Capturing the pathway logic of cell signaling for epithelial-mesenchymal transition in ovarian cancer cell lines.

Abstract

e18063 Background: Epithelial-to- mesenchymal transition (EMT) plays a key role not only in cancer invasiveness and progression but also in treatment resistance. Experimental models suggest that epithelial and mesenchymal phenotypes may represent steady states supported by the regulatory logic of intracellular signaling networks. Using known regulatory interactions between intracellular signaling molecules as a framework, we explored EMT in silico with the aim of identifying potentially novel means of inhibiting metastasis in ovarian cancer. Methods: A mathematical model connecting 27 genes involved in EMT through 153 regulatory interactions was assembled from the Mogrify, Reactome, String and Pathway Studio (PS) databases, the last of which used the MedScan natural language processing (NLP) engine to process over 8,000 full-text publications. Logic model parameters dictating the regulatory response dynamics of the EMT circuit were constrained to values that accurately predicted the proteomic profiles characterized as either epithelial or mesenchymal in 30 cell lines in the publicly available MaxQB database. Results: A total of 94,672 candidate models predicted the protein expression in stable epithelial and mesenchymal cells equally well, clustering into 24 classes of slightly different EMT transition kinetics. Discrete event simulations based on these models identified 7 unique combinations, involving the concurrent modulation of at least 3 gene products capable of effecting this transition successfully even in the presence of noise. Every combination required inactivation of ZEB2, along with 2 other manipulations. Inactivation of TP53, HOXA5, or IRF7 was required in 3 combinations, while inactivation of ICAM1 was required in 2 cases. Only BCL2 was required to be activated in any of these scenarios. Interestingly, BCL2 inhibition is now applied clinically in some hematologic cancers and proving effective in patient-derived cell lines for high-grade serous ovarian cancer when combined with MEK inhibition. Conclusions: This work suggests that EMT might at least in part be driven by normal regulatory interactions between specific proteins in a network of pathways, without the requirement for additional mutations or alterations in the underlying circuitry. Maintenance of ZEB2 or TP53 activation as well as BCL2 inhibition may thus represent promising avenues for future research in arresting EMT and associated chemo-resistance.