A Single Cell Transcriptomic View of Embryonic Development and Evolution

Abstract

A major goal of developmental biology is to define how the cell types of adult organisms emerge from the naive cells of the early embryo. This has been challenging because no methods have been available to comprehensively map the transient intermediate states of differentiation and the detailed molecular processes that occur during cell fate choices. We develop novel approaches for this purpose based on droplet-microfluidic single cell RNA-sequencing (“scRNA-seq”). We show that by recording the transcriptomes of hundreds of thousands of cells in parallel, over multiple points in time, and using methods in high-dimensional data analysis, it is possible to track differentiation from a pluripotent state into many distinct end points simultaneously. The resulting maps provide exquisite detail on the molecular dynamics of differentiation. Chapters 1-3 focus on mapping fate choices in embryos, and between species. We begin by optimizing experimental methods for scRNA-seq profiling of frog embryos. This allows us to collect a timeseries of >130,000 cell transcriptomes spanning the first 24 hours post-fertilization. We develop computational approaches to comprehensively annotate this dataset by defining all the cell types observed and reconstructing developmental sequences. Next, we compare the resulting differentiation map from frogs with a similar map from fish embryos. We ask how cell types change in gene expression across evolution, and how their developmental histories vary. Finally, we begin to ask systematic questions about patterns of cell fate choice including how cells transition from continua into discrete expression states, how overlapping gene expression refines at fate choices, and how transcription factors are reused across tissues. Chapters 4-5 apply similar approaches to identify principles on how artificial systems differentiate in culture. We test whether cell types are sensitive to the methods used to generate them. We show that differentiation is path-independent by identifying two trajectories leading to the same terminal state. And we infer and experimentally test a dose-dependent fate choice mechanism in human organoids. Overall our results establish general tools that allow data-driven dissection developmental programs, how they change across species, and how they can be engineered to produce cell types of interest in a dish.