Large-scale reconstruction of cell lineages using single-cell readout of transcriptomes and CRISPRCas9 barcodes by scGESTALT

Abstract

Lineage relationships among the large number of heterogeneous cell types generated during development are difficult to reconstruct in a high-throughput manner. We recently established a method, scGESTALT, that combines cumulative editing of a lineage barcode array by CRISPR-Cas9 with large-scale transcriptional profiling using droplet-based single-cell RNA sequencing. The technique generates edits in the barcode array over multiple timepoints using Cas9 and pools of single-guide RNAs introduced during early and late zebrafish embryonic development, which distinguishes it from similar Cas9 lineage tracing methods. The recorded lineages are captured along with thousands of cellular transcriptomes to build lineage trees with hundreds of branches representing relationships among profiled cell types. Here we provide details for (i) generating transgenic zebrafish; (ii) performing multi-timepoint barcode editing; (iii) building single-cell RNA-seq libraries from brain tissue; and (iv) concurrently amplifying lineage barcodes from captured single cells. Generating transgenic lines takes 6 months while performing barcode editing and generating single-cell libraries involve 7 days of hands-on time. scGESTALT provides a scalable platform to map lineage relationships between cell types in any system that permits genome editing during development, regeneration or disease.

EDITORIAL SUMMARY

This protocol describes how to generate transgenic zebrafish expressing a barcode array that can be edited by CRISPR/Cas9 at multiple developmental stages. Single cell RNA sequencing of edited barcodes and cellular transcriptomes allows reconstruction of lineage relationships.

INTRODUCTION

Single-cell technologies have rapidly progressed to enable large-scale profiling of cell-to-cell heterogeneity in tissues during development, homeostasis and disease^1–3. It is now feasible to characterize cell identities based on transcriptional, epigenetic and protein marker landscapes by capturing tens of thousands of single cells or nuclei in a short timeframe and at a low cost. These methods rely on droplet microfluidics, high density arrays or combinatorial indexing for massively parallel experiments^4–17. Among the various single-cell assays, single-cell RNA-sequencing (scRNA-seq) is widely used to categorize cell types using gene expression signatures in many tissues and animals, including humans, without requiring extensive prior knowledge of cell composition. Concomitant with technological advances, substantial efforts have been made to develop computational approaches to correct for batch effects, enable efficient data clustering, and compare datasets between different conditions or methods^18–21. The resultant cell type catalogs are revealing new insights into cellular heterogeneity such as de novo identification of rare cell types and gene markers, delineation of cellular dynamics via developmental trajectories and cell cycle states, and changes in structure of profiled cell types during disease^22–38.

Large-scale reconstruction of lineage relationships at single-cell resolution

Despite the ability to define cell types on a large scale, it has been challenging to determine the lineage relationships between cell types and characterize the progressive steps in their specification^39–41. Reconstruction of lineage trees that capture the patterns and timings of lineage segregation can provide information about the structural and functional organization of tissues and their dysregulation in congenital disease and cancer; this may eventually provide a means for translation into in vitro reprogramming and organoid systems. The realization of such lineage trees would require simultaneous measurements of cell type identities and lineage relationships at high throughput. Lineage tracing methods using barcoded viral libraries, Cre-loxP recombination, somatic mutations and transposon tagging have been linked with next-generation sequencing, but have not been combined with simultaneous cell type identification at single-cell resolution^42–48. Coupling transcriptomic and lineage readouts with scRNA-seq would provide a powerful assay to perform unbiased lineage tree reconstruction.

Lineage recording with CRISPR-Cas9 genome editing

The CRISPR-Cas9 nuclease system has become a widely-used tool in genome engineering with applications extending beyond gene editing to regulation of gene expression, epigenetic editing, base editing, knockout screening, in situ labeling, RNA tracking, protein localization, chromatin interaction profiling, and signal recording, among others^49–65. Recently, CRISPR-Cas9 has been adapted for use in several lineage tracing strategies including GESTALT, LINNAEUS, MEMOIR and ScarTrace^66–71. In general, these methods rely on the introduction of Cas9-induced stochastic mutations (insertions or deletions during DNA repair following double strand breaks) to predefined target sites in transgenes, and use the edited sequences as lineage barcodes for clonal tracing and lineage tree reconstruction.

We have established an approach, scGESTALT⁷², which extends the CRISPR-Cas9 lineage recording capabilities of GESTALT⁶⁶, with large-scale transcriptional profiling using droplet-based scRNA-seq, inDrops⁸, to reconstruct developmental lineage trees using zebrafish as a model system (Fig. 1). Our method makes use of combinatorial and cumulative additions of Cas9-induced mutations to a genomic CRISPR barcode array at multiple timepoints to write permanent records of lineage histories as cells divide⁷². The edited barcodes are expressed from a transgene, captured within cellular transcriptomes and sequenced at single-cell resolution with inDrops, which enables the indexing of >12,000 cells/h into nanoliter-size droplets⁷³. Using tools adapted from phylogenetics, branching lineage trees are generated using patterns of shared mutations between barcodes with the tips of the trees representing profiled cell types defined using associated scRNA-seq gene expression signatures.

Figure 1. Simultaneous recovery of lineages and cell types at single-cell resolution using scGESTALT.

Left panels, During early and late developmental time periods, CRISPR-Cas9 induced indels record cell lineages as mutated genomic barcode sequences. Middle panels, Tissues of interest, such as brain, are dissociated into a single-cell suspension and loaded into a microfluidics device (inDrops). Single cells are encapsulated in droplets and indexed using hydrogels (color-coded to indicate different cell identifier primers) that are coated with oligodT primers. Polyadenylated cellular transcriptomes and scGESTALT lineage barcodes bind to the oligodT sequences and are simultaneously extracted from the same cells. Transcriptome libraries are sequenced to generate gene expression matrices for thousands of single cells. Gene expression profiles are used to perform dimensionality reduction using principal component analysis and visualized in two dimensions on a t-distributed stochastic neighbor embedding (t-SNE) plot. Single cells are represented as grey dots on the shown plot. A modularity-based clustering algorithm (Louvain) is used to cluster cells into discrete cell types using significant principal components. A t-SNE plot of 58,492 cells from n = 22 animals is color-coded to show 63 distinct clusters that were identified from zebrafish juvenile brains⁷². Right panels, scGESTALT libraries are sequenced to obtain lineage barcodes of profiled single cells. The inDrops index sequences are used to match transcriptomes and lineage barcodes for the same cells. Cell lineage trees are generated using maximum parsimony based on patterns of shared edits. Black and red nodes represent early and late barcode edits, respectively. Dashed lines connect profiled single cells to nodes on the tree. Cells connected to the same node are clonal (i.e. contain the same lineage barcode). Each cell is categorized into a discrete cell type (color coded rectangles) based on prior transcriptional clustering analysis. Brown shades represent forebrain cell types, blue shades represent midbrain cell types, green shades represent hindbrain cell types, and pink shades represent progenitor cell types. This procedure was approved by the HU/FAS Committee on the Use of Animals in Research & Teaching under Protocol No. 25–08. Figure adapted from ref⁷².

Here we explain in detail how to perform scGESTALT multi-timepoint barcode editing with transgenic zebrafish. We describe the generation of scRNA-seq transcriptome and lineage barcode libraries using the juvenile brain as an experimental system. Due to the inherent scalability of droplet-based scRNA-seq, scGESTALT can be used to profile thousands of cells from a single animal. Furthermore, scGESTALT can be adapted for use in any organism or in vitro model system that is compatible with genome editing and single-cell technologies to characterize lineages in development, regeneration, cellular reprogramming and disease states.

Comparison with other lineage tracing methods

To generate large-scale lineage trees or perform large-scale clonal analysis, lineage tracing methods should possess high multiplexing capacities for unique and unambiguous labeling of ancestral cells. Approaches that incorporate high-throughput sequencing of diverse DNA or RNA lineage records are advantageous over the limited spectral labeling capacity of fluorescent reporters. For example, virally-delivered DNA barcode libraries can in principle encode high sequence complexity. However, the relative representation of each barcode in the library and its successful recovery can confound data interpretation. If subsets of barcodes are overrepresented in the viral library, it will result in coalescence of independent clones marked by the same barcode into a single clone, and rare barcodes would be underrepresented and potentially missed during data analysis. Inefficiencies in viral transduction of certain cell types can also preclude its use in various tissues or organisms. Furthermore, as the barcodes are not mutable (i.e. cannot be edited during development), only clonal analysis can be performed. In this case, all descendants of an ancestrally marked founder cell can be identified (all cells in a clone inherit the same barcode sequence), but lineage relationships between the clonal descendants cannot be inferred. Another approach known as Polylox barcoding, has been recently described to potentially generate hundreds of thousands in vivo barcode sequences (via Cre-induced recombination of a synthetic DNA cassette inserted into the genome), however its compatibility with single-cell analysis and use for generating lineage trees have not been demonstrated⁴⁴. Somatic mutations accumulated over an animal’s lifespan can be harnessed to generate detailed lineage trees and do not require building transgenics or introducing barcoding reagents⁴³. However, methods for retrieving somatic mutations from single cells involve whole genome amplification and sequencing strategies that can introduce technical artefacts and are currently cost-prohibitive for analyzing large single-cell populations^43,74,75. In contrast, scGESTALT barcodes can be recovered by targeted enrichment from a predefined genetic locus. Coupling scGESTALT to the inDrops scRNA-seq platform also provides several advantages such as high scalability, low library preparation cost (currently ~$0.06 per cell), high single-cell capture rates (>70% of cells), low cell input requirements, no cell capture size bias, and simultaneous transcriptome profiling for cell type or cell state identification⁷³. Recently, inDrops has also been coupled with Tol2 transposon-based barcoding (TracerSeq) for lineage tracing in zebrafish³⁷.

