Osteoarthritis (OA) is a common disease of the synovial joint characterized by cartilage loss and is often accompanied by alteration in the normal function of other tissues of the articulating joint [1]. Based on epidemiological studies, the disease is known to have a major genetic component, although elucidating this has proven challenging [2]. This is a reflection of the heterogeneity associated with the disease, with severe OA being a clinical end-point reached from a multitude of starting points. Genome-wide association scans (GWASs) are now being employed as powerful, objective tools for mapping susceptibility loci of complex diseases including OA [3, 4]. The resulting signals originate from regions of the genome that do not harbour obvious OA candidates and an example of this came from stage one of the recently published arcOGEN GWAS [5]. In an analysis of over 67,000 individuals of European descent this study identified, amongst other signals, an association to SNP rs2277831 (A/G), with a P-value of 2.9 × 10^-5 and an odds ratio for allele G of 1.07 in knee and/or hip OA.

rs2277831 is located within intron 32 of the MICAL3 gene, which codes for the microtubule-associated monoxygenase, calponin and LIM domain containing 3 protein. The MICAL3 protein is thought to be involved in the vesicle transport system in mammalian cells through the interaction with Rabs, which are small GTPases [6, 7]. Two additional genes map close by and are encompassed by the association signal. These are BCL2L13 (also known as BCL-Rambo), which codes for the BCL2-like 13 (apoptosis facilitator) protein, and BID, which codes for the BH3 interacting domain death agonist protein. The BCL2L13 protein has structural homology to BCL2, and is thought to induce apoptosis through the activation of caspase-3 [8]. The BID protein is also proapoptotic and is cleaved by Caspase-8. The cleaved form translocates to the mitochondria where it forms heterodimers with BAX or BCL2 and triggers the release of cytochrome c, leading to the activation of apoptotic pathways [9]. Scrutinizing public databases reveals that MICAL3 gene is expressed in various tissue types, including adipose tissue, blood, bone, brain, cartilage, heart, liver, kidney and muscle. BCL2L13 and BID expression are also observed in many tissue types.

As the number of susceptibility loci that are identified for common human diseases increases, it is becoming clear that many, if not the majority, of associated alleles contribute to disease risk by acting as expression quantitative trait loci (eQTLs), influencing the expression or stability of a transcript [10, 11]. In the majority of cases eQTLs act in an organ or even tissue specific manner. It is important therefore to study RNA extracted from tissues relevant to the disease when assessing the potential effect of eQTLs.

We hypothesized that the OA association to rs2277831 may be marking an eQTL. To assess this we have measured the overall expression of BCL2L13, BID and MICAL3 using real time reverse transcription PCR and stratified the expression by the genotype at rs2277831. We also tested for allelic expression imbalance of BCL2L13, BID and MICAL3 using transcript SNPs and assays that can accurately discriminate and quantify the mRNA synthesised from each allele. We examined RNA extracted from the joint tissues of patients undergoing elective joint replacement surgery.

Expression of BCL2L13, BID and MICAL3 was observed in the five joint tissues analysed; cartilage, fat pad, synovium, cancellous bone and osteophyte (data not shown).

We quantitatively measured the expression of BCL2L13, BID and MICAL3 in cartilage RNA in 32 OA patients and stratified the expression data by genotype at the OA associated SNP rs2277831. Two patients were homozygous GG, 9 were heterozygous and 21 were homozygous AA. There was no significant correlation between genotype at rs2277831 and the expression of BCL2L13, BID or MICAL3, with P-values of 0.09, 0.07 and 0.33, respectively (Figure 2, Additional file 5).

Twenty-eight of the 33 patients were heterozygous for one or more of the three BCL2L13 transcript SNPs. Twelve of the 28 demonstrated allelic expression imbalance (AEI) when applying the arbitrary threshold of a 20% or greater difference in expression between alleles (data in bold in Table 2). Of these 12 patients, three were heterozygous for the associated SNP rs2277831 (patients 35, 54 and 55), three were homozygous GG at rs2277831 and six were homozygous AA at rs2277831. AEI at BCL2L13 did not therefore correlate with genotype at rs2277831. The largest allelic ratio observed at BCL2L13 was 2.82 (P = 0.004) at transcript SNP rs2587100 for patient 13 (Figure 3), who is homozygous AA at rs2277831.

