SEARCH

SEARCH BY CITATION

To the Editor:

Linkage and association analysis provides evidence of a role for the TNFR2 gene in the pathology of several inflammatory diseases and pathologic reactions. Our group originally reported an association between the TNFR2 exon 6 (+196M/R) gene polymorphism and susceptibility to familial, but not sporadic, rheumatoid arthritis (RA) (1). However, it was not clear whether this polymorphism was functional. Thus, although it was associated with interleukin-6 production in systemic lupus erythematosus patients, it had no effect on the expression, shedding, or ligand binding of TNFR2 (2). Alternatively, this association may have been a result of linkage disequilibrium (LD) between the +196M/R polymorphism and the true, but as-yet-unidentified, susceptibility variant. The association between the TNFR2 gene and RA susceptibility, therefore, required further investigation. This report provides evidence that the original observation was spurious and resulted from errors in genotyping that occurred with the use of the earlier and less robust genotyping method.

We performed high-density single-nucleotide polymorphism (SNP) marker analysis in a Caucasian case–control cohort. We compared 192 RA probands from 96 affected sibling pairs with 183 unrelated controls. Details on these subjects are reported elsewhere (3). Twelve SNPs that are located within, or in close proximity to, the TNFR2 gene (Table 1) were selected and genotyped by either RealSNP assay (Sequenom, Cambridge, UK) or 5′ allelic discrimination assay (TaqMan; Applied Biosystems, Warrington, UK). In addition, the +196M/R polymorphism (SNP ID: rs1061622) was genotyped by Pyrosequencing (Biotage, Manchester, UK). Genotyping assays were performed according to the manufacturer's guidelines, or modified as described elsewhere (4). The primer and probe details are available by request from the authors. No deviation from Hardy-Weinberg expectations was observed in either cases or controls (P < 0.01). No significant association with RA susceptibility was detected for any of the polymorphisms (P > 0.2 in all comparisons) (Table 1). Extensive LD was exhibited across the centromeric subregion, but not the 5′ or 3′ region of the TNFR2 gene (Table 1).

Table 1. Comparison of genotype frequencies between rheumatoid arthritis affected sibling pair cases and unrelated controls, determined using a method that accounts for the use of related cases*
SNP IDPosition (1p36) (build 34)Distance from preceding SNP, bpGenotype frequency, no. (%)Trend testMean LD across the subregion
ControlsCases
n1/11/22/2n1/11/22/2χ2PR2 (SD)D′ (SD)
  • *

    Calculations based on published allele frequencies indicated that each test of association had 80% power to detect an effect size of 2 at the 5% significance level. The method used to compare genotype frequencies between cases and controls is described in reference 6. Linkage disequilibrium (LD) analysis was implemented with HelixTree software (Golden Helix,). SNP = single-nucleotide polymorphism.

5′ region            0.34 (0.1)0.83 (0.2)
 rs5903681193299713152 (48)63 (40)16 (12)16468 (41)70 (43)26 (16)0.03420.8532  
 rs9768811194330010,30316388 (54)54 (33)21 (13)14676 (52)53 (36)17 (12)0.00620.9373  
 rs3766730119502226,922176123 (70)49 (28)4 (2)172119 (69)47 (27)6 (4)0.05080.8217  
 rs474247119557211,146154106 (69)45 (29)3 (2)9469 (73)21 (22)4 (4)0.08750.7674  
Centromeric  subregion            0.79 (0.1)0.93 (0.06)
 rs603151119578372,11615266 (43)77 (51)9 (6)13965 (47)67 (48)7 (5)0.20060.6543  
 rs653667119613543,51715962 (39)80 (50)17 (11)18375 (41)81 (44)27 (15)0.04780.8269  
 rs1061622119625011,14713071 (55)54 (42)5 (4)180101 (56)71 (40)8 (4)0.00930.9233  
 rs590977119649062,40515399 (65)51 (33)3 (2)177113 (64)57 (32)7 (4)0.18910.6637  
 rs235249119677772,87117895 (53)74 (42)9 (5)174100 (58)65 (37)9 (5)0.32290.5699  
 rs5746053119718444,067175103 (59)66 (38)6 (3)166102 (61)59 (36)5 (3)0.20520.6505  
3′ region            0.1 (0.1)0.3 (0.2)
 rs235219119739012,057175140 (80)31 (18)4 (2)170140 (82)30 (18)0 (0)0.88080.348  
 rs3397119768382,93715570 (45)66 (43)19 (12)18468 (37)90 (49)26 (14)1.09680.295  
 rs1061631119780451,20715894 (59)57 (36)7 (4)186107 (58)69 (37)10 (5)0.10260.7487  

The absence of an association between the +196M/R polymorphism and RA susceptibility was unexpected because the subjects in the present cohort overlapped with those investigated in our original study (1). Comparisons between the raw data of both studies highlighted discrepancies in the genotypes obtained from 28% of samples common to both studies. Repeat genotyping of the present cohort by an alternative method, namely SNaPshot (Applied Biosystems), demonstrated 100% concordance with the Pyrosequencing data (data not shown). In contrast, our earlier study was based on genotyping by restriction fragment length polymorphism (RFLP), which was widely used at the time. There are 2 possible explanations for the erroneous genotyping. These are 1) incomplete digestion by the Nla III enzyme, and 2) the presence of a second, as-yet-unidentified SNP within the enzyme binding site. We investigated the latter hypothesis by sequencing the surrounding region in 8 samples wrongly genotyped as R/R by RFLP (Lark Technologies, Essex, UK). However, no unexpected polymorphisms were identified.

Hence, it is likely that incomplete enzyme digestion was responsible for the incorrect genotyping. Although incomplete enzyme digestion has long been recognized as a potential problem with RFLP-based genotyping methods, several quality control measures were included in the original study. First, no deviation from Hardy-Weinberg equilibrium was observed. Second, allele frequencies were compared in the control population and found to be similar to those described in previous and subsequent reports. Finally, and most importantly, the association with familial RA was replicated in a second set of RA patients with a family history of RA. Indeed, a subsequent independent European study demonstrated a similar association with familial RA (5). It remains unclear as to why a systematic bias occurred in genotyping the familial RA cases, but not in sporadic RA cases or controls. These findings serve to indicate the importance of robust genotyping platforms. Having tested the association of SNPs spanning the entire TNFR2 gene, we now conclude that the gene is not associated with RA.

Acknowledgements

Supported by the Arthritis Research Campaign. Dr. Barton's work is supported by a Wellcome Advanced Fellowship.

REFERENCES

  1. Top of page

Catherine Potter BSc*, Jane Worthington PhD*, Alan Silman FRCP*, Anne Barton MRCP*, * University of Manchester, Manchester, UK