Breast cancer is a highly heterogeneous disease at both the molecular and pathological levels. To understand this heterogeneity and 'cells of origin' of breast cancer, it is important to dissect the normal mammary epithelial hierarchy. Over the past decade, we have established master transcriptional and epigenetic regulators that act at specific points along the mammary differentiation hierarchy, including GATA-3 and EZH2. Both transplantation and lineage tracing strategies have proven to be essential for understanding the cell types that reside in breast tissue and their molecular regulators. Lineage tracing is a powerful strategy that enables the fate of stem and progenitor cells to be tracked in situ and their roles established in development, tissue maintenance and cancer. We have combined lineage tracing with a novel three-dimensional imaging strategy to explore the relative contributions of stem and progenitor cells to post-natal mammary gland development and tissue homeostasis. Cell lineage tracing studies also provide the current gold standard for identifying 'cells of origin' in cancer. Towards this end, we are utilising newly generated transgenic strains harbouring lineage-specific gene regulatory regions, to direct the expression of specific mammary oncogenic lesions to distinct epithelial cell types. These include Pten and p53 gene deletion, both of which are frequently inactivated in breast cancer. Interrogation of the molecular expression profiles of the diverse epithelial subsets in human breast tissue has provided insight into potential 'cells of origin' of the different subtypes of cancer. Recently, we have identified a target cell population that is perturbed in precancerous tissue from BRCA1 mutation carriers, who are highly predisposed to breast and ovarian cancer. Interestingly, the RANK receptor marks a small subset of luminal progenitors and this subset is expanded in precancerous tissue from BRCA1 mutation carriers (performed in collaboration with Amgen). RANK+ but not RANK- cells were shown to be highly proliferative, prone to DNA damage and exhibited a molecular signature similar to that of aggressive basal-like cancers, thus pinpointing them as a key target population for oncogenesis. Furthermore, inhibition of the RANK ligand (RANKL) could prevent or delay tumorigenesis in Brca1-deficient mouse models, thus implicating blockade of the RANK-RANKL signaling axis as a promising breast cancer prevention strategy. Finally, to test new therapies for breast cancer, we have generated an extensive bank of patient-derived xenografts (PDXs) from primary breast cancers that have proven to be valuable preclinical models for exploring new 'druggable' targets.