Two other methods have linked CRISPR-Cas9 lineage tracing with scRNA-seq in zebrafish^68,69. These techniques generate barcode edits in a multi-copy transgenic reporter (in tandem or in distinct genomic loci) by injecting Cas9 protein or mRNA along with one sgRNA targeting a reporter-specific sequence. This restricts lineage analysis to early embryonic stages as Cas9 editing terminates upon degradation of Cas9/sgRNA or mutation of all potential sites, and the observed number of distinct lineage barcodes is limited to ~1,000⁶⁸. Multi-copy integrations also require that sequences from all sites are recovered to obtain full lineage information, which can be challenging with sparse scRNA-seq data⁶⁹. The scGESTALT barcode array consists of 9 distinct CRISPR sites in tandem, each of which can be targeted by a different sgRNA. It also enables multi-timepoint editing by targeting subsets of the CRISPR sites during early and late embryogenesis using a combination of injection and transgenic Cas9/sgRNA expression. Such a modular system enables longer periods of lineage recording and increases barcode diversity such that >10,000 barcodes are observed⁷². Furthermore, the use of single-copy barcode integrations facilitates unequivocal profiling of the complete barcode array extracted from each cell without the need for inference of dropout events (i.e. partial loss of lineage information due to distribution of lineage recording sites across multiple loci).

Limitations of scGESTALT

Several considerations have to be made when choosing scGESTALT for lineage recording. First, the edited barcode is recovered from fewer than 30% of profiled single cells, possibly due to low expression level of the barcode, inefficient capture of barcode transcript within droplets, or amplification bottlenecks during sequencing library preparation. If using this method for profiling lineages of rare cell types, an enrichment step (e.g. FACS) for populations of interest may be required to obtain sufficient lineage barcodes. Future optimizations using stronger promoter sequences to drive barcode expression or improved library preparation could address these issues. Alternatively, the barcode could be extracted from genomic DNA rather than the transcriptome⁶⁸. Second, due to the tandem arrangement of CRISPR sites on the barcode array, some lineage records can be erased by large deletions that span multiple sites. Third, barcode diversity is limited to ~10,000 unique labels due to relatively quick saturation of Cas9 editing activities. To follow fate specification of progenitors over longer timeframes (e.g. days or weeks) and generate lineage trees with more branches, the timing and duration of editing will need to be further modulated, and the diversity of barcodes will need to be increased to millions of unique labels. Finally, tissue dissociation results in loss of precise spatial information of profiled cells, thus spatial distributions of clones and distantly related cells cannot be investigated.

Experimental design

The scGESTALT protocol is divided into four experimental sections: (i) generating two transgenic zebrafish lines: one for barcode expression and the other for inducible Cas9 and ubiquitous sgRNA expression; (ii) performing multi-timepoint barcode editing using a combination of injection and induction of barcoding reagents for early and late labeling of cells, respectively; (iii) transcriptome profiling using inDrops; and (iv) concurrently extracting lineage barcodes from captured single cells. Transcriptome and scGESTALT libraries are sequenced separately and gene expression and lineage information for each cell are matched using the inDrops cell identifier index.

Generating transgenics for barcoding.

Two transgenic lines expressing different components of the scGESTALT barcoding strategy are established using the Tol2 transposon system (Fig. 2). In this system, a DNA sequence of interest flanked by minimal cis-sequences from the left and right ends of the medaka Tol2 element is randomly integrated into the genome in the presence of Tol2 transposase by a cut-and-paste mechanism⁷⁶. Thus, a transgenic construct can be cloned in between the Tol2 ends and the resulting plasmid is injected along with mRNA encoding Tol2 transposase into zebrafish embryos. Translated Tol2 protein will mediate excision and stable integration of the transgenic construct into the genome of embryonic cells during early development, including some cells that give rise to germ cells. The injected fish are outcrossed and screened for successful germline transmission of the Tol2 transgenic construct into the F₁ generation to establish transgenic animals. The lineage barcode array construct comprises a heat shock-inducible promoter driving ubiquitous DsRed transgene expression with the scGESTALT CRISPR array inserted in the 3′UTR upstream of the SV40 polyadenylation signal. In addition, the construct contains a heart-specific GFP reporter as a transgenesis marker (cmlc2:GFP). Upon heat shock, the wildtype or edited barcode is expressed as part of the DsRed mRNA and can be extracted from the cellular transcriptome. The barcode transgenic line is screened by qPCR to determine transgene copy number; single-copy integrants are identified and outcrossed to expand and maintain the line. A separate transgenic line is established using a Tol2 construct comprising a heat shock inducible promoter driving Cas9-t2A-GFP and 5 independent U6 promoters driving 5 distinct sgRNAs targeting the barcode array. This line does not need to be surveyed for single-copy integration. The two lines are available upon request from the Schier or Gagnon labs (see REAGENTS). Depending on the user’s goal, it is possible to make changes to both of the above constructs and generate new transgenics. For example, cell- or tissue-specific promoters can be used to drive Cas9 or barcode expression. In addition, modifications can be made to integrate the constructs into other organisms or cell lines.

Figure 2. Strategy for a CRISPR-Cas9 system that enables early and late barcode editing.

Zebrafish with single-copy heat shock promoter-driven scGESTALT barcode (to promote ubiquitous barcode expression at stage of interest) are crossed to zebrafish that express heat shock-inducible Cas9 and U6-driven sgRNAs 5–9. The barcode is cloned downstream of the dsRed coding sequence and upstream of the SV40 polyadenylation sequence (pA). Resulting embryos are injected with Cas9 protein and sgRNAs 1–4 at the one-cell stage (blue bars; early editing). The embryos are screened for GFP positive heart transgenics (cmlc2 promoter drives heart-specific GFP expression) at 30 hpf to identify embryos containing the barcode transgene, and sorted embryos are heat shocked to induce transgenic Cas9 for a second round of editing (orange bars; late editing). The embryos are screened again for ubiquitous GFP expression (Cas9 is linked to GFP with a t2A self-cleaving peptide), which indicates successful Cas9 transgene induction. Double transgenic embryos are grown for downstream profiling, and heat shocked at time of interest (e.g. juvenile stage 23–25 dpf) to induce expression of the edited barcode array prior to scRNA-seq analysis. Protocol steps for each stage are indicated. Adapted with permission from ref⁷².

Early and late barcoding.

To enable recording of lineages during early and late development, the scGESTALT barcode transgenic is crossed to the Cas9/sgRNA transgenic. Single-cell embryos are injected with Cas9 protein and a pool of sgRNAs which target the first 4 sites of the CRISPR array. This strategy initiates an ‘early’ round of Cas9 editing activity that introduces mutations at target sites 1–4 of the barcode. Upon zygotic genome activation, sgRNAs targeting sites 5–9 of the array are expressed from U6 promoters. Embryos are then heat shocked at a time of choosing (e.g. 30 hours post-fertilization, hpf) to induce ubiquitous expression of transgenic Cas9, and initiate barcode editing at target sites 5–9 of the array. The sequential activation of this ‘early and late’ editing strategy enables longer lineage recording, higher barcode diversity, and encodes the relative order of mutation patterns in the two halves of the array, thereby facilitating lineage tree reconstruction⁷². Edited double transgenic embryos are selected by sequentially screening for strong heart-specific GFP expression to identify the barcode transgene, and ubiquitous GFP expression to identify the Cas9/sgRNA transgene, and are grown until the user is ready to perform scRNA-seq (Fig. 2). Due to the modular nature of the barcoding strategy, if the user is interested in tracing early lineage segregations, it is possible to barcode cells during early embryogenesis only by injecting Cas9 protein or mRNA together with all 9 sgRNAs targeting the CRISPR array. The sequences for the 9 sgRNA target sites are provided in Supplementary Table 1. Generally, the spectrum of barcode edits is dominated by deletions, which can be categorized into intra-site (small deletion restricted to one target site) and inter-site (larger deletions that span two or more sites). Insertions are observed at a low frequency. The combination of small indels and larger deletions in early and/or late editing generates high overall diversity of lineage barcodes (<0.5% of lineage barcodes were found to overlap between n=8 different embryos⁷²). We direct the reader to ref⁷² for further information about the nature and frequency of barcode repair products.

Single-cell transcriptome profiling.

scRNA-seq is performed to simultaneously extract edited lineage barcodes and cellular transcriptomes. For droplet-based scRNA-seq, samples are first dissociated into single-cell suspensions prior to loading on the device. We processed juvenile zebrafish brains to provide a snapshot of the heterogeneity of cell types and lineages in the vertebrate brain⁷², and here provide details for brain dissociation and scRNA-seq using inDrops. However, the user can perform similar analyses using any tissue or developmental stage of interest for which a single-cell dissociation protocol that yields high cell viability is available. The inDrops protocol has been described in detail previously⁷³. Briefly, it involves (i) obtaining a compatible microfluidic chip, (ii) synthesizing indexed hydrogel beads, (iii) encapsulating and indexing single cells in nanoliter-size droplets, and (iv) building transcriptome libraries for next-generation sequencing. As single cells are lysed during inDrops, polyadenylated cellular mRNA and edited scGESTALT lineage barcode mRNA are hybridized to oligodT primers on hydrogel beads (Fig. 3a). Following reverse transcription of hybridized mRNA, the cDNA is indexed with an inDrops cell identifier, subjected to second strand synthesis and linearly amplified by in vitro transcription (Fig. 3b). A fraction of the resulting amplified RNA is chemically fragmented for standard transcriptome library preparation, and converted to a DNA library by another round of reverse transcription followed by addition of Illumina-compatible sequencing adapters by PCR. The remaining unfragmented RNA is used for building scGESTALT libraries. scGESTALT can also be adapted for scRNA-seq with other droplet-based (e.g Drop-seq⁹, 10× Genomics¹⁰) or plate/array-based methods. However, amplification and library construction of lineage barcodes as detailed below will have to be empirically determined by the user.

