To investigate the role of the newly discovered epigenetic mark 5-hydroxymethylcytosine (5hmC) and its regulators in altered gene expression in osteoarthritis (OA).
To investigate the role of the newly discovered epigenetic mark 5-hydroxymethylcytosine (5hmC) and its regulators in altered gene expression in osteoarthritis (OA).
Cartilage was obtained from OA patients undergoing total knee arthroplasty and from control patients undergoing anterior cruciate ligament reconstruction. Global levels of 5hmC and 5-methylcytosine (5mC) were investigated using immunoblotting, enzyme-linked immunosorbent assays, and cellular staining. Gene expression changes were monitored by quantitative polymerase chain reaction (PCR) analysis. Levels of locus-specific 5hmC and 5mC at CpG sites in the matrix metalloproteinase 1 (MMP-1), MMP-3, ADAMTS-5, and hypoxanthine guanine phosphoribosyltransferase 1 (HPRT-1) promoters were quantified using a glucosylation and enzyme digestion–based method followed by quantitative PCR analysis. Global and locus-specific 5hmC levels and gene expression changes were monitored in normal chondrocytes stimulated with inflammatory cytokines to identify the effect of joint inflammation.
A global 5–6-fold increase in 5hmC concomitant with a loss of TET1 was observed in human OA chondrocytes compared to normal chondrocytes. Enrichment of 5hmC was observed in promoters of enzymes critical to OA pathology, MMP-1 and MMP-3. Short-term treatment of normal chondrocytes with inflammatory cytokines induced a rapid decrease in TET1 expression but no global or locus-specific 5hmC enrichment.
This study provides the first evidence of an epigenetic imbalance of the 5hmC homeostasis in OA leading to TET1 down-regulation and 5hmC accumulation. Our experiments identify 5hmC and its regulators as potential diagnostic and therapeutic targets in OA.
Osteoarthritis (OA) is an age-associated multifactorial disease characterized by joint dysfunction and cartilage degeneration that is widely prevalent in the elderly population (). Clinical management of this disorder is largely limited to pain management or an eventual total joint replacement. Various genes render susceptibility to OA; however, there is not a single consensus genetic basis for the disease (). Insight into the early epigenetic changes leading to the altered gene expression in OA can provide a novel target axis for OA pathology ([3, 4]).
DNA methylation is a key epigenetic mark associated with gene silencing (). Studies conducted during the last few years have brought about a paradigm shift in our understanding, elucidating the fact that active DNA demethylation is more dynamic and prevalent than was previously appreciated and involves DNA repair pathways (). DNA hydroxymethylation of the cytosine base (5hmC; currently referred to as the “sixth base”), has been discovered to be stably present in most tissues and particularly abundant in embryonic stem cells and neurons ([6, 7]). The TET family of proteins, consisting of TETs 1, 2, and 3, converts 5mC to 5hmC ([7, 8]) and can also further oxidize 5hmC to 5-carboxylcytosine (5caC) and 5-formylcytosine (5fC) (). All of these intermediates are substrates for thymine DNA glycosylase (TDG), leading to replacement by an unmodified cytosine and resulting in active DNA demethylation ([10, 11]). Another possible route for active DNA demethylation involves the activation-induced deaminase (AID) or apolipoprotein B messenger RNA–editing enzyme catalytic polypeptide–like (APOBEC) family of DNA deaminases that can act independently on 5mC or deaminate 5hmC after TET action, leading ultimately to DNA demethylation ().
A critical role of 5hmC and TET proteins has emerged in stem cell differentiation and embryonic development ([13, 14]) as well as in cancer. Mutations in TET1 and TET2 were initially associated with various forms of leukemia ([15, 16]). Intriguingly, a global loss of 5hmC was observed in multiple cancers, including hematologic disorders and solid tumors, such as colon, prostate, and breast cancers. In melanoma, the loss of 5hmC is a direct result of the loss of TET function (). Gain of TET2 activity was shown to not only restore the 5hmC epigenome, but also suppress the tumor, demonstrating that 5hmC homeostasis is dynamic and its perturbation is a causal factor in melanoma ().
