Fig. 4: SoptSC and STITCH analyses across timepoints suggest a branching dCPEC lineage tree.

A Heatmaps of the top-300 DEGs for dCPECs (“C”) and NECs (“N”) suggest two dCPEC subtypes at 3 of the 4 timepoints, but only one cluster at 55 div. Color scale of Z-score of DEGs. B SoptSC GRAPH charts and lineage relationships coded by subtype. NECs are more similar to C1 than C2 cells. See Supplementary Fig. 5C for pseudotemporal color coding. C Violin plots (with median and quartiles) of TTR expression in the cells of each cluster across the datasets, which increases in dCPECs over time with C1 > C2 and C1a > C1b. Every dot represents a cell. D UMAPs of aggregated dCPECs, color coded and encircled for simplified subtype groupings (left) or timepoint (right). C2 cells (lower right) are mainly present early, but also at later timepoints, while C1 cells shift over time. E UMAPs of dCPECs from individual timepoints, color coded for dCPEC subtype as in B. As C2 cells contract, C1 cells adopt a “hybrid” C1a-C1b profile at 55 div before specifying into C1a and C1b subtypes by 75 div. F–H STITCH dot plots (upper) and lineage trees (lower), color coded for all subtypes (F), dCPEC subtypes only (G), or timepoint (H). The dCPECs cluster towards the left of the roughly-triangular dot plots, with NECs at the other two corners. Each corner has cells from all four timepoints (H). C1 (pink-purple) and C2 cells (blues) belong to distinct lineages from the outset. The C2 lineage displays transience, while the C1 lineage evolves and diversifies. Small 55-div and 75-div C2 “clusters” were manually added to the lineage trees to indicate their presence in low numbers at these timepoints.