Breast cancer therapies traditionally involve targeting cell growth or expression markers unique to tumor cells to halt proliferation. However, they often fail due to the ease with which cancer can adapt new ‘escape’ cell states. Transcriptional hedging has emerged as a property of cancer that allows it to persist after the application of stressors such as chemotherapy. Using innate transcriptional heterogeneity present in biology across a population of cells, along with a high mutation rate, cancer cells are able to hedge their survival and live on through states that are more favorable under changing external stressors.In Chapter 1, we explore a specific cell state that is present in a single cell study of triple negative breast cancer tumor samples and cell line models under therapy conditions. We use computational methods to identify this drug-resistant inflamed cell population with upregulated cGAS/STING signaling that is consistently observed across post-treatment patient samples and also show that it is a result of transcriptional not genetic variation. We then experimentally show that it is selected for under chemotherapy treatment. In Chapter 2, we explore a method that cancer cells use to increase transcriptional heterogeneity beyond natural levels to increase survival. We observe using Kaplan-Meier analysis that patients with increased levels of RNF8 or MIS18A, most of whom are under treatment, have poor survival outcomes compared to those with lower levels. Looking at RNA-seq profiles of the tumors, we see that increased RNF8 or
MIS18A leads to higher transcriptional heterogeneity which is responsible for the difference in survival after accounting for genetic variation. We show experimentally that this heterogeneity is inducible by perturbing RNF8 and MIS18A expression by using computational methods to summarize the complex increase in transcriptomic heterogeneity seen across a cell population. Once this heterogeneity is induced, we observe that it can lead to increased cell survival after drug treatment. Finally, in Chapter 3, we create a tool to analyze novel transcriptional states induced in breast cancer cells by Decoy-seq, a scalable small RNA inhibition platform. Our analysis explores transcriptomic shifts in a holistic manner as opposed to direct pathway analysis like GSEA. Given the subtle transcriptomic shifts that small RNA perturbations create, an approach that maximized signal over noise was needed to observe these changes. Using a varimax rotation to traditional PCA gene shift explorations, we were able to link specific gene shifts to specific perturbations which then allowed for a whole transcriptome approach to pathway analysis as opposed to gene set analysis. We identified perturbations that deplete cells in G1 by speeding up the G1/S transition, upregulate the glycolytic process, downregulate mRNA 3’-end processing, and many others allowing one to potentially identify targetable induced cell states by small RNA inhibition.
In summary, methods in this work allow us to identify transcriptional states in breast cancer cells achieved through transcriptional heterogeneity or small RNA inhibition, and therefore create the starting point for successfully inhibiting their growth or survival. Use of therapies that directly target these identified processes may hold promise for improved patient outcomes.