Evidence has been accumulating with regard to a role of DNA methylation in the pathogenesis of OA ([3, 18]). Alterations in DNA methylation patterns have been observed in OA chondrocytes, particularly a loss of DNA methylation of the promoters of various OA-associated genes, including matrix metalloproteinase 3 (MMP-3), MMP-9, MMP-13, ADAMTS-4, and interleukin-1β (IL-1β) ([19-22]). Expression of growth differentiation factor 5, the OA susceptibility gene, is modulated by DNA methylation (). In addition, a recent study has reported DNA demethylation at cytokine-responsive enhancer elements of inducible nitric oxide synthase to be critical for gene induction (). In view of the recent advances in the dynamics of DNA methylation, we sought to investigate whether 5hmC and the active DNA demethylation machinery play a role in the pathogenesis of OA.
Normal adult (age 34 years) and juvenile (fetal age 24 weeks and natal ages 6 months, 18 months, and 6 years) articular chondrocytes were purchased from Lonza. Grossly normal cartilage pieces were obtained during notchplasty or debridement from 12 patients undergoing anterior cruciate ligament (ACL) reconstruction who had no history of OA symptoms (age range 18 months to 42 years [mean ± SD 30 ± 12 years]; 58% male and 42% female). Articular chondrocytes were harvested from OA cartilage samples obtained during total knee arthroplasty from 32 patients (age range 47–84 years [mean ± SD 67 ± 10 years]; 28% male and 72% female). All samples were obtained under approved Human Subjects Institutional Review Board protocols.
Cartilage was dissected, and the chondrocytes were dissociated and cultured in high-density monolayers for limited passages, as described previously ().
Total RNA was purified using an RNeasy Plus Mini kit (Qiagen), and complementary DNA (cDNA) was synthesized using a high-capacity cDNA reverse transcription kit (Applied Biosystems). Quantitative polymerase chain reaction (qPCR) was performed using TaqMan gene expression arrays for ADAMTS-4, ADAMTS-5, TET1, TET2, TET3, and TDG, with a universal master mix (Applied Biosystems). For analysis of MMP-1, MMP-3, MMP-9, and MMP-13, USB VeriQuest SYBR Green qPCR Master Mix (Affymetrix) was used with gene-specific primers. All analyses were performed using the ΔΔCt method, and expression was normalized to GAPDH or 18S. SYBR Green primer sequences were as follows: for MMP-1, 5′-CTGGAATTGGCCACAAAGTT-3′ (forward) and 5′-TCCTGCAGTTGAACCAGCTA-3′ (reverse); for MMP-3, 5′-GGTGTGGAGTTCCTGATGTTG-3′ (forward) and 5′-AGCCTGGAGAATGTGAGTGG-3′ (reverse); for MMP-9, 5′-CTGGGCAGATTCCAAACCT-3′ (forward) and 5′-TACACGCGAGTGAAGGTGAG-3′ (reverse); for MMP-13, 5′-TCAGGAAACCAGGTCTGGAG-3′ (forward) and 5′-TGACGCGAACAATACGGTTA-3′ (reverse); and for GAPDH, 5′-GTGTTCCTACCCCCAATGTGT-3′ (forward) and 5′-ATTGTCATACCAGGAAATGAGCTT-3′ (reverse).
Genomic DNA was extracted using a DNeasy Blood and Tissue kit (Qiagen), denatured (0.4M sodium hydroxide, 10 mM EDTA at 100°C for 10 minutes), and then neutralized (6.6M cold ammonium acetate, pH 7). A total of 200 ng of DNA was applied to a prewet Amersham Hybond-N+ membrane (GE Healthcare) using a gentle vacuum and then air-dried and hybridized by ultraviolet crosslinking. After blocking for 1 hour, the membrane was incubated overnight at 4°C in primary antibody (1:200 dilution of 5hmC and 5mC [Active Motif] and a 1:50 dilution of single-stranded DNA [ssDNA; EMD Millipore]), washed, and incubated with an appropriate secondary antibody (1:5,000 dilution; GE Healthcare). Blots were visualized using ECL Plus (Pierce) and analyzed using ImageJ software (National Institutes of Health; online at http://rsbweb.nih.gov/ij/).
