Abstract
Objective
Cartilage epigenetic changes are strongly associated with human osteoarthritis (OA). However, the influence of individual environmental OA risk factors on these epigenetic patterns has not been determined; herein we characterize cartilage DNA methylation patterns associated with aging and OA in a mouse model.
Methods
Murine knee cartilage DNA was extracted from healthy young (16‐week, n = 6), old (82‐week, n = 6), and young 4‐week post–destabilization of the medial meniscus (DMM) OA (n = 6) C57BL6/J mice. Genome‐wide DNA methylation patterns were determined via Illumina BeadChip. Gene set enrichment analysis was performed by Ingenuity Pathway Analysis. The top seven most differentially methylated positions (DMPs) were confirmed by pyrosequencing in an independent animal set. Results were compared to previously published human OA methylation data.
Results
Aging was associated with 20,940 DMPs, whereas OA was associated with 761 DMPs. Merging these two conditions revealed 279 shared DMPs. All demonstrated similar directionality and magnitude of change (Δβ 1.0% ± 0.2%, mean methylation change ± SEM). Shared DMPs were enriched in OA‐associated pathways, including RhoA signaling (P = 1.57 × 10−4), protein kinase A signaling (P = 3.38 × 10−4), and NFAT signaling (P = 6.14 × 10−4). Upstream regulators, including TET3 (P = 6.15 × 10−4), immunoglobulin (P = 6.14 × 10−4), and TLR7 (P = 7.53 × 10−4), were also enriched. Pyrosequencing confirmed six of the seven top DMPs in an independent cohort.
Conclusion
Aging and early OA following DMM surgery induce similar DNA methylation changes within a murine OA model, suggesting that aging may induce pro‐OA epigenetic “poising” within articular cartilage. Future research should focus on confirming and expanding these findings to other environmental OA risk factors, including obesity, as well as determining late OA changes in mice.
INTRODUCTION
Osteoarthritis (OA) is a chronic debilitating musculoskeletal disease characterized by progressive loss of joint function and includes pain, loss of mobility, increase in morbidity, and early mortality. It is the leading cause of chronic disability in the United States, affecting roughly half of adults over 65 years of age (1). Despite its impact, there are no disease‐modifying drug therapies available, due in no small part to an incomplete understanding of OA pathogenesis. Epigenetics, defined as heritable changes in gene expression/function that occur in the absence of mutations of underlying genomic DNA, is a common mechanism whereby organisms alter gene transcription in response to both external and internal environmental cues, for example, diet, mechanical stressors, obesity‐related adipokines, inflammatory cytokines, and aging. Epigenetic regulation occurs by a few common mechanisms, including cytosine genomic DNA methylation, histone modifications or variants, nucleosome spacing, and noncoding RNA. The most widely studied epigenetic control mechanism is DNA methylation.
Early epigenetic analyses of candidate genes in OA cartilage identified significant changes in DNA methylation signatures of several matrix metalloproteinase genes, as well as genes linked to obesity and inflammation (2). A previous genome‐wide methylation study of knee OA cartilage identified several differentially methylated genes in patients with OA compared to healthy controls (3), which also uncovered a cluster of patients with OA characterized by differential methylation of inflammation‐related genes. We reported significant alterations in DNA methylation patterns of eroded compared to intact OA cartilage from the same joint. Further, we revealed associations between DNA methylation and histopathologic score within these specimens and epigenetic alterations of roughly one third of known OA susceptibility genes (4), raising the possibility of epigenetics as a mediator of genetic risk in OA. We subsequently found similar changes among many more genes in hip OA subchondral bone underlying eroded and intact cartilage (5). Other groups have confirmed an altered, geographically distinct epigenome in cartilage from human knee and hip OA compared to control cartilage from patients with neck‐of‐femur fracture, as well as differences along the spectrum of OA disease severity (3, 6, 7).
Although these studies provide strong evidence that human OA is associated with epigenetic changes, it is impossible to deconvolute the specific epigenetic contributions to OA driven by various environmental factors. This is of particular importance given that the strongest risk factors for OA development are nongenetic, including age, obesity, and antecedent trauma (8). Additionally, epigenetic “priming” for disease development by aging has been described in a number of other age‐related diseases, including malignancy (9), although this has not yet been evaluated in OA. In the present study, we sought to determine epigenetic changes in articular cartilage–associated aging and trauma, and the overlaps thereof, in the destabilization of the medial meniscus (DMM) posttraumatic mouse model of knee OA.
MATERIALS AND METHODS
Ethics statement
The institutional animal care and use committee of the Oklahoma Medical Research Foundation (OMRF) approved this study (OMRF IACUC protocol numbers 16‐40, 19‐43, 20‐29, 19‐56, 18‐45, 18‐18). All animal husbandry procedures adhered to the National Institute of Health Guide for the Care and Use of Laboratory Animals. There were no unexpected adverse events during the course of these experiments.
Mouse husbandry, tissue harvest
Young (12 weeks) and old (78 weeks) C57BL6/J male mice were fed a chow diet (PicoLab Rodent Diet 20, LabDiet #5053) and were exposed to a 12‐hour light‐dark cycle. In a subset of young mice, DMM surgery was performed on a unilateral stifle (knee) joint at 12 weeks of age by an experienced surgeon (VI) 4 weeks prior to euthanasia. Only male mice were included in our study because female animals do not reliably exhibit an OA phenotype following DMM (10). Animals were euthanized at 16 weeks (young) or 82 weeks (old), and cartilage was collected. In DMM animals, samples were collected from both the operated and unoperated knees. Our final analysis included six young chow animals, seven old chow animals, and six young animals post‐DMM. Knee joints were dissected in a biosafety cabinet using sterilized and DNA‐decontaminated instruments. Articular cartilage was removed from the tibia and femur and flash frozen in liquid nitrogen, then cryogenically ground using a Precellys Cryolys instrument (Bertin, Bretonneux, France), and DNA was extracted using a DNEasy kit (Qiagen).
