Chick embryo chondrocytes express type II collagen; however, if they dedifferentiate, they cease their synthesis and begin to synthesize type I collagen, which is more typical in fibroblasts. Interestingly, in chick embryo chondrocytes, the Col2a1 gene exhibits reduced methylation in comparison with that of fibroblasts, but there is no change in methylation status when chondrocytes become dedifferentiated [17]. In human articular chondrocytes and in MSC subjected to chondrogenesis, all of the 74 CpG sites of a region around the Transcription Start Site (TSS) of the COL2A1 gene promoter are unmethylated. However, this does not correlate with COL2A1 expression because, in human articular chondrocytes, gene expression is low, whereas in MSC subjected to chondrogenesis, COL2A1 demonstrates high expression [18]. On the other hand, in chondrocytes from patients with OA and controls, all of the 21 CpG sites in the COL2A1 enhancer are nearly completely demethylated; however, COL2A1 expression is 9-fold higher in OA chondrocytes in comparison with controls [19]. All of these findings suggest that the regulation of COL2A1 could be independent of DNA methylation of the gene itself.

Type IX collagen is important for the formation of a stable collagen network and for the maintenance of cartilage organization and integrity. COL9A1 gene-expression levels in OA chondrocytes are 6,200-fold lower than in normal chondrocytes, and six of the eight CpG sites of the COL9A1 promoter are significantly hypermethylated [19]. Human articular chondrocytes are negative for type X collagen unless they become hypertrophic. In these cells, two CpG sites of COL10A1 are consistently methylated and there is no gene expression. In contrast, in chondrocytes produced from MSC, there is reduced methylation at these two CpG sites, with strong COL10A1 expression [18]. Aggrecan is an essential component of cartilage ECM that is reduced during aging; hence, it is associated with OA development; however, there are no significant changes in methylation levels at the CpG island of the Aggrecan (ACAN) gene promoter in normal aged or in OA chondrocytes [20]. Thus, it appears that the expression of ACAN is not modulated by changes of methylation in the promoter region.

MMP expression in healthy cartilage is low and, in contrast, it is elevated in OA, resulting in ECM degradation. Methylation analysis of the promoter region of MMP3, MMP9, and MMP13 genes in OA cartilage shows a significant loss of methylation in comparison with normal cartilage. However, not all CpG sites of these promoter genes are equally susceptible to methylation loss and, for each gene, there is a specific CpG site from TSS where demethylation is more significant: at -635 bp for MMP3; at -36 bp for MMP9, and at -110 bp for MMP13 [21]. In addition, methylation at the -110 bp site for MMP13 decreases the binding of Hypoxia-inducible Factor 2 Alpha (HIF-2α), a TF that regulates MMP13 expression [22]. On the other hand, demethylation of another specific region in the MMP13 promoter, the -104 bp site, correlates with increased MMP13 expression and avoids the binding of a TF, the cAMP Response Element Binding (CREB) [23]. ADAMTS5 is considered the major aggrecanase in OA; however, ADAMTS4 also contributes to aggrecan degradation. There is methylation loss at CpG sites in the ADAMTS4 promoter, but the -753 bp site comprises the most consistently demethylated site, with concomitant, strong ADAMTS4 expression up to 700-fold in the surface zone of OA cartilage [21, 24]. All of this is of interest because previously it was widely thought that the methylation of each CpG site was required to repress gene expression; however, these findings suggest that methylation of a single site may be sufficient to affect gene expression.

