Patients and methods

Subjects

Nine hundred and forty-five Dutch Caucasian patients with RA, all of whom fulfilled the American College of Rheumatology classification criteria for RA were studied and have been described elsewhere.30 31 32 Controls were 388 unrelated Dutch Caucasians with no history of RA.33 For both patients and controls informed written consent according to the Declaration of Helsinki was obtained. The Commissie Medische Ethiek, the Leiden institutional review board, approved all protocols.

ACPA status was available for 619 patients, and was positive in 58.8% (n = 364) of cases. rheumatoid factor status was available for 899 patients, and was positive in 64.8% (n = 583) of cases. Shared epitope status was available for 610 patients and was positive in 70.2% (n = 428) of cases. Serum ACPA was determined by ELISA (CCP2, Immunoscan RA Mark 2, Euro-Diagnostica, Arnhem, the Netherlands and Axis-Shield, Dundee, UK) and the cut-off level for ACPA positivity was set at 25 arbitrary units, according to the manufacturer’s instructions.

Genotyping

FcγRIIIA 158V/F (rs396991) was genotyped using the MassArray matrix-assisted laser desorption ionisation-time-of-flight mass spectrometry, according to the protocols recommended by the manufacturer (Sequenom, San Diego, California, USA). The sequences of PCR primers used in the assay were (ACGTTGGATGTTCACAGTCTCTGAAGACAC) and (ACGTTGGATGAAGCCACACTCAAAGACAGC) and the sequence of the extension primer was (ggagACTTCTGCAGGGGGCTT). SpectroCaller software supplied by the manufacturer was used to automatically call the genotypes. All doubtful calls were rechecked, and after manually evaluating their spectra, they were either accepted or recalled, and if still doubtful the calls were rejected. Ten per cent of samples were genotyped in duplo. The error rate of genotyping was 0%.

CNV

The CNV status of the FcγRIIIA gene was assessed using multiplex ligation-dependent probe amplification (MLPA); which is a sensitive method for copy number quantification.34 MLPA probe design and assay were performed as described by White et al.35 The MLPA probe sequences used were (GACTCCCACCTTGAATCTCATCCCCAGGGTCTCA) and (CTGTCCCATTCTTGGTGCTGGGTGGATCTAAATCCAGG). Because a relatively large amount of DNA is needed for MLPA (in our experiment 125 ng DNA for each sample to get accurate and reliable results), enough DNA was not available from all the patients with RA and controls genotyped for the FcγRIIIA 158V/F SNP. Additionally, not all the DNA samples used for SNP genotyping were extracted using the same method. According to the manufacturer protocols, the use of DNA samples extracted using different methods may influence the MLPA results. The presence of remnants of phenol in phenol-extracted DNA can inhibit MLPA PCR and impede ligase enzyme activity, and the use of old magnetic particles in automated DNA extraction devices may result in incomplete sample denaturation, subsequently influencing MLPA results and rendering them incomparable. Therefore DNA samples that were extracted using phenol were not used for MLPA. Consequently, the MLPA was performed on 456 patients with RA and 285 controls for whom we had enough DNA extracted using the same method.

Statistical analysis

The χ² test with two degrees of freedom (Epi Info v6, CDC, Atlanta, Georgia, USA) was used to compare the relation between genotypes and ACPA-positive and ACPA-negative RA. The MLPA results were analysed as described by White et al.36 The height of each probe-specific peak was divided by the sum of three control peaks to give a ratio. The median ratio for FcγRIIIA across all samples within an assay was calculated and used to normalise the ratios around a value of 1. The normalised ratio for each subject was calculated and plotted in a scatter plot (fig 2). Subgroups corresponding to different FcγRIIIA gene copy numbers were defined by eye and confirmed by cluster analysis (using R statistical software, version 2.5.0), and are delineated by vertical dotted lines. To minimise the possibility of the mis-genotyping of FcγRIIIA 158V/F polymorphism that can be caused by CNV (fig 1), we performed the analysis on the subgroup of subjects with no CNV (the middle cluster in fig 2), thus excluding genotypes from samples with either low or high copy number (the first and the third clusters in fig 2). p Values <0.05 were considered statistically significant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Copy number variation (CNV) of the FcγRIIIA gene. Cluster 1 represent samples with a low copy number and cluster 3 represents samples with a high copy number of FcγRIIIA compared with cluster 2 that represents samples without CNV of the FcγRIIIA gene. MLPA, multiplex ligation-dependent probe amplification.

