In this study, we examined the methylation profiles of chondrocytes from two regions of 8 primary OA knee joints. To overcome the limitations of small sample size and obtain more comprehensive methylation information, we performed the first genome-wide DNA methylation analysis by Illumina Human Methylation 850K BeadChip(13). In light of the results, several worthwhile findings were revealed. First, we found that the methylation profile that we obtained not only confirmed most of the known DM genes in OA but that more DM genes were detected for the first time, some of which were intimately involved in OA. These results indicate that epigenetic dysregulation plays a remarkable role in pathogenesis. Then, the results of GO and KEGG enrichment analysis suggested that the dysregulation of some of these genes disturbed the functions of ECM, and some of the genes participated in the inflammation of OA through signalling pathways.
Compared to previous studies such as 450K beadchip, many of the identified DM genes, such as GDF6, BMP6, SMAD3, ABI3BP, FZD1, MSX2, CHSY1 and MATN2(16–21), also appear in our methylation profile. Homeobox (hox) genes, known OA related DM genes, were also observed in our study; these genes include HOXB3, HOXB1, HOXBAS1, HOXA11-AS, and HOXA-AS3. Hox genes, which represent self-renewal capacity, encode transcription factors that regulate skeletal formation during development(22). A previous study(23) showed that discovering the regulatory mechanisms of Hox genes could not only help to elucidate the pathology and initiation of OA but also provide information for identifying biomarkers that reflect the early stage of OA. At the same time, the expression changes of Hox genes indicated the self-renewal capability of chondrocytes in the late stage of OA(12).
More importantly, some new DM genes that could lead to the progression of the disease were found in our study. RNF43 was one of the most hypermethylated genes. RNF43 plays an important role in the regulation of the Wnt/β-catenin pathway(24). Loss of function of RNF43 results in enhancement of the Wnt/β-catenin signalling pathway because of the decrease/lack of degradation of Frizzled. Meanwhile, the canonical Wnt/β-catenin signalling pathway has been implicated in the pathogenesis of OA, and an increased capacity for Wnt/β-catenin signalling might contribute to cartilage loss(25). In other words, RNF43, one of the potential target genes of Wnt/β-catenin signalling, could reduce cartilage damage by limiting Wnt/β-catenin signalling, which was suppressed in OA. Therefore, we conclude that continuing to study the role of RNF43 may provide new clues for the treatment of OA. In hypomethylated genes, the expression of semaphorin 4D (SEMA4D) was also likely to be associated with OA. SEMA4D/CD100 has multiple roles in immune activation, bone metabolism, and neural development. Recent findings(26) revealed the associations between an elevated systemic soluble fragment of SEMA4D (sSEMA4D) levels and the severity of infectious and inflammatory diseases. A study of rheumatoid arthritis (RA) (27) showed that elevated baseline levels of SEMA4D in RA patients were associated with radiographic progression, suggesting that SEMA4D may be an active mediator of joint damage caused by RA. Another study(28) confirmed that sSEMA4D levels were elevated in the synovium of RA patients and promoted the expression of proinflammatory cytokines, such as interleukin-6 (IL-6) and tumour necrosis factor α (TNFα). The inhibition of arthritis by the anti-SEMA4D antibody was also demonstrated in mice with collagen-induced arthritis (CIA). Since the same inflammatory cytokines are also involved in OA(29), whether we can use SEMA4D as a potential therapeutic target for OA requires further research to determine.
