Single cell RNA-seq reveals cell fate bias of multipotent lymphoid progenitors
Herman, J.S.1, Sagar1, Grün, D.1
1Max-Planck-Institut für Immunbiologie und Epigenetik, 79108 Freiburg
Cellular differentiation of multipotent progenitors towards alternative lineages is a stochastic process governed by interactions of thousands of genes subject to substantial random fluctuations of expression levels. Understanding the process of cellular fate decision thus requires quantification of gene expression within individual cells. Here, we apply highly sensitive single-cell transcriptome sequencing of thousands of lymphocytes by a robotic nanoliter-scale version of the CEL-seq2 protocol, to investigate cell fate emergence within bone-marrow derived progenitor populations. We substantially improved our previous StemID algorithm to resolve heterogeneity across lymphoid progenitors indicative of distinct lineage biases. We present a novel algorithm, called FateID, which is based on iterative supervised learning and enables the probabilistic quantification of cell fate bias. In contrast to existing methods for lineage tree inference, which uniquely assign each cell to a single lineage, FateID infers probabilities for alternative fates without any modeling assumption and permits the derivation of high-resolution, non-linear differentiation trajectories. We apply our method to elucidate the cell fate decision of common lymphoid progenitors (CLPs) towards distinct lineages. Our approach reveals the branching of CLPs into the B lineage and the plasmacytoid dendritic cell (pDC) lineage. Consistent with previous fate-mapping experiments, FateID predicts that conventional dendritic cells, in contrast to pDCs, differentiate from CLPs only at very low probability. Furthermore, our method derives a pseudo-temporal order of single cell transcriptomes along non-linear trajectories, which enables the identification of early cell fate markers within multipotent progenitor populations. By this approach, we identified lineage-determining factors within the heterogeneous CLP population, and, in particular, a novel cell surface marker that can be utilized for the purification of pDC progenitors.