Figure 3. Transcriptome and scGESTALT library preparation overview.

a. Single cells are encapsulated and indexed in droplets using the inDrops platform (left, Step 82). Upon lysis (cells shown with dotted lines), polyadenylated cellular transcripts and scGESTALT lineage barcodes hybridize to oligodT primers on hydrogels. Hydrogels are color-coded to indicate distinct indexing identifiers. Adapted with permission from ref⁷².

b. Overview of transcriptome library preparation steps from single cells (Steps 82–84). Hydrogels are coated with oligonucleotides containing T7 promoter, an adaptor sequence (PE1), a cell indexing identifier (Cell ID), unique molecular identifier (UMI), and oligodT sequence. Polyadenylated cellular mRNA hybridizes to the oligodT sequence and is reverse transcribed into cellular cDNA sequence. Second strand synthesis is carried out to generate double stranded DNA (dsDNA) that is used for T7 in vitro transcription and generates linearly amplified RNA (aRNA). The aRNA is chemically fragmented and then subjected to a second round of reverse transcription using an oligonucleotide that contains a random hexamer and an adaptor sequence (PE2). The resulting single stranded DNA (ssDNA) is PCR amplified with Illumina P5 and P7 adaptor sequences containing overlap with PE2 and PE1 sequences, and a limited number of cycles to generate the final sequencing-ready transcriptome library. ssDNA fragments that do not contain the PE1 adaptor sequence will not be amplified.

c. Overview of scGESTALT lineage barcode library preparation steps from single cells. Polyadenylated lineage barcode mRNA, consisting of dsRed and edited scGESTALT CRISPR array (see Fig. 2, Step 33), are expressed from a transgene upon heat shock of the animal prior to brain dissection and dissociation. The lineage barcode mRNA hybridizes to oligodT sequence on the hydrogel and is reverse transcribed into lineage barcode cDNA. Second strand synthesis and T7 in vitro transcription is carried out similar to the transcriptome library preparation (Step 82). However, the aRNA is not fragmented in order to preserve the full lineage barcode sequence (Step 83, part of the of reverse transcribed product is not fragmented and stored at −80 °C). Instead reverse transcription is carried out using random hexamers to generate full-length ssDNA (Steps 85–89). The edited scGESTALT CRISPR array sequence is selectively enriched from the full-length ssDNA using a nested PCR approach. The first round amplifies a longer piece of dsDNA using a primer that hybridizes in the dsRed sequence (GP6) and one that overlaps the PE1 adaptor sequence (PE1sa) (Steps 90–92). The resulting dsDNA is used in a second round to target the edited lineage barcode sequence using a primer that binds upstream proximal to the start of the CRISPR lineage array sequence (GP12) and a primer that overlaps the adaptor sequence (PE1sb) (Steps 93–95). The resulting dsDNA is PCR amplified (Steps 97–98) similar to the transcriptome library preparation to generate the final sequencing-ready scGESTALT lineage barcode library (Step 99).

Targeted amplification of lineage barcodes.

After in vitro transcription of indexed cDNA, a fraction of the full-length amplified RNA is reverse transcribed using random hexamers. scGESTALT lineage barcodes are amplified from the resulting cDNA using a nested PCR strategy (Fig. 3c). First, a larger fragment is enriched using primers that flank the DsRed transgene and a universal inDrops adaptor sequence located downstream of the cell identifier and UMI (unique molecular identifier, for counting number of unique transcripts of a gene) sequences. Following purification, the PCR product is used to amplify a shorter lineage barcode fragment using primers that flank the barcode array and a universal inDrops adaptor sequence. Finally, Illumina-compatible DNA libraries are generated following the same strategy for transcriptome libraries. If mixing multiple libraries in a sequencing run, it is critical that the same multiplexing primers are used for both transcriptome and lineage barcode libraries to ensure that these can later be matched to each other. In this protocol, we provide details for preparing “V3” inDrops transcriptome and scGESTALT libraries, which are compatible with standard Illumina sequencing primers. If using “V2” inDrops libraries, they will need to be sequenced with custom read and index primers as described previously⁷³.

Experimental controls.

A negative control for editing involves no injection of Cas9 protein and sgRNAs 1–4 followed by no heat shock Cas9 induction. In this case, no substantial editing of the barcode array at sites 1–9 should be observed by PCR amplification from genomic DNA or using next generation sequencing from genomic DNA and/or scRNA-seq libraries. Similarly, injection of Cas9 protein alone (no sgRNAs 1–4) followed by no heat shock Cas9 induction, should also result in little to no editing at sites 1–9. In some cases, low levels of editing may be observed at sites 5–9 due to cross-activity between injected Cas9 protein and U6-expressed sgRNAs 5–9. When no Cas9 protein and sgRNAs 1–4 are injected, but Cas9 is induced at 30 hpf by heat shock, editing is expected to be largely confined to sites 5–9. In contrast, performing two-timepoint editing by Cas9 protein and sgRNAs 1–4 injection followed by heat shock Cas9 induction, should result in edits observed throughout sites 1–9 of the barcode array. …

Data Analysis.

The data analysis consists of three major parts: (i) analysis of the transcriptome data for cell type identification, (ii) analysis of the scGESTALT lineage barcodes for identification of editing events, and (iii) generation of a single-cell resolved lineage tree with associated cell types. The inDrops.py pipeline (see Software in REAGENTS below) will generate a gene expression matrix that can be used as input for identification of cell types and gene markers using an analysis tool designed for processing sparse scRNA-seq data (e.g. Seurat^18,77). The inDrops.py pipeline can also be used to obtain sequence reads for scGESTALT lineage barcodes that are sorted by the inDrops cell identifier index (i.e. obtain lineage reads for single cells). Next, the sorted scGESTALT lineage barcode reads are further formatted using the scGestaltPrepFunc.R script (Supplementary Software 2) and processed with the scGESTALT analysis pipeline (see Software in REAGENTS below, also provided as Supplementary Software 1). This will generate a table containing the lineage barcode sequence for each cell as well as its inDrops cell identifier index sequence. Next, for each cell from which a lineage barcode was successfully recovered and processed, the matching transcriptome and hence cell identity is assigned using the Transcriptome-scGESTALTMatchPipe.R and MatchPipeFunc.R scripts (Supplementary Software 3–4). At this stage, a table will be generated (GestMaster.txt, see TestData in Supplementary Software 1) containing the scGESTALT lineage barcode, cell cluster identity (obtained using Seurat for cell type classification) and the inDrops cell identifier. This can be used as input for drawing the overall multi-timepoint edited lineage tree using the scGESTALT analysis pipeline.

MATERIALS

REAGENTS

Nuclease-free water (Life Technologies, cat. no. M5310–1L)

NotI-HF (NEB, cat. no. R3189)

E.Z.N.A. Cycle Pure Kit (Omega, cat. no. D6492–02) or equivalent

E.Z.N.A. Plasmid DNA Mini Kit I (Omega, cat. no. D6942–02) or equivalent

E.Z.N.A. Total RNA Kit I (Omega, cat. no. R6834–02) or equivalent

RNA Clean & Concetrator-25 (Zymo Research, cat. no. R1018)

mMESSAGE mMACHINE SP6 (Thermo Fisher Scientific, AM1340)

NaOH, 1M (Honeywell Research Chemicals, cat. no. 319511–1L)

Tris-HCl pH 7.5, 1M (VWR, cat. no. 75800–958)

Tris–HCl pH 7.0, 1 M (Thermo Fisher Scientific, cat. no. AM9851)

Tris–HCl pH 8.0, (Thermo Fisher Scientific, cat. no. 15568025)

EnGen sgRNA Synthesis Kit, S. pyogenes (NEB, cat. no. E3322)

CRITICAL sgRNA oligonucleotide sequences (sgRNA 1–4) provided (see DNA oligonucleotide sequences below) are designed for use with this kit

EnGen Cas9 NLS, S. pyogenes (NEB, cat. no. M0646)

iTaq Universal SYBR Green Supermix (Biorad, cat. no. 1725120)

Phusion High-Fidelity DNA polymerase (NEB, cat. no. M0530)

CRITICAL Amplification of the scGESTALT library has been optimized using this enzyme Q5 High-Fidelity DNA polymerase (NEB, cat. no. M0491)

CRITICAL Amplification of the scGESTALT library has been optimized using this enzyme KAPA 2× HiFi HotStart PCR mix (KAPABiosystems, cat. no. KK2601)

dNTP mix, 10 mM each (NEB, cat. no. N0447)

100 bp DNA ladder (NEB, cat. no. N3231)

Papain Dissociation System (Worthington, cat. no. LK003150)

CRITICAL Brain dissociation has been optimized using this kit Neurobasal media (Thermo Fisher Scientific, cat. no. 21103049)

CRITICAL Brain dissection and dissociation have been optimized using this media B27 supplement, serum free, 50× (Thermo Fisher Scientific, cat. no. 17504044)

CRITICAL Brain dissection has been optimized using this supplement for high cell viability DPBS, no calcium, no magnesium (Thermo Fisher Scientific, cat. no. 14190144) DBPS, calcium, magnesium, glucose, pyruvate (Thermo Fisher Scientific, cat no. 14287080)

35 μm cell strainer (e.g. VWR, cat. no. 352235)

20 μm cell strainer (e.g. Sysmex, cat. no. 04–004-2325)

Trypan blue, 0.4% (wt/vol) (Thermo Fisher Scientific, cat. no. T10282)

OptiPrep Density Gradient Medium (Sigma-Aldrich, cat. no. D1556–250ML)

Agencourt AMPure XP magnetic beads (Beckman Coulter, cat. no. A63881)

Random Hexamers, 50 μM (Fisher Scientific, cat. no. N8080127)

PrimeScript Reverse Transcriptase (Takara Clonetec, cat. no. 2680A)

RNaseOUT Recombinant Ribonuclease Inhibitor (Fisher Scientific, cat. no. 10777019)

Phenol red, 0.5% (wt/vol) (Sigma-Aldrich, cat. no. P0290)

Methylene blue, 0.1% (wt/vol) (Aqua Solutions, cat. no. 5761–500ML)

Instant Ocean salt (Instant Ocean)

Agarose (e.g. National Diagnostics, cat. no. EC-202)

Sylgard 184 (Electron Microscopy Sciences, cat. no. 24236–10)

Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, cat. no. Q33230)

Tricaine (Sigma-Aldrich, cat. no. E10521–50G)

CAUTION: Use gloves when handling (can cause skin irritation).

