Regulatory polymorphisms in extracellular matrix protease genes and susceptibility to rheumatoid arthritis: a case-control study

Many studies support an important role for genetic factors in rheumatoid arthritis (RA) susceptibility and progression [1]. Overall, the genetic component has been estimated to account for about 50% of the variance in disease susceptibility, the remainder being environmental and stochastic components. The best known RA genetic factor is the human leukocyte antigen (HLA) gene, where multiple alleles of the DRbeta1 chain that share a common epitope in the third hypervariable region determine disease susceptibility and severity. Other HLA molecules and several non-HLA genes have also been related with RA susceptibility. Among the many genes that have been studied, only two that encode extracellular matrix (ECM) proteases have been explored [2–6], despite the unequivocal involvement of this family of proteins in RA.

The ECM proteases comprise a large family of proteins grouped in several subfamilies, including the matrix metalloproteinases (MMPs), the most extensively studied in RA [7–9]. Many MMPs are expressed at increased levels in RA tissues and in synoviocyte cultures in response to inflammatory cytokines, show specificity for joint tissue components and affect the evolution of experimental models of arthritis. Drugs able to inhibit a wide array of MMPs have been tried for the treatment of RA and, although effective in experimental models, human clinical trials had to be discontinued due to intolerable side effects. It is expected that more specific protease inhibitors will retain therapeutic potential without the associated side effects. It is unclear what ECM proteases to target with these drugs, however, because it has been difficult to ascertain the specific participation of each of them in RA. As a group, they are the major actors in the degradation of ECM in RA cartilage and bone. In addition, they increase and perpetuate joint inflammation through the activation of cytokines, chemokines and other proteases by cleavage of their precursors at specific sites [7–10]. They can also contribute to inflammation by exposing cryptic epitopes in ECM components that have biological actions in angiogenesis, cell migration and proliferation [10]. The difficulty in discerning the specific role of each of the ECM proteases found in the joints stems from the apparent redundancy of their effects and the wide variety of targets that each could degrade. We expect that genetic studies will provide clues to the identities of proteases that are critical for the RA process.

We have searched bibliographic databases for ECM proteases and their specific inhibitor proteins that have been described as involved in cartilage homeostasis and joint pathology. About 35 were identified with very varied supporting evidence, including some with a putative protective effect. We looked for evidence of single nucleotide polymorphisms (SNPs) in the genes encoding these ECM proteases that have shown a regulatory effect on their transcription level, most often from reporter gene assays but also from electrophoretic mobility-shift assays and in some cases from ex vivo studies. Nine SNPs in eight genes that fulfilled these criteria were found. In addition, each of these SNPs has been associated with susceptibility to at least one from a wide list of diseases, including cardiovascular diseases, aneurysms, preterm rupture of amniotic membranes, and tumor metastasis, indicating that their effects in gene transcription have a significant in vivo repercussion. The nine SNPs were found in the promoters or enhancers of the genes encoding MMP1 [11], MMP2 [12], MMP3 [13], MMP7 (with two SNPs) [14], MMP9 [15], MMP13 [16], plasminogen activator, tissue type (PLAT) [17] and plasminogen activator inhibitor-1 (PAI-1) [18]. They were selected as appropriate candidates to participate in RA susceptibility and studied in a large case-control setting. Two of the SNPs, those in PLAT and MMP2, showed moderate association with RA. The alleles of these two SNPs found with increased frequency in RA patients are low-expression alleles, which is in agreement with previous data suggesting that these ECM proteases have a protective effect in RA.

The lack of association with RA of seven of the regulatory SNPs in ECM proteases was somehow unexpected because they affect proteases with a recognized role in RA [7–9] and because the size of the study allowed for detection of modest effects (for example, the post hoc power to detect an excess of the T allele of the MMP9 SNP was 87% for a risk ratio of 1.2). In fact, there is much more published evidence supporting the involvement of some of the ECM proteases that did not show association in our study than the two that were associated with RA. This is especially clear for the metalloproteinases MMP13, MMP1, MMP3 and MMP9. We also analyzed the possibility of cooperation or of cumulative effects between these SNPs with regard to their association with RA. It would be a mistake, however, to interpret too strongly these results as questioning the importance of these six ECM proteases in RA. At least two factors moderate a conclusion of this type. The multiplicity of control mechanisms of ECM protease activity, of which transcription regulation is only one of importance, compartmentalization by pericellular accumulation, activation by cleavage of latent pro-enzymes and inhibition by specific proteins being the others [7, 8]. Also, variation in these proteases could impinge on other aspects of disease progression different from disease susceptibility, although we did not find significant association with any of the clinical features available for study. They did not include disease activity indexes or quantitative assessment of bone erosions, however, which could be more informative of the possible involvement of these SNPs. This seems the case for the SNP in MMP3 found in previous studies to be associated with quantitatively evaluated RA erosions [3, 5], but not to RA susceptibility [3, 6]. It is also possible that the SNP in MMP1 predisposes to some RA features because there are reports of association with RA inflammatory activity [4] or with RA erosions [6] although these associations were not found in other studies [2].