A MethylFlash hydroxymethylated DNA quantification kit (Epigentek) was used for colorimetric detection of 5hmC per the manufacturer's guidelines. Briefly, 200 ng of genomic DNA and 100 μl of DNA binding solution were added to the assay microplate and incubated at 37°C for 90 minutes. The 5hmC capture antibody (1 μg/ml) was added for 60 minutes, followed by 0.2 μg/ml of the biotin-conjugated detection antibody for 30 minutes, and the absorbance was read at 450 nm. The amount of 5hmC was calculated based on a standard curve generated using the kit controls.
Chondrocytes were fixed in 4% paraformaldehyde and permeabilized with cold 0.4% Triton X-100 in phosphate buffered saline (PBS) for 15 minutes. A denaturation step with 2N HCl for 15 minutes was followed by neutralization in 100 mM Tris HCl (pH 8.5) for 10 minutes. The cells were then blocked for 1 hour in PBS containing 1% bovine serum albumin/10% fetal bovine serum/0.4% Triton X-100, incubated overnight with 5hmC primary antibody (1:100 dilution; Active Motif), and visualized with Alexa 488–conjugated goat anti-rabbit secondary antibody (1:250 dilution; Invitrogen). Cellular DNA was counterstained with NucBlue fixed cell stain (Molecular Probes).
Detection of 5hmC at a particular CCGG site was performed using an EpiMark 5hmC and 5mC analysis kit (New England Biolabs) ([26, 27]). Genomic DNA (2.5 μg) was treated with 30 units of T4 β-glucosyltransferase and 80 μM UDP-glucose at 37°C for 16 hours. The glucosylated DNA was then digested with 100 units of Msp I or 50 units of Hpa II at 37°C for 6 hours, followed by treatment with Proteinase K at 40°C for 30 minutes and inactivation at 95°C for 10 minutes. Real-time PCR was then performed on 1 μl of the glycosylated/digested DNA using site-specific primers. The percentages of 5hmC, 5mC, and cytosine were calculated using the EpiMark comparative Ct method.
Primer sequences were as follows: for ADAMTS-5 (–43 bp), 5′-TCTGGAGCACGAATCCAAAC-3′ (forward) and 5′-CACTTGCTTGCAGGATTGAG-3′ (reverse); for MMP-1 (–1,300 bp), 5′-GCACCAAGGAGCGAAGATAG-3′ (forward) and 5′-GAGAAGACCCCTCATCCACA-3′ (reverse); and for MMP-3 (–541 bp), 5′-GGAGGGGAAAAGGTTGAAAG-3′ (forward) and 5′-CCACGTAGCTGCTCCATAAATAG-3′ (reverse).
Nuclear protein was extracted using a NE-PER nuclear and cytoplasmic extraction kit (Pierce). The extracted protein (35 μg) was run on a 4–15% Tris–glycine gel (Bio-Rad), transferred onto a PVDF membrane, and incubated overnight with TET1 antibody (1:50 dilution; Santa Cruz Biotechnology). Membranes were washed and incubated with an anti-goat secondary antibody (1:5,000 dilution; Santa Cruz Biotechnology). Blots were visualized using ECL Plus (Pierce). Equal protein loading was verified by Ponceau S staining.
Chondrocytes were seeded at 104 cells/cm2 and cultured for 24 hours in complete medium. After 24 hours, cells were treated for 48 hours with control medium or with serum-free Dulbecco's modified Eagle's medium/F-12 medium containing 10 ng/ml of human IL-1β (PeproTech) or containing a combination of 20 ng/ml of human tumor necrosis factor α (TNFα) plus 10 ng/ml of human oncostatin M (PeproTech).