MiSeq v2 300 cycle kit (Illumina, cat no. MS-102–2002)

NextSeq 500/550 High Output v2 75 cycle kit (Illumina, cat. no. FC-404–2005)

PhiX Control v3 (Illumina, FC-110–3001)

Zebrafish (Danio rerio)

wildtype strains (e.g. TL/AB, TL or AB; available from ZIRC, http://zebrafish.org),

- Tg(hsp70:DsRed-barcodev7; myl7:eGFP) strain (allele number a167; scGESTALT barcode transgenic; available by request from Schier or Gagnon labs),

- Tg(hsp70:Cas9-t2a-GFP; 5×U6:sgRNA) strain (allele number a168; inducible Cas9 and U6-driven sgRNA transgenic; available by request from Schier or Gagnon labs)

CAUTION This procedure was approved by the Harvard University/Faculty of Arts & Sciences Standing (HU/FAS) Committee on the Use of Animals in Research & Teaching under Protocol No. 25–08. The HU/FAS animal care and use program maintains full AAALAC accreditation, is assured with OLAW (A3593–01), and is currently registered with the USDA.

Plasmids

pCS2-zT2TP (available from Koichi Kawakami lab⁷⁸)

pTol2-hspDRv7_scGstlt (Addgene, plasmid ID 108870)

pTol2-hsp70l:Cas9-t2A-GFP, 5×U6:sgRNA (Addgene, plasmid ID 108871)

DNA oligonucleotide sequences (purified by standard desalting, can be ordered from IDT)

Name	Sequence	Purpose	Step
qPCRctrl F	5′-TCAGTCAACCATTCAGTGGCCCAT-3′	PCR amplify ultraconserved genomic region	23
qPCRctrl R	5′-CAGGAAAGGGAATGCAGGGTTTGT-3′	PCR amplify ultraconserved genomic region	23
qPCRdsRed F	5′-GAGCGCGTGATGAACTTCGAGG-3′	PCR amplify dsRed transgenic region	23
qPCRdsRed R	5′-CAGCCCATAGTCTTCTTCTGCATTACG-3′	PCR amplify dsRed transgenic region	23
sgRNAl	5′TTCTAATACGACTCACTATAGACAGCAGTATCATCGACTAGTTTTAGAGCTAGA-3′	Synthesize sgRNA1 for CRISPR-Cas9 lineage barcoding by injection	27
sgRNA2	5′-TTCTAATACGACTCACTATAGAGAGCGCGCTCGTCGACTAGTTTTAGAGCTAGA-3′	Synthesize sgRNA2 for CRISPR-Cas9 lineage barcoding by injection	27
sgRNA3	5′-TTCTAATACGACTCACTATAGTCAGCA GTACTACTGACGAGTTTTAGAGCTAGA-3′	Synthesize sgRNA3 for CRISPR-Cas9 lineage barcoding by injection	27
sgRNA4	5′-TTCTAATACGACTCACTATAGACAGCAGTGTGTGAGTCTAGTTTTAGAGCTAGA-3′	Synthesize sgRNA4 for CRISPR-Cas9 lineage barcoding by injection	27
scGESTALT F	5′-TCGAGCTCAAGCTTCGG-3′	PCR amplify scGESTALT fragment for assessing editing efficiency	43
scGESTALT R	5′-CTGCCATTTGTCTCGAGGTC-3′	PCR amplify scGESTALT fragment for assessing editing efficiency	43
inDrops_GP6	5′-GAGGACTACACCATCGTGGAG-3′	PCR amplify long scGESTALT fragment from inDrops library	90
inDrops_PE1Sa	5′-CTCTTTCCCTACACGACGCTGGGTGTCGGGTGCAG-3′	PCR amplify long scGESTALT fragment from inDrops library	90
inDrops_GP12	5′-TCGTCGGCAGCGTCAGATGTGTATAA GAGACAG NNNNNNNNNTCGAGCTCAA GCTTCGGAC-3′, where N stands for a random base	PCR amplify short scGESTALT fragment from inDrops library	93
inDrops_PE1Sb	5′-CTCTTTCCCTACACGACGCT-3′	PCR amplify short scGESTALT fragment from inDrops library	93
R1-PCRix	5′-AATGATACGGCGACCACCGAGATCTACACxrefTCGTCGGCAGCGTC-3′, where xref is an index sequence for multiplexing (see Supplementary Table 2)	PCR amplify transcriptome or scGESTALT final sequencing library	97
R2-PCR	5′-CAAGCAGAAGACGGCATACGAGATGGGTGTCGGGTGCAG-3′	PCR amplify transcriptome or scGESTALT final sequencing library	97

EQUIPMENT

1.5 ml microcentrifuge tubes (e.g. USA Scientific, cat. no. 1415–2600)

1.5 ml microcentrifuge tubes (e.g. USA Scientific, cat. no. 1415–2500)

0.2 ml PCR tubes (e.g. VWR, cat. no. 53509–304)

50 ml tubes (e.g. VWR, cat. no. 352070)

15 ml tubes (e.g. VWR, cat. no. 352096)

5 ml sterile serological pipettes (e.g. VWR, cat. no. 89130–896)

NanoDrop (Thermo Scientific)

PCR and qPCR machines (e.g. Bio-Rad)

DNA electrophoresis system (e.g. Bio-Rad)

Benchtop centrifuge for 1.5 ml tubes (e.g. Eppendorf, cat. no. 54124)

Refrigerated centrifuge for 1.5 ml tubes (e.g. Eppendorf, cat. no. 5424 R)

Minifuge (e.g. Benchmark, cat. no C1008-C)

Water bath (e.g. VWR)

Heat block for 1.5 ml tubes (e.g. VWR)

Vortex (e.g. Scientific Industries, cat. no. SI-0236)

Magnetic rack for 0.2 ml PCR strip (e.g. Permagen, cat. no. MSR1224)

2100 Electrophoresis Bioanalyzer Instrument (Agilent, cat. no. G2939AA)

Laser-based micropipette puller (e.g. Sutter Instruments, cat. no. P-2000)

Microinjection glass capillaries (World Precision Instruments, cat. no. TW100F-4)

Fine forceps (Fine Science Tools, cat. no. 11254–20)

Fine scissors (Fine Science Tools, cat. no. 14040–10)

Microloader tips for glass capillaries (e.g. Eppendorf, cat. no. 930001007)

Injection molds (e.g. Adaptive Science Tools, model no. TU-1)

Pneumatic PicoPump (World Precision Instruments, model no. PV820) or equivalent

Piconozzle Kit v2 (World Precision Instruments, cat. no. 5430-ALL) or equivalent

Manipulator, magnetic stand and iron plate (Narishige, model. no. M-152, GJ-1, IP) or equivalent

Treadlite II Foot Switch (Lineamaster, model no. T-91-S) or equivalent

Micrometer for calibration (AmScope, cat. no MR100) or equivalent

Dissection microscope with light source (e.g. Stemi 2000 stereomicroscope, Zeiss)

Fluorescent microscope with GFP filter (e.g. Zeiss)

10-cm Petri dishes (e.g. VWR, cat. no. 25384–342)

Insect pins (Fine Science Tools, cat. no. 26002–15)

Breeding tanks and 0.5 L, 1 L and 2 L tanks (e.g. Techniplast)

95%O₂: 5%CO₂ gas tank (Airgas, X02OX95C2003102)

Gas regulator (e.g. VWR, 55850–392)

Tygon E-3603 tubing for bubbling gas (e.g. Component Supply)

Hemocytometer (e.g., Bulldog Bio, cat. no. DHCN420)

Qubit (Thermo Fisher Scientific, cat. no. Q33226)

Qubit Assay Tubes (Thermo Fisher Scientific, cat. no. Q32856)

REAGENT SETUP

Zebrafish

Adult fish are maintained at a density of 8–10 fish per 2 L tank with a 14-h light:10-h dark cycle at 28.5 °C. Wildtype TL/AB strain is bred in house by crossing wildtype TL strain with wildtype AB strain and is used in all experiments described in this protocol. Animals used for breeding are typically 3 to 18 months old.

CAUTION This procedure was approved by the HU/FAS Committee on the Use of Animals in Research & Teaching under Protocol No. 25–08.

Phenol red

Dilute 0.5% (wt/vol) phenol red stock 1:5 with nuclease-free water. Use this diluted stock for injections, and store at room temperature (23 °C – 26 °C) for at least one year.

Blue water

Combine 20 L of water in a carboy with 20 ml of 0.1% (wt/vol) Methylene blue and 5gof Instant Ocean sea salt, and adjust pH to 7.0 with sodium bicarbonate buffer. Store at room temperature for up to 1 month.

50 mM NaOH

Combine 19 ml of nuclease-free water with 1 ml of 1M NaOH. Store at room temperature for up to 3 months.

MESAB, 25×

Combine 489.5 ml of water with 2g tricaine powder and 10.5 ml of 1M Tris (pH 9.0), and adjust pH to 7. Make 10 ml aliquots and store at −20 °C for up to 1 year.

B27 supplement

Upon receiving the product, thaw the solution completely at room temperature. Make 500 μl aliquots and store at −20 °C for up to 1 year.

RNA elution buffer

Combine 9.9 ml of nuclease-free water, 100 μl of 1 M Tris–HCl (pH 7.0), and 2 μl of 0.5 M EDTA (pH 8.0). Store 1 ml aliquots at −20 °C for up to 1 year.

DNA elution buffer

Combine 9.9 ml of nuclease-free water, 100 μl of 1 M Tris–HCl (pH 8.0), and 2 μl of 0.5 M EDTA (pH 8.0). Store 1 ml aliquots at −20 °C for up to 1 year.

R1-PCRix/R2-PCR primer mix

For each variant of R1-PCRix, combine 25 μl of 10 μM R1-PCRix primer and 25 μl of 10 μM R2-PCR primer in DNA elution buffer. Store primer mixes at −20 °C for up to 1 year. See Supplementary Table 1 for a list of 24 R1-PCRix primer sequences.

EQUIPMENT SETUP

Preparation of injection needles

Pull glass capillaries for injection as described previously⁷⁹.

Embryo microinjection dish

Boil 3 g of agarose in 100 ml blue water in a microwave until the agarose dissolves. Let it cool slightly on bench, and pour 25–30 ml per 10-cm Petri dish. Gently layer an injection mold (pre-wetted with water to avoid introducing bubbles) on top of the agarose. Let the agarose solidify and remove the injection mold. The microinjection dish can be stored at 4 °C for two weeks.

