Reconstructing how zebrafish embryos develop, cell by cell

Profiling thousands of individual cells enables researchers to create an atlas of embryonic development at unparalleled resolution

Jeffrey Farrell, Schier lab, Harvard University.<p>A mid-gastrulation zebrafish embryo, with two axial mesoderm populations adopting their cell fate identities. (Blue: DNA. Red: Prechordal plate, <i>gsc</i> RNA <i>in situ</i> hybridization. Green: Notochord, <i>ntd5</i> RNA <i>in situ</i> hybridization.) <a href="">View full-size image.</a></p>
Credit: Jeffrey Farrell, Schier lab, Harvard University.

A mid-gastrulation zebrafish embryo, with two axial mesoderm populations adopting their cell fate identities. (Blue: DNA. Red: Prechordal plate, gsc RNA in situ hybridization. Green: Notochord, ntd5 RNA in situ hybridization.) View full-size image.

During embryonic development, a single cell eventually gives rise to all other cell types and tissues in the body. Now, by profiling gene expression in thousands of individual cells during the earliest stages of zebrafish development, a team of researchers from Harvard University and the Broad Institute of MIT and Harvard has provided a staggeringly detailed picture of how cells switch genes on and off as they divide and differentiate into a variety of tissues.

Measuring RNA in individual cells, rather than averaging millions of cells together, enabled the researchers to build a comprehensive gene-expression dataset of 38,731 zebrafish embryo cells. Using a new computational approach to analyze the data, the team mapped the developmental trajectories and relationships of 25 different emerging cell types in the early hours of zebrafish development.

The team describes these results in a recent issue of Science with co-senior authors Aviv Regev, core institute member and director of the Klarman Cell Observatory at Broad Institute and a Howard Hughes Medical Institute Investigator, and Alex Schier, associate member at Broad and professor at Harvard University. The paper is one of a trio focusing on single-cell analysis of vertebrate embryogenesis; the others were published by teams at Harvard Medical School.

“By determining which genes a cell is expressing at a certain point, we developed a way to predict which genes it will express in the future as it adopts a particular identity,” explains Jeff Farrell, co-first author on the study and a postdoctoral fellow in Schier’s lab at Harvard University. “The next step will be to investigate which of those genes actively regulate and drive development and how we might manipulate them.”

To individually profile tens of thousands of cells, the research team used Drop-seq — a technology developed in 2015 by a team based at the Broad Institute and Harvard Medical School, led in part by Regev, Broad associate member David Weitz, and Broad institute member Steve McCarroll.

From nearly 700 zebrafish embryos, Farrell and colleagues collected the 38,731 cells over a 12 hour period and recorded their individual gene expression profiles.

“Even just five years ago, no one would have imagined that you could reconstruct development this way,” says Schier. “We simply didn’t have the technology to study an individual cell. Now, we can do thousands at a time. The sheer scale and depth of information that you can get is astounding.”

To actually reconstruct the cells’ developmental trajectories from this data, Farrell led the development of a computational method capable of assessing which cell profiles were most similar and how they related to each other over time. The method, named URD for a Norse mythology figure, allowed the team to pin down genes that were expressed together at specific developmental stages — and to then trace back distinct trajectories that led to each cell type. (Explore the team’s data through the Single Cell Portal here.)

Surprisingly, the researchers also uncovered patterns in the data that indicated the presence of what they called “intermediate cells,” which expressed genes that appeared to mark them for two different cell fates. The team hypothesizes that these cells initially express markers for one developmental path, but later alter their gene expression and change trajectories, directed to their final type by the environmental cues between two neighboring tissues in an embryo.

In addition to illuminating the developmental processes in typical zebrafish embryos, the team hopes that this dataset will allow researchers to better understand what goes awry in the case of  genetic mutations or disease. “For embryos with genetic mutations, studies of RNA at the single-cell level can offer us a detailed view of the gene activity changes that are leading to their phenotypes,” says Yiqun Wang, co-first author and graduate student in Schier’s lab. “That information could help us develop interventions to bring the altered cell states back to normal.” 

As a proof of concept, the team compared the wild-type developmental tree to data collected from zebrafish embryos with a mutation in the Nodal signaling pathway (a molecular pathway that ensures certain tissues will develop in the right places). In these analyses, the team discovered that while certain cell types and tissues were completely missing from the Nodal mutants, the remaining cells expressed genes similarly to those in the wild-type embryos.

“We’ve been able to create this atlas of embryonic gene expression at an unparalleled resolution,” says Regev. “And our computational approach provides a framework that could be applied to study many developmental systems, without needing any prior knowledge of their gene expression patterns — enabling broad comparative studies of basic development and disease.”

This study was funded by the National Institutes of Health, Allen Discovery Center for Cell Lineage Tracing, Jane Coffin Childs Memorial Fund, Charles A. King Trust, Howard Hughes Medical Institute, and the Klarman Cell Observatory.

Paper(s) cited

Farrell JA and Wang Y et al. Single-cell reconstruction of developmental trajectories during zebrafish embryogenesis. Science. Online April 26, 2018. DOI: 10.1126/science.aar3131