From mechanism to medicine: The progress and potential of epigenetics in osteoarthritis

Abstract

Objective

Osteoarthritis (OA) is a chronic, degenerative disease of the articular joints. The disease presents an enormous clinical and economic burden globally, due in part to the lack of disease modifying therapies. For over a decade, OA researchers have been working to determine epigenetic mechanisms underlying the disease to better understand pathology, identify biomarkers of progression, and pinpoint novel targets for therapeutic intervention.

Design

This article presents a summary of the 3rd International Workshop on the Epigenetics of Osteoarthritis held in Toronto, Ontario, Canada, in September 2024. The purpose of this meeting was to gather the international community to discuss the status of OA epigenetic research and share expertise on innovative techniques for future.

Results

Since the two previous meetings, there has been increasing adoption of advanced single-cell and spatial sequencing technologies and bioinfomatic analyses. Furthermore, investigations of multiple joint tissues has highlighted the shifting paradigm from OA as a cartilage centric disease to the consideration of all joint tissues.

Conclusions

The workshop provided a unique opportunity for early-career researchers to expand their network, and for all participants to discuss new or improved approaches to advance the field, including international consortia and data sharing. The highlights and outcomes from this OA epigenetics workshop are described in this report.

1. Osteoarthritis and epigenetics

Osteoarthritis (OA) is a common, complex disease affecting all tissues of the articular joint, manifesting through molecular dysregulation which leads to anatomical changes including cartilage degradation, bone remodelling, and synovial inflammation [1]. OA affects more than 500 million individuals globally, with increasing prevalence, conferring significant economic cost as well as restricting mobility and quality of life [2,3]. No disease-modifying OA drugs exist to halt disease progression. Lifestyle modifications and pain management are recommended for symptom relief, particularly in early disease, yet end-stage disease commonly results in surgical replacement (arthroplasty) of the affected joint [4,5]. Understanding the mechanisms underpinning OA is vital to improve patient quality of life and clinical outcomes by slowing down, stopping, or even reversing the destructive pathological changes within the joint.

Epigenetic processes modulate gene expression without changing the underlying DNA sequence. There are three canonical epigenetic mechanisms: DNA methylation (DNAm), histone post-translational modifications (PTMs), and noncoding RNAs, including microRNAs. These epigenetic regulators impact both the genome and transcriptome, consequently influencing the proteome, signalling cascades, cellular phenotype, and cell function [6].

Over the last decade, all three epigenetic mechanisms have become inextricably linked to OA pathogenesis [7]. Distinct changes in histone PTMs, methylome patterns, and microRNA expression are now well-characterised between non-OA and OA cartilage [[8], [9], [10], [11]]. OA epigenetic research now focusses on understanding the conferred contribution to OA phenotypes and endotypes, the use of modifications as biomarkers and for therapeutic intervention.

2. 3rd international workshop on the epigenetics of OA

The first international workshop on the epigenetics of OA was held in October 2015 in Amsterdam, the Netherlands, with the primary goal of discussing OA epigenetic research in what was then a nascent field of research [12]. In November 2018, a second workshop was held in Dublin, Ireland, focussing on the latest breakthroughs in OA epigenetic research, which included emerging Cas9 genome and epigenome editing tools. The third workshop was recently held in Toronto, ON, Canada, on the 4th and 5th of September, 2024, marking the first workshop held outside of Europe and the first since the Covid-19 pandemic. This meeting aimed to highlight the latest technological advances in the OA epigenetics field, including the application of multi-omic analyses over the last six years, and provide a collaborative forum for “team science” approaches to tackling the field's prescient challenges. In addition, the workshop was preceded by a half-day “crash course” for early-career investigators on high-throughput sequencing technologies, equipping them with the skills they require to handle large (epi)genomic datasets. The workshop was co-organised by Mohit Kapoor, Yolande Ramos, and Sarah Rice, with oral presentations from 11 invited and 22 abstract-selected speakers. In total, 77 researchers from across the globe gathered to present and discuss their research.

In the six years since the last meeting, the OA epigenetics field and the technologies employed have advanced substantially. The speakers highlighted a shift from cartilage-focussed research to a “whole joint” perspective, including studies of synovium and bone. Furthermore, employed methodologies have moved towards single-cell (sc) and single-nuclei (sn) high-throughput sequencing with widespread applications including the identification of biomarkers, and implementation of predictive technologies such as machine learning. In this report, we provide a summary of the research presented at this workshop, illustrating recent advances, challenges we face as a research community, and the anticipated directions in the coming years.