DNA methylation quantitation and analysis
Following quantitation, 500 ng of genomic DNA was treated with sodium bisulfite (Zymo EZ‐DNA methylation kit) and loaded onto Illumina Infinium Mouse Methylation arrays, which quantify more than 285,000 methylation (CpG) sites per sample at single‐nucleotide resolution. It includes balanced coverage of CpG islands, transcription start sites, enhancers, imprinted foci, and other regions. DNA methylation β values (fraction mean methylation at a given CpG site, 0–1 scale) were extracted and normalized using the SeSAMe package (11), which implements a P value with out‐of‐band array hybridization (pOOBAH) approach to reduce spurious methylation results due to hyperpolymorphic regions and technical variation. CpG sites were considered statistically significant between groups if both the Benjamini‐Hochberg‐corrected group q ≤ 0.05 and the absolute difference in group mean DNA methylation values (|Δβ|) was 5% or more. CpG island comparisons were made based on a χ2 test with a Pearson P value calculation.
Gene set enrichment analysis
Functional properties, networks, pathways, and upstream regulators enriched in differentially methylated genes were assessed using the Ingenuity Pathway Analysis (IPA) system (Ingenuity Systems) using the Ingenuity Knowledge Base reference set. Direct and indirect relationships were calculated, and experimentally observed relationships were included; P ≤ 0.05 were considered significant. Gene set enrichment patterns were then compared to our previous database of human OA‐associated DNA methylation changes from cartilage and subchondral bone in paired lesioned and intact regions of human OA cartilage (5). The upstream regulator analysis within IPA is based on prior knowledge of expected effects between transcriptional regulators and their associated target genes as indexed by Ingenuity. The analysis calculates the number of known targets of each transciption regulator within the target gene set (in our case, differentially methylated genes). The definition of upstream transcriptional regulator is intentonally broad and includes any molecule that can affect the expression of other molecules (eg, transcription factors, microRNAs, kinases, etc).
Comparison of results with previously published OA methylation data
We then sought to compare our findings with other previously published OA methylation datasets. In addition to our previously published cartilage and subchondral OA dataset (5), we identified five additional comparable studies, including three human cartilage analyses (6, 12, 13) and two baboon cartilage analyses (14, 15). Data from a 2013 study by Fernandez‐Tajes et al (3) were based on a significantly older DNA microarray platform (Illumina 27k) and were not included in our comparison. Differential DNA methylation patterns between OA and control groups were determined as described above. As these studies were conducted using differing methylation arrays and among differing species than our mouse studies, we integrated them with our present murine methylation dataset using the associated analogous gene as a common index.
Pyrosequencing confirmation
Pyrosequencing assays were then designed as previously described (5) for 7 of the top 10 differentially methylated positions (DMPs) shared in both aging and DMM comparisons (three hypo‐ and four hypermethylated CpGs) (Supplementary Table 1); pyrosequencing primers were not able to be developed for the remaining three top DMPs. We then confirmed our array results in an independent cohort of six young, six old, and six young post‐DMM mouse cartilage samples.
RESULTS
Aging is associated with substantial differential methylation within murine cartilage tissue
Our final analysis included six young OA‐free mice, six young mice with OA after unilateral DMM (cartilage samples for analysis collected from both the DMM and non‐DMM sides), and six old OA‐free mice. Aging was associated with 20,940 CpG DMPs (Supplementary Table 2); of these, 10,264 were hypomethylated in aged mice compared to young mice, and 10,676 were hypermethylated.
DMM‐induced OA is associated with significant methylation changes in murine cartilage compared to both contralateral nonoperated and wild‐type mouse joints
To evaluate the epigenetic effects of early OA, we compared cartilage methylation patterns from the operated side of young mice 4 weeks post‐DMM to the unoperated side of the same animal (Supplementary Table 3). We identified 761 DMPs, 148 hypermethylated in OA, and 613 hypomethylated in OA. Although less statistically powerful, we also compared the operated side of post‐DMM mice to young control mice and found 59 DMPs, 5 hypermethylated and 54 hypomethylated (Supplementary Table 4). Of these, 38 were also differentially methylated in the DMM− versus DMM+ analysis, and all 38 demonstrated similar methylation changes in the two comparisons.
Aging‐ and OA‐associated DNA methylation changes overlap and are of similar magnitude and direction
We next compared aging‐associated (young vs. old) and OA‐associated (DMM− vs. DMM+) DMPs. We identified 279 DMPs shared among the two groups, representing 37% of age‐associated DMPs. All 279 overlapping DMPs demonstrated similar directionality in both OA and age and were of strikingly similar magnitude (difference in Δβ 1.0% ± 0.2%, mean ± SEM) (Table 1, Figure 1, Supplementary Table 5). Shared DMPs were less likely to be located in CpG islands (2.2% vs. 11.5%; P < 0.001) and less likely to be in CpG island shores (6.8% vs. 12.4%; P = 0.001) than would be expected based on the methylation array makeup.
Table 1.
