Abstract:
Epithelial-mesenchymal transition (EMT) is a developmental program aberrantly activated in pancreatic ductal adenocarcinoma (PDAC), promoting chemoresistance and metastasis. Antagonizing EMT may be an effective adjuvant therapy approach to promote tumor response to chemotherapy, but identifying the right drug targets is complicated by the high degree of heterogeneity with which EMT occurs in PDAC tumors. Even in PDAC cancer cells treated identically with different EMT agonists, EMT occurs heterogeneously, leading to problems with identifying the responsible druggable signaling pathways based on population-level measurements. This dissertation research describes the development and application of two data-driven modeling approaches to understand the transcriptional and signaling basis of EMT heterogeneity. In the first study, we developed a workflow integrating: 1. iterative indirect immunofluorescence imaging of PDAC cells to quantify the activities of up to seven purported EMT-regulating pathways and two EMT phenotypic markers, and 2. a mutual information (MI) computational model to predict the cooperating pathways that are most informative of the EMT phenotype. MI analysis of signaling states and EMT phenotypes in multiple cell lines and in tumors from a patient-derived xenograft model of PDAC identified ERK as the most important signaling node explaining EMT heterogeneity across the diverse agonists investigated. Inhibition and knockdown studies confirmed the role of ERK and revealed a compensatory, EMT-driving JNK activation in the context of MEK inhibition. In the second study, we determined the degree to which EMT heterogeneity may be transcriptionally primed for different EMT agonists. We first demonstrated lineage fidelity of EMT response in a series of single-cell sorting experiments. We then developed a workflow combining genetic barcoding with single-cell RNA sequencing to pair likelihood of EMT induction with baseline transcriptional states. We found conserved lineage-dependent propensities to undergo EMT and implemented a partial least-squares discriminant analysis (PLS-DA) using differentially expressed transcripts as model features to classify EMT state post-induction. We identified genes in a range of pathways whose variation was predictive of later EMT induction. Together, our results demonstrate the importance of ERK for explaining EMT heterogeneity in response to diverse agonists and the presence of durable EMT-priming within cancer cell populations. These findings nominate specific targeted inhibitors to combine with chemotherapy to prevent EMT-associated resistance.
To view the full dissertation, please click here.
Comments