(482f) Modeling Cell Signaling Control of Epithelial-Mesenchymal Transition
AIChE Annual Meeting
2016
2016 AIChE Annual Meeting
Food, Pharmaceutical & Bioengineering Division
In silico Systems Biology
Wednesday, November 16, 2016 - 10:00am to 10:18am
Epithelial-mesenchymal transition (EMT) is a normal cellular developmental program that occurs aberrantly in many cancers and has been associated with metastasis and chemoresistance. This understanding has motivated efforts to identify the central molecular drivers of EMT. This is a challenging undertaking, however, because EMT is clearly regulated by complex multivariate processes, including signaling perturbations due to oncogenic mutations, complex transcriptional regulatory programs, and mechanical stimuli. Growth factor-mediated signaling is also centrally involved, with the combined activation of growth factor receptors in different families typically leading to the most complete EMT. This complexity suggests that it is unlikely that any individual signaling pathway exerts exclusive control over EMT. We undertook an unbiased and systematic analysis of the cell signaling processes leading to EMT. Specifically, we developed a partial least squares regression (PLSR) model to predict the quantitative relationships between multivariate signaling events and EMT-associated phenotypes using a cell culture model of EMT in response to transforming growth factor beta, epidermal growth factor, hepatocyte growth factor, or combinations of these ligands. Using multiplexed Luminex measurements, the phosphorylation states or expression levels of 32 distinct signaling pathway proteins (initially identified by antibody microarray) were measured at six time points over a 48 hr time course in a human pancreatic cancer cell line treated with five different ligand treatment conditions. In parallel, phenotypic measurements were made including the expression or localization of vimentin and E-cadherin and measurements of cell scatter from epithelial clusters. Using these data, PLSR was implemented, with the initial model being refined through several rounds of trimming redundant or minimally informative information. The final PLSR model identified ten distinct signaling measurements predicted to be most important for EMT. The importance of these signaling proteins in the human cell line used to build the model was validated in two-dimensional cell culture and three-dimensional cell spheroid studies using targeted inhibitors and siRNA reagents. The results were further validated in primary mouse pancreatic carcinoma cell lines not used in model development. In both human and mouse cell lines, targeting many of the individual signaling proteins identified by the model partially inhibited EMT, but combinations of inhibitors were substantially more effective. Indeed, specific combinations of inhibitors demonstrated an extraordinary ability to antagonize spheroid invasion in collagen in response to ligands. EMT inhibition using the same combinations of inhibitors also increased the sensitivity of murine cell lines to the EGFR inhibitor gefitinib by an order of magnitude. Our results identify a network of critical cell signaling processes required for efficient EMT. The new understanding gained provides a basis for multiple lines of future investigation, including the analysis of cell-to-cell variability in EMT and the rational design of durable combination therapies for cancer.