Top 20 most differentially methylated positions shared among OA and aging groups
| ProbeID | Discovery cohort (DNA methylation microarrays) | Validation cohort (pyrosequencing) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| OA (q value) | Age (q value) | OA (Δβ) (non‐DMM side − DMM side) | Age (Δβ) (young − aged) | Gene symbol | CpG position relative to CpG islands | Gene transcript type | OA (Δβ) | OA P value | Age (Δβ) | Age P value | |
| cg38526099_TC11 | 0.03 | n/a2 | −0.18 | −0.25 | |||||||
| cg32648481_BC11 | 0.05 | n/a1 | −0.11 | −0.19 | Sorbs3 | Island | Protein coding | −0.07 | 0.02 | −0.09 | 0.04 |
| cg45614555_TC11 | 0.05 | n/a2 | −0.14 | −0.18 | Mtus1 | Protein coding | −0.09 | 0.04 | −0.10 | 0.1 | |
| cg34977518_BC21 | 0.04 | 0.01 | −0.15 | −0.17 | Trerf1 | Processed transcript | −0.17 | n/a3 | −0.23 | n/a7 | |
| cg38745200_BC21 | 0.05 | n/a4 | −0.10 | −0.16 | Slc1a2 | Protein coding | |||||
| cg31435747_BC11 | 0.04 | 0.03 | −0.18 | −0.16 | Ppp1r3g | Shore | Protein coding | ||||
| cg41186133_BC21 | 0.02 | n/a4 | −0.08 | −0.15 | Pgm1 | Protein coding | |||||
| cg31530167_BC21 | 0.04 | n/a6 | −0.11 | −0.15 | Atxn1 | Protein coding | |||||
| cg42239266_TC21 | 0.04 | 0.01 | −0.15 | −0.14 | Limch1 | Shelf | Protein coding | ||||
| cg35134520_TC21 | 0.04 | 0.01 | −0.09 | −0.14 | Ptprm | Protein coding | |||||
| cg37063318_BC11 | 0.03 | 0.02 | 0.24 | 0.22 | Gm28818 | lincRNA | |||||
| cg47488822_TC11 | 0.02 | 0.02 | 0.21 | 0.22 | |||||||
| cg30227623_BC21 | 0.04 | n/a4 | 0.24 | 0.22 | Rnf157 | Protein coding | |||||
| cg38801114_TC11 | 0.02 | 0.03 | 0.20 | 0.23 | |||||||
| cg39372503_BC21 | 0.03 | n/a1 | 0.20 | 0.23 | |||||||
| cg41450683_BC21 | 0.02 | 0.02 | 0.10 | 0.23 | 0.06 | n/a7 | 0.03 | 0.27 | |||
| cg40344526_BC11 | 0.03 | 0.02 | 0.24 | 0.24 | Gm29865 | lincRNA | 0.10 | 0.05 | 0.08 | 0.05 | |
| cg45366695_BC11 | 0.02 | n/a6 | 0.24 | 0.24 | Gm44956 | lincRNA | 0.06 | 0.02 | 0.05 | 0.04 | |
| cg35512683_TC21 | 0.03 | 0.01 | 0.26 | 0.26 | Gm18085 | Processed pseudogene | |||||
| cg42300124_TC21 | 0.04 | n/a1 | 0.22 | 0.27 | Lnx1 | Protein coding | 0.08 | 0.02 | 0.12 | 0.05 | |
Abbreviations: CpG, methylation; DMM, destabilization of the medial meniscus; OA, osteoarthritis.
Figure 1.
Differentially methylated CpG sites (DMPs) shared in aging (old animals vs. young) and OA (young animal DMM+ knee compared to contralateral DMM− knee). y‐axis represents mean Δβ (difference in mean group DNA methylation fraction), and x‐axis represents individual CpG sites arranged from most‐hypomethylated to most‐hypermethylated with aging. CpG, methylation; DMM, destabilization of the medial meniscus; DMPs, differentially methylated positions; OA, osteoarthritis.
Gene set enrichment analysis
We then investigated gene pathways and upstream regulators overrepresented among differentially methylated genes associated with age + OA DMPs from earlier, which corresponded to 195 unique genes. IPA identified 60 pathways (Supplementary Table 6) (Table 2) and 208 upstream regulators (Supplementary Table 7) associated with this gene set. Age + OA shared DMP pathways including RhoA signaling (P = 1.57 × 10−4), a tissue‐injury‐response regulatory pathway (16), protein kinase A signaling (P = 3.38 × 10−4), nuclear factor of activated T‐cells (NFAT) signaling (P = 6.14 × 10−4), involved in calcineurin signaling and previously linked to OA (17), and regulation of the epithelial‐mesenchymal transition (P = 8.79 × 10−3), among others. Upstream regulators associated with shared DMPs include tet methylcytosine dioxygenase 3 (TET3) (P = 1.15 × 10−4), a DNA glycosylase mediating the first step in active DNA demethylation (18), miR‐29b‐3p (P = 6.15 × 10−4), a chondrocyte apoptotic promoter (19), immunoglobulin (P = 6.14 × 10−4), toll‐like receptor 7 (TLR7) (P = 7.53 × 10−4), involved in OA pain (20), and the mitogen‐activated protein kinases (MAPK) 9 and 8 (P = 1.32 × 10−3 and 1.57 × 10−3, respectively), members of the MAP kinase family that are potential therapeutic targets in OA (21). Associated upstream regulators include the cytokine oncostatin M (P = 1.73 × 10−3) and the Wnt transcription regulator transcription factor 7 like 2 (TCF7L2) (P = 1.94 × 10−3).
Table 2.