Microinjection setup

Configure and connect the Pneumatic PicoPump, piconozzle, manipulator, magnetic stand, iron plate and foot switch according to the manufacturer’s instructions. Connect the Pneumatic PicoPump to a nitrogen tank according to the manufacturer’s instructions.

Preparation of 95%O₂: 5%CO₂ gas tank

Fit the gas tank with the gas regulator and Tygon tubing according to the manufacturer’s instructions.

Preparation of sylgard dissection dishes

Mix 45 ml of Sylgard elastomer base with 4.5 ml of Sylgard elastomer curing agent (10:1 ratio; both reagents provided in Sylgard 184) in a 50 ml tube. Mix thoroughly by inverting several times and transferring back and forth between two 50 ml tubes. Pour ~10 ml of the mix per 10-cm Petri dish. Gently tap the dish to force air bubbles to rise to the top. Allow to cure and set overnight. The dishes can be stored at room temperature for at least one year.

inDrops microfluidics platform setup

Detailed instructions for setting up the inDrops platform have been described previously⁷³. Note that the system should be primed and ready for cell barcoding prior to starting single-cell preparation (Step 48). Alternatively, if working in a pair, one person should get the device ready while the other is preparing cells for encapsulation.

Software

inDrops pipeline inDrops.py: https://github.com/indrops/indrops

scGESTALT pipeline: https://github.com/aaronmck/SC_GESTALT

R: https://www.r-project.org

R studio: https://www.rstudio.com

Seurat: https://satijalab.org/seurat/

PROCEDURE

Tol2 transposase mRNA synthesis TIMING 1 d

• 1Linearize 5 μg of pCS2-zT2TP plasmid using NotI enzyme according to the manufacturer’s instructions.
• 2Purify the linearized plasmid with a mini quick-spin DNA column (e.g. E.Z.N.A. Cycle Pure kit) according to the manufacturer’s protocol and elute in 50 μl nuclease-free water.
• 3
  In vitro transcribe Tol2 mRNA using mMESSAGE mMACHINE SP6 kit. Assemble the reaction below (20 μl total) and follow the manufacturer’s protocol.

  Component	Amount (μl)	Final Concentration
  2× NTP/CAP	10	1×
  10× buffer	2	1×
  Linear, purified pCS2-zT2TP	6	~20–30 ng/μl
  Enzyme mix	2
  Total	20

• 4Purify Tol2 mRNA using a mini quick-spin RNA column (e.g. E.Z.N.A. Total RNA kit I) according to the manufacturer’s protocol and elute in 50 μl nuclease-free water. Measure mRNA concentration using NanoDrop. Successful in vitro transcription reactions will yield mRNA at concentrations of 300–500 ng/μl. Dilute Tol2 mRNA to 100 ng/μl using nuclease-free water. PAUSE POINT Store the mRNA in 3 μl aliquots at −80 °C for at least one year.
• 5Purify DNA from minipreps of pTol2-hspDRv7_scGstlt and pTol2-hsp70l:Cas9-t2A-GFP, 5×U6:sgRNA plasmids using a plasmid purification kit (e.g. E.Z.N.A. Plasmid DNA Mini Kit I). To remove residual RNase A, purify ~2 μg of miniprep DNA with a mini quick-spin DNA column (e.g. E.Z.N.A. Cycle Pure kit) according to the manufacturer’s protocol and elute in 30 μl nuclease-free water. Measure DNA concentrations using NanoDrop. Dilute each of the plasmids to 25 ng/μl using nuclease-free water.

  PAUSE POINT Store diluted plasmids at −20 °C for at least one year.

Zebrafish transgene microinjection TIMING 3 h

• 6The day prior to performing microinjections, set up 6 to 10 overnight tanks of wild type zebrafish for mating as described previously⁷⁸.

  CAUTION All vertebrate animal work must be performed in accordance with relevant guidelines and regulations.

• 7
  On the morning of microinjections, assemble reagent mixes (3 μl total) as follows for each plasmid. Vortex to mix, centrifuge briefly, and hold on ice.

  Component	Amount (μl)	Final Concentration
  pTol2-hspDRv7_scGstlt OR pTol2-hsp70l:Cas9-t2A-GFP, 5×U6:sgRNA plasmid (25 ng/μl)	1.5	12.5 ng/μl
  Tol2 mRNA (100 ng/μl)	1.0	33.3 ng/μl
  Phenol red (0.1% wt/vol)	0.5	0.017%
  Total	3

• 8Stagger mating of fish to enable several rounds of injections to be performed at the one-cell stage (~20–30 min after fertilization). For example, mating can be initiated in two tanks at once, followed by ~35 min-spaced intervals between each subsequent mating pairs.
• 9While fish are mating, prepare an injection needle by breaking the tip of a pulled glass capillary with fine forceps. Load 1.5 μl of the pTol2-hspDRv7_scGstlt injection mix (Step 7) into the needle with a microloader tip and attach it to the piconozzle and manipulator. Set the pressure to 20 psi and calibrate the needle with a micrometer as described previously⁷⁹.
• 10Collect fertilized eggs into Petri dishes with blue water. Transfer ~100 one-cell stage embryos to the agarose ramp in the injection dish (distribute the embryos across 4 rows of the ramp), and add enough blue water to just cover the embryos. Pierce the chorion with the calibrated needle, move the needle into the cytoplasm, and inject 1 nl of the reagent mix. Repeat until all embryos are injected. Transfer the injected embryos with blue water into a new Petri dish and incubate at 28.5 °C. Keep some uninjected control embryos to assess the health of the clutch.
• 11Repeat steps 8–10 for the pTol2-hsp70l:Cas9-t2A-GFP, 5×U6:sgRNA injection mix (Step 7).
• 12At the end of the day, remove any unfertilized (embryos will appear as if still at single-cell stage rather than having undergone multiple rounds of cell division) or damaged (embryos are squashed or have become deformed from the injection procedure) embryos from the injected batches.

Sorting and rearing transgenic fish TIMING 1 d sort, 2–3 months rear

• 13Heat shock fish injected with pTol2-hsp70l:Cas9-t2A-GFP, 5×U6:sgRNA at 24 hpf. Transfer 20–30 healthy looking embryos into a 1.5 ml microcentrifuge tube and add 500 μl of blue water. Use as many tubes as required depending on how many embryos survived injection (20–30 embryos per tube). Incubate the tubes in a heat block set at 37 °C for 30 min. Transfer heat shocked embryos back into Petri dishes with blue water and incubate at 28.5 °C. Alternatively, one can perform the heat shock in 50 ml tubes (with 5 ml blue water) in a water bath at 37 °C for 30 min.
• 14GFP expression starts to become apparent ~2 h after heat shock and becomes bright enough for easy screening and sorting ~4 h after heat shock. Place the petri dish from Step 13 on the stage of a fluorescent microscope fitted with a GFP filter. Swirl the petri dish to bring the embryos to the center and turn on the fluorescent light. Look through the eyepiece to identify fish that exhibit mosaic GFP fluorescence throughout the body. Using a plastic or glass transfer pipette, transfer the brightest GFP positive fish into a new Petri dish containing 20 ml blue water. Incubate at 28.5 °C until 5 dpf and then transfer into 2 L tanks (8–10 larvae per tank) in a zebrafish facility to raise to adulthood. CRITICAL STEP: Seeding tanks at a low density will enable fish to grow and reach sexual maturity faster.
• 15Screen and sort fish injected with pTol2-hspDRv7_scGstlt at 30 hpf. Check for bright GFP fluorescence in the heart using a fluorescent microscope. Using a plastic or glass transfer pipette, transfer GFP positive fish into a new Petri dish containing 20 ml blue water. Incubate at 28.5 °C until 5 dpf and then transfer into tanks (8–10 larvae per tank) in a zebrafish facility to raise to adulthood.

Screen for germline transmission TIMING 2 d, then 2–3 months to rear

• 16Identify adult founder fish with germline integration by outcrossing individual injected F₀ animals to wildtype fish of the opposite sex. Collect embryos the next morning in blue water and incubate at 28.5 °C. Keep potential founders separate until screening of clutches is complete.
• 17Screen clutches from pTol2-hsp70l:Cas9-t2A-GFP, 5×U6:sgRNA F₀ fish by heat shocking embryos at 24 hpf as described in Steps 13–14. The screened F₁ are henceforth referred to as “hsp:Cas9, U6:sgRNA” transgenic line.
• 18Screen clutches from pTol2-hspDRv7_scGstlt F₀ fish by GFP heart expression at 30 hpf as described in Step 15. The screened F₁ are henceforth referred to as “scGESTALT” transgenic line.
• 19Incubate sorted F₁ embryos at 28.5 °C until 5 dpf and then transfer into 2 L tanks (8–10 larvae per tank) in a zebrafish facility to raise to adulthood.

  PAUSE POINT Maintain stable lines by outcrossing. Adults can be maintained for up to 2 years.

Determine copy number of scGESTALT barcode TIMING 1 d

• 20Fin clip adult F₁ pTol2-hspDRv7_scGstlt fish. Anesthetize fish in batches of 3–4 with 100 ml of 1× MESAB and snip a small section of the tail fin using fine scissors. Transfer the tail fin into a PCR strip tube and hold on ice. Transfer each fin clipped fish to individual 0.5 L tanks to recover and number them uniquely. Repeat until all F₁s have been fin clipped.
• 21Prepare genomic DNA using a modified HotSHOT method⁸⁰. Add 50 μl of 50 mM NaOH to each fin clip and incubate at 95 °C for 20 min in a thermocycler.
• 22Add 5 μl of 1M Tris-HCl pH 7.5 and vortex to mix. Centrifuge briefly. PAUSE POINT genomic DNA can be stored at 4 °C for at least 3 months.
• 23
  Set up qPCR reactions in triplicate. For each genomic DNA sample, set up qPCR reactions with two sets of primers, one for DsRed (to determine copy number of scGESTALT barcode) and another for an ultraconserved control region that is known to be present in 2 copies⁸⁰. Thus, each sample will have a total of 6 reactions. In addition, set up qPCR with reference standards using genomic DNA from fish with no DsRed integration, and from fish with known single and/or double copy DsRed integrations.

