Zebrahub

10 stages,

40 embryos,

~120000 cells,

more to come...

Zebrahub logo showing a zebrafish embryo imaged by lightsheet microscopy
Single-embryo
Single-cell
Atlas
of vertebrate development using zebrafish as a model organism.

About

Zebrahub is a single-cell RNA sequencing timecourse dataset of zebrafish development at single-embryo resolution.

The first dataset of around 120,000 cells spans 10 developmental stages: from end-of-gastrulation embryos to 10-day larvae (bud-stage, 5-, 10-, 15-, 20-, 30-somites stages, as well as 2-, 3-, 5- and 10-days post-fertilization). Four embryos were sequenced per time point. We strive to achieve the highest possible quality; in that context, the dataset will evolve to include more stages and more cells.

The Royer Group leads this work in collaboration with CZ Biohub’s Data Science and Genomics Platforms. We aim to provide an easy-to-navigate and high-quality single-embryo resolved timecourse dataset, providing a comprehensive picture of development by leveraging the latest single-cell technologies.

For this first dataset we used a combination of the 10x Chromium (standard and HT) platform for library preparation and the NovaSeq 6000 for sequencing. This dataset was produced to complement upcoming preprints, but in the spirit of open and accelerated science, we have made this dataset available ahead of time (see Data release policy for details). Redistribution of these data should include the full text of the data use policy.

Zebrahub datasets have been key for our own upcoming and future work on deciphering vertebrate development, and we hope that they will also prove useful for the zebrafish community at large.

Full Dataset

Full zebrahub single cell dataset, rotating 3D umap,

Full Dataset

10 stages, 40 embryos

Launch cellxgene

Timepoints

zebrafish embryo,  10 hpf, bud stage

10 hpf

Bud stage

Launch cellxgene
zebrafish embryo,  12 hpf, 5 somites

12 hpf

5 somites

Launch cellxgene
zebrafish embryo,  14 hpf, 10 somites

14 hpf

10 somites

Launch cellxgene
zebrafish embryo,  16 hpf, 15 somites

16 hpf

15 somites

Launch cellxgene
zebrafish embryo,  19 hpf, 20 somites

19 hpf

20 somites

Launch cellxgene
zebrafish embryo,  1 dpf, 30 somites

1 dpf

30 somites

Launch cellxgene
zebrafish embryo,  5 dpf, Larva

5 dpf

Larva

Lauch cellxgene
zebrafish embryo,  10 dpf, Larva

10 dpf

Larva

Launch cellxgene

hpf = hours post fertilization
dpf = days post fertilization

Method

To obtain high-quality data, we optimized a single-embryo single-cell dissociation protocol. This protocol varies over the 10 developmental stages considered, with very gentle dissociation for early stages and harsher dissociation for later stages. We used a combination of chemical and mechanical methods to achieve the highest dissociation efficiency possible while maximizing cell viability. Details of these protocols will be released with the preprint(s).

Single-embryo single-cell dissociation protocol from embryo collection, dissociation,  sequencing, and analysis

Cluster Annotation

We generated UMAPs for each developmental stage by combining the data of all individually sequenced embryos (four embryo replicates per stage). Per time point, we computed Leiden clusters, which we annotated based on the expression of specific enriched genes followed by a literature search using ZFIN, as well as existing published and annotated scRNAseq data (Farrell et al. 2018; Wagner et al. 2018; Farnsworth et al. 2020; Raj et al. 2020).

For the global UMAP we integrated all the developmental stages together to get a temporally coherent cell-type annotation. Annotations were done using information from previously annotated single UMAPs, cross-validated with a literature search based on enriched genes per cluster groups. We leveraged the Zebrafish Anatomy Ontology (ZFA) to provide the community with cell-type annotations that use the controlled vocabulary provided by the ZFA.

Notes on Annotation

Zebrahub is an ongoing project and we are working on improving the resolution of our annotations over time. We welcome collaboration to improve the quality of the cell-type annotations and/or other metadata variables. Therefore if you detect any ambiguity in the current data objects or want to help us on a specific region or cluster please contact us.

Credits

Conceptualization: Loïc A. Royer, Merlin Lange, Angela Oliveira Pisco, Alejandro Granados

Methodology:  Merlin Lange, Shruthi VijayKumar, Alejandro Granados, Yang-Joon Kim

Cell-type annotation: Merlin Lange, Shruthi VijayKumar, Xiang Zhao, Keir Balla, Adrian Jacobo, Rachel Banks, Tiger Lao, Angela Oliveira Pisco

Data Analysis: Alejandro Granados, Sarah Ancheta, Aaron McGeever, Hirofumi Kobayashi

Data portal: Kyle Awayan, Samuel D’Souza

Experiments: Merlin Lange, Shruthi VijayKumar, Michael Borja

Sequencing: Sheryl Paul, Honey Mekonen, Angela Detweiler, Norma Neff

Data Curation: Merlin Lange, Alejandro Granados, Loïc A. Royer

3D Visualizations: Hirofumi Kobayashi, Kyle Awayan, Samuel D’Souza

Supervision: Loïc A. Royer, Angela Oliveira Pisco, Norma Neff, Sandra Schmid

Organization: Loïc A. Royer, Merlin Lange

Funding: Chan Zuckerberg Biohub

Acknowledgements

Thanks to the CZI Science Tech team for creating CZ CELLxGENE, the tool that makes the visualization of this research possible. For the 3D rotating UMAPs on this website, we use the deck.gl library.

The static renderings of 3D rotating UMAPs were made using the napari viewer. Some of the figures, and all images of zebrafish embryos and larvae on this website, were created with BioRender. Finally, we thank the Chan Zuckerberg Biohub and its donors for funding this work.