Under normal conditions, there is no expression of IL-1β gene (IL1B) in human articular chondrocytes; however, when these are experimentally exposed to IL-1β and TNF-α there is a loss of DNA methylation at a specific CpG site (-299 bp) in the promoter and IL1B expression increases 100- to 1,000- fold [25]. Interestingly, in OA chondrocytes, the same IL1B CpG site is demethylated [22], which suggests that inflammatory cytokines can change methylation status at specific CpG sites during the OA process. Interleukin 8 (IL-8), also known as Chemokine (C-X-C motif) Ligand 8 (CXCL8), is a pro-inflammatory chemokine that mediates the activation and migration of neutrophils into tissue from peripheral blood, as well as the release of MMP13 [26]. In OA chondrocytes, IL8 expression is considerably high, and its promoter region is significantly demethylated at three CpG sites: -116 bp; -106 bp, and -31 bp, with -116 bp CpG site the strongest predictor of IL8 expression [27]. Suppressors of Cytokine Signaling proteins (SOCS) are inhibitors of cytokine signaling, and include SOCS1-SOCS7 and Cytokine-Inducible SH2-domain-1 (CIS-1), with SOCS1, SOCS2, SOCS3, and CIS-1 the best characterized of these [28]. The SOCS2 promoter possesses 28 CpG sites in OA chondrocytes; 16 of these CpG sites, located between -920 and -641 bp, are hypermethylated, while 13 CpG sites, localized between -419 and -15 bp, are demethylated. However, there is no difference in SOCS2 promoter methylation status between OA and normal chondrocytes or in human chondrocytes after cytokine stimulation. On the other hand, the gene expression of SOCS2 is reduced in OA chondrocytes [29], suggesting that this is not regulated due to DNA methylation of the gene itself.

Leptin (LEP) is a cytokine-like peptide secreted by white adipose tissue that can regulate bone growth through collagen synthesis, mineralization, osteoblast proliferation, and the stimulation of endochondral ossification [8]. LEP is directly correlated with the OA grade, and its upregulation increases MMP9 and MMP13 expression [30]. In healthy chondrocytes and in the chondrocytes of minimal OA, LEP promoter is highly methylated with a consequent very low gene expression; this downregulation of LEP exerts an effect on MMP13 expression by reducing its levels significantly, whereas in advanced OA, LEP promoter possesses very low levels of methylation and the gene is highly expressed [31].

Chondrocytes express Nicotinamide adenine dinucleotide phosphate-Oxidase (NOX) and Nitric Oxide Synthase (NOS) family members, which generate the Reactive Oxygen Species (ROS): NO and anion superoxide. The increase of NO in articular chondrocytes suppresses energy metabolism, activates MMP, and represses the synthesis of collagen type II and aggrecan in ECM [32]. To prevent damage due to ROS accumulation, chondrocytes produce the Superoxide Dismutase (SOD) enzymes SOD1, SOD2, and SOD3, which are downregulated in OA chondrocytes. In OA cartilage, there is significantly increased methylation at CpG sites of the SOD2 promoter, with low expression levels of the gene [33]. NO production in OA is also the consequence of upregulation of inducible NOS (iNOS), which is activated by IL-1β and TNF-α [34]. In the promoter and enhancer regions of iNOS, between -1400 bp and +117 bp from the TSS, there are 13 CpG sites (seven in the promoter, and six in exon 1 nearest to the TSS). Nevertheless, in OA and normal chondrocytes, six of the seven CpG sites of the iNOS promoter are predominantly methylated, and the remaining CpG sites of the promoter closest to the TSS (-289 bp) and the six sites of the exon 1 are demethylated [35]. These data indicate that epigenetic regulation of iNOS in human chondrocytes does not involve the promoter region between -1400 bp and -117 bp.

iNOS expression is regulated by NF-κB, a signaling factor activated by tissue damage and inflammation. The iNOS enhancer region contains NF-κB binding sites, and the CpG sites of that region in OA chondrocytes are significantly demethylated. It appears that the loss of methylation at specific CpG sites in the iNOS promoter could not account for its abnormal expression; however, demethylation of specific NF-κB enhancer elements could explain the increased iNOS expression in OA [35]. Interestingly, glucosamine and an NF-κB inhibitor inhibit cytokine-induced demethylation at a specific site in the IL1B gene promoter, resulting in decreased gene expression [36].

