(a) Retinal sections stained with L1CAM (red), which was used to enrich for RGCs at early stages, and the pan-RGC marker RBPMS, (green) at E13, E14, E16, and P0. Nuclei are counterstained by the Hoeschst dye (blue). (b) Relative proportions (y-axis) of major cell classes shown in Figure 1c–e at each combination of age and enrichment method. Both anti-Thy1 and anti-L1cam were used to enrich RGCs at E13, E14, E16, and P0, but only anti-Thy1 was used at P5, because L1cam becomes localized to axons postnatally. AC, amacrine cells; RPC, retinal progenitor cells. (c) Box-and-whisker plots show gene expression levels of key markers by RGCs as a function of age and enrichment method. Markers shown are two pan-RGC markers, Rbpms, Nefl, and the two cell surface proteins used for enrichment, Thy1 and L1cam. Note that Thy1 expression is poor at E13, consistent with low RGC yield in anti-Thy1-enriched cells in panel (b). Black horizontal line, median; bars, interquartile range; vertical lines, range; dots, outliers. (d) Dotplot showing genes (columns) that are selectively expressed in RGCs and RPCs. The size of each circle is proportional to the percentage of cells expressing the gene, and the color depicts the average log-normalized expression. (e) Co-embedding analysis of E14, E16, and P0 data collected in this study with whole retina single-cell transcriptomes in independent studies: E14, E16, and P0 data from Clark et al., 2019 and E15.5 data from Lo Giudice et al., 2019. Cells (points) are visualized in Uniform Manifold Approximation and Projection (UMAP) and colored by study of origin. (f) Same as (e), with cells colored by the expression level of Nefl, an RGC marker. This shows the higher enrichment of RGCs in our study compared to Clark et al., 2019 and Lo Giudice et al., 2019. (g) Same as (e), with cells colored by expression level of Fgf15, an RPC marker. (h) Relative proportions of major cell classes across different datasets analyzed in panel (e) separated by age. T, this study. (i) Confusion matrix showing post hoc transcriptomic correspondence between P5 RGC types identified in Rheaume et al., 2018 (y-axis) and P5 RGC clusters reported in this study (x-axis). In each case, RGCs from Rheaume et al. were assigned P5 cluster labels using an XGBoost classifier trained on P5 RGCs profiled in this study.