The two regulatory SNPs that showed moderate association with RA susceptibility cause lower transcription of their respective genes, PLAT and MMP2. Both of them seem genuine associations because they have been found in a hypothesis driven case-control study and have shown modest effects, in the range that is expected in complex diseases and specifically in RA [1]. The size of the study, 550 cases and 642 controls, effectively prevents interference from random variation of allelic frequencies, which is the major cause of false positive results in genetic association studies [24]. In this regard, it is reassuring that about 90% of the associations shown in studies involving more than 150 cases and controls have subsequently been replicated, as pointed out in a recent meta-analysis [24]. Finally, population stratification, another widely claimed cause of spurious association results, is not a significant concern in this study as all cases and controls comprised a very homogenous population. They resided in a largely rural area where immigration has been very restricted. Specifically, 71.8% of the RA patients and 74.0% of the controls had all known ancestors from the same province (Corunna), and 95.4% of the RA patients and 95.3% of the controls were from the same historic region (Galicia, composed of four of the 52 Spanish provinces). In addition, analysis of data restricted to study participants with all known ancestors from Galicia or from Corunna gave similar results to those from the whole study. Nevertheless, circumspection should be exercised in interpreting these associations until the results can be replicated.

PLAT participation in RA has been related to fibrin accumulation in RA synovial cavities [25–27] and the association found here could be involved in this process. A contributing factor in the increase of fibrin could be the lower or unchanged expression of PLAT in RA synovium compared to healthy synovium [28, 29], which would cause the availability of plasmin, the major fibrinolytic enzyme, to be restricted. A putative protective role for PLAT in arthritis has also been shown in experimental models of RA [30, 31]. As the T allele of PLAT -7351 disrupts a GC box in the PLAT enhancer [17, 32] it could contribute to the insufficient fibrinolysis and, thus, to RA. Other processes in which hypofibrinolysis is a contributing disease mechanism, such as myocardial infarction [33] and lacunar stroke [34], have also been associated with the T allele of the PLAT -7351 C/T SNP and these two diseases are observed at increased rate in RA.

Very few studies have addressed the role of MMP2 in RA. It has been assumed that MMP2 promotes RA by participating in cartilage degradation and by activating pro-inflammatory mediators based on its in vitro reactivity [8, 9]. However, MMP2 levels and activity seem to be unaltered in human RA [35–37] and in experimental models of RA [38]. In addition, MMP2 plays a suppressive role in the pathogenesis of antibody-induced arthritis, as shown by an exacerbated disease in MMP2 deficient mice [39]. A likely mechanism for MMP2-mediated protection against RA involves the inactivation of chemokines (CCL7 and SDF1), thereby limiting inflammatory infiltration [40, 41]. Consistent with a protective role for MMP2 in RA, the allele associated with increased RA susceptibility in our study is the low-expression allele [42, 43]. The in vivo relevance of this change has been demonstrated by the association of the MMP2 -1306 C/T SNP with several types of cancer [43–48], although a recent report showed no association with chronic periodontitis [49], a disease with many similarities to RA with respect to inflammation and ECM proteases [50].

Our study indicates that both SNPs act independently in their contribution to RA liability. This conclusion is consistent with the independent roles proposed for the two proteases for their putative protective effect in RA. Interaction with other gene polymorphisms should be explored as it is likely that variants of cytokine genes that are important in RA and that trigger ECM protease gene expression, especially tumor necrosis factor and interleukin-1 but also interleukin-6, epidermal growth factor, platelet-derived growth factor, basic fibroblast growth factor and transforming growth factor-beta, potentiate the effect of the SNPs studied here. In the same way, it is possible that variants of the genes encoding chemokines that are cleaved by MMP2 could interact with the -1306 C/T SNP in determining increased RA susceptibility.

It seems that genetic variants affecting transcription of ECM proteases are not major contributors to RA susceptibility. It is possible, however, that some play a minor role as shown here for the PLAT and MMP2 SNPs. These associations need to be confirmed, although it is already possible to see that the likely effects of these SNPs are consistent with previous evidence supporting a protective effect of the two proteases in arthritis. Confirmation of these associations will lend support to this hypothesis and will show how important it is to define the participation of each ECM protease in joint pathology before trying to manipulate them therapeutically.