Single-cell transcriptomics reveal the heterogeneity and dynamic of cancer stem-like cells during breast tumor progression

Breast cancer stem-like cells (BCSCs) play vital roles in tumorigenesis and progression. However, the origin and dynamic changes of BCSCs are still to be elucidated. Using the breast cancer mouse model MMTV-PyMT, we constructed a single-cell atlas of 31,778 cells from four distinct stages of tumor progression (hyperplasia, adenoma/MIN, early carcinoma and late carcinoma), during which malignant transition occurs. We identified that the precise cell type of ER^low epithelial cell lineage gave rise to the tumors, and the differentiation of ER^high epithelial cell lineage was blocked. Furthermore, we discovered a specific signature with a continuum of gene expression profiles along the tumor progression and significantly correlated with clinical outcomes, and we also found a stem-like cell cluster existed among ER^low epithelial cells. Further clustering on this stem-like cluster showed several sub-clusters indicating heterogeneity of stem-like epithelial cells. Moreover, we distinguished normal and cancer stem-like cells in this stem-like epithelial cell cluster and profiled the molecular portraits from normal stem-like cell to cancer stem-like cells during the malignant transition. Finally, we found the diverse immune cell infiltration displayed immunosuppressive characteristics along tumor progression. We also found the specific expression pattern of cytokines and their corresponding cytokine receptors in BCSCs and immune cells, suggesting the possible cross-talk between BCSCs and the immune cells. These data provide a useful resource for illuminating BCSC heterogeneity and the immune cell remodeling during breast tumor progression, and shed new light on transcriptomic dynamics during the progression at the single-cell level.