MULTISCALE MODELLING OF THE KNEE ARTICULAR CARTILAGE TO STUDY CELL-TISSUE COMMUNICATION DYNAMICS

Osteoarthritis is a degenerative joint condition driven by complex biological and mechanical interactions in chondrocytes and articular cartilage (AC). Assessing these factors simultaneously in experimental set-ups is challenging, making computational models valuable. Regulatory network-based models (RNM) describe chondrocyte regulation, while finite-element models (FEM) quantify tissue mechanics. However, chondrocyte-AC model integration is not fully tackled. Hence, a multi-scale model is proposed to simulate single-cell regulation coupled with the transport of diluted species in AC, validated with in vitro models. A 2D FEM represented an AC plug with chondrocytes randomly placed by local density. A chondrocyte RNM translated into 21 reaction terms formed a reaction-diffusion model. FEM nodes with RNM terms (R i ) were restricted to chondrocyte surfaces (C i surface ), in a pericellular matrix (PCM) with reduced diffusion (Figure 1). Diffusion was zero in bone, and synovium concentrations were null. The model, initialized as fixed, healthy, or pro-inflammatory RNM-based phenotypes, evolved until steady AC concentrations were predicted. Each chondrocyte RNM (X i ) was then regulated for a month by local FEM concentrations. An in vitro cartilage degradation model used 5 mm full-depth knee AC plugs from three young calves with two replicates per animal. Explants were exposed to 10 ng/mL Interleukin-1b for two weeks, and key OA gene expression was compared to controls. Results & Discussion. Diffusion streamlines revealed chondrocyte communication, with varying gradients among cell clusters. Initial chondrocyte phenotypes persisted, according to literature findings illustrating the difficulty to regenerate the AC in OA states. Translation of RNM chondrocyte profiles into quantitative tissue concentrations enabled intercellular communication through the reaction-diffusion restricted model. Moreover, predicted pro-inflammatory chondrocyte activity matched the experimental model tendency after 2 weeks, supporting the validity of the computational multi-scale integration of models. Discrepancies between experimental and computational models were found for MMP3, IL-6 and ITGA5, being qualitative only for MMP3 and ITGA5, probably because of initial extraction injury of the plug. Next steps include the incorporation of mechanical stimulation, both in the experimental and computational models, which is ongoing work, as well as the incorporation of patient data such as synovial liquid concentrations.