Association between GDF5 rs143383 polymorphism and knee osteoarthritis: an updated meta-analysis based on 23,995 subjects

Osteoarthritis (OA) is the most prevalent form of arthritis in the worldwide and is regarded as a disorder of the whole joint [1, 2]. One of the most frequently affected joints is the knee, with a prevalence of 30% in those over 65 years old [3]. There is a strong genetic component of OA with heritability estimates showing that genetic components account for 39-65% of the risk for the development of knee OA [2, 4, 5]. However, the identification of specific genes has been problematic with some genes associated with pain [6, 7] and others with joint structures [8]. Overall, there is a lack of consistency of associations.

One of the most comprehensively studied candidate genes is growth differentiation factor 5 (GDF5). GDF5, also known as cartilage-derived morphogenetic protein 1, is a member of the transforming growth factor-β superfamily and closely correlated with bone morphogenetic proteins. GDF5 has been shown to be involved in musculoskeletal processes including the development, maintenance and repair of bone, cartilage and other tissues of synovial joint as well as tendon [9, 10]. In light of the important functions of GDF5, any changes affecting a reduction in the expression of this protein could increase the risk of OA.

GDF5 mutations in humans have been implicated in several disorders of skeletal development [11]. Single-nucleotide polymorphisms (SNPs) have been identified in the 5′-untranslated region (5′-UTR) of GDF5 which is involved in the regulation of GDF5 transcriptional activity [12]. As for one of the most common polymorphisms (rs143383), T to C substitution in the promoter region of GDF5 has an effect on the expression of GDF5 production, with lower GDF5 expression of the OA-associated T allele [13]. Several studies have suggested that GDF5 rs143383 polymorphism may be related to an elevated risk of OA in certain ethnic groups [12, 14–16]. However, these positive associations have not been consistently replicated. For instance, two studies from Korea and Greece failed to detect any association with knee OA [17, 18]. Two earlier meta-analyses suggested that an increase in the risk of knee OA was associated with GDF5 rs143383 polymorphism in Asians and Caucasians [11, 19]. Since then, multiple studies on the relationship of knee OA with GDF5 have been published. Therefore, the aim of this study was to determine the overall association between GDF5 rs143383 polymorphism and knee OA risk and whether the association varies by ethnicity.

To our knowledge, this is the largest and most comprehensive meta-analysis to assess the association of GDF5 rs143383 polymorphism with knee OA, including data from 20 studies in 8,709 knee OA cases and 15,286 controls. Overall analysis of pooled results demonstrated a statistically significant association between the variant genotype of GDF5 and knee OA risk in all comparisons. When stratification by ethnicity, significant associations were found in Asian as well as in Caucasian populations with a greater effect sizes in Asian population, suggesting that GDF5 rs143383 polymorphism is a determinant for knee OA risk and shared between Asian and Caucasian populations.

GDF5, an extracellular signalling molecule, plays a critical role in the development, maintenance and repair of synovial joint tissues, and it has been suggested that deficiency of GDF5 is one of the most important risk factors for the pathogenesis of OA [10]. The expression of the GDF5 protein is modulated by the GDF5 gene, and rare deleterious mutations in the GDF5 gene cause several disorders of skeletal development, such as chondrodysplasias and brachydactyly, suggesting this gene has a crucial role in joint homeostasis and repair [17]. Several animal models have further confirmed the evidence supporting a critical role of GDF5 [31–34]. In mice with GDF5 mutation, a number of abnormalities of joint were found including the decrease in appendicular skeleton and the limb long bones, soft tissue deformities and tendon anomaly. Taken together, these results imply that GDF5 polymorphism may have an important function in the aetiology and pathogenesis of OA.

In this study, we found that C allele of GDF5 was protective for knee OA susceptibility (OR = 0.85, 95% CI = 0.80-0.90, P <0.001), and T allele of GDF5 was associated with a higher risk for knee OA development. These findings seem to be biologically plausible. The T allele of the rs143383 SNP has been shown to be associated with a reduction in GDF5 transcriptional activity, thereby increasing the risk of developing knee OA, compared with the GDF5 C allele [12, 14]. In the subgroup analysis by ethnicity, effects sizes were consistently greater in Asian populations as compared to Caucasian populations, indicating that the same gene polymorphism may have different roles in knee OA susceptibility among different racial backgrounds, and the difference in linkage disequilibrium patterns may exist [19].

