Mathematical modeling of signaling pathways in osteoarthritis

Purpose: Cartilage homeostasis relies on an intricate balance between anabolic and catabolic processes. In osteoarthritis this balance is shifted towards catabolism, leading to hypertrophy and a gradual degradation of cartilage tissue. So far, drug-based intervention in this process has shown limited progress. We propose to construct a mathematical model of the molecular network that governs key processes in articular cartilage homeostasis. This model can be used as a platform for model expansion by introduction of new experimental findings and hypotheses. Methods: We recently developed ANIMO (Analysis of Networks with Interactive Modeling), an intuitive software tool for modeling molecular networks. Here, we demonstrate a mathematical model of growth plate cartilage using a combination of literature and experimental data. We show how ANIMO allows for intuitive exploration of the model, despite the size and complexity of the model. Results: We constructed a network model of regulatory processes in growth plate chondrocytes. In this model the effects downstream of extracellular growth factors FGF, WNT, IGF-1, PTHrP, Ihh, BMP, and TGF-β are integrated into a cellular response. In silico experiments predict the phenotypic outcome for different inputs and starting states of the model. Osteoarthritic chondrocytes and hypertrophic growth plate chondrocytes show strong parallels in their gene expression profile. We have examined the gene expression profiles of growth plate and articular cartilage. In articular cartilage the expression of the WNT and BMP antagonists DKK1, FRZB and GREM1 is over 10-500 fold higher than in growth plate cartilage. This leads us to think that DKK1, FRZB, GREM1 could act as gatekeepers for preventing hypertrophy. We are currently investigating the range of conditions under which these proteins exert their stabilizing effect on the articular cartilage phenotype in the model. Furthermore, we are interrogating the model to obtain in silico leads to targets for novel combination therapies. Such therapies could be used to intervene in the osteoarthritic state of the network and restore the balanced situation of healthy cartilage. Conclusions: Traditionally, modeling efforts in the realm of molecular cell biology have been the exclusive domain of researchers with a thorough training in mathematics or computer science. We show here that a complex model that is intuitively amenable to exploration and adaptation by biologists is an invaluable asset in cartilage research. Expansion of an existing model with DKK1, FRZB and GREM1 provided evidence for their role in preserving the articular cartilage phenotype.