Purpose: Epigenetic alterations are one of the common cellular hallmarks of aging. Several studies have identified CpG sites whose methylation levels change with age in different tissues, and DNA methylation at 353 ‘clock’ CpG sites can be used to calculate epigenetic age independently of tissue, cell type and disease status. Several human diseases are associated with accelerated epigenetic aging, including Down syndrome and obesity, and it has been hypothesised that the age-related loss of normal epigenetic control may be responsible for the late onset of common human diseases. Given that age is the strongest risk factor for OA, we decided to investigate if OA is associated with accelerated epigenetic aging in cartilage. Methods: A meta-analysis was performed on three cartilage methylation datasets, which used identical sample types and were generated using the Infinium HumanMethylation450 BeadChip. The raw IDAT data files were normalised using the preprocessFunnorm function from the Minfi package. The following CpG probes were removed from the analysis: probes with a detection P value threshold of <0.01 in at least 50% of samples; those on the sex chromosomes and; probes containing SNPs with a MAF of greater than 5%. This resulted in a total of 422,069 CpG probes for the age analysis. After QC and normalisation, 179 cartilage samples with an age range of 45 to 95 years (mean age of 71.7 years) were included in the analysis; these were composed of non-OA hip cartilage, preserved and lesioned OA hip cartilage, and preserved and lesioned OA knee cartilage. Age regression tests were performed using the Limma package with linear models that controlled for the different origins of the datasets. A Benjamini-Hochberg adjusted P value cut-off of <0.05 was used to identify age-associated CpGs (age-CpGs). Gene ontology analysis was performed using Ingenuity Pathways Analysis. Results: No correlation was observed between mean probe methylation of all probes included in the analysis and age. 716 and 345 age-CpGs were identified when all samples (n = 179) or all OA samples (n = 144) were analysed, respectively. The majority of age-CpGs were hypermethylated with age (91.3% of cartilage and 93.3% of OA-only age CpGs). Age-CpGs were enriched relative to non-age CpGs within CpG islands (2.0 fold and 2.15 fold, respectively) and depleted in the regions >2kb away from a CpG island (p<0.05, Pearson Chi2 test). Significantly fewer age-CpGs were located in genic regions than expected, but those within gene loci were enriched in the region spanning 200bp upstream of the transcription start site (1.6 fold and 1.3 fold for cartilage age-CpGs and OA age-CpGs, respectively) and depleted in gene body and 3ʹUTR. Age-CpG sites were significantly enriched for binding sites of the polycomb proteins EZH2, SUZ12 and the co-repressor protein CTBP2, and depleted for sites where the histone demethylase KDM5A transcription factor binds (p<0.001, Pearson Chi2 test). After removing age-CpG sites identified from other studies in other tissues, we identified 605 age-CpGs specific to cartilage and 126 specific to OA cartilage. The cartilage-specific age-CpGs were enriched for in genes involved in cancer, embryonic and organismal development, and neuronal migration and function. OA-specific age-CpGs were enriched for in genes involved in cancer, analgesia, synaptic transmission, and embryonic development; relevant canonical pathways included circadian rhythm signalling and neuropathic pain signalling in dorsal horn neurons. There was limited overlap in age-CpGs between the five cartilage subtypes when analysed separately, especially between hip and knee cartilage. We used the 353 ‘clock’ CpGs to calculate DNA methylation age for each sample, and observed accelerated epigenetic aging in males relative to females (p = 0.019, Kolmogorov-Smirnov 2 sample t test), and in OA hip samples relative to OA knee samples independently of gender (p = 0.012). DNA methylation age was different between preserved and lesioned cartilage from the same joint, although the direction of this difference varied between individuals. Conclusions: We have identified CpG sites whose methylation levels significantly correlate with chronological age in cartilage, over 90% of which are hypermethylated with age. The limited age-CpG overlap and accelerated epigenetic aging observed in OA hip cartilage relative to OA knee cartilage suggests that the aging processes may be different in cartilage from different joint sites, although cell communication and development pathways are common targets of age-related methylation changes for both joints.

