Automated tracking of in vitro morphogenesis

Embryogenesis is a critical period in the developmental life cycle of organisms, but the processes that enable multicellular coordination of germ layer formation and the subsequent dynamics of tissue structure assembly are still poorly described at the single cell level. Recent advances in microscopy and computer vision have enabled whole embryo tracking of cell migration, lineage commitment, and tissue formation in model organisms such as fly, zebrafish, frog, and chick, but it is unknown to what extent these organisms recapitulate the complex processes of early human development. Direct interrogation of the development of post-implantation human embryos carries technical and ethical limitations that render in vivo study of human embryogenesis largely observational, but recent advances in human induced pluripotent stem cells (hiPSCs) derived organoid models provide a powerful in vitro platform to elucidate developmental processes. The four studies described in this dissertation developed biological and computational pipelines to study the interdependencies between cell migration, lineage specification, and morphogenesis during early human development. First, several automated cell tracking algorithms were designed and deployed to study individual cell migration events in models of mosaic pattering and differentiation. Sparsely labeled cell populations were tracked during pluripotency, yielding measures of cell migration pre-differentiation, then cell behavior change kinetics were quantified in a mosaic Wnt differentiation model, demonstrating that automated cell tracking can non-destructively detect signatures of cell fate transition. Second, collective cell migration was interrogated using a novel ensemble of deep neural cell networks to generate dense spatio-temporal descriptions of hiPSC colony behaviors. Dense cell tracks revealed colony position dependent cell-cell interactions, including correlated migration, changes in cell density, and coordinated multicellular structure formation. Cell behaviors were found to be dependent on substrate, media, and colony size, and striking behavioral transitions were detected during differentiation to germ lineages using multiple protocols, including divergent cell migration patterns induced by Wnt and BMP signaling that nonetheless resulted in convergent germ layer organization. These results demonstrate how robust multicellular pattern formation is achieved through integration of cell intrinsic, local, and global signals. Third, collective migration and body axis emergence was studied in an organoid model of gastrulation and neural tube formation, where initially round organoids spontaneously assembled an organizer like structure and then underwent several days of elongation. Automated tracking of multi-day time lapse videos and segmentation of culture images spanning weeks of differentiation, facilitated quantification of organoid growth, elongation, and stratification, enabling optimization of culture conditions to improve differentiation robustness and repeatability across multiple stem cell lines. Elongation was demonstrated to depend on the emergence of a subpopulation of neuromesodermal progenitors (NMPs) which formed signaling centers reminiscent of the node structures that direct early primitive streak formation in vivo. Ablation of the early NMP transcription factor TBXT did not abrogate tail formation, but rather caused multiple spontaneous extensions to form, suggesting a key role for TBXT expressing cells in enabling the spontaneous symmetry breaking events underlaying the formation of a head-tail axis. Fourth and finally, a mechanistic simulation of the observed TBXT symmetry breaking and tail axis formation events was developed. Cell-cell interaction forces, cell migration parameters, and cell growth were derived from 2D measurements of re-plated organoids, shown to explain observed 3D organoid structure, growth, and elongation rates, and then applied to a 3D particle simulation of tail formation. Coalescence of structures matching the TBXT+ organizers were found to depend on both percentage of TBXT+ cells and the directionality of cell migration, demonstrating that a combination of chemotaxis and population heterogeneity is sufficient to produce organizer-like structures from initially randomly mixed populations. Overall, these studies demonstrate the critical activity of cell migration in coordinating multicellular organization during early embryogenesis, and provide a kinetic, mechanistic platform to interrogate how tissues form at multiple scales across the developmental landscape