Normal chondrocytes were treated for 48 hours with control medium, IL-1β, or TNFα plus oncostatin M as previously described. Cytotoxicity was estimated using the Live/Dead Viability/Cytotoxicity kit (Invitrogen) that uses a 2-color fluorescence-based assay. Cell monolayers were incubated for 30 minutes at room temperature with 2 μM calcein AM and 4 μM ethidium homodimer 1 and were then viewed under a fluorescence microscope. Manual cell counts were performed to determine the percentages of live cells (calcein positive, green fluorescence) and dead cells (ethidium homodimer 1 positive, red fluorescence).
To identify DNA methylation changes associated with OA pathogenesis, we used a set of normal and OA chondrocytes harvested from the articular cartilage of the knee joints of non-OA patients and OA patients, respectively, who were undergoing surgery (see Patients and Methods). In addition, control samples included commercially available articular chondrocyte preparations (fetal age 24 weeks and natal ages 6 months, 18 months, 6 years, and 34 years) that were similar in phenotype and gene expression to chondrocytes derived from non-OA patients undergoing ACL surgery. Gene expression profiles of normal and OA chondrocytes were measured using qPCR analysis. OA chondrocyte populations showed the characteristic up-regulation of MMPs 1, 3, 9, and 13 as compared to healthy control chondrocytes (but no consistent up-regulation of ADAMTS-4 or ADAMTS-5, in accordance with published reports ) (Figure 1).
The relative global levels of 5mC and 5hmC in the DNA of normal and OA chondrocytes were assessed using previously described specific antibodies with no demonstrated cross-reactivity ([29, 30]). A dot-blot analysis of 5hmC, 5mC, and total DNA, using an antibody against ssDNA, showed a dramatic increase in 5hmC levels in OA chondrocytes as compared to normal chondrocytes, while the 5mC and ssDNA levels remained unchanged (Figure 2A). The results shown are for a representative set of 8 normal samples (mean ± SD age 20 ± 16 years) and 13 OA samples (mean ± SD age 69 ± 9 years). Quantification of the dot-blot signal by normalization to ssDNA showed a 5–6-fold up-regulation of 5hmC, while the 5mC levels were not significantly different (Figures 2B and C).
Using an alternate 5hmC-specific antibody in an ELISA-based assay, we observed a similar 5-fold increase in the percentage of 5hmC in the DNA of OA chondrocytes as compared to normal chondrocytes. The age distribution of the normal subjects and OA patients was 25.5 ± 14 and 66 ± 5 years, respectively (mean ± SD) (Figure 2D). Immunostaining of chondrocytes showed the expected nuclear localization of 5hmC. An intense staining in representative OA chondrocytes (from a 64-year-old patient) as compared to normal chondrocytes (from an 18-year-old patient without OA) reflected the global increase in 5hmC levels observed in the dot-blot experiments (Figure 2E). Taken together, these findings show a dysregulation of the 5hmC homeostasis in OA chondrocytes leading to a global increase in 5hmC levels.
An age-dependent increase in 5hmC levels in the mouse brain has previously been reported (), but the effects of aging on 5hmC levels in chondrocytes have not previously been studied. The normal chondrocytes did not show any significant age-dependent changes in 5hmC levels. Comparison of chondrocytes from normal subjects and OA patients of similar age groups (ages 34, 39, 41, and 42 years for normal chondrocytes and ages 49, 51, 56, and 57 years for OA chondrocytes [Figures 2A and B]) revealed that 5hmC levels in OA patients increased abruptly and not gradually with an increase in the subject's age. Therefore, the high 5hmC levels are likely associated with OA pathology and are not merely a reflection of normal aging in the older OA patients.