  Component	Amount (μl)	Final Concentration
  Biorad iTaq mix	10	1×
  Primer (F+R mix) 10μM	0.8	0.4 μM
  Nuclease-free water	7.2
  Genomic DNA (1:5 dilution)	2
  Total	20

  Primer pairs: 1. qPCRctrl F+R 2. qPCRdsRed F+R (see DNA oligonucleotide sequences)

• 24
  Run qPCR as follows

  Cycle number	Denature	Anneal/Extend	Final
  1	95 °C, 3 min
  2–36	95 °C, 10 s	55 °C, 30 s (with fluorescence readout)
  37			8 °C, hold

• 25
  Calculate copy number with the equations below

  $$\mathrm{\Delta }\left(\text{Ct}\right)=\left(\text{Ct for dsRed PCR}\right)-\left(\text{Ct for cons\_ctrl PCR}\right)$$

  • Perform calculations for all unknown samples and positive and negative reference standards)

  $$\mathrm{\Delta }\mathrm{\Delta }\text{Ct}=\mathrm{\Delta }\left(\text{Ct},\text{ sample}\right)-\mathrm{\Delta }\left(\text{Ct},\text{ reference standard}\right)$$

  • Perform calculations for all unknown samples and positive and negative reference standards. Δ(Ct, sample) is the Δ(Ct) for each sample and positive and negative reference standards calculated using the first equation. Use the Δ(Ct) of the known single copy reference standard as Δ(Ct, reference standard). Thus, ΔΔCt for the single copy reference standard will be 0.

  $$\text{Copy number in the sample}=2^{-\mathrm{\Delta }\mathrm{\Delta }\text{Ct}}$$

  • Perform calculations for all unknown samples positive and negative reference standards. The value for the known single copy reference will be 1. The values for the known negative and double copy reference standards should be around 0 and 2 respectively.

  See ANTICIPATED RESULTS and Supplementary Data for examples of single-copy calculations

• 26Keep single-copy scGESTALT F₁ animals for all downstream experiments. Identify females and maintain them in separate 1 L tanks for barcoding experiments described below.

  PAUSE POINT Maintain stable lines by outcrossing single-copy F₁ animals. Adults can be maintained for up to 2 years.

scGESTALT sgRNA synthesis TIMING 2 h

• 27Generate sgRNAs targeting sites 1–4 of the scGESTALT barcode using EnGen sgRNA synthesis kit according to the manufacturer’s guidelines. Oligonucleotide sequences are provided in the Reagents list (see DNA oligonucleotide sequences).
• 28Purify sgRNAs with RNA Clean & Concetrator-25 kit according to the manufacturer’s guidelines and elute with 50 μl nuclease-free water.
• 29Quantify sgRNAs using NanoDrop and dilute each sgRNA to ~350 ng/μl with nuclease-free water. Combine equal volumes (e.g. 5 μl) of sgRNAs 1–4 to make an equimolar pool. PAUSE POINT Store the sgRNA pool in 1.5 μl aliquots at −80 °C for at least one year. CRITICAL STEP Avoid freeze-thaw of sgRNA pools. Each aliquot is sufficient for 1 injection mix.

scGESTALT ‘early’ barcode editing TIMING 2 h

• 2The day prior to performing microinjections, set up 4–6 overnight tanks with one scGESTALT F₁ female and one hsp:Cas9, U6:sgRNA F₁ male per tank for mating.

  CRITICAL STEP Male hsp:Cas9, U6:sgRNA must be used in the ‘early and late’ barcoding experiments to avoid maternal contribution of Cas9 mRNA and sgRNAs, and enable barcode editing at later developmental timepoints. Female scGESTALT animals must be verified to have single-copy transgene integrations.

• 3
  On the morning of microinjections, assemble the following reagent mix (3.3 μl total). Vortex to mix and centrifuge briefly.

  Component	Amount (μl)	Final Concentration
  Engen Cas9 NLS buffer	1
  Phenol red (0.1%, wt/vol)	0.3	0.009%
  sgRNA pool (~350 ng/μl each)	1	~106 ng/μl
  EnGen Cas9 NLS protein	1
  Total	3.3

• 4Incubate the mix at room temperature for 5 minutes to enable Cas9-sgRNA complexes to assemble. Hold the mix on ice until ready to start injections.
• 5Cross scGESTALT and hsp:Cas9, U6:sgRNA fish as described in Step 8.
• 6While fish are mating, prepare and calibrate needles for microinjection as described in Step 9. Load 1.5 μl of the Cas9-sgRNA injection mix into the needle.
• 7Collect fertilized eggs into Petri dishes with blue water and arrange 1-cell stage embryos for injections as described in Step 10. Inject 1.5 nl of Cas9-sgRNA mix through the chorion.
• 8Transfer injected embryos with blue water into a new Petri dish and incubate at 28.5 °C. Keep some uninjected control embryos to assess the health of the clutch. At the end of the day, remove any unfertilized or damaged embryos from the injected batches.

scGESTALT ‘late’ barcode editing TIMING 6 h

• 2At 30 hpf, screen embryos from Step 36 for heart GFP expression as described in Step 15 to identify ones carrying the scGESTALT transgene (~50% of injected progeny).
• 3Heat shock the sorted embryos for 30 min at 37 °C as described in Step 13.
• 4Screen embryos 2–4 hours after heat shock for ubiquitous GFP expression as described in Step 14 to identify ones carrying the hsp:Cas9, U6:sgRNA transgene.

Rearing double transgenic fish TIMING ~25 d

• 2Incubate sorted double transgenic embryos from Step 39 (~25% of injected progeny) at 28.5 °C until 5 dpf and then transfer into 2 L tanks (8–10 larvae per tank) in a zebrafish facility to raise to juveniles (23–25 dpf).

  PAUSE POINT The rearing time is user-dependent. Animals can be used in experiments at earlier stages or grown to adults.

Identify barcode edited juveniles TIMING 1 d

• 30The day prior to performing droplet encapsulation for scRNA-seq, identify individual juveniles from Step 40 with edited scGESTALT barcodes. Fin clip juveniles as described in Step 20. Transfer each fin clipped fish to individual 0.5 L tanks to recover and number them uniquely.
• 31Prepare genomic DNA as described in Steps 21–22.
• 32
  Set up PCR reactions (10 μl) as follows

  Component	Amount (μl)	Final Concentration
  5× HF Phusion buffer	2	1×
  10 mM dNTP	0.2	0.2 mM
  Nuclease-free water	6.2
  scGESTALT primer mix (10 μM)	0.5	0.5 mM
  Phusion polymerase	0.1
  Genomic DNA	1
  Total	10

• 33
  Run the following program on a thermocycler

  Cycle number	Denature	Anneal	Extend	Final
  1	98 °C, 30 s
  2–34	98 °C, 10 s	63 °C, 20 s	72 °C, 20 s
  35			72 °C, 3 min
  36				8 °C, hold

• 34Load PCR products on 1% agarose gel along with 100 bp DNA ladder according to standard procedures.
• 35Run the gel and visualize bands. Successfully edited scGESTALT barcodes will appear as smears ranging from ~120–250 bp. Generally, >80% of injected embryos are efficiently edited. See ANTICIPATED RESULTS for examples of successful editing.

  ?Troubleshooting

• 36Select 2–4 juveniles showing sufficient barcode editing for inDrops experiments the next day, and keep each juvenile in a separate 0.5 L tank overnight.

Prepare single-cell suspension of whole brain TIMING 2.5 h

• 9The next day, pre-warm 40 ml zebrafish facility water in a water bath at 37 °C in a 50 ml tube.
• 10Heat shock one 23–25 dpf juvenile selected in Step 47 to induce expression of the barcode array. Incubate the animal in a 50 ml tube with 20 ml of pre-warmed fish facility water at 37 °C for 40 min. This will promote expression of scGESTALT mRNA barcodes just prior to inDrops encapsulation. While the fish is being heat shocked, proceed with preparing dissociation solutions as described below (Steps 50–54).
• 11Transfer 25 ml Neurobasal media to a 50 ml tube. Add 500 μl B27 supplement. Oxygenate the solution by passing 95%O2:5%CO2 gas through a tubing attached to a sterile 5 ml serological pipette for 2 min (Supplementary Fig. 1). Hold Neurobasal/B27 media on ice.
• 12Prepare papain solution for dissociation. Add 5 ml Neurobasal media to Vial 2 (papain) of the Papain Dissociation System (Worthington). Add 0.5 ml Neurobasal media to Vial 3 (DNase). Gently resuspend the DNase and transfer 250 μl to Vial 2. Mix the papain/DNase solution by inverting the vial several times.
• 13Oxygenate the papain/DNase mix by passing 95%O2:5%CO2 gas through a tubing attached to a sterile 5 ml serological pipette for 2 min (Supplementary Fig. 1 and Supplementary Movie). CRITICAL STEP Do not bubble gas directly into the solution as it will cause frothing and protein denaturation, and reduce papain activity. Instead, slowly move the pipette in a circular motion above the liquid’s surface being careful not to cause the liquid to spill over (adjust the gas pressure if necessary).
• 14Prepare ovumucoid protease inhibitor (Vial 4) as described by the manufacturer.
• 15Incubate papain/DNase mix in a 34 °C water bath while preparing the single-cell brain suspension as described below (Steps 55–61).
• 16Anesthetize the heat shocked juvenile from Step 49 by adding 800 μl 25× MESAB.
• 17Add 1 ml DPBS (containing calcium, magnesium, glucose, pyruvate) to a 1.5 ml microfuge tube and hold on ice.
• 18Add 10 ml ice-cold Neurobasal/B27 media to a 10-cm sylgard dish. Add 400 μl 25× MESAB.
• 19Transfer the anesthetized fish to the sylgard dish (Supplementary Fig. 2). Pin the fish just posterior of the head, in the middle of the trunk and near the tail using 3 insect pins.
• 20Using a pair of fine forceps, remove the jaw, eyes, heart and gut tissues. Pierce the skin on top of the head and gently peel it back to expose the brain. Using fine forceps, gently scoop the brain out taking care not to lose part of the hindbrain in the process. Rip the brain tissue into 6–8 small pieces using two pairs of fine forceps. Typically, dissection takes ~5 min per brain.
• 21Transfer the brain tissue pieces to the 1.5 ml microcentrifuge tube with DPBS from Step 56.
• 22Briefly spin down the tube using a benchtop minifuge (2 second spin at room temperature). Discard DPBS and add 1 ml of ice-cold DPBS (no calcium, no magnesium). Hold the tube at room temperature.
• 23Remove the papain/DNase mix from the water bath (Step 54). Transfer 1 ml of the papain/DNase solution to a new 1.5 ml tube and add 1 ml Neurobasal media. This dilutes the papain to 10 units/ml. Note the concentration of papain can be adjusted if needed (generally 10–20 units/ml are equally effective for dissociation).
• 24Discard DPBS from the brain tissue pieces (step 61) and add 1 ml of the diluted papain/DNase mix from Step 62. Transfer the entire contents of the 1.5 ml microcentrifuge tube to a 15 ml tube, and oxygenate the solution as described in Step 52.

