Background: Without the availability of disease-modifying drugs, there is an unmet therapeutic need for osteoarthritic patients. During osteoarthritis, the homeostasis of articular chondrocytes is dysregulated and a phenotypical transition called hypertrophy occurs, leading to cartilage degeneration. Targeting this phenotypic transition has emerged as a potential therapeutic strategy. Chondrocyte phenotype maintenance and switch are controlled by an intricate network of intracellular factors, each influenced by a myriad of feedback mechanisms, making it challenging to intuitively predict treatment outcomes, while in silico modeling can help unravel that complexity. In this study, we aim to develop a virtual articular chondrocyte to guide experiments in order to rationalize the identification of potential drug targets via screening of combination therapies through computational modeling and simulations. Results: We developed a signal transduction network model using knowledge-based and data-driven (machine learning) modeling technologies. The in silico high-throughput screening of (pairwise) perturbations operated with that network model highlighted conditions potentially affecting the hypertrophic switch. A selection of promising combinations was further tested in a murine cell line and primary human chondrocytes, which notably highlighted a previously unreported synergistic effect between the protein kinase A and the fibroblast growth factor receptor 1. Conclusions: Here, we provide a virtual articular chondrocyte in the form of a signal transduction interactive knowledge base and of an executable computational model. Our in silico-in vitro strategy opens new routes for developing osteoarthritis targeting therapies by refining the early stages of drug target discovery. Graphical Abstract: [Figure not available: see fulltext.]

During standard expansion culture (i.e., plasma osmolarity, 280 mOsm) human articular chondrocytes dedifferentiate, making them inappropriate for autologous chondrocyte implantation to treat cartilage defects. Increasing the osmolarity of culture media to physiological osmolarity levels of cartilage (i.e., 380 mOsm), increases collagen type II (COL2A1) expression of human articular chondrocytes in vitro, but the underlying molecular mechanism is not fully understood. We hypothesized that TGF-β superfamily signaling may drive expression of COL2A1 under physiological osmolarity culture conditions. Human articular chondrocytes were cultured in cytokine-free medium of 280 or 380 mOsm with or without siRNA mediated TGF-β2 knockdown (RNAi). Expression of TGF-β isoforms, and collagen type II was evaluated by RT-qPCR and immunoblotting. TGF-β2 protein secretion was evaluated using ELISA and TGF-β bioactivity was determined using an established reporter assay. Involvement of BMP signaling was investigated by culturing human articular chondrocytes in the presence or absence of BMP inhibitor dorsomorphin and BMP bioactivity was determined using an established reporter assay. Physiological cartilage osmolarity (i.e., physosmolarity) most prominently increased TGF-β2 mRNA expression and protein secretion as well as TGF-β bioactivity. Upon TGF-β2 isoform-specific knockdown, gene expression of chondrocyte marker COL2A1 was induced. TGF-β2 RNAi under physosmolarity enhanced TGF-β bioactivity. BMP bioactivity increased upon physosmotic treatment, but was not related to TGF-β2 RNAi. In contrast, dorsomorphin inhibited COL2A1 mRNA expression in human articular chondrocytes independent of the osmotic condition. Our data suggest a role for TGF-β superfamily member signaling in physosmolarity-induced mRNA expression of collagen type II. As physosmotic conditions favor the expression of COL2A1 independent of our manipulations, contribution of other metabolic, post-transcriptional or epigenetic factors cannot be excluded in the underlying complex and interdependent regulation of marker gene expression. Dissecting these molecular mechanisms holds potential to further improve future cell-based chondral repair strategies.