It is known that 5hmC is an intermediate in DNA demethylation as well as an epigenetic mark by itself. OA is the first example of a pathologic situation with a global increase in 5hmC levels; however, the exact tissue-specific distribution of 5hmC and its effects on gene expression are just beginning to be unraveled (). We therefore initially focused on characteristic genes known to be critical to the pathogenesis of OA in order to understand the role of 5hmC in OA.
Previous studies have shown significant DNA demethylation at specific sites in the promoters of MMPs 3, 13, and 9 in OA patients as compared to control groups (). Those experiments used either bisulfite sequencing or restriction enzymes that cannot distinguish between 5mC and 5hmC (). We therefore used a recently reported method based on glucosylation and restriction enzyme digestion that can accurately distinguish 5mC from 5hmC ([26, 27]). This method is more sensitive than antibody-based methods for the locus-specific detection of 5hmC in gene promoters with a medium-to-low CpG density. Restriction enzymes Msp I and Hpa II both recognize the sequence CCGG. While Hpa II cleavage is blocked by any cytosine modification (i.e., 5mC or 5hmC), Msp I cleavage is selectively blocked by glucosylation of the 5hmC residues. Glucosylation followed by digestion with Msp I or Hpa II and quantitative PCR therefore allows the detection of fractional 5hmC and 5hmC plus 5mC content at the specific CCGG site.
We examined the 5hmC and 5mC content at CCGG sites in the promoters of genes associated with OA pathology (MMP-1, MMP-3, and ADAMTS-5) and in the promoter of the housekeeping gene HPRT-1. The CCGG sites interrogated were chosen based on their proximity to the transcription start site (TSS) and to the specific CpG sites reported to be significantly demethylated in the promoters of MMP-3 (–635 bp from the TSS) and MMP-1 (–1,538 bp from the TSS) ([20, 34]) (primer sequences are shown in Patients and Methods). In addition, the methylation status at the –110-bp CpG site in the proximal MMP-13 promoter has been demonstrated to regulate gene expression ([21, 35]). We were, however, unable to interrogate the 5hmC content in the MMP-13 promoter using this restriction digestion technique due to the lack of CCGG sites.
Using glucosylation and digestion followed by quantitative PCR analysis, we observed that 5hmC is indeed increased in OA chondrocytes as compared to normal chondrocytes in the promoters of both MMP-3 and MMP-1 (Figures 3A and B). The percentage of 5hmC in the OA chondrocytes varied from 20% to 50% in the MMP-3 promoter while remaining undetected for normal samples (Figure 3A). A corresponding decrease in 5mC was observed, with values of ∼40–60% in normal chondrocytes and 15–30% in OA chondrocytes in the MMP-3 promoter. A 20–60% increase in 5hmC at the MMP-1 promoter site in OA chondrocytes was observed, with a concomitant decrease in 5mC (Figure 3B). A similar increase in 5hmC (20–35%) was observed at another site (–1,024 bp from the TSS) in the MMP-1 promoter (data not shown).
In contrast, no 5hmC was detected at the ADAMTS-5 promoter site in either the normal or the OA samples (Figure 3C). Correspondingly, no consistent change in ADAMTS-5 expression was observed in OA chondrocytes (Figure 1F). As a housekeeping control, we investigated the 5mC and 5hmC content of HPRT-1 at 2 promoter sites (–511 bp and –710 bp), and as expected, no significant difference in the 5hmC, 5mC, or unmethylated cytosine content was observed between the normal and OA samples (data not shown).
Statistical analysis of the averaged data for normal subjects and OA patients, using Student's 2-tailed t-test, demonstrated that the average 5hmC level was significantly increased in OA patients as compared to normal subjects at both the MMP-3 and the MMP-1 promoter CCGG sites (P = 0.0014 and P = 0.0038, respectively). The average 5mC level, on the other hand, was significantly decreased at the MMP-3 and MMP-1 sites in OA patients (P = 0.001 and P = 0.0208, respectively). There was no significant difference in the average unmethylated cytosine levels for MMP-3 or MMP-1, or in the 5hmC, 5mC, or unmethylated cytosine levels for ADAMTS-5, between normal and OA samples.
