Distinct neurogenic progenitor cell states balance cell type production in the embryonic mouse retina

A central focus of developmental biology is how cellular diversity is generated during embryogenesis. The mammalian retina is a particularly striking instance of this problem, as over 100 types of retinal neurons are reliably produced from a common pool of multipotent progenitor cells in the embryo. While the types of retinal cells being generated shift over time, it remains unclear how retinal progenitor cells generate diverse cell types during a single temporal window. We address this problem by focusing on the early embryonic mouse retina, when retinal progenitor cells are generating retinal ganglion cells, amacrine cells, cone photoreceptors, and horizontal cells. Prior studies suggest that individual progenitor cell fate choices may be largely stochastic, but how a stochastic mechanism reliably produces all retinal cell types in their correct proportions is unknown. We characterized retinal progenitor cell heterogeneity by integrating single cell RNA-sequencing datasets, which revealed transcriptionally distinct populations of neurogenic progenitor cells (NPCs) about to undergo a terminal division. We used a Cre-based labeling approach to identify Galanin-expressing NPCs as a population biased to produce retinal ganglion cells and amacrine cells. This population complements the previously described Olig2-expressing NPCs, which produce cones and horizontal cells. By analyzing embryonic clones, we determined that Galanin+ and Olig2+ NPCs are largely born in independent, asymmetric progenitor cell divisions that also produce a self-renewing daughter cell. We found that the production of distinct types of NPCs is directed by differential levels of Notch signaling, and that the pattern of Notch signaling components is consistent with a mechanism of lateral inhibition between neighboring NPCs. These findings support a model of embryonic retinogenesis in which fate choices are organized through discrete, fate-biased NPC populations and buffered by local feedback between them. This refines the model of retinal neurogenesis and provides a new understanding of the clonal origins of amacrine and retinal ganglion cells in the mammalian retina.