Comparison of shared OA and aging differentially methylated genes with previously published studies
| Gene | Δβ mouse aging | Δβ mouse OA | Mean Δβ in other published OA studies | Consistency of current data with previously published OA data | Jeffries 2016 (5) | Fan 2021 (12) | Rushton 2014 (6) | Aref‐Eshghi 2015 (13) | Housman 2018 (15) | Housman 2020 (14) |
|---|---|---|---|---|---|---|---|---|---|---|
| Study design | Human OA: eroded vs. intact | Human OA vs. non‐OA | Human OA vs. non‐OA | Human OA vs. non‐OA | Baboon OA vs. non‐OA | Baboon OA vs. non‐OA | ||||
| Ablim1 | −0.06 | −0.03 | 0.06 | No | N/A | N/A | 0.18 | 0.16 | N/A | N/A |
| Apobec2 | −0.10 | −0.03 | −0.03 | Yes | N/A | N/A | −0.17 | N/A | N/A | N/A |
| Aqp1 | 0.10 | 0.09 | 0.03 | Yes | N/A | N/A | 0.15 | N/A | N/A | N/A |
| Arsb | 0.13 | 0.08 | −0.03 | No | N/A | N/A | −0.17 | N/A | N/A | N/A |
| Atxn1 | −0.15 | −0.03 | −0.01 | Yes | 0.18 | N/A | −0.27 | −0.16 | N/A | 0.19 |
| Bicc1 | 0.09 | 0.14 | 0.05 | Yes | 0.16 | N/A | N/A | 0.17 | N/A | N/A |
| Cacna1c | 0.12 | 0.10 | −0.01 | No | N/A | 0.25 | −0.16 | −0.20 | 0.02 | N/A |
| Capzb | 0.11 | 0.06 | 0.04 | Yes | −0.15 | N/A | 0.22 | 0.20 | N/A | N/A |
| Cd82 | −0.12 | −0.05 | −0.04 | Yes | N/A | N/A | N/A | N/A | 0.02 | −0.28 |
| Cfh | 0.21 | 0.16 | −0.03 | No | N/A | N/A | −0.20 | N/A | N/A | N/A |
| Clip1 | 0.13 | 0.09 | 0.03 | Yes | 0.20 | N/A | N/A | N/A | N/A | N/A |
| Cmip | 0.09 | 0.08 | −0.06 | No | N/A | N/A | −0.17 | N/A | N/A | −0.18 |
| Dst | 0.06 | 0.04 | 0.06 | Yes | N/A | 0.18 | N/A | 0.17 | N/A | N/A |
| Dync1i1 | 0.10 | 0.05 | 0.04 | Yes | N/A | N/A | N/A | 0.23 | N/A | N/A |
| Dyrk3 | 0.09 | 0.05 | 0.04 | Yes | N/A | N/A | N/A | N/A | 0.25 | N/A |
| Dysf | −0.09 | −0.06 | −0.08 | Yes | −0.21 | N/A | −0.21 | 0.16 | N/A | −0.20 |
| Epas1 | −0.13 | −0.10 | 0.03 | No | N/A | N/A | N/A | 0.17 | N/A | N/A |
| Epm2a | 0.13 | 0.12 | 0.04 | Yes | N/A | N/A | N/A | 0.22 | N/A | N/A |
| Fblim1 | 0.18 | 0.11 | 0.03 | Yes | N/A | N/A | N/A | 0.17 | N/A | N/A |
| Galnt2 | 0.10 | 0.07 | 0.01 | Yes | N/A | N/A | −0.18 | 0.23 | N/A | N/A |
| Gas7 | 0.06 | 0.04 | −0.08 | No | N/A | ‐0.15 | −0.18 | N/A | N/A | −0.15 |
| Glp2r | 0.08 | 0.07 | 0.03 | Yes | N/A | N/A | N/A | 0.15 | N/A | N/A |
| Gng7 | −0.07 | −0.07 | 0.06 | No | N/A | N/A | 0.15 | N/A | N/A | 0.18 |
| Hivep2 | 0.13 | 0.07 | 0.05 | Yes | N/A | N/A | 0.20 | N/A | 0.09 | N/A |
| Ift122 | 0.19 | 0.17 | 0.03 | Yes | N/A | N/A | N/A | 0.18 | N/A | N/A |
| Itgb5 | 0.18 | 0.19 | 0.03 | Yes | N/A | N/A | 0.18 | N/A | N/A | N/A |
| Jdp2 | 0.14 | 0.15 | 0.03 | Yes | 0.16 | N/A | N/A | N/A | N/A | N/A |
| Lgals3bp | 0.08 | 0.04 | 0.04 | Yes | N/A | N/A | 0.25 | N/A | N/A | N/A |
| Limch1 | −0.14 | −0.08 | −0.07 | Yes | N/A | −0.22 | −0.20 | N/A | N/A | N/A |
| Lnx1 | 0.27 | 0.18 | −0.04 | No | N/A | N/A | N/A | −0.22 | N/A | N/A |
| Ly6d | −0.06 | −0.03 | 0.02 | No | N/A | N/A | N/A | N/A | 0.11 | N/A |
| Mtus1 | −0.18 | −0.10 | 0.05 | No | N/A | N/A | ‐0.18 | 0.21 | N/A | 0.29 |
| Mylk | −0.07 | −0.04 | 0.08 | No | N/A | N/A | 0.16 | 0.17 | N/A | 0.16 |
| Nav1 | 0.18 | 0.10 | −0.01 | No | N/A | N/A | −0.20 | 0.16 | N/A | N/A |
| Ninj2 | 0.06 | 0.04 | −0.03 | No | N/A | N/A | N/A | −0.19 | N/A | N/A |
| Npas3 | 0.13 | 0.05 | −0.07 | No | N/A | N/A | −0.19 | −0.23 | N/A | N/A |
| Nrp2 | 0.09 | 0.06 | 0.15 | Yes | 0.17 | 0.35 | 0.25 | 0.16 | N/A | N/A |
| Osbpl3 | 0.12 | 0.09 | 0.04 | Yes | N/A | N/A | N/A | N/A | 0.22 | N/A |
| Pgs1 | 0.14 | 0.09 | 0.03 | Yes | N/A | N/A | N/A | 0.18 | N/A | N/A |
| Pik3r1 | 0.05 | 0.01 | 0.06 | Yes | N/A | N/A | 0.19 | N/A | N/A | 0.18 |
| Ppp1r3g | −0.16 | −0.05 | 0.02 | No | N/A | N/A | N/A | N/A | 0.12 | N/A |
| Prtg | −0.10 | −0.07 | 0.03 | No | N/A | N/A | N/A | 0.15 | N/A | N/A |
| Ptpn22 | 0.07 | 0.08 | 0.03 | Yes | N/A | N/A | N/A | N/A | N/A | 0.18 |
| Ptprg | −0.08 | −0.03 | 0.03 | No | N/A | N/A | N/A | 0.16 | N/A | N/A |
| Rapgef2 | −0.09 | −0.03 | 0.07 | No | N/A | N/A | N/A | 0.26 | N/A | 0.16 |
| Rnf157 | 0.22 | 0.19 | −0.03 | No | −0.21 | N/A | N/A | N/A | N/A | N/A |
| Scara5 | 0.20 | 0.10 | 0.05 | Yes | N/A | N/A | N/A | 0.16 | 0.11 | N/A |