In order to explore the mechanism by which 5hmC accumulates, we tested whether any of the DNA demethylation regulators implicated in 5hmC maintenance and turnover were altered in OA (Figure 4A). We first used qPCR analysis to examine the relative levels of TETs 1, 2, and 3 in normal chondrocytes. All 3 TET family members were expressed in chondrocytes, but the levels of TET2 were 2–4-fold higher than the levels of TET1 and TET3 (Figure 4B). Gene expression of the base excision repair glycosylase, TDG, remained relatively constant between normal and OA chondrocytes, showing that the 5hmC accumulation was not simply a result of loss of TDG function (Figure 4C). Surprisingly, OA chondrocytes showed a significant down-regulation of TET1 gene and protein expression, while no significant differences in the expression of TETs 2 and 3 were observed between normal subjects and OA patients (Figures 4D, E, and F). Gene expression of AID/APOBEC family members was similar in normal and OA chondrocytes (data not shown). It was a little unexpected that, unlike in cancer ([17, 36]), a loss of TET1 expression corresponded to an increase in 5hmC levels, rather than a decrease, in OA chondrocytes. Increased conversion of 5mC to 5hmC was also not a simple effect of increased TET2 or TET3 levels. Taken together, the increase in 5hmC and loss of TET1 function demonstrate a clear perturbation of the 5hmC homeostasis in OA chondrocytes.
End-stage OA (like rheumatoid arthritis [RA]) is characterized by an inflammatory microenvironment in the joint. To gain a mechanistic understanding, we investigated the effect of inflammation per se on normal chondrocytes. Normal human chondrocytes were treated with IL-1β or with TNFα plus oncostatin M to induce the cytokine-triggered damage observed in the inflammatory microenvironment in OA ([37, 38]). To determine cell viability upon cytokine treatment, we used a 2-color fluorescence–based assay that simultaneously identified live and dead cells based on intracellular esterase activity and membrane permeability (see Patients and Methods).
Chondrocytes were viable upon cytokine treatment, and the mean ± SD percentage of live cells was determined to be 99.6 ± 0.35 with control treatment, 99.6 ± 0.43 with IL-1β treatment, and 97.7 ± 2.0 with TNFα plus oncostatin M treatment (Figure 5A). Treatment with either IL-1β or TNFα plus oncostatin M did, however, cause a dramatic up-regulation of MMPs 1, 3, and 13 (but not MMP-9 [data not shown]), mimicking an OA-like response in the normal chondrocytes (Figure 5B). TET2 expression was unchanged with either cytokine treatment, while TET3 expression was variable, showing down-regulation in response to IL-1β, but essentially no change with TNFα plus oncostatin M (Figure 5C). Importantly, cytokine-treated chondrocytes showed a significant down-regulation of TET1 messenger RNA (mRNA) expression, similar to that in the OA patients, and decreases in TET1 protein expression with both IL-1β (10%) and TNFα plus oncostatin M (50%) treatment (Figures 5D and E). The effect of IL-1β treatment in down-regulating both TET1 mRNA and protein was modest as compared to the effect of TNFα plus oncostatin M treatment. Accordingly, up-regulation of MMPs 1, 3, and 13 was many times higher upon TNFα plus oncostatin M treatment as compared to IL-1β treatment (Figure 5B).
Upon testing global DNA methylation and hydroxymethylation after treatment of normal chondrocytes with the inflammatory cytokines (both IL-1β and TNFα plus oncostatin M), we did not observe any significant increase in global 5hmC or 5mC levels in immunoblot experiments (Figures 6A and B). Hence, unlike in OA chondrocytes, TET1 down-regulation in normal chondrocytes upon treatment with IL-1β or TNFα plus oncostatin M was not accompanied by an appreciable increase in 5hmC levels (Figures 6A and B). Accordingly, no significant locus-specific enrichment of 5hmC was observed at the MMP-3 or MMP-1 promoter, in contrast to the OA chondrocytes (Figures 6C and D). Percentages of 5hmC at the ADAMTS-5 promoter were also unchanged (Figure 6E). Taken together, these experiments showed that although even a short-term treatment with the inflammatory cytokines is sufficient to induce TET1 loss, it does not lead to an increase in 5hmC.
