Data-driven modeling of T cell morphodynamics during migration

T cells can spontaneously migrate rapidly through the extracellular matrix using an 'amoeboid' mechanism, which is believed to aid their search for antigens. While modeling and experiments have in part addressed motility parameters and the impact of matrix characteristics, the morphodynamics during locomotion remains not well understood. We consider a data-driven approach using low–spatial resolution time-lapse fluorescence microscopy videos of activated T cells migrating in collagen matrix or under agarose gel. We analyze the resulting cell shapes by using an autoencoder to extract a low-dimensional 'shape space'. We find time-irreversible motion in shape space, as expected from Purcell's theorem for a swimmer at low Reynolds number. We also find evidence of distinct signatures of turning behavior as a result of cell–matrix interactions. Our statistical analysis allows us to generate artificial trajectories of cells and their shapes using a coarse-grained morphology. This approach confers the possibility of inferring predictive dynamical laws which would inform biophysical models of cell morphology during migration and interaction with the environment.