Charting and Navigating Fate Decisions in Directed Differentiation of Stem Cell-Derived Human Beta Cells

Abstract

Beta cells are front-runners in the fields of regenerative medicine and directed differentiation. Significant progress has been made in producing beta cells in vitro from pluripotent human cells and using these cells in animal models of diabetes. Despite these successes, relatively little is known about the developmental trajectory followed during differentiation or the genes that govern this process at the molecular level. In this thesis, we use a combination of descriptive and perturbative approaches to expand our understanding of in vitro beta cell differentiation. We first present our work on applying single-cell RNA sequencing to characterize the cell types present in the human and murine pancreas (Baron et al., 2016). In this study, we generate whole-transcriptome profiles of all endocrine cell types present in the islets (beta, alpha, gamma, delta and epsilon cells) as well as other cell types present within the pancreas, including acinar, ductal, stellate, endothelial and immune cells. By providing these definitions, this study helps frame the goals and criteria for success of in vitro differentiation. We then apply single-cell RNA sequencing in the context of in vitro beta cell differentiation to comprehensively characterize the cell types produced by, and during, differentiation. Leveraging our earlier study, we compare in vitro and in vivo cell types. We identify stem cell derived SC-beta and SC-alpha cells that resemble their in vivo counterparts in the human islets. We also discover a novel in vitro endocrine cell type resembling enterochromaffin cells (SC-EC), and show that a late lineage bifurcation separates SC-beta and SC-EC cells in vitro. Finally, we identify and characterize a new cell surface marker allowing for specific isolation of SC-beta cells. In the following chapter, we introduce consensus non-negative matrix factorization (cNMF) as a method for the identification of cell identity and cell activity signatures in single-cell RNA sequencing data (Kotliar et al., 2018). While this method was initially developed to analyze in vivo and in vitro pancreatic cell types, we realized that this approach for matrix factorization was particularly well suited to identifying gene expression profiles related to cellular activities such as replication or activation. Using both simulated and experimental data, we show cNMF can accurately infer these programs and provide an additional layer of interpretation for single cell data. Finally, we present unpublished results on whole-genome knock-out screening in the context of beta cell in vitro differentiation. In this chapter, we apply CRISPR-based lentiviral libraries to systematically perturb human genes and measure the effect of these perturbations on the capacity of the cells to reach the endocrine cell types first identified by single-cell RNA sequencing. The goal of this project is to produce bespoke cell lines via targeted genetic modifications that maximize the production of SC-beta cells. Secondarily, results may suggest either small molecules or signaling factors that can mimic certain knockouts and increase the efficiency of SC-beta differentiation. In summary, the work presented here advances our understanding of in vitro beta cell production, through a deeper characterization of stem cell derived beta cells, other lineages that are produced in parallel. Further, this work sets the stage for the next generation of strategies for creating maximally efficient differentiation protocols.