Introduction

Primary osteoarthritis (OA, MIM 165720) is the most common form of arthritis and is a major cause of morbidity and disability, especially amongst the elderly. It is recognised histologically by focal cartilage loss within the articulating joint (Haq et al. 2003). This loss is accompanied by osteophytes, sclerosis and often some inflammatory response. Numerous epidemiological studies have provided strong evidence for a major genetic component to OA susceptibility (for reviews, see Loughlin 2001, 2002). They have also revealed that this susceptibility shows both gender and skeletal joint-site heterogeneity. Through linkage analysis, we have identified a locus on chromosome 16p with a maximum multipoint LOD score of 1.7 in 146 families with affected female sibling pairs concordant for hip OA ascertained by total hip replacement (female-THR families; Forster et al. 2004). This locus is located at 16p12.3-p12.1, with a genetic position between 40 cM and 52 cM from the 16p-telomere. This interval directly overlaps with the locus identified in an Icelandic study in which an extensive pedigree with an early-onset form of hip OA was investigated (Ingvarsson et al. 2001). This concordant study gave us added confidence to pursue a gene-based association strategy.

A database search of the 12-cM interval highlighted the interleukin 4 receptor α chain gene (IL4R, 16p12.1, MIM 147781) as a particularly strong candidate. Cartilage chondrocyte function and, thus, cartilage integrity are regulated by mechanical stimulation and the cytokine interleukin 4 (IL4) has been demonstrated to be the active autocrine or paracrine signalling molecule in this mechanotransduction pathway (Millward-Sadler et al. 1999; Salter et al. 2001). In order to initiate signalling, IL4 binds to the IL4R chain, which then forms heterodimers either with the common gamma (cγ) chain to form the Type I receptor or with the IL13Rα1 subunit to form the Type II receptor. Formation of the different receptor types results in the activation of alternative downstream signalling pathways. Studies of chondrocytes from OA cartilage have revealed a number of differences in response to mechanotransduction compared with chondrocytes from normal tissue. In OA chondrocytes, IL4 signalling appears to be via the Type I receptor and results in the activation of tetrodoxin-sensitive sodium channels with a subsequent membrane depolarisation. This is as opposed to the signalling via Type II receptors, the activation of apamin-sensitive SK channels and the membrane hyperpolarisation observed in normal chondrocytes (Salter et al. 2001, 2002). In addition, IL4 has been demonstrated to be chondroprotective directly, by regulating the turnover of tissue inhibitors of metalloproteinases and that of matrix metalloproteinases (MMPs; Cawston et al. 1996; Lubberts et al. 1999; Nemoto et al. 1997). It has also been shown to work indirectly, by suppressing IL1- and IL17-mediated MMP3 production and proteoglycan synthesis inhibition and by inhibiting tumour-necrosis-factor-α-induced prostaglandin E2 production (Alaaeddine et al. 1999; Lubberts et al. 2000). IL4 is therefore a strong chondroprotective cytokine and it is reasonable to speculate that polymorphism within its receptor could be a risk factor for OA. We therefore investigated the IL4R gene as the potential 16p OA susceptibility locus by the association analysis of nine common IL4R single nucleotide polymorphisms (SNPs).

Materials and methods

Nomenclature

Gene mutation nomenclature used in this article follows the recommendations of den Dunnen and Antonarakis (2001).

OA cohorts and SNP genotyping

The nine IL4R SNPs (Table 1) were genotyped by conventional polymerase chain reaction/restriction fragment length polymorphism analysis. Digestion products were electrophoresed through 3% agarose and scored following ethidium-bromide staining. The accuracy of the genotyping was confirmed by the direct sequencing of each SNP in a panel of 48 individuals. We genotyped the SNPs in two stages. In stage 1, the SNPs were genotyped in the 146 female probands from our 146 female-THR families. In stage 2, the SNPs were genotyped in a separate cohort of 310 unrelated female-THR cases. SNP allele frequencies in both stages were then compared with those in a cohort of 399 age-matched female controls. All probands and cases had been ascertained through the Nuffield Orthopaedic Centre in Oxford and had undergone total joint replacement of the hip only (mono- or bi-lateral) for primary OA. The primary status was supported by clinical, radiological, operative and histological findings. We excluded any secondary forms of the disease (i.e. osteochondrodysplasia) or non-OA cases (i.e. rheumatoid arthritis). The female controls had not undergone joint replacement surgery and had no overt signs of OA. All cases and all controls were aged 55 or over and were of UK Caucasian origin. Ethical approval for the study was obtained from appropriate ethics committees and informed consent was obtained from all subjects.

Table 1 Association analysis of the IL4R SNPs
Full size table

Statistical analysis

Genetic association and the Hardy-Weinberg equilibrium for the distribution of genotypes were tested by χ2 analysis with Yates’ correction. Odds ratios (ORs) were calculated with 95% confidence intervals (CI). The pair-wise linkage disequilibrium (LD) coefficient r2 (Ardlie et al. 2002) was calculated by using the GOLD program (Abecassis and Cookson 2000). Haplotype frequencies amongst SNPs in LD were calculated by using the estimate haplotype program EH-PLUS (Zhao et al. 2000; http://www.iop.kcl.ac.uk/IoP/Departments/PsychMed/GEpiBST/software.stm). Haplotype frequency differences between the probands and the controls were then compared by means of χ2 and contingency table analysis.

