Dynamic plasticity within the EMT spectrum, rather than static mesenchymal traits, drives tumor heterogeneity and metastatic progression of breast cancers

The Epithelial-to-Mesenchymal Transition (EMT) is a developmental cellular program frequently coopted by cancer cells 1 and is a key contributor to both heterogeneity in solid tumors 2–4 and later stage chemo-resistance and metastasis 5,6 . Rather than being a switch from an epithelial to a mesenchymal state, increasing evidence points to the existence of intermediate EMT states, wherein cells co-express both epithelial and mesenchymal traits 7–13 . Multiple stable intermediate EMT states possessing unique characteristics exist across the EMT spectrum 7,8,14,15 , contributing to the complex heterogeneity of tumors to promote metastasis 16 . While much work has been carried out identifying and characterizing EMT-inducing transcription factors 17–20 , the transcriptional and epigenetic networks responsible for the stability and maintenance of the midpoints along the EMT spectrum are poorly defined. In addition, there are currently no approaches to identifying and quantifying intermediate EMT subpopulations within patient tumors to evaluate their prognostic significance. Using clonally isolated derivatives of the SUM149PT breast cancer cell line, we systematically interrogate how each EMT state independently contributes to heterogeneity and influences metastatic progression, uncovering the role of RUNX2 in stabilizing certain intermediate states. Using SUM149PT-derived tumors as a training set, we develop an entropy-based model to quantify phenotypic heterogeneity and EMT status. Remarkably, the majority of cell states captured in the SUM149PT model are represented in a selection of patient tumors, laying the foundation for quantification of epithelial-mesenchymal heterogeneity and understanding the role of the intermediate EMT state in tumor progression.