Enrichment terms highlighted by GO revealed that certain DM genes were related to the control of chondrogenesis, especially ECM molecules and CAM. TGFBR1 receptor of transforming growth factor-β (TGF-β), fibroblast growth factor (FGF), bone morphogenetic protein (BMP), nuclear transcription factor SOX9, runt-domain transcription factor (RUNX2) and MMPs are methylated genes related to the chondrogenesis (4), and have been confirmed in our methylation profile again. NID1, F11R, COL19A1, ITGAE, ITGB1, and EDIL3 were newly discovered from the DM genes. The ECM protein interacts with integrin to initiate intracellular signalling. For instance, EDIL3 stimulated osteoblast differentiation by increasing the expression of RUNX2 through α5β1 integrin(30). This gene was also found to be a susceptibility locus for ankylosing spondylitis in a genome-wide association study in a Han Chinese population(31). Junctional adhesion molecule 1 (JAM1/JAM-A/F11R), a kind of CAM, was hypomethylated in OA, which is involved in leukocyte adhesion and migration, and its knockdown decreased cell-matrix adhesion and Rap1 activity(32). Research has shown that F11R mRNA is highly expressed in peripheral blood mononuclear cells (PBMCs) of RA patients and speculated that the F11R promoter − 688 C may be a protective factor for the development of anti-CCP antibodies(33). Coincidentally, another study of systemic lupus erythematosus revealed that the blockade of F11R might have therapeutic potential in patients with lupus(34). Since these diseases are all rheumatic diseases, these results lead us to speculate and investigate the roles that F11R and EDIL3 play in OA.
In this study, we examine the information provided by the enrichment of KEGG pathway. The PI3K/Akt signalling pathway, Rap1 signalling pathway, and phospholipase D signalling pathway are all connected with inflammatory factors. The PI3K/Akt signalling pathway is one of the most conspicuous terms involved in many diseases and is essential in regulating the pathophysiology of OA. As the PI3K/Akt pathway is a classic inflammatory signalling pathway, animal experiments have found that the inhibition of this pathway could attenuate the inflammatory response and was able to prevent aberrant bone formation(35). Several DM genes enriched in this pathway merit further study. PKN1 was hypermethylated in OA, but its role in OA has not been reported to date. PKN1 is a protein kinase whose catalytic domain is highly homologous to the protein kinase C (PKC) family. Tumour necrosis factor alpha (TNF-α) receptor-associated factors (TRAF) are the major mediators of transducing TNF signalling to a variety of functional targets, including activation of NF-κB, c-Jun NH2-terminal kinase (JNK), and apoptosis. PKN1 can silence NF-κB kinase (IKK) and JNK by phosphorylating TRAF1 and regulating the competitive binding between TRAF1 and TRAF2 to TNF receptor type 2 (TNFR2)(36, 37). PKN1, as a key factor that can directly interact with TNFR2, plays an important role in regulating the activity of NF-κB. In OA, hypermethylated PKN1 may be responsible for NF-κB activation after self-expression inhibition, or it may be the result of long-term stimulation of chronic inflammation, which requires further research to explain. At the same time, PKN1 inhibits Wnt/β-catenin signalling in a human melanoma cell research(38), suggesting that the role of PKN1 in the pathogenesis of OA should not be underestimated. FLT-1, vascular endothelial growth factor (VEGF) receptor 1, is a hypomethylated DM gene in this pathway. Genome-wide studies have shown that VEGF is associated with OA, and increased expression of VEGF is positively associated with disease severity and pain. At present, the treatment of targeting VEGF to resist angiogenesis and enhance cartilage repair, particularly blocking or reducing FLT-1 expression, has been extensively studied in animal models of OA and RA(39, 40). Notably, supported by both the findings of a large number of studies and our genome-wide screening, this result points to the role of VEGF in the treatment of OA. We are looking forward to the development of a safer and more efficacious VEGF-targeted OA therapy in clinical practice.
Phospholipase D is a regulator of intercellular signalling and metabolic pathways, particularly in cells under stress conditions. Most studies have revealed the role of the phospholipase D signalling pathway in cancer, but the relationship between this protein and OA has not been determined(41). PTPN11 encodes the tyrosine phosphatase SHP-2; this hypomethylated DM gene, enriched in the phospholipase D pathway, may demonstrate its role in some aspects. PTPN11 was overexpressed in RA fibroblast-like synoviocytes (FLS) compared with OA FLS, and inhibition of SHP-2 in mice reduced RA FLS invasion, migration, and reduced arthritis severity, suggesting that targeting SHP-2 may be a therapeutic strategy for RA(42). Therefore, what is the role of SHP-2 in OA? The differential methylation of PTPN11 in our report suggests that we should reconsider whether the impact of OA disease status on research is neglected in RA studies with OA as a control and whether some of the effective treatment methods for RA are equally applicable to OA.