| Shroom3 | 0.07 | 0.05 | 0.08 | Yes | 0.19 | −0.28 | 0.17 | 0.16 | 0.24 | N/A |
| Sipa1l3 | 0.06 | 0.06 | 0.06 | Yes | N/A | N/A | N/A | 0.18 | N/A | 0.18 |
| Slc12a8 | 0.17 | 0.15 | −0.03 | No | −0.16 | N/A | N/A | N/A | N/A | N/A |
| Slit3 | 0.11 | 0.07 | −0.01 | No | N/A | N/A | −0.24 | 0.21 | N/A | N/A |
| Smurf1 | 0.07 | 0.07 | 0.03 | Yes | N/A | N/A | N/A | 0.18 | N/A | N/A |
| Sorbs3 | −0.19 | −0.08 | 0.05 | No | 0.17 | N/A | N/A | 0.15 | N/A | N/A |
| Stk32c | −0.10 | −0.08 | −0.03 | Yes | N/A | N/A | ‐0.16 | N/A | N/A | N/A |
| Ston2 | 0.13 | 0.15 | −0.03 | No | ‐0.20 | N/A | N/A | N/A | N/A | N/A |
| Susd1 | 0.15 | 0.12 | 0.06 | Yes | N/A | N/A | N/A | 0.21 | 0.14 | N/A |
| Syngap1 | −0.09 | −0.04 | −0.05 | Yes | N/A | N/A | −0.17 | −0.15 | N/A | N/A |
| Tarbp2 | 0.16 | 0.15 | 0.03 | Yes | N/A | N/A | N/A | N/A | N/A | 0.17 |
| Tcf7 | 0.06 | 0.07 | −0.03 | No | N/A | N/A | −0.21 | N/A | N/A | N/A |
| Tmco4 | −0.09 | −0.03 | 0.03 | No | N/A | N/A | N/A | 0.17 | N/A | N/A |
| Tmod3 | 0.11 | 0.10 | 0.03 | Yes | N/A | N/A | N/A | N/A | N/A | 0.19 |
| Trerf1 | −0.17 | −0.11 | 0.15 | No | 0.15 | N/A | 0.35 | 0.20 | N/A | 0.21 |
| Trpm1 | 0.17 | 0.12 | 0.03 | Yes | 0.19 | N/A | N/A | N/A | N/A | N/A |
| Usp6nl | 0.12 | 0.08 | −0.03 | No | N/A | N/A | −0.17 | N/A | N/A | N/A |
| Vps13d | 0.11 | 0.10 | 0.03 | Yes | N/A | N/A | N/A | 0.17 | N/A | N/A |
| Wdfy2 | 0.14 | 0.11 | 0.03 | Yes | N/A | N/A | N/A | 0.21 | N/A | N/A |
| Zbtb42 | −0.11 | −0.06 | −0.04 | Yes | N/A | N/A | −0.24 | N/A | N/A | N/A |
| Angpt2 | 0.18 | 0.09 | 0.10 | Yes | 0.17 | N/A | 0.17 | 0.25 | N/A | N/A |
| Plekhg4 | 0.14 | 0.14 | 0.08 | Yes | N/A | N/A | N/A | N/A | 0.28 | 0.22 |
| Dlgap2 | 0.14 | 0.18 | 0.06 | Yes | N/A | N/A | 0.20 | 0.17 | N/A | N/A |
| Mical2 | 0.15 | 0.11 | 0.12 | Yes | 0.21 | N/A | 0.16 | 0.15 | N/A | 0.22 |
Note: Δβ = mean methylation values in OA − mean methylation values in control.
Abbreviations: NA, not applicable; OA, osteoarthritis.
Comparison with other OA DNA methylation datasets
We then compared these results with six previously published OA DNA methylation datasets (5, 6, 12, 13, 14, 15). These included our previously published OA cartilage methylation study (5), three additional human OA cartilage studies (6, 12, 13), and two analyses of baboon OA cartilage (14, 15). Given the different species analyzed, our comparisons were limited to the gene level rather than the specific chromosomal location. In the present study, shared differential DNA methylation sites in aging and OA were identified in 193 distinct genes. Of these, 71 genes had significant DNA methylation changes identified in at least one comparison study. The mean methylation change from these studies was concordant with our mouse findings in 61% (43 of 71) (Table 2).
Gene set enrichment analysis comparison with other OA DNA methylation datasets
Next, we performed gene set enrichment analysis (GSEA) on the previously published OA DNA methylation datasets aforementioned and compared our present mouse findings. A number of overlaps in both gene pathways and upstream regulators were identified (Figure 2, Supplementary Table 8). Of the 336 canonical pathways identified among mouse‐aging DMPs in the present study, 100 were also significantly associated with mouse OA. Many of these pathways were shared among the bulk of previously published human and baboon OA methylation studies as well, including four (extracellular signal‐regulated kinase (ERK)/MAPK signaling, cardiac hypertrophy signaling, cardiac β‐adrenergic signaling, and insulin receptor signaling) that were shared among all studies. Similarly, many upstream regulators were shared among the current study and previous datasets (Figure 2, Supplementary Table 9). Among the 1299 age‐associated upstream regulators, 213 were shared with the mouse OA. Nine upstream regulators (ETS transcription factor (ERG), GLI family zinc finger 1 (GLI1), histone H3, enhancer of Zeste 2 polycomb repressive complex 2 (EZH2), bone morphoegenic progein 7 (BMP7), transforming growth factor beta 1 (TGFB1), foxf2 (F2), helicase with RNase motif DICER1, and PR/SET domain 5 (PRDM5)) were shared among all OA datasets analyzed.