  CRITICAL STEP Re-oxygenation maintains cell viability during dissociation.

• 25Incubate the brain tissue pieces in a 34 °C water bath for 20 min.
• 26Gently triturate the tissue 10 times with a p1000 pipette set at 800 μl.

  CRITICAL STEP Do not triturate too vigorously and avoid introducing bubbles. Generally, pipetting at a rate of 1 sec per up and down motion of the pipette is desirable.

• 27Re-oxygenate the solution as described in Step 52 for 1 min. Incubate in a 34 °C water bath for an additional 5–6 min.
• 28Gently triturate the tissue 15–20 times with a p1000 pipette as described in Step 65. CRITICAL STEP Typically, this range of trituration is sufficient for efficient tissue dissociation. However depending on the strength of the papain, if large pieces of tissues are still visible, it may be necessary to perform additional pipetting or increase incubation time by 5–10 min. If longer incubation time at 34 °C is necessary, repeat Step 66 just prior to re-incubation.
• 29Transfer the entire contents of the 15 ml tube to a 1.5 microcentrifuge tube. Spin at 500 g for 4 min at 4 °C.
• 30While tube is spinning, prepare inhibitor solution as follows. Transfer 0.9 ml EBSS (Vial 1) solution to a 15 ml tube. Add 100 μl of resuspended ovumucoid protease inhibitor from Step 53. Add 50 μl of resuspended DNase (Vial 3) from Step 51. Oxygenate the mix as described in Step 52 and hold on ice.
• 31Following centrifugation in Step 68, discard the papain supernatant. A clear cell pellet should be visible. Resuspend the pellet with 1 ml of the prepared inhibitor solution.
• 32Spin at 500 g for 4 min at 4 °C.
• 33Discard the supernatant and gently resuspend the pellet with 1 ml ice-cold DPBS (no calcium, no magnesium).
• 34Sequentially filter the resuspension first through a 35 μm cell strainer, followed by a 20 μm cell strainer. Perform all steps on ice with pre-chilled tubes.
• 35Spin at 500 g for 4 min at 4 °C.
• 36Discard the supernatant and add 500 μl ice-cold DBPS (no calcium, no magnesium). Do not pipette the pellet. This step is to remove any residual debris.
• 37Spin at 500 g for 4 min at 4 °C.
• 38Discard the supernatant and resuspend the pellet with 400 μl ice-cold DPBS (no calcium, no magnesium). Hold on ice.
• 39Determine cell concentration and viability. Add 10 μl of the cell suspension to 10 μl of Trypan blue in a 1.5 ml microcentrifuge tube. Mix by gentle pipetting and load 10 μl onto a hemocytometer (e.g. Bulldog Bio). Count cells and determine the concentration as described by the manufacturer. Cell viability should be >70% (ideally >80%).

  CRITICAL STEP High cell viability is required to proceed with scRNA-seq experiments. Poor viability will result in fewer cells passing data quality filtering steps and can introduce more ambient RNA (i.e. non-cell associated RNA, such as free floating RNA from dead or dying cells) into cell suspension, compromising single-cell measurements.

  ?Troubleshooting

• 40Dilute cells to ~200,000 cells/ml with DPBS (no calcium, no magnesium).
• 41Add 300 μl 36% (vol/vol) OptiPrep to 300 μl diluted cells and gently resuspend the cell suspension. Final cell concentration is 100,000 cells/ml in 18% optiprep/DPBS (vol/vol). Proceed immediately to inDrops cell encapsulation.

inDrops single-cell encapsulation TIMING 30–60 min

• 5Run the inDrops device as described previously⁷³. Depending on the desired number of cells to be profiled, cells can be collected at a rate of 10,000–20,000 cells per hour. Generally, single-cell transcriptomes are obtained for ~70% of cells introduced into the device.

Reverse transcription in droplets TIMING 3 h

• 3Proceed to reverse transcription as described previously⁷³. If more than 3,000 cells were captured, split the emulsion into fractions containing ~3,000 cells.

  PAUSE POINT: Libraries post-reverse transcription can be stored at −80 °C for at least 3 months

inDrops transcriptome library preparation TIMING 2 d

• 37Generate single-cell transcriptome libraries as described previously⁷³ with the following modification. After in vitro transcription (IVT, for linear amplification) of the second-strand synthesis product, purify the reaction with 1.3× volume of AMPure beads according to the manufacturer’s instructions, and elute with 25 μl RNA elution buffer.
• 38Use 9 μl of the purified post-IVT product to proceed with fragmentation of amplified RNA and complete library preparation as described previously⁷³.

  PAUSE POINT Store remaining unfragmented RNA at −80 °C for at least 3 months.

scGESTALT library preparation: reverse transcription TIMING 2.5 h

• 42
  Assemble the following reaction (11 μl total) on ice using unfragmented RNA from Step 84.

  Component	Amount (μl)	Final Concentration
  Unfragmented RNA	5
  50 μM random hexamer	1.5	3.75 μM
  10 mM dNTP	1	0.5 mM
  Nuclease-free water	3.5
  Total	11

• 43Vortex the mixture and briefly centrifuge. Incubate the reaction at 70° C (lid 105 °C) for 3 min and immediately cool on ice.
• 44
  Add the following reverse transcription reagents on ice and mix by pipetting (20 μl total). Briefly spin down.

  Component	Amount (μl)	Final Concentration
  Mixture from Step 86	11
  5× PrimeScript buffer	4	1×
  Nuclease-free water	3.5
  RNaseOut (40 U/μl)	1	2 U/μl
  PrimeScript RT enzyme (200 U/μl)	0.5	5 U/μl
  Total	20

• 45Incubate the reaction mix in a thermocycler at 30 °C for 10 min, followed by 42 °C for 1 h, and then 70 °C for 15 min.
• 46Purify the reverse transcription product with 1.2× volume AMPure beads (24 μl) and elute with 20 μl DNA elution buffer.

  PAUSE POINT Purified cDNA can be stored at −20 °C for at least 3 months.

scGESTALT library preparation: PCR round 1 TIMING 1.5 h

• 6
  Set up a PCR reaction (25 μl) for targeted amplification of scGESTALT barcodes as follows

  Component	Amount (μl)	Final Concentration
  5× Q5 reaction buffer	5	1×
  10 mM dNTP	0.5	0.2 mM
  10 μM inDrops_GP6	1.25	0.5 μM
  10 μM inDrops_PE1Sa	1.25	0.5 μM
  cDNA from Step 89	7.5
  Q5 DNA polymerase	0.25
  Nuclease-free water	4.25
  Q5 enhancer	5
  Total	25

  CRITICAL STEP This PCR amplifies a longer scGESTALT fragment using a forward primer that binds in the DsRed transgene upstream of the scGESTALT barcode. We have found that using a nested PCR strategy improves specificity of amplification.

• 7
  Run the following program on a thermocycler

  Cycle number	Denature	Anneal	Extend	Final
  1	98 °C, 30 s
  2–16	98 °C, 10 s	61 °C, 25 s	72 °C, 30 s
  17			72 °C, 2 min
  18				8 °C, hold

• 8Add 25 μl nuclease-free water to the PCR product and purify with 0.6× volume AMPure beads (30 μl). Elute the product with 20 μl DNA elution buffer.

  PAUSE POINT Purified PCR product can be stored at −20 °C for at least one week.

scGESTALT library preparation: PCR round 2 TIMING 1.5 h

• 4
  Set up a second PCR reaction (25 μl) for targeted amplification of scGESTALT barcodes as follows

  Component	Amount (μl)	Final Concentration
  5× HF Phusion reaction buffer	5	1×
  10 mM dNTP	0.5	0.2 mM
  10 μM inDrops_GP12	1.25	0.5 μM
  10 μM inDrops_PE1Sb	1.25	0.5 μM
  PCR product from Step 92	8
  Phusion DNA polymerase	0.25
  Nuclease-free water	8.75
  Total	25

  CRITICAL STEP This PCR amplifies the desired scGESTALT fragment using a forward primer that binds 5′ of the scGESTALT barcode sequence. The primer contains 9×N sequences, where N is a random base, to introduce sequence variety.

• 5
  Run the following program on a thermocycler

  Cycle number	Denature	Anneal	Extend	Final
  1	98 °C, 30 s
  2–9	98 °C, 10 s	60 °C, 25 s	72 °C, 30 s
  16			72 °C, 2 min
  17				8 °C, hold

• 6Add 25 μl nuclease-free water to the PCR product and purify with 0.6× volume AMPure beads (30 μl). Elute the product with 15 μl DNA elution buffer.

  PAUSE POINT Purified PCR product can be stored at −20 °C for at least one week.

• 7Check concentration of 1 μl of eluate using Qubit according to the manufacturer’s guidelines. Generally, concentrations range from 1–2.5 ng/μl. If concentrations are substantially higher, repeat steps 93–94 with lower DNA input and/or fewer PCR cycles.

scGESTALT library preparation: final PCR TIMING 1.5 h

• 39
  Set up a PCR reaction (25 μl) for final library amplification as follows

  Component	Amount (μl)	Final Concentration
  2× KAPA HiFi HotStart PCR mix	12.5	1×
  Eluate from step 95 (1:10 dilution)	5
  R1-PCRix/ R2-PCR primer mix (5 μM)	2.5	0.5 μM
  Nuclease-free water	5
  Total	25

  CRITICAL STEP If multiplexing libraries, use different R1-PCRix/R2-PCR primer mix variants. For each scGESTALT library, ensure that the same primer pairs as those used for building the transcriptome libraries are used. For example, if multiplexing two libraries (e.g. A and B) in the same sequencing run, use R1-PCRix1/R2-PCR (for library A) and R1-PCRix2/R2-PCR (for library B) for both transcriptome and scGESTALT sequencing.