Results

This study included 945 patients with RA and 388 healthy controls. Table 1 shows the genotype frequencies of the FcγRIIIA 158V/F SNP in patients with RA and controls. The genotype frequencies of FcγRIIIA 158V/F polymorphism (rs396991) were in accordance with HWE (p value is 0.6 in cases and 0.4 in controls). No statistically significant differences in the genotype or allele frequencies between patients with RA and controls were seen. However, after stratifying for ACPA status, the 158VV genotype was more frequent in ACPA-positive patients with RA than in controls (p = 0.05, OR = 1.5, 95% CI 0.99 to 2.27). Similarly, the frequency of the 158V allele in the ACPA-positive RA group (n = 358) was higher than in controls (p = 0.034, OR = 1.3, 95% CI 1.01 to 1.55). No differences were found in the ACPA-negative group (n = 252) (table 1).

The FcγRIIIA gene shows CNV in 6.6% of patients with RA and in 9.5% of controls (table 2). Since the presence of CNV might lead to skewing of the genotype frequencies by causing genotyping errors (fig 1), we assessed whether CNV of the FcγRIIIA gene has an influence on the association of FcγRIIIA 158V/F polymorphism and RA. Therefore, we determined the association between FcγRIIIA 158V/F and ACPA-positive and ACPA-negative RA in subjects with no CNV of the FcγRIIIA gene. Genotypes from subjects with no CNV were selected (cluster 2 in fig 2), thus excluding samples that showed either low or high copy number of FcγRIIIA gene (clusters 1 and 3, respectively in fig 2). Without stratifying for ACPA, the 158VV genotype was significantly more frequent in patients with RA than in controls (17.2 vs 11.7, respectively, p = 0.05, OR = 1.6, 95% CI 0.97 to 2.6), but the frequency of the 158V allele was not significantly higher in patients with RA than in controls. Subsequently, stratifying for ACPA status showed an increased risk of RA as the 158VV genotype was more frequent in the ACPA-positive patients with RA than in controls (21.5 vs 11.7, respectively, p = 0.009, OR = 2.1, 95% CI 1.2 to 3.8) and the presence of the 158V allele was also associated with ACPA-positive RA (p = 0.039, OR = 1.4, 95% CI 1 to 1.9). No association was found in the ACPA-negative group (table 3). Comparing the data without and with stratification for CNV (table 1 and 3, respectively) shows that almost similar results were observed for the effect of the 158V allele on the risk of ACPA-positive RA. In contrast, the odds ratio for the effect of the 158VV genotype on the risk of RA became higher after correcting for the presence of CNV, although the confidence intervals overlapped.

All the subjects identified as having low copy number (a single copy) were genotyped as homozygous for either the 158V or the 158F alleles (as suggested by fig 1).

The third aim of this study was to investigate whether the difference in FcγRIIIA gene copy number confers risk for RA. The distribution of CNV was not significantly different between patients and controls with and without stratifying for ACPA status (table 2).

Discussion

In this study we investigated the association between the FcγRIIIA 158V/F SNP and ACPA-positive as well as ACPA-negative RA and explored the effect of CNV of the FcγRIIIA gene on the association of FcγRIIIA 158V/F SNP with RA. We observed that the association between the FcγRIIIA 158V allele and RA is confined to the ACPA-positive group. In addition, after correction for the effect of the presence of CNV on genotypes, the strength of the association was slightly increased. To our knowledge, this is the first study to consider the effect of disease heterogeneity (presence/absence of ACPA) and genetic heterogeneity (effect of CNV on SNP genotyping) on the association between FcγRIIIA 158V/F polymorphism and RA.