3. Omic technologies and applications to OA

Where targeted approaches such as CRISPR-Cas9 dominated previous meetings, this workshop marked a shift in focus towards single-cell and multi-omic analyses, and the generation of large-scale datasets. Terence Capellini outlined the application of functional genomics in developing human joint tissues to reveal the developmental processes underlying joint shape and their links to OA. Most recently, his laboratory has optimized protocols to extract all joint cell types from developing human hips and knees to enable scATAC-seq and scRNA-seq, along with spatial transcriptomics, and more traditional bulk sequencing. He presented the results of these multi-omic analyses on cartilage samples from more than 40 sites across the post-cranial skeleton during a developmental period when skeletal elements grow markedly and differentially [13]. Notably, the studies revealed novel regulatory and transcriptomic profiles for developing joints, and transcription factor networks that appear unique to individual joints in development. Together, these studies offer a valuable framework to study the complexity of OA through integration of multi-omic analyses.

Multi-omic approaches have also been applied to OA synovium and infrapatellar fat pad, presenting a move away from the cartilage-centric view of OA epigenetic research, as compared to the previous two meetings. In synovium, work presented by Tom Appleton focused on scRNA-seq analysis of synovial tissue macrophages associated with worse pain experiences in patients with knee OA. Interestingly, synovial tissue biopsies from patients with worse pain exhibited inflammation with more severe synovial tissue damage [14]. Worse pain was linked to altered macrophage profiles, with a decrease in immune-regulatory and interferon-stimulated macrophages, and an increase in LYVE1+ macrophages. These findings highlight the importance of distinguishing between inflammation and tissue damage when assessing synovial pathology and that targeting transcriptional regulatory processes related to innate immune inflammation could be a treatment target to prevent synovial tissue damage and pain in OA [15].

The application of snRNA-seq in synovium from human knee OA (early and advanced stage disease) and from mice following destabilisation of the medial meniscus surgery was also discussed [16]. In studies of the infrapatellar fat pad, comprehensive maps of cellular and transcriptomic diversity of major cell populations (including fibroblasts, macrophages, adipocytes and endothelial cells), and subtypes unique to knee OA patients have been outlined using snRNA-seq [17]. Stratification of snRNA-seq fibroblast data by OA severity scoring, sex, and obesity also revealed distinct expression profiles. Spatial transcriptomics and bioinformatic analyses have further been applied to determine the distribution of cellular subpopulations and cell-cell interactions using bioinformatic tools such as CellChat [18]. Downstream metabolomic analyses revealed unique expression profiles in isolated fibroblasts marked by alterations in secreted metabolites.

In addition to chromatin accessibility and transcriptomic studies, the 3D organization of the genome plays an integral role in understanding the epigenetic underpinnings of OA. In 2024, the first chromosome conformation analysis (Hi-C) map of primary OA chondrocytes was published [19]. Douglas Phanstiel illustrated how an integrated multi-omic approach (Hi-C, ATAC-seq, and RNA-seq) in human primary chondrocytes can be used to interpret OA GWAS variants. Genetic variation, gene expression, and OA risk loci were linked to identify putative effector gene (eGene)s influencing OA risk. Chromatin accessibility and 3D chromatin structure helped explain how and when these genetic variants affect transcriptomic activity. Moving forward, the Phanstiel laboratory plans to extend this research by investigating other cell types and using CRISPR gene editing tools to study how these genes are underlying OA-related characteristics [20,21].

Progressing our understanding of the functional epigenomic mechanisms underlying OA across joint tissues is vital for clinical translation, including strategies such as cartilage regeneration. Neocartilage produced via chondrogenic differentiation of human induced pluripotent stem cells (hiPSCs) has previously been shown to be remarkably similar in quality compared to that generated by human primary chondrocytes, with high concordance between their methylomic and transcriptomic profiles post-differentiation [22]. Building upon these findings, Rick Mulders presented a map of hiPSC differentiation to chondrocytes projecting a detailed progression of expression and chromatin conformation, including cell fate commitments. To investigate this, snRNA-seq and snATAC-seq of the same single nuclei were applied across multiple timepoints and data were combined to characterise cell populations for different fates and stages of differentiation at multimodal level. By identifying points of cell fate decision, combinatory analysis of expression and open chromatin can inform modeling and be used to adapt differentiation accordingly, with potential implications for clinical utility in the future.