Several meta-analyses have been performed to identify the association between the GDF5 variant and knee OA risk. In a previous meta-analysis by Chapman et al. [11] including 2,207 cases and 4,356 controls, a significant association of GDF5 rs143383 polymorphism with knee OA was observed in Asians as well as Caucasians. In another meta-analysis, Evangelou et al. [19] included 5,085 cases and 8,135 controls and found that GDF5 rs143383 polymorphism was associated with the risk of knee OA. In the subgroup analysis, the same association was observed across different populations. Although two recent meta-analyses also suggested that GDF5 rs143383 polymorphism was associated with knee OA susceptibility, their results should be interpreted with caution [35, 36]. In the study performed by Liu et al. [35], all OA cases were pooled into their study and subgroup analyses by joint site, ethnicity, and sex were not performed. In the study by Hao and Jin [36] comprising 6 studies with 2,744 cases and 4,518 controls, there was incomplete identification of publications, which may distort the results [37]. Valdes et al. [30] also performed the meta-analysis with the largest sample size (7,579 cases and 11,947 controls), reporting that the T allele of the GDF5 polymorphism is associated with a 17% elevated risk for knee OA. Consistent with this, the present meta-analysis with a larger sample size showed a similar effect size of T allele for knee OA in the overall population. However, we found a slightly lower risk estimate for the T allele (OR = 1.15) in Caucasian population as compared to the Valdes’ paper (OR = 1.16). This discrepancy may be due to inconsistency of reporting data from the Rotterdam I study. In the current paper, data from this study was extracted from one of the original papers reporting Rotterdam I study [28] rather than from a previous meta-analysis [19], thereby leading to a slight data variation. Additionally, the GDF5 polymorphism was found to be consistently associated with knee OA risk in Asian population. This further provides strong evidence of GDF5 rs143383 polymorphism to knee OA risk across different populations.

Heterogeneity is a potential problem in the understanding the results of meta-analyses. In this study, significant heterogeneity between different studies was observed in the overall population. To clarify the source of heterogeneity, ethnicity and sex were used to stratify the studies, finding part of this heterogeneity can be effectively attenuated or removed when stratification by sex. This indicates that it is important for meta-analyses of genetic association studies to perform subgroup analyses by sex. After subgroup analysis by source of controls, the heterogeneity was also decreased; therefore, it can be assumed that the heterogeneity partly results from difference of source of controls. That may be because potential confounding factors in many epidemiologic studies may result from the difference in control types [38]. In addition, different studies used different criteria to define the cases, which might be one of sources of heterogeneity. Some centres defined their cases using the K/L classification and/or ACR criteria, whereas other centres used a TKR to define their cases. These differences between studies in the control group as well as key characteristics of the participants might lead to heterogeneity in the magnitude of the genetic effects [19]. Therefore, a broad consensus should be reached about OA phenotype definitions and how to enrol an ideal control group. Furthermore, other factors also should be explored to identify the source of heterogeneity if more data was available.

Of note, several potential limitations of this study should be acknowledged. Firstly, knee OA is a multifactorial disease with complex associations between genetic factors and environmental factors, and is a polygenic disease that could not be conferred significantly by no loci individually [39, 40]. Hence, some environmental factors or other polymorphic loci should be taken into account together to arrive at a true effect of GDF5 gene. Secondly, in view of our results from unadjusted estimates, a more accurate assessment should be performed according to age, body mass index, smoking status, and other lifestyle factors if more detailed data were available. Thirdly, publication bias was found in two models, which may give rise to biased results, in particular potentially an overestimate of the effect. However, unpublished studies would need to have a large negative association to have sufficient weight to substantially change our results.

This meta-analysis suggests that GDF5 rs143383 polymorphism is highly associated with the susceptibility to knee OA with protective associations for the C allele and CC genotype across different populations.