These observations suggest that either TET1 down-regulation and 5hmC enrichment are independent and distinct events in OA pathology or a prolonged TET1 down-regulation is required for 5hmC accumulation. However, these experiments demonstrate clearly that even short-term treatment with inflammatory cytokines can lead to TET1 down-regulation in chondrocytes, thereby linking inflammation and TET1 function.
OA has a complex pathogenesis, affected by both genetic and environmental factors, and the early causal and sequential events during OA development are still unclear (). Our understanding of the epigenetic changes that take place during the initiation and progression of OA is also limited. Previous work has suggested a role of DNA demethylation in “unsilencing” OA-associated enzymes ([3, 18]). Our study provides the first evidence of a dysregulation of the DNA methylation dynamics in OA. It is surprising that a significant down-regulation of TET1 is concomitant to an increase in 5hmC in OA chondrocytes, considering that a loss of TET1 in melanoma, prostate, and breast cancer leads to an opposite effect, a global loss of 5hmC ([17, 36]). These observations, however, underscore the dynamic nature of the global 5hmC homeostasis and its profound effect on transcription and cellular fate in pathologic conditions.
All of the TET family members, TETs 1, 2, and 3, are capable of the conversion of 5mC to 5hmC, as well as further oxidation to 5fC and 5caC intermediates (). TET1 function, however, appears to be distinct from that of TETs 2 and 3 in OA pathogenesis, since only TET1 expression was down-regulated in OA chondrocytes. However, TET function involves multiple known and hitherto unknown cofactors, and hence, it remains possible that the function of TET2 or TET3 is perturbed in OA chondrocytes independently of their gene expression. A decrease in TET1 function can potentially lead to 5hmC accumulation if TET1 is the major contributor to the conversion of 5hmC to 5fC or 5caC; however, the exact role of the TET proteins in the stepwise generation of 5hmC as well as its removal (by conversion to 5fC and 5caC) is still undetermined.
MMPs play important roles in diverse biologic processes and pathologies, including cancer and arthritis (both OA and RA), and their transcriptional regulation has thus been widely studied (). The global increase in 5hmC translates to locus-specific 5hmC enrichment in both MMP-1 and MMP-3, which play significant roles in OA pathology. The promoters of most MMPs, including MMPs 3, 9, and 13, do not contain dense CpG regions. Although initially, differential methylation of CpG islands was mainly thought to influence gene expression, it is now known that low-density CpG regions contribute extensively to tissue-specific gene regulation ([40, 41]). In addition, while 5mC enrichment appeared to be regulatory mostly in promoters and distal regions, the presence of 5hmC in gene bodies shows a strong positive correlation with high levels of gene expression in embryonic stem cells and neurons ([31, 42-45]).
Our studies demonstrate that 5hmC enrichment at specific CpG sites in the MMP-1 and MMP-3 promoters correlates with increased gene expression in OA chondrocytes. These observations are consistent with studies showing that 5hmC enrichment at promoters and gene bodies correlate with high levels of gene expression (). A study in OA patients had previously demonstrated significantly higher DNA demethylation associated with the –36-bp site for MMP-9, the –635-bp site for MMP-3, and the –110-bp site for MMP-13 promoters in OA patients (). However, the restriction enzymes used in this study did not distinguish between 5mC and 5hmC; therefore, the changes in 5mC were likely to have been understated.
