Automated time-lapse data segmentation reveals in vivo cell state dynamics

Abstract

Embryonic development proceeds as a series of orderly cell state transitions built upon noisy molecular processes. We defined gene expression and cell motion states using single-cell RNA sequencing data and in vivo timelapse cell tracking data of the zebrafish tailbud. We performed a parallel identification of these states using dimensional reduction methods and a change point detection algorithm. Both types of cell states were quantitatively mapped onto embryos, and we used the cell motion states to study the dynamics of biological state transitions over time. The time average pattern of cell motion states is reproducible among embryos. However, individual embryos exhibit transient deviations from the time average forming left-right asymmetries in collective cell motion. Thus, the reproducible pattern of cell states and bilateral symmetry arise from temporal averaging. In addition, collective cell behavior can be a source of asymmetry rather than a buffer against noisy individual cell behavior.