Model-Based Approach for Multivariate Signaling Regulation of Epithelial-Mesenchymal Transition in Pancreas Cancer Cells

Epithelial-mesenchymal transition (EMT) is a normal cell developmental program that occurs aberrantly in numerous carcinomas and promotes chemoresistance. EMT can occur in response to activation of growth factor and adhesion receptors, low-oxygen tension (hypoxia), and exposure to chemotherapeutic drugs. We hypothesize that these cellular perturbations ultimately drive EMT through a conserved set of kinase-regulated signaling pathways and that the strength and duration of signaling through a critical set of those pathways dictate the degree to which EMT will occur. In this work, we model the response of human pancreas cancer cells to a variety of EMT agonists (growth factors) by considering the dynamics of protein phosphorylation along 32 distinct signaling nodes. The most important signaling nodes for EMT were selected through a partial least square regression (PLSR) approach, and those pathways were then used to cast the regulation of EMT in response to cellular perturbations as a control problem.

The overall ligand–signal–phenotype system is decomposed into two subsystems in series: (i) a “signal response” subsystem (f₁), whose inputs are the ligand concentrations and the responses are the relative abundance of the signaling molecules; and (ii) a “phenotype response” subsystem (f₂), for which the signals—the responses from f₁—serve as the stimulus which then stimulates the mesenchymal phenotypes as the response of f₂. This natural decomposition of the system is convenient for modeling the overall system as a convolution of the two subsystems, which also allows us to understand the dynamics of the intermediate signal responses.

Such representation allows us to formulate the determination of the optimal ligand dosage for EMT inhibition as a control problem, where the manipulated variables are concentrations of the ligands—epidermal growth factor (EGF), hepatocyte growth factor (HGF), and transforming growth factor beta (TGF-β); the controlled variables are the indicators of EMT such as the expression and localization of E-cadherin and vimentin, and the circularity of the tumor cell clusters.

In this presentation, we discuss our preliminary results on the identification of the ligand–signal–phenotype system model and on the solution of the control problem. We estimated the model parameters of f₁ (an autoregressive model with exogenous input) and f₂ (a PLSR model) using time-series data of the responses of signaling proteins and of the final phenotype measurements obtained from experiments performed with five different ligand dosages. Subsequently, we determined the optimal ligand dosage trajectory and implemented the prescribed solution via simulation in MATLAB. The results of our study provide a new set of experimentally testable hypotheses for validating the identity of the most important signaling pathways that govern EMT. They also provide a new paradigm for integrating data-driven modeling approaches such as PLSR with dynamic control system models to develop a predictive understanding of how multivariate signaling processes control complex cell phenotypes. The primary benefit of such a paradigm is that it provides a quantitative, model-based framework for using the indicated predictive understanding, in reverse, to determine how best to manipulate the signaling processes to achieve desired phenotypic responses optimally.