Together, these highlighted talks illustrate the emergence of large epigenomic datasets that researchers are now utilising to enhance understanding of the complex mechanisms affecting risk of OA onset, with the capacity to capture multiple mechanistic processes at once (Fig. 1). These developments will only be bolstered by the advancement of multi-omic approaches that are now becoming more common, providing greater insight for heterogenous tissues such as those affected in OA, including the identification of functional cellular subtypes. In addition, the application of proteomics and metabolomics will also aid in characterising functionally important changes in OA [23].

Fig. 1.

An overview of omic techniques used and discussed during the 3rd International Workshop on the Epigenetics of Osteoarthritis. Assay for Transposase-Accessible Chromatin, ATAC; Nuclear Magnetic Resonance, NMR; Quantitative Trait Loci, QTL. Figure was created using Biorender.

4. Epigenetic mechanisms of OA

Technological advancements have facilitated variant-to-function analyses of GWAS in OA. Rik Lories summarised more than a decade of focussed research investigating the pivotal role of the hip OA risk gene DOT1L encoding the histone methyltransferase DOT1L in OA pathogenesis [24]. The loss of DOT1L function exacerbates OA severity due to the hyperactivation of Wnt signalling [25]. The team has further discovered that hypoxia plays a crucial role in sustaining DOT1L levels and activity [26], offering potential insights into tissue-level regulatory mechanisms. In their most recent work, researchers identified insulin-like growth factor 1 as a downstream effector of Wnt hyperactivation, contributing to chondrocyte hypertrophy [27]. These findings connect altered DOT1L function and Wnt signalling to molecular and cellular changes driving cartilage degeneration.

Variant-to-function analyses have been conducted to elucidate the mechanistic impact of two GDF5 OA risk variants. The OA risk-conferring alleles of these two variants exhibit demonstrable impact on the formation of the tibial plateau (rs6060369) and acetabulum (rs4911178), increasing risk of knee OA and developmental dysplasia of the hip, respectively [28]. Multiple regulatory elements are required to enhance, repress, and restrict tissue-specific expression of GDF5 during development, influencing joint shape and risk of knee OA and developmental dysplasia of the hip [29]. These findings implicate the important role of joint development for the onset of OA in adulthood.

DNA methylation (DNAm) quantitative trait loci (mQTLs) have now been identified in multiple human OA tissues, including cartilage, synovium and fat pad [10,30,31]. Targeted epigenetic studies also showed that these mQTLs, which are enriched in enhancers [11], can modulate OA eGene expression in-vitro [[32], [33], [34]]. In 2023, Rice and colleagues reported that mQTLs previously identified in OA cartilage were present in human developmental limb tissues, demonstrating that these putative functional epigenetic mechanisms can be active from the onset of skeletogenesis [35]. More recently, an epigenome-wide analysis revealed that ∼25 ​% of OA GWAS signals colocalize with mQTLs in human developmental cartilage [36], including CpGs mapping to GDF5 and the carbohydrate sulfotransferase CHST3 [36]. These findings suggest that a proportion of OA genetic risk loci mediate their effects via the functional intermediary of DNAm at specific timepoints: solely during development, across the life course, or in aged tissue only [37].

Associations between splicing patterns and OA risk variants (splicing QTLs, sQTLs) in chondrocytes have now been reported at six GWAS loci [38]. Of note, two of the six sQTLs also map to genes previously linked to OA cartilage mQTLs: COLGALT2, encoding a collagen galactosyltransferase enzyme [32], and WWP2, encoding an E3 ubiquitin ligase [32,34,39]. It is plausible that these mechanisms are interconnected, with DNAm known to modulate splicing activity [40]. However, functional studies are required to validate and biologically interpret these recently discovered developmental cartilage mQTLs and OA cartilage sQTLs [11,37].

