis than FCGR3B-NA1 homozygotes 45. This polymorphism has important biological consequences, especially in the development of blood transfusion reactions, autoimmune neutropenias and the severity of renal disease in systemic vasculitis 646. Individuals with duplications and deletions of FCGR3B have been reported 3047, with the estimated frequency of the FCGR3B deletion being 0.001–0.08 in various Caucasian populations 48. Standard genotyping assays, as performed in the present study, do not allow a calculation of the gene copy number. This may provide an explanation for a failure of our control populations to conform to Hardy–Weinberg equilibrium and the previously reported non-Mendelian segregation in some Caucasian families 49.

FcγRIIb plays a crucial role in the regulation of antibody production and susceptibility to several spontaneous and induced murine autoimmune diseases 505152. We found no evidence of an association between FCGR2B- or FCGR2B-containing haplotypes and RA in our cohorts, unlike previous observations in a Japanese cohort in which an alternative SNP in FCGR2B was investigated 15.

There is good data that FcγRs may be critical modulators of inflammation within the synovium and that subtle changes in either expression or structure of these receptors may influence both the susceptibility to RA and the development of nodules. The analyses performed in this study have strengthened our original observation that the FCGR genetic locus is associated with RA, particularly in a UK Caucasian population with nodular disease. Our haplotype data, together with the stepwise regression analysis, suggest that additional polymorphic variants within FCGR3A or in linkage disequilibrium with the FCGR3A–FCGR3B 158V-NA2 haplotype may contribute to RA pathogenesis.

ARMS = amplification refractory mutation system; BAC = bacterial artificial chromosome; bp = base pairs; BLAST = basic local alignment search tool; CI = confidence interval; FcγR = Fcγ receptor; HTR = haplotype trend regression; NA = neutrophil antigen; OR = odds ratio; PCR = polymerase chain reaction; RA = rheumatoid arthritis; RF = rheumatoid factor; SE = shared epitope; SNP = single nucleotide polymorphism; UTR = untranslated region.

The authors declare that they have no competing interests.

AWM participated in the design of the study, undertook all database searches, oversaw all aspects of the laboratory work, analyzed the data and prepared the manuscript. JHB gave additional statistical support and performed the haplotype analysis. BG, RDS and PE participated in the collection of clinical data and the recruitment of patients into the study. DS, JR and VK undertook some of the genotyping assays on DNA prepared in the laboratory of EAJ and RWO, who participated in the original design of the study. FP and MA gave invaluable advice during the retrieval of sequence data from the public databases and during the optimization of some genotyping assays. AWB, AFM, PE and JDI participated in the design of the study, interpretation of the results and writing of the final manuscript.