Results

SNP analysis

All nine SNPs conformed to the Hardy-Weinberg equilibrium in the control group (P>0.05). Two of the nine SNPs (S411L and S727A) were associated (P≤0.05) in stage 1, with P-values of 0.03 and 0.02, respectively (Table 1). These associations were attributable to an increased frequency in the probands of the minor allele at each SNP. These two SNPs were also associated in stage 2, with P-values of 0.04 and 0.007, respectively. Again, the minor allele at each SNP was associated. Positive associations in both stages strongly suggested that these results were genuine and not type I errors. SNPs N142N and Q551R, which were not associated in stage 1, were associated in stage 2, both with P-values of 0.04. The minor allele at each SNP was associated. When stages 1 and 2 were combined, SNPs N124 N, S411L and S727A demonstrated association, with the association at S727A being the most significant (P=0.004; OR=2.1, 95%CI: 1.3–3.5). Two other SNPs approached significance in this combined data, viz. S478P (P=0.08) and Q551R (P=0.06).

LD and the identification of two SNP groups

The pair-wise LD coefficient r2 was calculated for all nine SNPs by using data from all cohorts genotyped (probands, cases and controls combined; Table 2). This revealed moderate LD between the 5’ end of the gene and exon 6, with an r2 of 0.52 between (−1914) and I50V. LD then decreased between exons 6 and 12. There were two r2 values that suggested strong LD between two pairs of SNPs located in exon 12, with the first pair comprising SNPs S411L and S727A (r2=0.79) and the second pair comprising SNPs S478P and Q551R (r2=0.79). The two SNP pairs are not in LD with each other, with the highest r2 value between them being 0.25 for SNPs S727A and S478P. The first pair of SNPs (S411L and S727A) were both associated with OA in the 146 female-THR probands, with P-values of 0.03 and 0.02, respectively, whereas neither of the second pair (S478P and Q551R) showed association with OA, with P-values of 0.4 and 0.5, respectively (Table 1). However, SNP Q551R from this second pair was associated in the 310 female-THR cases (P=0.04), whereas its partner SNP, S478P, approached significance (P=0.07). Overall, these data imply that we have detected two independent associations between IL4R and OA, the first involving SNP pair S411L and S727A, the second involving SNP pair S478P and Q551R. The exon 7 SNP N142N was also associated in our study, in the cases and in the combined data (P-values of 0.04 and 0.03, respectively). This SNP is in weak to moderate LD with both SNPs from the first SNP pair, with an r2 value of 0.31 between N142N and S411L and an r2 value of 0.35 between N142N and S727A. We suspect therefore that the association at N142N is not independent of the association at S411L and S727A.

Table 2 Pair-wise linkage disequilibrium (r2) values for the nine IL4R SNPs
Full size table

To determine whether either of the two SNP groups that we had identified (N142N-S411L-S727A and S478P-Q551R) were markers for variant alleles that we had not directly tested, we used the EH-Plus program to construct haplotypes for the two SNP groups. The eight possible haplotypes for SNPs N142N-S411L-S727A were constructed for all female-THR probands and cases combined (affected females) and then compared with the eight possible haplotypes in the female controls (data not shown). The haplotype frequencies between the affected females and the female controls were significantly different (P≤0.05). However, the difference was no more significant than that observed when the three SNPs that constituted this SNP group had been tested individually. The same result was obtained for the four possible haplotypes for SNPs S478P-Q551R. These results imply that the associations that we have detected are not attributable to LD between the associated SNPs and as yet untested polymorphisms.

Inheriting SNP alleles from both SNP groups

We next assessed whether inheriting a copy of an associated allele from both of the two SNP groups was a particular risk factor. For each member of the first SNP group (N142N, S411L and S727A), we determined the number of individuals who had inherited at least one copy of the associated allele at that SNP whilst also inheriting at least one copy of the associated allele at one of the two SNPs from the second SNP group (S478P and Q551R). A comparison was then made between the affected females and the female controls (Table 3). This analysis revealed that inheriting associated alleles at N142N (first SNP group) and S478P (second SNP group) or at N142N (first SNP group) and Q551R (second SNP group) were particular risk factors, with ORs of 2.4 (P=0.0008) and 2.3 (P=0.002), respectively. These results were much more significant than the results obtained when these three SNPs had been tested individually, when the most significant P-value was 0.03 for N142N (Table 1). All six comparisons in Table 3 are significant. However, these are not independent tests because of the LD that exists between the SNPs that constitute each of the two SNP groups.

Table 3 Proportion of individuals who have inherited associated alleles at both SNP groups
Full size table

Discussion

We have found significant evidence that polymorphism within the functional candidate gene IL4R is associated with OA in a Caucasian population. We targeted this gene based on its known role in cartilage homeostasis and that it resides within our linkage interval on chromosome 16p. Investigations into the functional effects of the S478P and Q551R substitutions have been undertaken by others. Residues 478 and 551 are located in the intracellular portion of IL4R and the alleles associated in our study (P478 and R551) can effect the binding and phosphorylation of the intracellular substrates STAT6 and IRS (Kruse et al. 1999). Functional studies into the S411L and S727A substitutions have not been reported. Residue 411 resides in a region that may be involved in interaction with JAK1, whereas residue 727 lies within the terminal cytoplasmic domain of the protein, in the region involved in the negative regulation of IL4 signalling. Both SNPs result in the substitution of an aliphatic hydrophobic residue for a polar hydrophilic one, a change that could have profound effects on protein function.

Polymorphism within the IL4R gene has been associated with several common diseases, including type I diabetes (Bugawan et al. 2003) and atopy (Kruse et al. 1999). Understanding the effects that the different allelic forms of this highly polymorphic receptor have on cell signalling should be of considerable value to investigators studying complex diseases.