As discussed above, progress has been made in the epigenetic studies of non-cartilaginous tissues. Subchondral bone and marrow adipose tissue in OA undergo substantial remodelling, marked by increased formation of hypo-mineralised bone and vascular and fibrotic changes in the marrow. While the hypo-mineralisation phenotype appears imprinted in bone-forming cells [41,42], it remains unclear whether this is regulated by epigenetic mechanisms. Epigenetic profiling of subchondral bone has been limited to two studies [43,44], which suggest unique molecular profiles (endotypes) between knee and hip OA. Findings presented by Jeroen Geurts indicate that OA stromal vascular fractions are more affected than marrow adipose tissue, with histone methyltransferases including DOT1L showing specific downregulation in knee OA. Future research should focus on the interplay between the epigenome and transcriptome and investigate the relationship between obesity and DNAm in bone marrow adipocytes.

5. Epigenetics as biomarkers for OA

Since the first workshop, microRNAs have emerged as promising candidates for precision medicine, given their potential as biomarkers, mechanistic drivers, and therapeutic targets [45]. Amanda Ali outlined the importance of studying microRNAs as biomarkers for OA. Advances in sequencing technology and bioinformatic tools have accelerated the discovery of circulating microRNAs as biomarkers for OA [46], uncovering profiles that are associated with specific phenotypes, including early-stage knee OA [47] and fast-progressing knee OA [48]. Notably, miR-126-3p was identified through secondary analysis of existing microRNA-seq datasets to classify individuals with radiographic knee OA [49] using an optimized analysis pipeline [50]. This highlights the importance of tailored approaches for microRNA discovery when using big data. MiR-126-3p is thought to serve a pro-resolving role in knee OA, inhibiting angiogenesis associated with joint structural damage and pain even when delivered systemically in mice, underscoring the importance of biomarker discovery in easily accessible biofluids, and understanding potential tissue crosstalk mechanisms with local joint environments.

Multi-omics approaches have bolstered biomarker identification. Plasma microRNAs and metabolites have been investigated longitudinally following anterior cruciate ligament reconstruction post-injury, identifying time- and sex-dependent changes in select microRNAs and metabolite signatures 2-weeks post-surgery [51]. Jason Rockel presented a study investigating 414 subjects undergoing total knee arthroplasty. Using a deep-learning variational autoencoder modeling approach, three independent clusters of subjects were identified based on baseline microRNA profiles in plasma, urine and synovial fluid, and metabolites in plasma, with each cluster having a unique endotype [52]. Subsequently, using a machine learning approach to investigate classification performance of Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and function responders (> 33% improvement from baseline) and non-responders (≤ 33% improvement from baseline) to total knee arthroplasty in each cluster, improved modeling performance was observed when molecular and clinical features were integrated, compared to each molecular/clinical domain alone. The top 20 predictors of model performance for each cluster were also found to be molecular and not clinical features. Thus, it is important to further explore the potential of epigenetic and molecular features in a multi-fluid, multi-omic approach for endotyping subsets of OA patients, to potentially help identify individuals more likely to respond to therapy.

In addition to microRNAs and metabolites, DNAm levels in blood can also be used as biomarkers of OA development and progression. DNAm arrays of blood buffy coat from 554 subjects from the Osteoarthritis Biomarkers Consortium were able to uncover early systemic immune DNAm signatures from peripheral blood [53]. These signatures were able to classify future radiographic or pain progression in the study sample, with validation cohort studies showing similar performance by machine-learning modeling. Likewise, sex-specific DNAm signatures in human T cells can discriminate chronic low back pain sufferers from healthy controls, with most of the CpGs being hypomethylated in women, and significant enrichment for genes encoding extracellular matrix proteins, for immune-related genes such as cytokines, and for epigenetic regulators [54].

Overall, to realise the full potential of epigenetic biomarkers of OA, continued efforts are needed to conduct validation studies in large cohorts, investigate tissue-specific mechanisms of action, and optimise in-vivo modulation for potential precision medicine approaches.

6. Modulation of the OA epigenome

Epigenetic dysregulation in OA pathogenesis and the dearth of disease-modifying OA drugs highlight the epigenome as a potential therapeutic target and raises the important question: how best to modulate the OA epigenome? In a presentation by Nidhi Bhutani, DNA hydroxymethylation (5hmC) enrichment in gene bodies and putative regulatory regions was reported to have a key role in shaping the OA epigenome in early- and late-stage OA [55]. Notably, inhibition of TET1-mediated 5hmC via intra-articular injection of 2-hydroxyglutarate into the knee joints of mice could prevent or reverse OA [56]. While these studies highlight a potential therapeutic axis, the challenge will be targeting this widespread epigenetic regulator in a controlled manner to avoid untoward off-target effects. Future studies are being directed at identifying specific co-factors of TET1 that help orchestrate OA pathology [57].

