The epithelial (E) to mesenchymal (M) transition (EMT) is fundamental to development, wound-healing and cancer progression. Despite decades of research, cellular and molecular dynamics underpinning EMT are ill-defined, although the predominant view is that EMT arises from plastic conversion of individual epithelial cells into mesenchymal cells. Additionally, while EMT is controlled by growth factor and cytokine signalling, it is unclear whether such incoming signals are perceived differently across cell sub-populations spanning diverse (pre-existing) E/M-states. These questions are vital to potential modulation of EMT in cancer, and to how E/M-state diversity between and within tumours may compromise the efficacy of therapies targeting growth factor or cytokine signalling.
To better elucidate the cellular dynamics underpinning EMT, we performed highly multiplexed fluorescence imaging of >60 (phospho)protein markers (defining E/M-state, phospho-signalling, cell cycle and fate (death, senescence)), extracting ~26k quantitative features per cell, across ~80k HCC38 TNBC cells spanning a 7-day TGF-beta-driven EMT-induction. This unprecedented analysis revealed that pre-existing E- and M-state cell subpopulations within the HCC38 line undergo divergent fate regulation in response to TGF-beta, with suppression of epithelial cell proliferation (likely via p21) while, conversely, mesenchymal cells see enhanced proliferation given TGF-beta. This shifts proliferative 'preference' from epithelial (before TGF-beta) to mesenchymal cells (with TGF-beta), substantially re-balancing E-to-M states over time. We also trace putative plasticity dynamics, suggesting that EMT comprises a mixture of proliferative control and plasticity; raising new questions about why these processes co-exist and what each contributes to resulting cell populations.
In parallel, we also examined how pre-existing E/M-state diversity shapes 'perception' of incoming ligand signals. We employed epithelial- and mesenchymal-like clonal populations from heterogeneous A549 NSCLC cells and stimulated them (and the mixed parental line) with five distinct ligands over a 5-60 min time course. Using multiplexed imaging, we concurrently measured 31 cell state markers (E/M, cell cycle, stemness) and 27 signalling components across 9 pathways, generating >6k features per cell in >600k cells. This revealed clear differences between E- and M-state cells in the magnitude, kinetics and subcellular spatial organisation of acute (phospho)signalling responses, defining distinctive signalling signatures and 'perception-profiles' between E and M cells.
These studies significantly advance understanding of the dynamics underpinning EMT and also reveal how E/M-states in-turn shape cellular perception of incident signals. Beyond leveraging advanced multiplexed imaging, this presentation will highlight novel deep learning methods for molecular marker 'virtual labelling' and 'unmixing' that democratise multiplexed methods for the wider research community.