Simultaneous lineage tracing and cell-type identification using CRISPR-Cas9-induced genetic scars

Abstract

A key goal of developmental biology is to understand how a single cell is transformed into a full-grown organism comprising many different cell types. Single-cell RNA-sequencing (scRNA-seq) is commonly used to identify cell types in a tissue or organ. However, organizing the resulting taxonomy of cell types into lineage trees to understand the developmental origin of cells remains challenging. Here we present LINNAEUS (lineage tracing by nuclease-activated editing of ubiquitous sequences)-a strategy for simultaneous lineage tracing and transcriptome profiling in thousands of single cells. By combining scRNA-seq with computational analysis of lineage barcodes, generated by genome editing of transgenic reporter genes, we reconstruct developmental lineage trees in zebrafish larvae, and in heart, liver, pancreas, and telencephalon of adult fish. LINNAEUS provides a systematic approach for tracing the origin of novel cell types, or known cell types under different conditions.

Figure 1. Using the CRISPR/Cas9 system for massively parallel single cell lineage tracing.

(a) Cas9 creates insertions or deletions in an RFP transgene. These genetic scars can be used as lineage barcodes. Using the fish line zebrabow M, which has 16-32 integrations of the RFP transgene, enables us to record complex lineage trees with a single sgRNA. Simultaneous transcriptome profiling by scRNA-seq allows unbiased cell type identification. (b) Sketch of the experimental protocol. Injection of Cas9 and sgRNA for RFP into the zygote marks cells with genetic scars at an early developmental stage. Scars can be read out together with the transcriptome by scRNA-seq at a later stage. (c) Approach for simultaneous detection of scars and transcriptome from single cells. Cells are captured by droplet microfluidics, followed by lysis, reverse transcription, and amplification. After amplification, the material is split and processed into a whole transcriptome library and a targeted RFP library for scar detection. (d) t-SNE representation of scRNA-seq data and identified cell types for dissociated zebrafish larvae (5 dpf, n=7 animals). Cell types were grouped into 8 categories as indicated by the color code. (e) Probability distribution of scars, measured in bulk experiments on the DNA level. Pie chart shows fractions of different types of scars (deletion, insertion, single nucleotide polymorphism (SNP), complex scars). (f) Length distributions for deletions and insertions for the data shown in (e). (g) Scarring dynamics as measured on the DNA and RNA level, with exponential fit.

Figure 2. Computational reconstruction of lineage trees on the single cell level.

(a) In a developmental lineage tree (top), each scar can be identified by a unique number corresponding to its ranking in the bulk scar frequency distribution (Fig. 1e). Newly created scars are indicated in black font. The resulting scar tree (middle), a reduced representation of the order of scarring events, can be represented as a network graph (bottom). In a scar network graph, each node corresponds to a different scar, and pairs of scars that are co-expressed in single cells are connected by gray lines. In LINNAEUS, we experimentally measure scar network graphs, based on which we computationally reconstruct the underlying lineage tree. (b) Cartoon of the computational approach. Network graphs allow reconstructing the order of scar creation events in an iterative approach. The first scar is determined as the one with the highest connectivity (red arrow). Upon removal of the first scar and its connections, the following scars are identified as the most highly connected ones in the reduced network. For details see Online Methods. (c) After the scar tree has been built, we position all individual cells in the tree according to their scar profile. Incomplete scar detection efficiency may lead to loss of information in single cells (black numbers: detected scars; gray crossed out numbers: missed scars). As a consequence, some cells cannot be placed all the way down to the lowest branch of the tree (example: red cell, in which scar 41 and 75 were not detected). However, some missing scars can be reconstructed (example: blue cell, in which scar 41 can be inferred). See also Supplementary Fig. 12. (d) Sketch of a simple single cell lineage tree with two cell types (red, blue). Single cell lineage trees can be represented in a condensed form by indicating fractions of cell types as pie charts (cumulative with respect to the branches below). (e) Lineage tree for one 5 dpf larva. Pie charts are plotted small for n<50, medium for n≥50, and large for n≥1000. Color code for cell types as in Fig. 1d. Scars with creation probability ≥0.001 and scars that were detected in more than 1 larva were excluded from the analysis. In general, developmental lineages separate well in the tree. However, since scarring ends at ~10 hours post fertilization, the end points of the branches may still give rise to multiple cell types in multiple tissues. (f) Lineage tree for one 5 dpf larva, zoomed into lateral plate mesoderm (see color code). The tree structure was determined based on the whole dataset (e).

Figure 3. Single cell lineage analysis of adult organs reveals hierarchies of cell fate decisions.

(a) t-SNE representations of scRNA-seq data for dissociated organs from adult zebrafish (red: heart, green: pancreas + liver, blue: telencephalon; n=3 animals). (b) Lineage tree for organs from one adult. Pie charts are plotted small for n<50, medium for n≥50, and large for n≥1000. Scars with creation probability ≥0.01 were excluded from the analysis. Color code as in (a). (c) Lineage tree zoomed into immune cell types from same adult as (b) (see color code). As expected, immune cells from different organs cluster together in the lineage tree, even though the sequencing libraries for the different organs were prepared separately. This observation is an additional important validation of the scar filtering pipeline, since it shows that even small cell populations such as these immune cells do not acquire scars from other cells types in their organ of origin by mechanisms such as cell doublets or sequencing errors.