Simultaneous single-cell profiling of lineages and cell types in the vertebrate brain by scGESTALT

Bushra Raj, Daniel E. Wagner, Aaron McKenna, Shristi Pandey, Allon M. Klein, Jay Shendure, James A. Gagnon, Alexander F. SchierbioRxiv 2017 · DOI

Abstract

Hundreds of cell types are generated during development, but their lineage relationships are largely elusive. Here we report a technology, scGESTALT, which combines cell type identification by single-cell RNA sequencing with lineage recording by cumulative barcode editing. We sequenced ~60,000 transcriptomes from the juvenile zebrafish brain and identified more than 100 cell types and marker genes. We engineered an inducible system that combines early and late barcode editing and isolated thousands of single-cell transcriptomes and their associated barcodes. The large diversity of edited barcodes and cell types enabled the generation of lineage trees with hundreds of branches. Inspection of lineage trajectories identified restrictions at the level of cell types and brain regions and helped uncover gene expression cascades during differentiation. These results establish scGESTALT as a new and widely applicable tool to simultaneously characterize the molecular identities and lineage histories of thousands of cells during development and disease.

Introduction

Biological development is a complex process where a few progenitor cells give rise to hundreds of specialized cell types. To understand this, scientists need to know two things: what a cell is (its transcriptomic state) and where it came from (its lineage). Previously, these were studied separately. Single-cell RNA sequencing (scRNA-seq) is excellent at identifying cell types but provides only a "snapshot" in time, losing historical context. Conversely, lineage tracing (using dyes or CRISPR "scars" in DNA) shows ancestry but often requires destroying the cell or losing its molecular identity. Furthermore, earlier versions of CRISPR barcoding were limited to very early embryonic stages, making it impossible to track later differentiation events. scGESTALT bridges this gap by recording lineage information directly into the RNA, allowing both identity and history to be read in a single sequencing run.

Methods

The authors developed scGESTALT (single-cell Genome Editing of Synthetic Target Arrays for Lineage Tracing) using several clever engineering steps to capture developmental history at high resolution.

Overview of the scGESTALT workflow for simultaneous lineage recording and cell type identification.
  1. The Barcode: They used a synthetic "barcode" transgene consisting of multiple CRISPR-Cas9 target sites.
  2. Two-Stage Editing: To capture both early and late development, recording is split into two windows. Cas9 and sgRNAs are injected at the one-cell stage (T1) to mark early progenitors. Later, a heat-shock inducible Cas9 system introduces a second round of "scars" at 30 hours post-fertilization (T2), capturing later lineage branching.
  3. RNA-Based Readout: The barcode is placed in the 3’ UTR of a fluorescent protein (DsRed) transgene. Because the barcode is transcribed into mRNA, it can be captured by standard droplet-based scRNA-seq (inDrops).
  4. Computational Reconstruction: A maximum parsimony approach builds lineage trees from the CRISPR "scars," while transcriptomic data is clustered to define cell types.

Results

The study focused on the juvenile zebrafish brain (23-25 days post-fertilization). A comprehensive transcriptomic atlas of ~60,000 single-cell transcriptomes identified 102 transcriptionally distinct cell types, including rare populations like Purkinje neurons and specific perivascular cells.

By focusing on three specific zebrafish (ZF1, ZF2, ZF3), the authors recovered barcodes and transcriptomes from ~3,731 cells to build lineage trees. The system generated highly branched trees; for example, one fish (ZF1) showed 25 major clades that diversified into 193 late-stage nodes. The two-timepoint strategy was highly efficient, with 100% of recovered barcodes being edited and a median of 1,504 distinct barcodes per embryo.

Key insights from the lineage analysis included:

  • Spatial vs. Type Restriction: Most lineage clones were spatially restricted (e.g., staying within the forebrain or hindbrain). However, single progenitors were often "multipotent," giving rise to many different neural cell types within that specific region.
  • Early Divergence of Non-Neural Cells: Non-neural lineages, such as microglia and endothelial cells, diverged from neural lineages very early, during the first round of barcode editing.
  • Tracking Differentiation Cascades: By overlaying lineage data onto "pseudotime" trajectories, the authors identified waves of gene expression as progenitors transitioned into mature neurons, particularly observable in the cerebellum.

Discussion

scGESTALT establishes a powerful framework for simultaneously mapping molecular identities and lineage histories. However, some limitations remain. Currently, the edited barcode is recovered in fewer than 30% of the captured transcriptomes; improving this capture efficiency is vital for building truly comprehensive trees. Additionally, while two timepoints are a major step forward, development is continuous; future iterations aim for recording systems that can add edits continuously over days or weeks. Finally, dissociating the brain into single cells loses precise spatial coordinates, a gap that could be filled by integrating scGESTALT with spatial transcriptomics to map family trees back onto a 3D physical model of the brain.

Summary updated

Abstract

ABSTRACT Hundreds of cell types are generated during development, but their lineage relationships are largely elusive. Here we report a technology, scGESTALT, which combines cell type identification by single-cell RNA sequencing with lineage recording by cumulative barcode editing. We sequenced ~60,000 transcriptomes from the juvenile zebrafish brain and identified more than 100 cell types and marker genes. We engineered an inducible system that combines early and late barcode editing and isolated thousands of single-cell transcriptomes and their associated barcodes. The large diversity of edited barcodes and cell types enabled the generation of lineage trees with hundreds of branches. Inspection of lineage trajectories identified restrictions at the level of cell types and brain regions and helped uncover gene expression cascades during differentiation. These results establish scGESTALT as a new and widely applicable tool to simultaneously characterize the molecular identities and lineage histories of thousands of cells during development and disease.
BibTeX
@misc{raj2017simultaneous,
  author = {Bushra Raj and Daniel E. Wagner and Aaron McKenna and Shristi Pandey and Allon M. Klein and Jay Shendure and James A. Gagnon and Alexander F. Schier},
  title = {{Simultaneous single-cell profiling of lineages and cell types in the vertebrate brain by scGESTALT}},
  year = {2017},
  doi = {10.1101/205534},
  howpublished = {bioRxiv},
}
Figure