Reverse engineering of gene regulatory networks governing cell-cell communication in the microenvironment of pancreatic cancer

Abstract

Background: Pancreatic ductal adenocarcinoma (PDAC) is one of the leading causes of cancer death, with a five-year survival rate of <5% and a median survival of 6 months. Extensive desmoplastic reaction is a characteristic feature and a prognostic factor of PDAC, which conveys its resistance. Desmoplastic stroma accounts for approx. 90% of tumor volume and consists predominantly of non-malignant fibroblasts (pancreatic stellate cells, PSC). Previous studies have revealed the PSC mesenchymal origins, capacity to switch between quiescent and activated states, proinflammatory features, expression of soluble factors, ability to migrate, and phagocytize. State of the art: Abundance of stroma has sparked previous attempts to dissect the interactions between PSC and tumor cells (TC) producing a common picture of a microenvironment supporting PDAC development. Unfortunately, focus on snapshot-like analysis has proven difficult to translate into therapeutical advances, as it discards the dynamic interactions in the microenvironment, as well as the temporal dynamics of gene expression itself. Gene regulatory networks (GRN) adapt to environmental cues by rewiring connections between genes, those induced modulations effectively lead to state-transitions e.g. PSC activation, or produce mutually exclusive cell-fate decisions e.g. differentiation, senescence, or death. We recognize that cell-specific assignment of stimuli, identification of genes forming the GRNs, as well as the identification of cellular state-changes remain undiscovered. We hypothesize that at an early stage, the quiescent → activated PSC transition yields a steady state PSC gene regulatory network (GRN), but the subsequent succession of impulse responses along TC→PSC→TC interaction axis drives both cell types into unstable states maintained only for the duration of the direct TC-PSC contact. Aims: Through the application of a high-throughput complexity reduction approach and in silico modeling I aim to reconstruct the GRNs underlying the cell-cell communication, and identify key soluble factors shaping the double-paracrine interactions. I aim to use the models to gain a mechanistic and functional insight into how the cues are integrated and how they affect GRN maintenance. I hope to capture cell-fate decisions and identify key dynamic changes with the ultimate goal of finding genetic markers to aid development of novel therapeutic options for this deadly malignancy. Results: We have individually stimulated PSC and TC with conditioned supernatant from the respective other cell type and recorded a time-series (1-24h) from which genome-wide microarray expression data has been generated. In this dissertation I used the time-resolved expression profiles to identify significant gene kinetics through an approach-involving gene ranking, filtering, and clustering followed by gene ontology and pathway analysis. I identified key gene interactions using a genetic algorithm embedded in a continuous time recurrent neural network (CTRNN) modeling scheme. Then I used the derived GRN’s to produce a picture of unique intercellular interactions. Through in silico simulations with the created models, and subsequent data analysis and interpretation I delivered targets for experimental testing on the inter- as well as intra-cellular levels. Experimental validation of the selected gene targets using gene silencing and qRT-PCR confirmed the in silico predicted TC network behavior; validation of the intercellular connections confirmed their dependence on the identified networks.