Purpose: It has become increasingly clear that DNA methylation is eminently involved in osteoarthritis (OA) pathology, as reflected by the large number of differentially methylated CpGs that have been reported between healthy, preserved and lesioned articular cartilage. However, as with the vast number of differentially expressed genes observed in OA affected articular cartilage, it remains unclear whether these observed differences are either cause or consequence of the disease. By combining genome wide methylation, expression and single nucleotide polymorphism (SNP) datasets we aim to achieve directional insight in the observed epigenetic and transcriptional changes in OA affected articular cartilage and report on putative protective and susceptibility loci. Methods: Transcriptional activity of CpGs (t-CpGs) was assessed using genome wide gene expression and DNA methylation data of respectively 33 and 31 pairs of preserved and lesioned articular cartilage. Disease responsive t-CpGs were identified by means of differential methylation between preserved and lesioned cartilage. Additionally, we addressed proximal SNPs near the OA responsive t-CpGs. Statistical analyses were corrected for age, sex, joint and technical covariates, while a random effect was included to correct for possible correlations between paired samples. Results: Of the 9838 transcribed genes in articular cartilage, 2324 correlated significantly with the methylation status of 3748 t-CpGs, both canonically negative (N=1741) as positive (N=2007) correlations were observed. Hypomethylation and hypermethylation (FDR<0.05, |Δβ|>0.05) were observed for 92 and 59 t-CpGs, respectively, covering 117 unique genes. Significant enrichment for developmental and ECM maintenance pathways was observed, indicating possible reactivation of endochondral ossification. Finally, we observed 11 and 84 OA cartilage relevant genes of which, respectively, methylation and expression is additionally affected by genetic variation. Among others, we here present the potential disease driving genes ROR2 and CAV1 (figure 1), as reflected by differential expression between preserved and lesioned cartilage and, moreover, epigenetic and genetic transcriptional consequences. Conclusions: We have shown that OA related epigenetic differences need to be integrated with other sources of molecular data, such as genomic and transcriptomic, to enhance our understanding of the pathophysiological processes of OA. Furthermore, by integration of multiple layers of genome wide data we have identified genes, such as ROR2 and CAV1, which are likely functionally involved in OA pathophysiology, as opposed to sheer consequence. Although targeting DNA methylation seems unlikely to stand at the basis for developing treatments, it serves to deepen our understanding of the complex transcriptomic changes in OA affected articular cartilage. Finally, our results comprise an important step in understanding the reported widespread epigenetic changes occurring in OA affected articular cartilage.

OBJECTIVES:To elucidate the functional epigenomic landscape of articular cartilage in osteoarthritis (OA) affected knee and hip joints in relation to gene expression. METHODS:Using Illumina Infinium HumanMethylation450 BeadChip arrays, genome-wide DNA methylation was measured in 31 preserved and lesioned cartilage sample pairs (14 knees and 17 hips) from patients who underwent a total joint replacement due to primary OA. Using previously published genome-wide expression data of 33 pairs of cartilage samples, of which 13 pairs were overlapping with the current methylation dataset, we assessed gene expression differences in differentially methylated regions (DMRs). RESULTS:Principal component analysis of the methylation data revealed distinct clustering of knee and hip samples, irrespective of OA pathophysiology. A total of 6272 CpG dinucleotides were differentially methylated between the two joints, comprising a total of 357 DMRs containing 1817 CpGs and 245 unique genes. Enrichment analysis of genes proximal of the DMRs revealed significant enrichment for developmental pathways and homeobox (HOX) genes. Subsequent transcriptomic analysis of DMR genes exposed distinct knee and hip expression patterns. CONCLUSIONS:Our findings reveal consistent DMRs between knee and hip articular cartilage that marked transcriptomic differences among HOX genes, which were not reflecting the temporal sequential HOX expression pattern during development. This implies distinct mechanisms for maintaining cartilage integrity in adulthood, thereby contributing to our understanding of cartilage homeostasis and future tissue regeneration approaches.