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

We generated an atlas of 489,686 single-cell transcriptomes from 62 closely spaced developmental stages spanning the first 5 days of zebrafish development, encompassing the first transcriptional events after zygotic genome activation to a freely-swimming, feeding animal. Overall design: To profile the molecular cell states that occur during wild-type zebrafish development, we generated single-cell transcriptomes from entire wild-type (TL/AB) zebrafish embryos and larvae across 50 closely spaced developmental stages ranging from 14–120 hours post-fertilization (hpf). To capture underrepresented populations and allow robust cluster identification, we sampled heavily every 12 hours; to maintain temporal continuity necessary to investigate gene expression dynamics, we additionally sampled smaller cell numbers every 2 hours (Fig. 1a). In order to profile a large number of cells at an acceptable cost, we employed the cell hashing technique MULTI-seq (McGinnis et al. 2019), which barcodes cells prior to sample collection. Input cell concentrations were then increased, which profiled more cells, but at the cost of increased artefactual cell doublets (where multiple cells are incorrectly identified as 'one cell'). The MULTI-seq hashing allowed computational identification and removal of doublets, which had multiple barcodes. We mapped reads to the GRCz11 genome, annotated using the Lawson Lab Zebrafish Transcriptome Annotation (v4.3.2) that harmonizes Ensembl and Refseq annotations, includes improved 3' UTR models, and proposes additional gene models (Lawson et al. 2020). We remapped published single-cell data from wildtype (TL/AB) zebrafish encompassing 3–12 hpf (Farrell et al. 2018) and merged the two datasets to generate a continuous single-cell time-course of 489,686 cells from 62 closely-spaced developmental stages spanning 3 to 120 hpf.