TIME-COURSE SINGLE-CELL TRANSCRIPTOMIC ANALYSIS OF CONTROLLED NEURAL CONVERSION OF HUMAN IPS CELLS
P.H. Chu1, C. Malley1, J. Braisted1, C. Tristan1, R. Bargaje1, and I. Singeç1
1National Center for Advancing Translational Sciences (NCATS), Stem Cell Translation Laboratory (SCTL), Rockville, MD
Presented at the National Institutes of Health Molecular Biology in Single Cells Symposium, March 23-24 2018, Bethesda, MD
The elucidation of cancer pathogenesis has been hindered by limited access to patient samples and lack of reliable model organisms. Recent progress in epigenetic reprogramming leveraged the development of new strategies in regenerative medicine, disease modeling and stem cell therapy. One key translational challenge in the field is to differentiate induced pluripotent stem cells (iPS) in a controlled, scalable, and reproducible fashion into distinct functional cell types at high purity under chemically defined conditions. Neural induction of pluripotent cells can be achieved by small molecule inhibition of specific pathways and previous studies used microarray and bulk RNA-seq strategies to characterize this process. However, such datasets cannot resolve how individual cells exit the pluripotent state and commit toward specific multipotent progenitor states, owing to the heterogeneous and asynchronous nature of the differentiation process. Identification and characterization of subpopulations and their lineage dynamics during differentiation is essential to define cell type signatures and optimize protocols. Here, we applied single-cell transcriptomics to analyze the transition from pluripotency to neural lineage entry by blocking BMP and TGF-beta pathways separately and in combination. By estimating gene expression correlations in differentiating cells across seven days, we identified critical state transitions. We constructed an unsupervised pseudotime model and clustered cells by nonparametric tSNE to identify significantly differentially expressed genes between differentiating subpopulations. Dissection of critical expression transition points suggests the existence of a stepwise regulatory process towards neural commitment in relationship to cell cycle states.
Notes from Claire
This poster was a true team science effort! We performed a 7 day differentiation of iPSCs with dual-SMAD inhibitor treatment to guide cells towards the neuroprogenitor lineage. About 200 cells per day were harvested for ddSeq single-cell sequencing (1385 cells total). I normalized, scaled, and visualized UMI count data in the R packages Monocle and Seurat. I found Seurat was fastest for the differential gene expression tests and heatmaps, while Monocle’s best feature is pseudotime tree construction. Additional analysis was done in SCPattern to find dynamic gene changes in the time-course; gene set enrichment in NCATS BioPlanet; and with in-house scripts for calculating the Critical Transition Index. I’m expanding this research with my colleagues and will present an updated version at an NIH bioinformatics poster day later this month. Seurat code to come soon.