Epigenetic changes occur during OA and histone methylation, in particular, appear to imbalance chondrocyte metabolism25,26,. Our data now associate expression of histone eraser UTX with OA development in humans and correlate UTX and H3K27 trimethylation in articular cartilage with chondrocyte marker gene expression in vitro. This is in agreement with the only other study demonstrating high H3K27me3 abundances in human hip osteoarthritic cartilage14. We further, for the first time, manipulated Utx activity through RNAi and lentiviral overexpression, established a murine Utx KO model and evaluated intraarticular effects of local Utx-specific inhibition in vivo. Our comprehensive study reveals a novel non-canonical role of the H3K27me2/3-specific demethylase Utx, in concert with PRC2 core components, in articular chondrocytes homeostasis and the progression of OA.
We used Chip-Seq to screen global changes in gene expression between WT and Utx KO chondrocytes to identify molecular mechanisms protecting cartilage against degradation. GO terms already revealed major expressional changes occurring in gene regulation networks related to mineralization, ossification and senescence. Candidates like Sox9, Igf-2 and Wnt10a arising from enrichment scores were then further analyzed to reveal an aberrant H3K27me3 occupation at their respective promoter loci between WT and Utx KO chromatin. IGF-2 and Wnt signaling components were known to regulate chondrocyte function and cartilage development23,24. Our enrichment scores now revealed upregulation of many markers associated with a proper articular chondrocyte phenotype, like relevant collagens of the cartilage ECM (e.g., Col9a2) or the cartilage-specific acidic protein 1 (CRTAC1)27 and that of BMP antagonists like GREM1. Col9a2 is considered to be a chondrocyte-specific marker gene, in addition to Col2a1, Col11a2 and aggrecan28.
Sox9 is a prominent activator of Col2a1 transcription, making it a master regulator of the chondrocyte phenotype29. To show a Utx-dependent compromised chondrocytic activity, we used key chondrocyte marker genes Sox9, Col2a1 and Acan as readouts, and quantified proteoglycan production as a measure of ECM quality. Using UTX-specific inhibitor GSK-J4, shown to inhibit H3K27 demethylation of UTX target genes30, and confirmed anabolic effects on ECM level in vitro and being in line with the observed overall improved relative thickness of the articular cartilage in vivo. BMP signaling ultimately results in unwanted hypertrophy in articular chondrocytes. Leijten et al. already showed that decreased GREM1 and DKK1 gene expression in cartilage were correlated with OA31. Harmonization of BMP and WNT signaling are important to maintain articular cartilage integrity. While BMPs32 and canonical WNTs33 may exert pro-hypertrophic actions under certain conditions, the parallel upregulation of BMP antagonists and WNT antagonists, like sFRP-1 or DKK1, may thus explain the overall chondroprotective net effect of Utx loss in vitro and in vivo.
Interestingly, adult articular cartilage has been long considered a post-mitotic tissue with terminally differentiated chondrocytes34,35, but this textbook dogma has been questioned recently. A ‘phenotypic plasticity’ of articular chondrocytes has then been associated with OA36 in which chondrocytes de-differentiation towards a more fibroblast-like phenotype37. To this end, Wnt signaling is an important regulator of cell plasticity38 and the Wnt–frizzled–β-catenin pathway appears activated in OA39,40. Expression of Wnt4 and Wnt10, in particular, can stimulate osteogenesis41, and excessive Wnt signaling can also lead to increased osteophyte formation42, which is in agreement with our findings. Recently, mouse genetic analysis also revealed that Wnt/β-catenin signaling components are important to regulate Sox9 and Runx2 and Wnt signaling can thus suppress chondrogenic differentiation, while increasing osteoblastic differentiation43. In line with our hypothesis, H3K27me3 demethylase UTX has earlier been shown to be essential to tissue development and stem cell plasticity44. In contrast, Wnt inhibition attenuates OA development through anti-catabolic and anti-fibrotic effects on chondrocytes and synovial fibroblasts, respectively45. Wnt signaling plays a context-dependent role in the development of OA and upregulated expression of canonical Wnts, like Wnt10, and non-canonical, like Wnt4 and Wnt7,46, together with Wnt inhibitors like sFRPs and Dkks hint towards a delicate balance of this system.
Currently developed Wnt inhibitors as disease-modifying osteoarthritis drugs (DMOAD)40 underscore their potential. Local intraarticular modulation of this pathway through Utx inhibition is another effective means to change the course of this joint disease. To the best of our knowledge, no further information on a direct involvement of Utx in regulating the IGF cell signaling axis in OA exists, while IGF-2 is known to regulate cartilage development47. IGF-2 also compromises ECM underproduction in inflamed chondrocytes and preserves cartilage integrity even in a model of experimental osteoarthritis48. This may, partly, explain the beneficial effects of Utx KO on ECM synthesis in chondrocytes. However, two groups independently reported that homologue Utx-1 in C. elegans regulates its life span through targeting IGF-1 pathway49,50. This may link age-related changes in Utx activity to age-related chondrocyte senescence and onset of primary OA2. IGF signaling also regulates cell proliferation, differentiation and apoptosis in cartilage51. While IGF signaling is considered a promising drug target, complex negative feedback regulation and the systemic importance of glucose homeostasis helps explaining the failures of single target therapies aiming at regulating IGF signaling51. Modulating IGF signaling indirectly and intraarticularly through pharmacological Utx modulation may thus be more promising.
