Intratumor heterogeneity is a major obstacle toward understanding and treatment of cancers. We have analyzed cellular genetic and phenotypic heterogeneity in breast tumors and found that higher pre-treatment genetic diversity predicts therapy resistance and distant metastases are the most diverse. We also developed an experimental model of clonal heterogeneity and shown that polyclonal tumors grow faster and are more metastatic, the tumor-driver clone can be a minor subpopulation acting via non-cell-autonomous mechanisms, a dominant clone can outcompete the tumor-driver minor clone leading to tumor collapse, and that cancer therapies intensify clonal competition potentially leading to inadvertent acceleration of disease progression. These data questions current views of how to define cancer-driving events in clinical samples and how to design treatment based on this knowledge. Based on our preliminary data we hypothesize that clonal heterogeneity within tumors drives metastatic progression and therapeutic resistance and that understanding the molecular and cellular mechanisms underlying clonal interactions within tumors will improve the clinical management of breast cancer patients. The goal of this proposal is to test these hypotheses using a multidisciplinary approach applied to clinical samples and experimental models. We will utilize a model of clonal heterogeneity of breast cancer that we have developed and will generate new ones from patient samples. We will analyze the composition of primary and metastatic tumors in unperturbed states and following therapies and investigate molecular and cellular mechanisms by which clonal interactions promote tumorigenesis using comprehensive molecular and mathematical approaches. We will explore the role of non-cell-autonomous tumor drivers and clonal interference in clinical breast tumors by analyzing tumors at different progression stages and pre- and post-treatment samples using comprehensive and single cell profiling approaches coupled with mechanistic studies in xenograft models. We will design novel treatment strategies for heterogeneous tumors. Tumor progression is a somatic evolution following Darwinian principles. Despite being universally accepted, current approaches to the study and treatment of cancers do not utilize and build on these evolutionary principles. The proposed studies will address this void by the combined analysis of experimental and clinical breast tumors. Our results will help the design of more effective therapeutic strategies for heterogeneous tumors.