Single-cell analysis of shared signatures and transcriptional diversity during zebrafish development

Key results

1. 01

   ShownThe atlas resolves roughly 300 cell types or states and gives temporally resolved expression for two-thirds of annotated genes, most of which vary across cell types or stages rather than being ubiquitous.

   489,686 cells from 62 stages in 19 tissues and 521 clusters; 24,216 of 36,250 genes (67%) detected; ~3.6% ubiquitous with low fluctuation (logCV ≤ –0.85), ~18% cell-type specific (≤25 clusters), 78% variable across many tissues; computationally predicted and XLOC genes detected at 28% and 37%. · Temporally dense sampling of zebrafish…; Spatiotemporal variation of gene expression…; Remapping and integration of 2018 Drop-seq data

2. 02

   ShownOnly a limited catalog of expression programs is reused across tissues, and shared cores are usually customized per cell type (e.g., a Megalin-linked reabsorption program in gut LREs and pronephros, a pan-epithelial program, a mucin O-glycosylation program).

   Mfuzz on kNN-smoothed data: 200 modules requested, 147 retained after QC, 90 shared across tissues (8–735 member genes, mean 95); 82 annotated (20 ubiquitous, 62 restricted to cell types in 2 or more tissues); GEP-193 split megalin paralogs lrp2a (pronephros) and lrp2b (intestine); GEP-106 spanned all epithelia incl. lens, liver, RPE. · Identification of shared developmental gene…; Gene module identification

3. 03

   ShownMost whole-transcriptome states persist for at most 36 hours; long-term states are mostly non-cycling late cells, while long-term cycling states include known stem cells and a few uncharacterized candidate progenitor populations.

   ε-neighborhood (ε = 35) stage differences: 18% of cells in states lasting ≤12 h, 58% ≤24 h, 81% ≤36 h, 93% ≤48 h, 7% >48 h; of long-term (≥36 h) cells, 68% non-cycling (74% of these at 72 hpf or older) and 32% cycling, including radial glia, satellite cells, HSCs plus oit3+/lyve1a+ lymphatic cells and myoseptal fibroblasts. · Mapping the duration of transcriptional states…; Identifying short-term and long-term…

4. 04

   ShownFocused non-skeletal muscle analysis reveals three zebrafish pericyte subpopulations, one (epas1a+) spatially restricted to hindbrain vessels, and distinct molecular profiles for circular versus longitudinal intestinal smooth muscle layers.

   3,866 cells yielded three pericyte clusters (C4, C9, C20); HCR placed epas1a+/ndufa4l2a+ pericyte-2 cells at the hindbrain posterior cerebral vein, almost none in forebrain, eye or arches; 5000 simulated doublet pairs per group failed to recreate subtype signatures; in situ mapped C8 (kcnk18, fsta) to inner circular, C10 (il13ra2) to outer longitudinal SMCs. · Focused analysis of zebrafish non-skeletal…; Molecular characterization of zebrafish…; Simulation of artificial pericyte doublets

5. 05

   ShownEndoderm sub-clustering yields the first specific markers for the pneumatic duct and identifies zebrafish homologs of human best4+ enterocytes, with URD trajectories nominating candidate progenitors and transcriptional regulators.

   12,592 endodermal cells in 33 clusters; C32 (arising 2–3 dpf) specifically expresses sftpba and sim1b, confirmed by in situ in pneumatic duct and anterior swim bladder bud; best4+ cells (first seen 3 dpf) share best4, otop2, cftr, notch2 with human counterparts; branchpoint progenitors express atoh1b, ascl1a, tnfrsf11a, followed by dacha and pbx3a. · Focused analysis of endodermal development…; Comparison of gene expression between species