Sex-determining region Y-box 1 (SOX) belong to a family of TF that carries a characteristic High-Mobility-Group (HMG) domain that binds DNA in a sequence-specific manner [37]. SOX are expressed in all chondrogenic cells, with the exception of hypertrophic chondrocytes, and play a significant role in the maintenance of chondrocytic phenotypes [37]. SOX9 is a key factor for chondrogenesis because it is required for controlling the expression of essential ECM genes, such as COL2A1, COL9A1, COL11A1, and ACAN [37, 38]. Methylation analysis of the promoter region of SOX9 and Runt-related transcription factor 2 (RUNX2) in MSC subjected experimentally to chondrogenesis demonstrates low methylation and increased gene expression, regardless of the differentiation status during chondrogenesis [39], whereas in OA chondrocytes, there is downregulation of SOX9 and promoter methylation is increased, reducing the binding affinity of TF such as CREB and CCAAT-Binding Factor/Nuclear Factor-Y (CBF/NF-Y) [40]. These data indicate that promoter regions of chondrogenesis-related genes SOX9 and RUNX2 are maintained at low levels during chondrogenesis to favor their expression; however, a change in the methylation status of SOX9 promoter could be associated with OA development.

Bone Morphogenetic Protein 7 (BMP7), or Osteogenic protein 1 (OP-1), is a member of the TGF-β superfamily of regulatory molecules. BMP7 is involved in cartilage maintenance and repair by regulating genes of ECM, of anabolic pathways and of bone formation, as well as genes of regulation of cytokines and of various catabolic pathways responsible for ECM degradation and apoptosis [41]. In aged chondrocytes, a positive correlation between age and methylation of the BMP7 promoter is observed, with concomitant decreased expression of BMP7, as well as of genes regulated by BMP7, such as IGF-1, IGF-1 receptor (IGF-1R), and ACAN [42]. This age-related increased methylation in the BMP7 promoter may contribute to the cartilage loss observed in aging and during the progression of OA. Sclerostin (SOST) is a BMP antagonist that modulates mitogenic activity through sequestering BMP [43]. In OA, SOST promotes subchondral bone sclerosis and inhibits cartilage degradation [44]. In OA chondrocytes, the CpG region of the SOST promoter is hypomethylated and gene expression is upregulated compared with normal chondrocytes. Interestingly, demethylation of the gene promoter favors Smad 1/5/8 binding, increasing the expression of SOST [45].

Growth Differentiation Factor 5 (GDF5) is also a member of the TGF-β superfamily and is implicated in chondrogenesis and in chondrocyte proliferation [46]. A Single Nucleotide Polymorphism (SNP) located in the 5´Untranslated Region (5´UTR) of the GDF5 gene, the rs143383 (C/T), is strongly associated with OA and exerts a functional effect [47]. In OA chondrocytes, the OA-risk T allele exhibits lower expression of the gene than that of the C allele, a phenomenon known as Differential Allelic Expression (DAE) [48]. The effect of rs143383 is dependent on a second SNP also localized within the 5´UTR of GDF5, the rs143384 (C/T), and decreased expression of the T allele of rs143383 is only observed in individuals who are compound heterozygous for both SNP. In cell lines and joint tissues, there is demethylation of GDF5 correlating with its increased expression, and interestingly, CpG sites formed by the C alleles of both SNP are variably methylated, and demethylation of the heterozygous cell line increases the DAE imbalance between C and T alleles [49]. This indicates that the DAE variability, thus the OA susceptibility conferred by rs143383, is regulated by DNA methylation.

Similar findings to those for SOCS2 are found in the Cyclin-dependent kinase inhibitor 1 (p21^WAF1/CIP1) gene; this is not an inflammation-related gene, but an inhibitor of cell proliferation that is highly expressed in non-proliferating chondrocytes [50]. In OA chondrocytes, p21^WAF1/CIP1 is downregulated regardless of its promoter methylation status [51]; thus, it appears that at least in chondrocytes, its downregulation is not due to hypermethylation of the promoter.