Figure 2.
Gene set enrichment analysis of DMPs from the current study (mouse aging and OA) and previously published OA DNA methylation human studies, including Jeffries et al (5), Fan et al (12), Rushton et al (6), Aref‐Eshghi et al (13), and two previously published baboon OA DNA methylation studies using human methylation microarrays, Housman et al (15) and Housman et al (14). A, Canonical pathways overrepresented among differentially methylated genes, intensity of color represents statistical significance (see legend), dot represents not significant. B, Predicted upstream regulators with bindings sites overrepresented among differentially methylated genes. DMM, destabilization of the medial meniscus; DMPs, differentially methylated positions; OA, osteoarthritis.
Pyrosequencing confirmation
Finally, to confirm our results, we then performed bisulfite pyrosequencing on 7 of the top 10 DMPs from our shared OA + aging DMP list in a separate validation cohort of mice following the same interventions as our discovery cohort (Table 1). Of these seven, all confirmed statistical significance with OA, whereas five of seven confirmed statistical significance with aging.
DISCUSSION
In this study, we performed a pilot analysis of alterations in cartilage epigenetic changes associated with aging and OA using the DMM mouse model. Our study represents the first detailed investigation of DNA methylation patterns associated with murine OA specifically and posttraumatic OA generally. We chose the 4‐week time point following DMM to capture relatively early changes within the OA joint; 4 weeks post‐DMM is the earliest time point that histologic changes are typically seen in this model, including initial ossification of osteophytes and early cartilage lesions (22), although later time points (eg, 12 weeks) have been shown to have additional DNA methylation changes (23), as discussed below.
Nevertheless, our analysis uncovered many DMPs associated with both aging and early OA post‐DMM. Three genes associated with top 20 DMPs have been previously associated with OA. Differential methylation of TRERF1 has been shown in human knee OA (24), whereas protein expression changes in PGM1 have been noted among Chinese patients with OA (25). Genetic variation in Lnx1 correlates with posttraumatic OA outcomes in animal models (26). Ingenuity analysis indicated several gene pathways shared in both aging and early OA with previous links to OA. The RhoA/ROCK pathway has been suggested in OA pathology, and inhibitors of RhoA have anti‐OA activity in vivo in animal models (27). NFAT signaling, critical for bone and joint remodeling, has been associated with OA in several studies (17). The cardiac hypertrophy signaling pathway is curious, particularly given that we previously identified strong association of this pathway among DMPs in human OA cartilage and subchondral bone. Although counterintuitive, some studies have suggested that cardiac disease and OA may share similar pathways; particularly, Wnt signaling, previously identified in OA pathogenesis, is also thought to play a role in cardiac hypertrophy and atherosclerosis (28). Cardiovascular disease is increased among patients with OA, and investigators have postulated that chronic inflammation may drive both diseases (29).
The upstream regulators we found among shared DMPs in age + OA also include several previous OA associations. The ten‐eleven translocation family is involved in active DNA demethylation, and Tet1 knockout mice are protected from OA (30). MicroRNA‐29 is a key regulator of WNT‐related genes and the vascular endothelial growth factor (VEGF) pathway and is itself dysregulated during OA development (31, 32). TLR7 is associated with OA pain, and interrupting the miRNA‐21‐TLR7 axis produces long‐lasting analgesia in animal OA models (20). MicroRNA‐29b‐3p is a noncoding RNA previously demonstrated to promote chondrocyte apoptosis; indeed, intraarticular injection of an antagomir of miR‐29b prevents cartilage loss in a rat knee OA model via interacting with progranulin (19). Intriguingly, progranulin‐deficient mice exhibit an age‐dependent spontaneous OA phenotype (33). Finally, the MAPKs play central roles in regulating several mediators of articular damage in OA; indeed, MAPK inhibitors have been suggested as potential therapeutic targets for the treatment of knee OA (21).
When comparing our present mouse results to previously published OA datasets, including our own, several commonalities emerge. Comparing DNA methylation sites across species is statistically challenging, given that the CpG probes included in the mouse and human arrays are not directly analogous; therefore, we chose to merge data by the associated gene. The mean methylation changes in previously published human and baboon data were concordant with our current mouse methylation data in approximately 61% of overlapping genes; however, this must be interpreted cautiously, because a DMP within a gene body may have a significantly different effect on gene transcription when compared to a DMP within a gene promoter. Unfortunately, the level of granularity required for genomic location annotation across these studies is available in the current analysis. A better indication of overlap in these datasets is to compare differentially methylated genes by GSEA. As illustrated in Figure 2 (sorted by the top GSEA pathways and regulators in mouse aging), a number of canonical pathways are shared among most previously published studies. These include several pathways previously associated with OA, including the NFAT pathway (17), protein kinase A signaling/RhoA signaling (16), white adipose tissue browning pathway (not specifically associated with OA, although much recent research has gone into elucidating adipose tissue as a key regulator of OA) (34), and nitric oxide signaling (35). Similarly, upstream regulators were mostly shared among the various comparison DNA methylation datasets, including P53 (36) and the TGFβ family members TGFBR2 and TGFB1 (37). Importantly, genetic variation within TGFβ family members have also been previously associated with OA in large human datasets (38).
We chose to compare our present findings with previous human and baboon OA DNA methylation datasets, as there has been only one previous report detailing epigenetic changes in mouse models of OA with disease progression. This 2021 study by Singh et al evaluated differential DNA methylation and hydroxymethylation changes (via reduced representation bisulfite and oxidative bisulfite sequencing, respectively) within mouse cartilage at early (4 weeks) and late (12 weeks) post‐DMM time points (23). They identified 842 DMPs at 4 weeks, increased to 3614 DMPs at 12 weeks post‐DMM, with Lrrc15 emerging as a key differentially methylated, and expressed gene in their analyses. Although our Illumina arrays included nine CpG sites within Lrrc15, we did not find differential methylation of any of these in either OA or aging. It should be noted, however, that this previous study used a different method (reduced representation bisulfite sequencing), with substantially broader coverage than our array‐based method.
