Without any prior linkage data to guide them toward particular chromosomal locations, a number of investigators have used their biological understanding of OA to select candidate genes for association studies. The initial candidate gene studies focused on genes that code for cartilage extracellular matrix (ECM) structural proteins. These included COL2A1, which codes for the α1 polypeptide chain of type II collagen, the principal collagenous component of articular cartilage. Other cartilage collagen genes studied included the type IX and type XI collagen genes and genes coding for noncollagenous components of the ECM, such as the cartilage oligomeric matrix protein gene (COMP) and the aggrecan gene AGC1. On the whole, these studies did not yield convincing data to support a role for common, nonsynonymous mutations in cartilage ECM structural protein genes as risk factors for primary OA.^1,2 It must be concluded, therefore, that these genes have never undergone a mutational event within their amino acid coding sequence that predisposes to primary OA or that if such mutations have occurred, these mutations do not have a sufficiently high frequency to confer a significant population risk.³

Concurrent with the analysis of ECM structural protein genes, a number of investigators focused on genes coding for proteins influencing bone density. This was driven by the observation that subchondral sclerosis is an early observation in some OA joints and the subsequent suggestion that this sclerosis might damage the cartilage by adversely affecting the transmittance of mechanical load. The vitamin D receptor gene VDR and the estrogen receptor gene (ESR1) have both been investigated by a number of groups. Early studies demonstrated association of both genes with OA. However, these associations were not replicated in all subsequent investigations.^1,2 This is to be expected for a common complex disease such as OA and highlights the fact that any single locus will have at most only a moderate effect on disease occurrence in the population under investigation. A reasonable number of positive reports are now being published, particularly for ESR1.⁴ Common variants within this gene are therefore likely to be OA risk factors.

As with other complex disease investigations, OA association studies often suffer from a number of shortcomings. These include the analysis of cohorts of relatively small size and the genotyping of only a few of the known DNA variants within the gene under investigation. The former should be unnecessary for a common disease like OA and makes many studies underpowered and liable to false positives. The latter is becoming avoidable as the cost of genotyping falls and as databases of common variants become more comprehensive in their coverage. Future OA association studies should be designed with these two considerations in mind.

Candidate gene studies have shed some light on OA genetic susceptibility. However, they are constrained by our incomplete knowledge of the biology of OA, which makes candidate selection a fallible act. Investigators have therefore taken a more systematic approach, including genome-wide linkage scans.

The first OA genome-wide linkage scan was published by an Oxford group in 1999 and was performed on 481 pedigrees that each contained at least one OA-affected sibling pair (ASP) ascertained by total joint replacement of the hip or the knee.^5,6 With this ascertainment, the investigators were treating the disease as a discreet trait, since subjects either had or had not undergone joint replacement. The investigators were also focussing on pedigrees that had severe, end-stage OA. Linkages were initially reported to chromosomes 2 and 11, and these were found to be particularly relevant to ASPs concordant for hip OA (chromosome 2) and to female ASPs concordant for hip OA (chromosome 11). A subsequent stratification of the genome scan by sex and by joint replaced (hip or knee) uncovered additional linkages, on chromosome 4 in female ASPs, chromosome 6 in hip ASPs, and chromosome 16 in female ASPs.⁶ These were important findings not only because they pointed toward areas of the genome that may encode for OA susceptibility but because they also suggested differences in the nature of the susceptibility between males and females and between different skeletal sites, something that had been suggested by previous epidemiological studies.^1,2

This original genome scan employed an average of just one microsatellite marker every 15 centiMorgan (cM). This medium density meant that the linkage intervals were relatively large. The Oxford investigators therefore subjected each locus to finer linkage mapping in an expanded cohort of 571 OA pedigrees. This analysis, which employed an average marker density of one marker every 5 cM, succeeded in narrowing each of the linkages and also confirmed or refined their restriction to particular strata. For example, the chromosome 6 linkage was narrowed from a 50-cM interval in hip ASPs to a 12-cM interval in female-hip ASPs. The results of the finer linkage mapping have been published^{11,12,13,14,15} and are summarized in Table 6-1. The most significant evidence for linkage from the Oxford study is a logarithm of the odds (LOD) score of 4.8 on chromosome 6.

