Mammary morphogenesis is an orchestrated process involving differentiation, proliferation and organization of cells to form a bi-layered epithelial network of ducts and lobules embedded in stromal tissue. We have engineered a 3D biomimetic human breast that makes it possible to study how stem cell fate decisions translate to tissue-level structure and function. Using this advancement, we describe the mechanism by which breast epithelial cells build a complex three-dimensional, multi-lineage tissue by signaling through a collagen receptor. Discoidin domain receptor tyrosine kinase 1 induces stem cells to differentiate into basal cells, which in turn stimulate luminal progenitor cells via Notch signaling to differentiate and form lobules. These findings demonstrate how human breast tissue regeneration is triggered by transmission of signals from the extracellular matrix through an epithelial bilayer to coordinate structural changes that lead to formation of a complex ductal-lobular network.

Phenotypic heterogeneity in cancers is associated with invasive progression and drug resistance. This heterogeneity arises in part from the ability of cancer cells to switch between phenotypic states, but the dynamics of this cellular plasticity remain poorly understood. Here we apply DNA barcodes to quantify and track phenotypic plasticity across hundreds of clones in a population of cancer cells exhibiting epithelial or mesenchymal differentiation phenotypes. We find that the epithelial-to-mesenchymal cell ratio is highly variable across the different clones in cancer cell populations, but remains stable for many generations within the progeny of any single clone—with a heritability of 0.89. To estimate the effects of combination therapies on phenotypically heterogeneous tumours, we generated quantitative simulations incorporating empirical data from our barcoding experiments. These analyses indicated that combination therapies which alternate between epithelial- and mesenchymal-specific treatments eventually select for clones with increased phenotypic plasticity. However, this selection could be minimized by increasing the frequency of alternation between treatments, identifying designs that may minimize selection for increased phenotypic plasticity. These findings establish new insights into phenotypic plasticity in cancer, and suggest design principles for optimizing the effectiveness of combination therapies for phenotypically heterogeneous tumours.

There is widespread interest in finding therapeutic vulnerabilities by analyzing the somatic mutations in cancers. Most analyses have focused on identifying driver oncogenes mutated in patient tumors, but this approach is incapable of discovering genes essential for tumor growth yet not activated through mutation. We show that such genes can be systematically discovered by mining cancer sequencing data for evidence of purifying selection. We show that purifying selection reduces substitution rates in coding regions of cancer genomes, depleting up to 90% of mutations for some genes. Moreover, mutations resulting in non-conservative amino acid substitutions are under strong negative selection in tumors, whereas conservative substitutions are more tolerated. Genes under purifying selection include members of the EGFR and FGFR pathways in lung adenocarcinomas, and DNA repair pathways in melanomas. A systematic assessment of purifying selection in tumors would identify hundreds of tumor-specific enablers and thus novel targets for therapy.

Supplementary table (.xlsx) with the phenotypic ratios and estimated population sizes of barcoded clones

Genes marking different glioblastoma subtypes. Drawn from DOI:10.1016/j.ccr.2009.12.020 and DOI:10.1126/science.1254257

Data used for simulations of clones under different treatments. Contains growth rate, estimated cell count, and log2(epithelial/mesenchymal) ratio for each clone.

The search for genes that regulate stem cell self-renewal and differentiation has been hindered by a paucity of markers that uniquely label stem cells and early progenitors. To circumvent this difficulty we have developed a method that identifies cell-state regulators without requiring any markers of differentiation, termed Perturbation-Expression Analysis of Cell States (PEACS). We have applied this marker-free approach to screen for transcription factors that regulate mammary stem cell differentiation in a 3D model of tissue morphogenesis and identified RUNX1 as a stem cell regulator. Inhibition of RUNX1 expanded bipotent stem cells and blocked their differentiation into ductal and lobular tissue rudiments. Reactivation of RUNX1 allowed exit from the bipotent state and subsequent differentiation and mammary morphogenesis. Collectively, our findings show that RUNX1 is required for mammary stem cells to exit a bipotent state, and provide a new method for discovering cell-state regulators when markers are not available.