Utx knockout largely protected against signs of cartilage degeneration in primary OA and in an experimental murine model of induced secondary OA development. We then used a pharmacological Utx inhibitor in the latter model to demonstrate protection against gonarthrotic changes in the joint, confirming anabolic effects on ECM level in vitro and being in line with the observed overall improved relative thickness of the articular cartilage in vivo. Pain is one of prominent symptoms of OA52. We thus used catwalk analyses as a well-accepted model to assess pain in rodents53, like that resulting from joint tissue degeneration due to deregulated joint kinematics20 in collagenase-treated limbs. Upon intraarticular injection, Utx inhibitor GSK-J4 not only protected against joint tissue degradation, but also against pain. The latter was evident from derived from restoration of normal gait profiles and mobility in animals with injured joints. To the best of our knowledge, only a single other study very recently studied Utx in chondrocytes ex vivo, also using GSK-J417. These earlier results also hinting towards a cartilage-protective effect of Utx inhibition, but it is well accepted that the complex TGF-β signaling in adult cartilage in vivo cannot be fully appreciated by in vitro models of chondrogenesis used in that study.
UTX, a member of the Jumonji C family of histone erasers, usually removes di- and tri-methyl groups on H3K27 to promote target gene activation54. Surprisingly, loss-of-function of Utx now activated chondroprotective pathways and mRNA expression of Igf2 and Sox9 in particular, while that of Wnt10a was significantly reduced (Fig. 7). Additionally, Utx KO contra-intuitively reduced the H3K27me3 occupation at the promoter loci of these genes. Apparently, other regulatory pathways potentially contribute and UTX loss can indeed enhance the EZH2-induced H3K27 trimethylation55. We thus postulate that PRC2 core components participate in the UTX deletion-induced H3K27 hypomethylation in chondrocytes. Our data now reveal that upon UTX loss, EZH2 appears to curtail chondrocytic metabolism as EZH2 RNAi-mediated suppression maintained H3K27 hypomethylated and improves ECM synthesis, whereas downregulation of EED or SUZ12 appears to stimulate matrix anabolism, as restoring EED or SUZ12 results in H3K27 hypermethylation and compromised ECM synthesis. In agreement with our data, chondrocyte-specific EZH2 or EED knockout mice show poor cartilage development and defective bone growth13,15. EED thus appears functionally indispensable for trimethylation of H3K27 by EZH256. Our study thus revealed a new paradigm in which opposing action of PRC2 components are responsible for UTX loss-mediated H3K27 hypomethylation to maintain proper ECM homeostasis in cartilage. Weak EED and SUZ12 signaling blocked EZH2-mediated H3K27 trimethylation, driving chondrocytes in KO mice to produce abundant extracellular matrices.
Utx is further a member of MLL2 H3K4 methyltransferase complex and got additional demethylase independent roles in chromatin remodeling through an interaction with SWI/SNF complex57,58. While activation of non-canonical Wnt signaling may promote osteogenic differentiation through H3K9 methylation, WNT10A facilitated β-catenin stabilization potentially also causes cartilage mineralization. In mammals, PRC2 core components EZH2 and EZH1 are important for writing trimethylation of H3K2759. Of note, H3K27 methyltransferase EZH2 is known to repress Wnt signaling components60. PRC2 and H3K27me3 are involved in bivalent control of transcription activation and repression during stem cell fate commitment61, in line with the earlier discussed plasticity of chondrocytes in OA. To this end, our data is in agreement with studies demonstrating that UTX KO decreases H3K27 trimethylation to alter mesenchymal stem cells differentiation62. Although histone demethylase Jmjd3 is found to remove trimethyl group of H3K27 in various cell types11, its expression in chondrocytes was not changed in UTX KO mice. In general, Polycomb-group proteins together with their target genes control differentiation program in a dynamic manner. Co-localization of PRC2 with H3K27me3 is required to catalyze trimethylation61. As Polycomb-group proteins regulate gene silencing, repressing transdifferentiation in a H3K27me3 dependent manner63 and the latter appears to be the link between inflammation and reprogramming of the epigenome63, this may - at least partly - explain our findings.
In conclusion, loss of function of Utx appears to be chondroprotective. However, a seemingly contradictory trimethylation status was observed at the selected gene loci of key chondrocyte markers: loss of histone eraser Utx caused a depletion of H3K27me3 occupation at these domains. To this end, we identified a novel interaction between Utx and PRC2 core complex components. Our current model of the epigenetic regulation of cartilage metabolism is illustrated in Fig. 9; on bivalent loci, PRC2 activity acts in concert with Utx to either stimulate cartilage anabolism or activate phenotypic de-differentiation and ECM deterioration through activating Wnt signaling.
Our study sheds new light onto epigenetic causes of OA development. A better understanding of the molecular composition, but also of the interactions between, the different canonical and non-canonical PRC complexes with histone erasers of the Jumonji family is still needed. Altogether, the complexity of biological functions assigned to Utx in general, and in cartilage in particular, has just started to emerge. We showed, for the first time, in a multimodal approach using human samples and animal models as well as RNAi and pharmacological intervention that targeted manipulation of Utx activity appears to hold a lot of potential for the development of future anti-OA therapies. While repetitive intraarticular injections with these inhibitors may be required to treat chronic diseases like OA, biomaterial-based controlled delivery systems may be a realistic clinical treatment option.