Eleven patients were heterozygous for the BID transcript SNP rs11538. Three of the eleven demonstrated AEI at the 20% threshold, patients 37, 57 and 58 (data in bold in Table 2). There was no correlation between genotype at rs2277831 and AEI at BID, with all three patients being homozygous AA at this SNP. The largest allelic ratio observed at BID was 2.09 (P = 0.001) in patient 57 (Figure 3).

Twenty-seven patients were heterozygous for one or more of the four MICAL3 transcript SNPs. Nineteen of the 27 patients demonstrated AEI at the 20% threshold (data in bold in Table 2). Of these 19, five were heterozygous for the associated SNP rs2277831 (patients 4, 35, 45, 55 and 59), one was homozygous GG at rs2277831 and 13 were homozygous AA. AEI at MICAL3 did not therefore correlate with genotype at rs2277831. The largest allelic ratio observed at MICAL3, and at any of the three genes studied, was 5.47 (P = 0.001) at transcript SNP rs5992854 for patient 38 (Figure 3), who is homozygous AA at rs2277831.

We noted that in some individuals who were heterozygous for more than one transcript SNP from a particular gene, the AEI pattern observed was not always consistent between the transcript SNPs. For example, for BCL2L13 patient 1 shows AEI at rs2587100 (0.67) and at rs9967 (0.71), but not at rs4488761 (1.04). BCL2L13 has 14 transcript isoforms and rs4488761 exists in nine of these whereas, rs2587100 and rs9967 exist in five. Such variability could account for such AEI differences. We therefore plotted the average allelic expression ratios for the heterozygotes at the eight transcript SNPs and stratified the data by genotype at rs2277831 (Figure 4). At any of the transcribed SNPs, the Kruskal-Wallis test showed no significant evidence of a difference in the allelic expression ratios between the different genotypes at rs2277831 (P-values of 0.45 at rs4488761, 0.11 at rs2587100, 0.20 at rs9967, 0.15 at rs11538, 0.88 at rs5992854, 0.35 at rs11917, 0.41 at rs1057721 and 0.53 at rs4819639).

The averaged log2 of the normalised allelic ratios are shown for each patient, categorised according to the genotype at the associated SNP, rs2277831. The shaded area represents a 20% fold difference in expression between the two alleles

Overall, we have studied hip cartilage, knee cartilage, fat pad, synovium, cancellous bone and osteophyte and although we observed AEI for one or more of the tissues in each gene, the AEI does not correlate with genotype at the OA associated SNP rs2277831.

It has become increasingly apparent that most common traits are influenced by large numbers of alleles that individually have small effects [13]. It is also becoming clear that many if not the majority of complex trait alleles influence the expression of the gene by modulating transcription or transcript stability [11]. In OA, the GDF5 SNP rs143383 is a very good example of this. GDF5 codes for growth and differentiation factor 5, an extracellular signalling molecule that participates in the development and maintenance of the synovial joint. rs143383 is a T to C transition located in the 5' untranslated region of GDF5 and it shows compelling association to OA in European and Asian cohorts (reviewed in [2]). Functional studies have suggested that rs143383 may itself be the polymorphism influencing OA susceptibility at this locus, with the OA associated T allele mediating reduced GDF5 transcription relative to the C allele in all joint tissues so far tested [14]. We therefore hypothesised that the recently reported knee and hip OA association signal at chromosome 22q11.21, detected at SNP rs2277831, may mediate its effect on OA susceptibility by modulating the expression of one or more of the three genes from within the signal. We chose to focus our expression studies on joint tissues from patients who had undergone joint replacement surgery rather than a tissue unrelated to OA, such as blood.

We first looked at overall expression of BCL2L13, BID and MICAL3 in cartilage tissue RNA stratified by genotype at rs2277831, with no significant correlations observed.

We subsequently tested each gene for the activity of cis-regulatory polymorphisms by measuring AEI. Our results revealed that AEI was common and was not restricted to a particular tissue type. If our hypothesis were correct, namely that the OA associated G allele of rs2277831 was a marker for altered expression of one or more of the three genes tested then we would predict that a heterozygote for this SNP would tend to demonstrate AEI. However, we did not observe any correlation between genotype at rs2277831 and AEI, such that an individual heterozygous for rs2277831 was no more likely to demonstrate AEI at one of the three genes than an individual homozygous at this SNP.

Overall our study reveals that BCL2L13, BID and MICAL3 are subject to cis-acting polymorphism/s that influence the expression of the genes in the joint tissues of patients who have undergone surgery. Our data do not however support a role for such cis-regulation at these genes as accounting for the OA association signal at chromosome 22q11.21. Other functional effects are presumably involved.