The Finnish scan was performed on 27 pedigrees that each contained at least two affected siblings with radiographic distal interphalangeal (DIP) OA.⁷ Nine genomic regions supported linkage and the genotyping of these in additional pedigree members confirmed linkage to chromosomes 2q, 4q, 7p, and the Xcen (Table 6-1). The 2q and 4q linkages do not show overlap with the finer linkage mapped 2q and 4q intervals from the Oxford study and therefore probably represent different loci. The Finnish 2q linkage encompasses the interleukin-1 (IL-1) receptor and ligand gene cluster at 2q12-q13. This cluster has now been associated with OA (discussed later).

A genome-wide linkage scan utilizing 1000 microsatellites was performed on 329 Icelandic families containing 1143 individuals with primary hand OA and 939 of their relatives.⁸ Each family contained at least two affected individuals related to each other at or within five meioses. Individuals were classed as having hand OA if they exhibited at least two nodes at the DIP joints of each hand or if they demonstrated squaring or dislocation of the first carpometacarpal (CMC1) thumb joint. The highest LOD
score observed was on chromosome 4 (LOD = 2.61), followed by chromosomes 3 (LOD = 1.79) and 2 (LOD = 1.48) (Table 6-1). The scan was subsequently stratified by site into a DIP cohort (274 families), a CMC1 cohort (204 families), and a DIP and CMC1 cohort (142 families). Linkage in the stratified analysis highlighted the same linkages as those observed in the unstratified scan. However, the evidence for linkage increased at chromosomes 2 and 4, with a LOD score of 4.97 for the CMC1 stratum on chromosome 2, and a LOD score of 3.29 for chromosome 4 in the DIP stratum (Table 6-1). The Icelandic group have gone on to identify an associated variant within the matrilin 3 gene MATN3 located within their 2p linkage.

The USA study involved 296 small, two-generational families composed of 684 parents and 793 of their offspring from the Framingham cohort.^9,10 Hand OA was characterized radiographically and was assessed in the late 1960s and again in the early 1990s for the parents, and in the 1990s for the offspring. The investigators performed a quantitative linkage analysis that separately tested the degree of joint space narrowing (JSN) and the frequency of osteophytes. They also investigated an overall radiographic score based on the Kellgren and Lawrence (K/L) grading scheme. Linkage was reported to eight chromosomal regions, with the highest LOD score being 2.96 for chromosome 1 in the JSN criteria (Table 6-1). For the overall radiographic score, no LOD exceeded 1.9, and so phenotypic definition was reconsidered by stratifying the linkage scan according to OA disease status in particular joints of the hand that are most susceptible to OA, namely, the DIP joints, the thumb interphalangeal (IP) joint, the proximal interphalangeal (PIP) joint, the metacarpophalangeal (MCP) joints, the wrist joints, and the CMC1 joint at the base of each thumb.¹⁰ This revealed two new loci, one on chromosome 7q in the DIP stratum with a LOD score of 3.06, and one on chromosome 15q in the CMC1 stratum with a LOD score of 6.25 (Table 6-1). Stratification by joint and by sex revealed two further loci in females, one on chromosome 1q in the DIP stratum with a LOD score of 3.03, and one on chromosome 20p in the CMC1 stratum with a LOD score of 3.74. Overall, the USA study had identified at least 12 loci, several of which are restricted to particular joints of the hand.

Intriguingly, the chromosome 2p locus identified in the USA study directly overlapped with the 2p locus identified in the Iceland scan. This might therefore represent an independent confirmation, although a skeptic could argue that
so many loci were detected in the USA study, many with relatively low LOD scores, that an overlap is likely to happen by chance alone. Nevertheless, the fact that the Icelandic group has gone on to identify an associated variant within MATN3 at the 2p locus means that an analysis of this gene in the USA cohort is merited.