The MMP-3 site used in our study is the Hpa II site (similar to the –686-bp site in the previous study; now designated –541 bp based on updated sequence) that is not demethylated in OA chondrocytes. Significant hydroxymethylation at this site was observed in 20–40% of OA chondrocytes, with a concomitant decrease in methylation. Consistent with a previous report (), we did not observe any increase in unmethylated cytosines. This site is proximal to the CpG site undergoing significant DNA demethylation in OA chondrocytes (–635 bp), suggesting a role of 5hmC in active DNA demethylation. Both the Hpa II sites we studied in the MMP-1 promoter showed an increase in hydroxymethylation with a decrease in methylation.
A contradiction to the observations in OA chondrocytes is that inflammatory cytokines can cause an up-regulation of MMPs 1 and 3 in normal chondrocytes without any associated increase in 5hmC in their promoters. However, the exact gene expression levels of MMPs 3 and 13 were many times higher (100–1,000-fold) in the OA patients as compared to the short-term cytokine-stimulated chondrocytes (Figure 1). Therefore, it remains possible that 5hmC enrichment in OA leads to higher MMP gene expression. Another possibility is that longer-term incubation with the cytokines and interrogation of their effects on the OA chondrocytes that may respond differently from the “normal” chondrocytes used in our study, is required to recapitulate the 5hmC dynamics in OA. In addition, it is important to analyze other CpG sites in the MMP genes, especially in the gene bodies and exons, to clarify the role of 5hmC in regulating MMP gene expression in OA.
One of the interesting insights of this study is that inflammatory cytokines can modulate TET1 expression. Analyses in OA chondrocytes as well as in short-term cytokine-treated chondrocytes, clearly showed that down-regulation of TET1 expression (both mRNA and protein) correlates with an up-regulation of the expression of MMPs 1, 3, and 13. These observations implicate TET1 as a novel modulator of OA pathology. Further analyses of the role of TET1 in normal and OA chondrocytes would be required to understand whether MMPs constitute direct or indirect targets of TET1 or whether TET1 and MMPs are modulated by a common upstream regulator. Regulation of TET1 expression by inflammatory cytokines could be a common nexus for epigenetic changes observed in cancer, OA, and possibly RA and other inflammation-associated disorders. The effect of aging on 5hmC levels is another area that needs further experimentation. Although there did not appear to be an age-associated increase in the 5hmC levels in the normal chondrocytes we tested, the disparity in the age range of the normal subjects and OA patients would not allow us to completely rule out an effect of aging based on the present observations. Detailed experimentation in a mouse model, as has been conducted for assessments of postnatal neurodevelopment and aging (), will be useful for addressing how normal aging affects 5hmC regulators and homeostasis in cartilage.
Enrichment in 5hmC in a few genes associated with OA suggests that it is a potential regulator of altered gene expression in OA; however, a broader understanding of its role will be provided by in-depth genome-wide analyses of the distribution of 5hmC and the subsequent effects on gene expression in normal and OA chondrocytes. Antibody-based approaches, using DNA enrichment with specific anti-5hmC and anti-5mC antibodies followed by high-throughput sequencing, have already been used in embryonic stem cells and neurons to map the distribution of 5hmC ([31, 42, 44-46]). Enrichment of 5hmC in euchromatic regions, at promoters with intermediate (not high) CpG density and in gene bodies of actively transcribed genes, are the common features observed in these studies. The precise effects of the location of 5hmC on transcriptional regulation and its modulation by other chromatin modifications in a context- and tissue-dependent manner are areas of vigorous investigation. Given the intriguing association of 5hmC with fundamental processes such as stem cell differentiation and pathologies such as cancer and OA, future studies will provide fundamental insights into the biology of 5hmC as well as suggest approaches for therapeutic interventions.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Bhutani had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Taylor, Smeriglio, Bhutani.
Acquisition of data. Taylor, Smeriglio, Dhulipala, Rath.
Analysis and interpretation of data. Taylor, Smeriglio, Bhutani.
We are grateful to Drs. Stuart Goodman and Jason Dragoo for their kind assistance in the procurement of cartilage samples from patients undergoing surgery. We thank Alex H. Harris for consulting on the statistical analyses and Professor R. L. Smith for insightful discussions.