One possibility for overcoming off-target effects of epigenome modulation is provided by recent advances in epigenome engineering. Precision epigenetic editing has now been used to identify the eGenes of six OA risk variants, including RWDD2B [58], TGFB1 [33], and most recently WWP2 [34]. These studies, which utilise a catalytically inactive ‘dead’ Cas9 (dCas9) fused to the epigenome modulators TET1 or DNMT3A (to de-methylate or methylate specific proximal CpG sites, respectively), have validated DNAm within regulatory elements as a functional intermediary between OA risk variants and eGenes in multiple joint cell types. At this meeting, a detailed account of the application of this technology to validate WWP2 as an OA eGene was presented by Jack Roberts [34]. dCas9 epigenetic editing has also been applied to reveal a functional link between DNAm and COLGALT2 expression that operates inversely between OA cartilage and OA synovium [59], highlighting tissue-specific biological pleiotropy even within the joint environment.

An alternative emerging strategy to epigenome modulation gaining traction in the field is the “nutraceutical” approach, with lifestyle and dietary changes associated with positive outcomes for OA patients [60]. Changes to diet have been shown to impact post-translational modifications and microRNA expression. Obese mice switched to a ketogenic diet after surgical induction of OA exhibit improved metabolic fitness and increased histone lysine β-hydroxybutyrylation in the liver and kidney in comparison to mice fed a high fat diet [61], suggesting that epigenetic changes may mediate positive outcomes in metabolic fitness. Adherence to the mediterranean diet, including intake of extra virgin olive oil, has previously been linked to lower OA incidence, potentially driven by anti-inflammatory effects in cartilage including downregulation of MMP13 [62]. This may be influenced by epigenetic regulation, with mice fed an extra virgin olive oil diet exhibiting lower expression of pro-catabolic miR-34a-5p [63]. However, epigenetic programming laid down by recent ancestry may impede the effectiveness of some nutraceutical approaches. Maternal obesity (induced by a high fat diet) has been shown to impair bone quality and strength in offspring, but not cartilage in ageing mice, with the authors proposing epigenetic programming as potential explanation for this observation [64].

Laura Stone described how environmental factors, including early-life adversity, become embedded in the methylome, influencing supraspinal expression of DNAm-regulated genes in mice and risk of chronic pain in later life [65]. Chronic pain has been shown to drive epigenetic changes in the prefrontal cortex, including altered DNAm levels and dysregulated gene expression. Such changes were found to be restrained by treatment with the methyl donor S-adenosyl methionine (SAM) or with increased physical activity [66,67].

7. Computational, data analytics and machine learning

Yenisel Cruz-Almeida described the relationship between OA, epigenetics and high-impact pain that significantly limits daily activities. Specifically, the study showed that in blood of knee OA patients with the highest pain grades compared to individuals with no pain or low pain disability, there were 13,951 hypermethylated CpG and 5,759 hypomethylated CpGs [68]. Of note, Ingenuity Pathway Analysis demonstrated that pain-related differentially methylated regions were enriched for cellular signalling processes related to immune responses, and upstream regulators of these pathways included both transcription factors and small molecule modifiers. In addition, the Cruz-Almeida laboratory found that specific DNAm patterns could predict which individuals would experience decreased pain impact over a two-year period. In a separate study, methyl-seq analysis of blood DNAm was used to generate polygenic risk scores to define facet joint OA patients who exhibited positive responses to surgery (>70 ​% improvement in pain) compared to non-responders (<30 ​% improvement in pain) at 12-months post-surgery [69].

Colocalization analysis [70] has been employed to characterise genetic associations between microRNAs and OA outcomes, with implications in identification of regulatory mechanisms and targets for clinical applications. A recently published review article describes such methodology to map microRNA expression QTLs before colocalization analysis, aiding in the determination of relationships between gene expression and genetic variants [71].

