The biggest challenges clinicians face during treatment of breast cancer are tumor heterogeneity and tumor metastasis. With breast cancer tumor heterogeneity, the problem is that the genomic variability within tumors and between patients limits the efficacy of breast cancer therapy. Directed therapies for specific types of breast cancer improved breast cancer survival times, yet due to the molecular complexity of breast cancer, treatment is still inadequate; with tumors initially regressing only to reoccur and become resistant to therapy. Many reoccurring tumors manifest as distant metastasis. It is these metastases that lead to breast cancer lethality. To simplify the molecular complexity of breast cancer, researchers have taken advantage of mouse models where different cancer-causing events found in human breast cancer are used to initiate mammary tumors in mice. However, the degree to which mouse models are reflective of the heterogeneity of human breast cancer needed to be demonstrated. If mouse models with relationships to individual types of human breast cancer could be identified, such a finding would represent a major breakthrough and enhance the research of mechanisms and treatments for drivers of breast cancer progression using mouse models. To test the hypothesis that genomic similarities exist between mouse models of breast cancer and human breast cancer, I characterized MMTV-Myc initiated tumors that had demonstrated histological heterogeneity. Using bioinformatic analysis of tumor gene expression data from MMTV-Myc mouse mammary tumors and human breast cancer samples, molecular similarities and mouse human counterparts were identified. As a result, I hypothesized that molecular similarities between mouse and human breast cancer are widespread. To this end, I generated and analyzed a database of gene expression data from over 1000 mouse mammary tumors and over 1000 human breast tumors. I detected relationships between individual mouse model tumors and specific types of human breast cancer through gene expression patterns. This was extended to predict which signaling pathways were activated in both human breast cancer and the mouse models. This novel approach established relationships between individual mouse mammary tumors and human breast cancer, identifying shared pathways that may contribute to tumor progression in mouse and human breast cancer. Using this database as a predictive resource, I developed the hypothesis that the E2F transcription factors regulate breast cancer metastasis. Using a genetic test in the MMTV-PyMT mouse model, I show that E2F1 and E2F2 are critical for progression through multiple stages of metastasis. Using predictive informatics and gene expression analysis, I show that multiple pro-metastatic features are impacted with E2F loss, including tumor angiogenesis and activation of the pro-metastatic hypoxia response gene expression program. As part of uncovering E2F1’s role in tumor metastasis, I uncover new regulators of metastasis: Adm and Fgf13. Collectively, the work in this dissertation demonstrates that integrating gene expression analysis, bioinformatics, mouse models and multiple experimental techniques provide the unique capacity to study the complex molecular differences and mechanisms across the spectrum of human breast cancer. Importantly, these strategies have allowed us to credential mouse models for relevance to human breast cancer and identify mechanistic features of breast cancer metastasis.
Read
