Metastasis is the major cause of poor outcome in breast cancer. Understanding the processes involved in metastasis is critical for developing new preventative and therapeutic strategies. The cancer cells within tumors are heterogeneous and can have properties as divergent as self-renewal, tumor initiation and repopulation potential, dormancy, evasion of cell death and metastasis. Cancer initiating cells are believed to reside in niches as is the case for normal tissue stem cells. Proliferation and differentiation are deregulated in these niches. We have first analyzed normal breast epithelium by single cell RNA sequencing as a powerful strategy to identify cell types and states in a given population of cells. We discovered previously unrealized expression patterns of known epithelial markers as well as identifying new markers. We then exploited patient-derived xenograft (PDX) models of human breast cancer to analyze metastatic cells in peripheral tissues and to characterize of the metastatic niche. To identify and isolate metastatic cells from PDX models, we developed a highly sensitive FACS-based assay, which allowed us to enumerate metastatic cells in mouse peripheral tissues as well as to isolate the murine stromal and inflammatory cells. We compared gene signatures by microfluidic PCR and RNA-seq in metastatic cells and stromal cells from tissues with low vs. high metastatic burden. Indeed, the inflammatory microenvironment of metastatic sites was distinct from that of the parental tumor and the normal tissue, even before metastasis takes hold. We showed at the single-cell level that early stage metastatic cells possess a distinct stem-like gene expression signature. Metastatic cells from low-burden tissues were distinct due to their increased expression of stem cell, EMT, pro-survival, and dormancy-associated genes. In contrast, metastatic cells from high-burden tissues were similar to primary tumor cells, which were more heterogeneous and expressed higher levels of luminal differentiation genes. These findings support a hierarchical model for metastasis, where metastases are initiated by stem-like cells that proliferate and differentiate to produce advanced metastatic disease. Thus metastatic niches may be sensitive to novel therapies and targeting them may prevent metastatic disease from advancing and presenting a poor outcome. (Funded by the National Cancer Institute (USA)