The FcγRIIIA is expressed on NK cells and on macrophages, the expression by macrophages being limited to only a few tissues which correlate with the sites of pathology seen in patients with RA (synovium, dermis under stress, lungs, pericardium and liver).37 The FcγRIIIA expression on NK cells and the number of FcγRIIIA-IgG binding sites per NK cell correlate with the antibody-dependent, cell-mediated cytotoxicity function of these cells.38 The presence of the FcγRIIIA 158V/F SNP, which is a T to G substitution at nucleotide 559 in the FcγRIIIA gene that results in a switch from phenylalanine to valine at amino acid position 158 in the immunoglobulin binding domain, has functional consequences. It was shown that this 158V/F SNP affects the binding affinity of FcγRs to IgG: the 158V allele is associated with higher NK cell IgG binding affinity than the 158F allele, with a gene-dosage effect.39 In addition, IgG stimulation of NK cells from 158VV subjects resulted in higher Ca²⁺ influx, higher concentrations of interleukin 2 (IL2) receptor (CD25) expression and reduced survival of NK cells after activation-induced cell death in comparison with 158FV or 158FF subjects.40

So far, the results of the published studies concerning the FcγRIIIA 158V/F SNP in RA vary markedly. They differ in the presence or absence of its association with RA as well as in the allele frequencies within similar ethnic populations. Table 4 summarises the results of these studies. In our study, the genotype and minor allele frequency in controls and patients with RA were almost identical to those previously reported in Dutch Caucasians.23

It was suggested that these contradictory results originated from methodological difficulties, owing to the extreme homology of the FcγRIIIA to FcγRIIIB gene that might lead to falsely detecting an FcγRIIIB sequence as the FcγRIIIA 158V variant, leading to false overpresentation of the 158V allele.29 As indicated, there are two additional explanations that may have been overlooked. First, we now provide evidence that the FcγRIIIA 158VV genotype associates with ACPA-positive subset of RA. A previous meta-analysis23 showed an odds ratio of 1.3 for the FcγRIIIA 158VV genotype to increase the risk of RA; the percentage of ACPA-positive patients with RA in those studies is unknown. Our data suggest the need for an additional meta-analysis in ACPA-positive patients with RA specifically. Second, the presence of CNV in this gene cluster may previously have led to skewing of the genotype frequencies, subsequently affecting disease associations. Since the frequency of CNV has been reported to vary significantly in different ethnic populations,36 CNV may be another cause of the different allele frequencies of FcγRIIIA 158V/F polymorphism and different associations with RA seen in different populations.

In our study, after controlling for the FcγRIIIA gene copy number, the association between patients with RA in general (without considering the ACPA status) and presence of the 158VV genotype became borderline significant (p = 0.05, OR = 1.6, 95% CI 0.97 to 2.6 vs p = 0.2, OR = 1.3, 95% CI 0.87 to 1.78 without controlling for the CNV). Similarly in the ACPA-positive group before correcting for the presence of CNV, the presence of the 158VV genotype was associated with RA susceptibility with a borderline significant p value (0.05) and an odds ratio of 1.5 (95% CI 0.99 to 2.27). In the ACPA-positive group without CNV this association had an odds ratio of 2.1 (95% CI 1.2 to 3.8). Although these confidence intervals are overlapping and although the presence of CNV of FcγRIIIA did not significantly change the genotype frequencies (probably because the frequency of CNV was relatively low in comparison with the 158V allele), correction for the presence of CNV affected the association between the 158VV genotype and RA. In our opinion, these data underline the need to take the CNV into consideration while performing analysis on SNPs.

The CNV of other FcγRs genes has been shown to associate with susceptibility to several autoimmune diseases such as lupus nephritis16 and idiopathic thrombocytopenic purpura.41 This study evaluated CNV in the FcγRIIIA gene. We confirmed the presence of CNV in the FcγRIIIA gene with a frequency of 9.5% in healthy controls and 6.6% in patients with RA. A comparable frequency of CNV was recently reported in another study with a smaller sample size (116 patients with idiopathic thrombocytopenic purpura and 100 healthy controls).41 We did not find an association between CNV of the FcγRIIIA gene and susceptibility to RA, although, because of the low frequency of CNV in this gene, we were underpowered to conclude formally that the CNV of the FcγRIIIA gene is not associated with RA susceptibility.

In conclusion, the FcγRIIIA gene shows a CNV which was not differently distributed between patients with RA and healthy controls. The analysis of the association between the FcγRIIIA 158V/F polymorphism and RA, stratified for ACPA status and CNV of the FcγRIIIA gene, showed that the FcγRIIIA 158VV genotype confers risk for ACPA-positive RA. The fact that a SNP in the FcγRs gene associates with ACPA-positive RA points to the relevance of antibodies in the pathophysiology of ACPA-positive RA.