The Deionidase Iodothyronine type 2 (DIO2) gene encodes a selenoprotein that catalyzes the conversion of an inactive thyroid hormone (T4) to its active form (T3). T3 drives the terminal maturation of growth plate chondrocytes, leading to cell hypertrophy, ECM destruction and mineralization, and the formation of bone [52]. Two SNPs of DIO2 have been associated with OA, with a protective association conferred by the T allele of rs12885300 (T/C), and a clear, predisposing association with the C allele of rs225014 (C/T) and with the haplotype C-C formed by these two SNP [53]. DIO2 also exhibits DAE, with the OA-risk C allele more abundantly present in articular joint tissues than the T allele [54]. In the methylation analysis of DIO2, a CpG site 2031 bp upstream of the TSS is methylated in OA cartilage, with a surprisingly positive association between methylation and DIO2 expression. This effect appears to be driven by the risk C allele, because the homozygosity and heterozygosity for this allele show hypermethylation at the CpG 2031 bp site with increased expression of DIO2, while in the homozygous T allele, there are no differences in methylation or gene expression [55]. This positive correlation between methylation and DIO2 expression in articular cartilage among carriers of the rs225014 OA-risk C allele does not act in accordance with the typical inverse relationship between CpG methylation and gene expression.

Most studies have explored the role of C methylation; however, recently a role for 5-hydroxymethylcytosine (5hmC) as an epigenetic mark has been described. The Ten-eleven translocation cytosine dioxygenases (TET) comprise TET1, TET2, and TET3. These proteins catalyze 5mC oxidation and generate 5mC derivatives, including 5hmC, which is stably present in the majority of tissues and might function as an epigenetic mark (Fig. 1) [56]. A recent study demonstrated a global increase in 5hmC levels up to 5-6-fold in OA chondrocytes, and locus-specific analysis showed a significant increase of 5hmC content in MMP1 and MMP3 promoters. With regard to TET proteins, there are no differences in TET2 and TET3 expression, but surprisingly, significant downregulation of TET1 is observed in OA chondrocytes, as well as in normal chondrocytes exposed to IL-1β and TNF-α [57]. It is surprising that significant downregulation of TET1 is concomitant to an increase in 5hmC in OA chondrocytes, considering that a loss of TET1 in some cancers leads to an opposite effect: a global loss of 5hmC [58]. These observations suggest that the increase in 5hmC and downregulation of TET1 demonstrate clear perturbation of 5hmC homeostasis in OA chondrocytes.

Lately, genome-wide methylation and genome-wide 5-hydroxymethylation profiles in OA have been developed with very interesting results (Table 2) [59-66]. The main findings of these studies can be summarized as follows: 1) OA can be distinguished undoubtedly from controls according to their DNA methylation profile; 2) OA of the hip and the knee possesses different DNA methylation profiles; 3) Differentially Methylated Regions (DMR) are represented more consistently in genes participating in TGF-β signaling, developmental pathways (specifically the homeobox family of TF), inflammation, and ECM degradation pathways; 4) Differentially Methylated Loci (DML) more consistently include RUNX1, genes associated with the degradation of ECM (ADAMTS and MMP), and genes that are members of the TGF-β signaling pathway (ACRV1B, SMAD2, SMAD3, TGFBR2, TGFB1, and BMP6); 5) there are DMR between OA and patients with Osteoporosis (OP), with an inverse methylation relationship of methylation in both groups; 6) OA can be differentiated from controls according to their 5-hydroxymethylome, observing a global increase of 5hmC, at least for OA of the knee, and 7) Differentially hydroxymethylated Regions (DhMR) are related to Wnt-signaling and bone-remodeling pathways, as well as with genes previously related to OA (MMP3, GDF5, and COL11A1).