Our study has limitations. First, we included a relatively small number of animal subjects, likely reducing our power to detect smaller‐magnitude changes in DNA methylation, although our confirmation of several most differentially methylated CpG sites in an independent confirmation cohort is reassuring. Our inclusion of only a single time point post‐DMM is also a limitation, as our data reflect only relatively early changes within the OA joint; further analyses should expand on these findings to determine longitudinal changes in cartilage methylation patterns at expanded time points following DMM. Furthermore, our pilot analysis only included a single tissue type (cartilage); future studies should expand this analysis to include additional articular and nonarticular tissues important to OA pathology, with the caveat that obtaining adequate DNA for epigenetic analysis from sparse mouse tissues (eg, synovium) is technically quite challenging. Finally, we did not include a transcriptomic analysis in our current study. Demethylation of regulatory regions alone is not sufficient to induce gene expression; rather, it simply creates a chromatin microenvironment conducive for subsequent gene expression, a concept known as epigenetic poising (39). Rather than evaluating the direct consequences of epigenetic changes in aging and post‐DMM, in this pilot study we were rather more interested in determining whether aging and DMM induce similar epigenetic poised states. Future investigations should investigate whether similar poising exists in other nongenetic OA risk factor states (eg, obesity and metabolic syndrome).
In summary, in this study we compared changes in cartilage epigenetic patterns associated with aging and OA development in the DMM mouse model. We identified many DMPs associated with aging (n = 20,940) and a relatively small number of DMPs associated with early OA (n = 761), 4 weeks after DMM surgery. Roughly one third of OA‐associated DMPs were also differentially methylated with aging (n = 279). Remarkably, aging and OA induced both a similar magnitude and direction of methylation changes in each of these shared DMPs. Methylation changes were clustered in several known OA‐associated pathways and upstream regulators and largely overlapped with our previous end‐stage human cartilage and subchondral bone OA methylation data. This pilot study represents the first comparison of epigenetic changes induced by individual OA risk factors (age and trauma) and suggests that aging may predispose cartilage to OA development via epigenetic poising of OA‐related genes.
Supporting information
Disclosureform
Table S1 Supplementary Tables
The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.
This work was supported by NIH grants K08‐AR‐070891, P20‐GM‐125528, R61‐AR‐078075, R01‐AR‐076440, and Department of Defense CDMRP grant PR191652. The funding source was not involved in the writing of this article.
Author disclosures are available at https://onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2Facr2.11506&file=acr211506-sup-0001-Disclosureform.pdf.
REFERENCES
- 1. Centers for Disease Control and Prevention (CDC) . Prevalence of doctor‐diagnosed arthritis and arthritis‐attributable activity limitation–United States, 2010‐2012. MMWR Morb Mortal Wkly Rep 2013;62:869–73. [PMC free article] [PubMed] [Google Scholar]
- 2. Barter MJ, Bui C, Young DA. Epigenetic mechanisms in cartilage and osteoarthritis: DNA methylation, histone modifications and microRNAs. Osteoarthritis Cartilage 2012;20:339–49. [DOI] [PubMed] [Google Scholar]
- 3. Fernández‐Tajes J, Soto‐Hermida A, Vázquez‐Mosquera ME, al. Genome‐wide DNA methylation analysis of articular chondrocytes reveals a cluster of osteoarthritic patients. Ann Rheum Dis 2014;73:668–77. [DOI] [PubMed] [Google Scholar]
- 4. Jeffries MA, Donica M, Baker LW, et al. Genome‐wide DNA methylation study identifies significant epigenomic changes in osteoarthritic cartilage. Arthritis Rheumatol 2014;66:2804–15. [DOI] [PubMed] [Google Scholar]
- 5. Jeffries MA, Donica M, Baker LW. Genome‐Wide DNA methylation study identifies significant epigenomic changes in osteoarthritic subchondral bone and similarity to overlying cartilage. Arthritis Rheumatol 2016;48:1403–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Rushton MD, Reynard LN, Barter MJ, et al. Characterization of the cartilage DNA methylome in knee and hip osteoarthritis. Arthritis Rheumatol 2014;66:2450–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Moazedi‐Fuerst FC, Hofner M, Gruber G, et al. Epigenetic differences in human cartilage between mild and severe OA. J Orthop Res 2014;32:1636–45. [DOI] [PubMed] [Google Scholar]
- 8. Roos EM, Arden NK. Strategies for the prevention of knee osteoarthritis. Nat Rev Rheumatol 2016;12:92–101. [DOI] [PubMed] [Google Scholar]
- 9. Rowbotham DA. Epigenetic changes in aging and age‐related disease. J Aging Sci 2015;3. Available at: 10.4172/2329-8847.1000130. [DOI] [Google Scholar]
- 10. Ma H‐L, Blanchet TJ, Peluso D, et al. Osteoarthritis severity is sex dependent in a surgical mouse model. Osteoarthritis Cartilage 2007;15:695–700. [DOI] [PubMed] [Google Scholar]
