Computational reconstruction of mouse development using single cell transcriptomics

Abstract

Single cell transcriptomics has significantly contributed to our understanding of cell types across species, organs and developmental processes. The rapid development of technologies, protocols and computational methods reflects a highly dynamic field over constant improvement. Particularly for Developmental Biology, being able to study at single-cell level and over time is required for a more detailed understanding of the underlying molecular mechanisms associated with cell differentiation and fate choice. Longitudinal cohort studies, ideal for capturing the dynamic nature of developmental mechanisms delineate the temporal relationship between cell type diversity and developmental dynamics. The efforts to build cell atlases from different model organisms already include extensive transcriptomic profiling of embryonic development, where detailed non-human models are essential to compensate the limitations arising from ethical reasons. In this context, the transcriptomic landscape of mouse development is perhaps the most complete among mammals. However, as cells are destroyed when measuring their transcriptomic profiles, only snapshots of the dynamical system are effectively captured. Thus, computational reconstructions of cell differentiation trajectories are essential even when lineage tracing experiments can be performed. The work presented in this dissertation uses a time course large-scale single-cell transcriptomic experiment of mouse embryos to present an effective mathematical and computational strategy that adequately describes the dynamics of gastrulation and early organogenesis on a variety of cell types during mouse embryonic development. The experimental data generated for this project, incorporates new time points into the mouse gastrulation and early organogenesis atlas already publicly available (Pijuan-Sala et al., 2018). That is, an extended version of the existing atlas. Results show a pipeline of sophisticated computational strategies to integrate the new time points. Then, to overcome the challenge of identifying and reconstructing the dynamics of all cell lineages, previously generated knowledge was combined with a variety of state-of-the-art computational methodologies. The analysis of differentiation trajectories is then taken to a deeper level in regards to the emergence of haemato-endothelial lineages, with emphasis on resolving the so-called Primitive and Definitive waves of blood production. Furthermore, perturbations to the system using mouse embryonic chimaera KO models are included. The analysis of haemato-endothelial lineages, presents a detailed reconstruction of the in vivo developmental process not reached by previous studies with single cell transcripomics. Lastly, predicted cell fates are compared to experimental observations by leveraging lineage trace experiments using cell grafting. In summary, this work highlights the complexity associated to generating developmental atlases, how to overcome the corresponding computational challenges and leverage this resource in a variety of contexts.