The development of an organism requires sequential state transitions towards more specialised cell types. Many state transitions coincide with changes in cell shape, with emerging evidence suggesting strong feedback between shape and state. An example of transitions where state and shape are tightly coupled is epithelial-to-mesenchymal transition (EMT) which plays a crucial role in development and pathogenesis. While the changes in gene expression driving EMT have been extensively studied, the cell shape dynamics during EMT remain poorly understood.

To address this challenge, we developed a morphometric pipeline employing spherical harmonics descriptors to represent 3D cell morphodynamics in a low-dimensional morphospace quantitatively. Combining live-cell imaging with this pipeline, we characterised the cell shape trajectories associated with EMT. We inferred the underlying stochastic morphodynamics by modelling shape dynamics as a Langevin process and characterised the cell shape noise. Our findings reveal a peak in noise coinciding with a transition from epithelial to mesenchymal attractor states. Molecular perturbation experiments and mathematical modelling suggest that an increase in actin protrusivity and a decrease in membrane tension account for the cell shape noise during EMT. Together, our study suggests that EMT-associated cell spreading can be described as a transition between morphospace attractors.