- 11. Zhou W, Triche TJ Jr, Laird PW, et al. SeSAMe: reducing artifactual detection of DNA methylation by Infinium BeadChips in genomic deletions. Nucleic Acids Res 2018;46:e123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Fan Y, Gao D, Zhang Y, et al. Genome‐wide differentially methylated region analysis to reveal epigenetic differences of articular cartilage in kashin‐Beck disease and osteoarthritis. Front Cell Dev Biol 2021;9:636291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Aref‐Eshghi E, Zhang Y, Liu M, et al. Genome‐wide DNA methylation study of hip and knee cartilage reveals embryonic organ and skeletal system morphogenesis as major pathways involved in osteoarthritis. BMC Musculoskelet Disord 2015;16:287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Housman G, Quillen EE, Stone AC. An evolutionary perspective of DNA methylation patterns in skeletal tissues using a baboon model of osteoarthritis. J Orthop Res 2021;39:2260–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Housman G, Havill LM, Quillen EE, et al. Assessment of DNA methylation patterns in the bone and cartilage of a nonhuman primate model of osteoarthritis. Cartilage 2019;10:335–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Deng Z, Jia Y, Liu H, et al. RhoA/ROCK pathway: implication in osteoarthritis and therapeutic targets. Am J Transl Res 2019;11:5324–31. [PMC free article] [PubMed] [Google Scholar]
- 17. Greenblatt MB, Ritter SY, Wright J, et al. NFATc1 and NFATc2 repress spontaneous osteoarthritis. Proc Natl Acad Sci U S A 2013;110:19914–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Wu X, Zhang Y. TET‐mediated active DNA demethylation: mechanism, function and beyond. Nat Rev Genet 2017;18:517–34. [DOI] [PubMed] [Google Scholar]
- 19. Chen L, Li Q, Wang J, et al. MiR‐29b‐3p promotes chondrocyte apoptosis and facilitates the occurrence and development of osteoarthritis by targeting PGRN. J Cell Mol Med 2017;21:3347–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Hoshikawa N, Sakai A, Takai S, et al. Targeting extracellular miR‐21‐TLR7 signaling provides long‐lasting analgesia in osteoarthritis. Mol Ther Nucleic Acids 2020;19:199–207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Loeser RF, Erickson EA, Long DL. Mitogen‐activated protein kinases as therapeutic targets in osteoarthritis. Curr Opin Rheumatol 2008;20:581–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Loeser RF, Olex AL, McNulty MA, et al. Disease progression and phasic changes in gene expression in a mouse model of osteoarthritis. PLoS One 2013;8:e54633. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Singh P, Wang M, Mukherjee P, et al. Transcriptomic and epigenomic analyses uncovered Lrrc15 as a contributing factor to cartilage damage in osteoarthritis. Sci Rep 2021;11:21107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Zhang Y, Fukui N, Yahata M, et al. Identification of DNA methylation changes associated with disease progression in subchondral bone with site‐matched cartilage in knee osteoarthritis. Sci Rep 2016;6:34460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Ma W‐J, Guo X, Liu J‐T, et al. Proteomic changes in articular cartilage of human endemic osteoarthritis in China. Proteomics 2011;11:2881–90. [DOI] [PubMed] [Google Scholar]
- 26. Chinzei N, Rai MF, Hashimoto S, et al. Evidence for genetic contribution to variation in posttraumatic osteoarthritis in mice. Arthritis Rheumatol 2019;71:370–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Takeshita N, Yoshimi E, Hatori C, et al. Alleviating effects of AS1892802, a Rho kinase inhibitor, on osteoarthritic disorders in rodents. J Pharmacol Sci 2011;115:481–9. [DOI] [PubMed] [Google Scholar]
- 28. Fernandes GS, Valdes AM. Cardiovascular disease and osteoarthritis: common pathways and patient outcomes. Eur J Clin Invest 2015;45:405–14. [DOI] [PubMed] [Google Scholar]
- 29. Rahman MM, Kopec JA, Cibere J, et al. The relationship between osteoarthritis and cardiovascular disease in a population health survey: a cross‐sectional study. BMJ Open 2013;3:e002624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Smeriglio P, Grandi FC, Davala S, et al. Inhibition of TET1 prevents the development of osteoarthritis and reveals the 5hmC landscape that orchestrates pathogenesis. Sci Transl Med 2020;12:eaax2332. [DOI] [PubMed] [Google Scholar]
- 31. Le LTT, Swingler TE, Crowe N, et al. The microRNA‐29 family in cartilage homeostasis and osteoarthritis. J Mol Med 2016;94:583–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Ko J‐Y, Lee MS, Lian W‐S, et al. MicroRNA‐29a counteracts synovitis in knee osteoarthritis pathogenesis by targeting VEGF. Sci Rep 2017;7:3584. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Zhao Y, Liu B, Tian Q, et al. Progranulin protects against osteoarthritis through interacting with TNF‐α and β‐Catenin signalling. Ann Rheum Dis 2015;74:2244–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Collins KH, Lenz KL, Pollitt EN, et al. Adipose tissue is a critical regulator of osteoarthritis. Proc Natl Acad Sci U S A 2021;118:e2021096118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Abramson SB. Osteoarthritis and nitric oxide. Osteoarthritis Cartilage 2008;16 Suppl 2:S15–20. [DOI] [PubMed] [Google Scholar]
- 36. Dai S‐M, Shan Z‐Z, Nakamura H, et al. Catabolic stress induces features of chondrocyte senescence through overexpression of caveolin 1: possible involvement of caveolin 1‐induced down‐regulation of articular chondrocytes in the pathogenesis of osteoarthritis. Arthritis Rheum 2006;54:818–31. [DOI] [PubMed] [Google Scholar]
- 37. Wang C, Silverman RM, Shen J, et al. Distinct metabolic programs induced by TGF‐β1 and BMP2 in human articular chondrocytes with osteoarthritis. J Orthop Translat 2018;12:66–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Tachmazidou I, Hatzikotoulas K, Southam L, et al. Identification of new therapeutic targets for osteoarthritis through genome‐wide analyses of UK Biobank data. Nat Genet 2019;51:230–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Coit P, Jeffries M, Altorok N, et al. Genome‐wide DNA methylation study suggests epigenetic accessibility and transcriptional poising of interferon‐regulated genes in naïve CD4+ T cells from lupus patients. J Autoimmun 2013;43:78–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Disclosureform
Table S1 Supplementary Tables