Integrative computational biology, artificial intelligence (AI) and machine learning algorithms can aid improving treatment of complex diseases by building explainable models. Igor Jurisica described the main challenges and opportunities in precision medicine using such workflows. Whilst publicly available datasets are essential for this, existing papers and datasets need to be curated to ensure proper format, annotation, and correctness [72]. It is therefore paramount to make raw data freely available, not only to facilitate uncovering potential errors or fraud, but to enable independent training and validation of AI or statistical models. Curated data, linked to clinical information and biological annotations, such as OsteoDIP [73], are an essential resource for translational research. Intertwining computational prediction and modeling with biological experiments and preclinical studies will lead faster to more useful findings, delivering a path for true patient-centric, precision medicine. Of note, the pathDIP [74] portal for comprehensive pathway enrichment analysis was described for use in better understanding the biological meaning behind epigenetic modifications. This platform integrates multiple pathway databases, enhancing gene annotation coverage and reducing bias in pathway enrichment analyses. To predict relationships of signatures of epigenetic regulators with OA pathology and outcomes, improved machine learning and AI computational biology approaches will be necessary.

8. Limitations and current challenges in the field

In addition to showcasing recent advances, the workshop also provided a unique and informal opportunity for our global community to discuss challenges the field faces. Cindy Boer highlighted the growing impact of businesses known as “paper mills” that publish low-quality research or even fabricated data, often featuring heavily plagiarised text [75,76]. These publications are often hard to identify, though guidance is available to distinguish papers originating from paper mills [77]. The impact of paper mills has been compounded by the advent of generative AI, polluting available literature and threatening scientific progress and integrity [78]. It is therefore vitally important that researchers utilise generative AI responsibly to prevent the proliferation of fraudulent datasets that may catastrophically impact future research direction [79].

Generation of large OA epigenomic datasets in multiple cell types provides the potential to advance research. However, it is fundamental that these datasets are made publicly available to allow for reusability and reproducibility, in concordance with FAIR principles of data management (Findable, Accessible, Interoperable, and Reusable) [80]. A recent review by Ramos and colleagues provides a strategy for epigenomic data collection, sharing, and access, as well as a list of key epigenomic repositories [81].

9. Conclusions and future perspectives

Despite the challenges facing the OA epigenetics field, the 3rd International Workshop on the Epigenetics of OA showcased many examples of success and research progress. The application of multi-omic datasets to OA epigenetic research has intensified in recent years, providing a deeper understanding of the molecular mechanisms underpinning OA. Advances in machine learning and data analytics are having a beneficial impact on data integration, and with more data, we may be able to move towards personalised medicine approaches. Since this is a fast-evolving field, we anticipate hearing more on the progress in machine learning approaches and its application in the near future. Meanwhile, single-cell transcriptomic analyses are paving the way to understand cellular subpopulations, whilst also underscoring heterogeneity of the tissues implicated in disease. In addition, research focus has shifted from a “cartilage-centric” focus to a “whole joint” perspective. These advances provide a platform for deepening our understanding of OA.

In 2015, the first workshop report concluded that a tremendous amount of work remained to be done [12]. This remains the case, yet the third iteration of the OA epigenetics workshop demonstrated that impressive strides have been made in understanding OA epigenetic mechanisms, the identification of biomarkers, in epigenome modulation, and in data analytics and machine learning. However, there remains a long road ahead. One prescient challenge is how best to modulate the OA epigenome for therapeutic intervention. Recent advances using in-vitro epigenetic editors are yet to be translated to in-vivo contexts, and safety implications surrounding this potential is still unclear. Furthermore, epigenetic drugs have been shown to exhibit nonspecific changes to DNAm patterns with potentially deleterious effects on other tissues, highlighting heterogeneity of OA across tissues [82]. To tackle the prescient challenges in the field, a consensus towards a shift in research culture towards “team science” was established as a key theme of the workshop. The sharing of resources, knowledge and expertise will accelerate our understanding of the intertwined relationship of epigenetics and OA. The importance of advocacy, knowledge translation, and involvement of patients in research will be crucial to better understand the impacts of OA pathogenesis, as indicated by Maureen Quigley, a patient partner. Though the challenges facing the field are substantial, we now look forward to advances driven through the continued efforts of scientists, clinicians and trainees across the globe to accomplish translation of this work to the clinic. Through continued epigenetic investigations of heterogeneous OA tissues, cells, and biofluids, future studies will unravel more of the complexity of OA disease. This will further aid our understanding of patient endotypes and push towards precision medicine approaches, which may require the need for development of novel epigenome-modifying therapies.

Declaration of competing interest

The authors declare that they have no competing interests to disclose.

Handling Editor: Professor H Madry