• 40
  Run the following program on a thermocycler

  Cycle number	Denature	Anneal	Extend	Final
  1	98 °C, 30 s
  2–3	98 °C, 20 s	55 °C, 30 s	72 °C, 30 s
  4–9/11a	98 °C, 20 s	65 °C, 30 s	72 °C, 30 s
  12			72 °C, 2 min
  13				8 °C, hold

  ^a

  The number of cycles performed at annealing temperature of 65 °C typically ranges from 6–8. This number can also be determined empirically by qPCR as described previously⁷⁰.
• 41Add 25 μl nuclease-free water to the PCR product and purify with 0.6× volume AMPure beads (30 μl). Elute the product with 25 μl RNA elution buffer. The eluate is the final sequencing-ready library.

  PAUSE POINT Libraries can be stored at −20 °C for at least a month.

• 42Check concentration of 1 μl of eluate using Qubit Generally, concentrations range from 2.5–6 ng/μl.

scGESTALT library preparation: library quality check TIMING 2 h

• 47Run a BioAnalyzer DNA HS assay to determine library size with 1 μl of eluate from Step 99. See ANTICIPATED RESULTS for examples of successful libraries.
• 48Determine the molar concentration of the library. Dilute the library to 10 nM. If sequencing multiple libraries together, combine equal volumes of libraries (each at 10 nM) to be multiplexed in 1.5 ml low adhesion microcentrifuge tubes.

  PAUSE POINT Libraries can be stored at −20 °C for at least a month or at −80 °C for a longer period.

Sequencing transcriptome and scGESTALT libraries TIMING 1 d

• 9Sequence transcriptome libraries using NextSeq 75 cycle high-output kit with standard Illumina primers.

  Read 1 (transcript read) is 61 cycles

  Index 1 (part 1 of inDrops cell barcode) is 8 cycles

  Index 2 (library index) is 8 cycles

  Read 2 (part 2 of inDrops cell barcode [8bp,] followed by 6 bp UMI) is 14 cycles

• 10Sequence scGESTALT libraries using MiSeq 300 cycle kits and 20% PhiX spike-in with standard Illumina primers.

  Read 1 (scGESTALT lineage barcode) is 250 cycles

  Index 1 (part 1 of inDrops cell barcode) is 8 cycles

  Index 2 (library index) is 8 cycles

  Read 2 (part 2 of inDrops cell barcode [8bp,] followed by 6 bp UMI) is 14 cycles

Sequencing data processing TIMING 1 d

• 8Process raw transcriptome reads as described previously⁷³. The inDrops bioinformatics pipeline (inDrops.py) is available at https://github.com/indrops/indrops. Bowtie version1.1.1 is used with parameter −e 200; UMI quantification is used with parameter −u 2.
• 9Process scGESTALT reads in parallel using the inDrops.py script and stop just prior to the transcriptome mapping step. Modify the Trimmomatic settings to LEADING: “10”; SLIDINGWINDOW: “4:5”; MINLEN: “16”. This will generate files with inDrops cell identifiers and the scGESTALT lineage barcode sequence. The data can be further processed using the scGestaltPrepFunc.R script in Supplementary Software 2 or a similar custom script.
• 10Process the resulting files using the pipeline available at https://github.com/aaronmck/SC_GESTALT (Supplementary Software 1). This will generate tables containing the lineage barcode sequence for each cell.

Downstream data processing TIMING 3 d

• 43Using Seurat (or related scRNA-seq analysis tool), identify cell clusters (i.e. different cell types) using the transcriptome scRNA-seq data (for tutorial refer to: https://satijalab.org/seurat/). In brief, gene expression profiles are used to perform dimensionality reduction using principal component analysis with highly variable genes. A modularity-based clustering algorithm (Louvain) is used to cluster cells into discrete cell types using significant principal components. The result is visualized in two dimensions on a t-distributed stochastic neighbor embedding (t-SNE) plot (Fig. 1). For further details of how clustering is performed for the zebrafish juvenile brain dataset, we refer the reader to the Methods section of ref⁷². See ANTICIPATED RESULTS for examples of clustering analyses from zebrafish brain scRNA-seq.
• 44Match transcriptome (from Step 108) and lineage barcodes (from Step 107) for each cell using the inDrops cell identifier index sequence (Supplementary Software 3–4) and generate lineage trees (based on maximum parsimony) using the pipeline available at https://github.com/aaronmck/SC_GESTALT. See ANTICIPATED RESULTS for an example lineage tree.

TIMING

Zebrafish transgenesis

Steps 1–5, Tol2 mRNA synthesis: 1 d

Steps 6–12, microinjection: 1 d

Steps 13–15, screening and rearing fish: 2–3 months

Steps 16–19, germline transmission: 2 d, wait for adults 2–3 months

Steps 20–26, copy number determination: 1d

scGESTALT barcode editing

Steps 27–29, sgRNA synthesis: 2 h

Steps 30–37, early timepoint Cas9 editing: 2 h

Steps 38–39, late timepoint Cas9 editing: 6 h

Step 40, growing to juveniles: 25 d

Steps 41–47, identifying edited fish: 1d

inDrops transcriptome and lineage profiling

Steps 48–80, single-cell processing of brain tissue: 2.5 h

Step 81, cell encapsulation: 30–60 minStep

82, reverse transcription: 3 h

Steps 83–84, transcriptome library preparation: 2 d

Steps 85–89, scGESTALT reverse transcription: 2.5 h

Steps 90–92, scGESTALT PCR round 1: 1.5 h

Steps 93–96, scGESTALT PCR round 2: 1.5 h

Steps 97–100, scGESTALT final library PCR: 1.5 h

Steps 101–102, check library with BioAnalyzer: 2 h

Sequencing and raw data processing

Steps 103–104, Miseq and NextSeq runs: 1 d

Steps 105–107, raw data processing: 1d

Downstream data processing

Steps 108–109, clustering analysis: 3 d (preliminary analysis)

TROUBLESHOOTING

Extensive troubleshooting pointers for running the inDrops device and preparing inDrops transcriptome libraries have been described previously⁷³.

There are two main parts in the scGESTALT protocol where issues may arise.

No or inefficient editing of the barcode (Step 46).

This usually indicates problems with the editing reagents. Ensure that the Cas9 protein used for injection is active by testing it for successful editing of endogenous genes e.g. tyr gene encoding tyrosinase results in loss of pigmentation by 48 hpf when mutated (see ref⁸¹ for details and tyr sgRNA sequences). If Cas9 is active, then the sgRNA pool is likely compromised. Remake the sgRNAs, aliquot in single-use amounts and store at −80 °C. Do not freeze/thaw sgRNAs and discard each aliquot after use. After heat shock of the embryos at 30 hpf, GFP expression is used as a proxy for successful Cas9 transgene induction. If weak or no GFP is observed, it indicates that heat shock was not successful. Test the heat shock conditions empirically and if necessary increase time of heat shock by 10 min. Do not heat shock for >45 min as it can result in large deletions in the lineage barcode array. Alternatively, screen F1 transgenic animals to identify ones with robust GFP expression following heat shock.

Low cell viability (Step 78).

High quality single-cell preparations are critical for obtaining successful single-cell transcriptome and scGESTALT libraries. If cell viability is <70%, repeat dissociation with another sample. Ensure that all media is properly oxygenated (this is the most critical part of the protocol), especially during the papain incubation at 34 °C (steps 63–64). Do not incubate tissue for >30 min at 34 °C and do not triturate the sample vigorously after papain digestion. Gently pipette cells during DBPS washes and final resuspension.

ANTICIPATED RESULTS

The protocol described herein enables the user to apply CRISPR-Cas9 for cumulative, combinatorial and heritable genomic barcode editing at multiple timepoints to generate permanent records of cell lineages during development. As the mutated barcodes are also expressed as mRNA, both lineage information and cell type identity can be simultaneously extracted from single cells at high throughput using droplet-based scRNA-seq. Cell relationships at the level of gene expression and lineages can then be represented on lineage trees, which can be explored to identify timing and patterns of lineage segregation within or between tissue regions and cellular subtypes.

Below, we describe results obtained at various stages of the protocol that can be used to gauge the experiment’s success.

Barcode copy number determination by qPCR (Step 25).

The number of integrations of the scGESTALT lineage barcode in transgenic F1 can be determined by qPCR. Figure 4 shows representative qPCR results from this assay. Animals with a calculated copy number of less than or equal to 1 (below 0.6 are not considered integrants) are considered to be single integrants, and can be used for downstream experiments and further expanded to maintain single-copy transgenic lines. Raw data for the calculations are included in Supplementary Data.

Figure 4. Barcode copy number determination.

Copy number of scGESTALT barcode integration was determined using qPCR. Samples 1–13 represent different genomic DNA samples (n = 13 animals). Reference standards: c1 is negative control (n = 1 animal), c2 is known to contain 1 copy (n = 1 animal), c3 is known to contain 2 copies (n = 1 animal). This procedure was approved by the HU/FAS Committee on the Use of Animals in Research & Teaching under Protocol No. 25–08.

Identification of successful barcode editing (Step 46).

Genomic DNA from edited fish is extracted, and lineage barcodes are PCR amplified and visualized on an agarose gel. As shown in Figure 5, band patterns that appear as smears ranging from 120–250 bp are indicative of efficiently edited barcodes. Depending on whether barcodes were edited at only ‘early’ timepoints by injections, only ‘late’ timepoints by Cas9 induction, or both ‘early and late’ timepoints by combination of Cas9 injection and induction, the gel band patterns have slight distinctions. If the pattern is dominated by short bands at ~100–150bp, it is advisable to avoid using the corresponding animal in scRNA-seq experiments, since it is likely that this individual’s barcode edits will be predominantly large deletions, reducing the diversity of edits and the resolution of lineage trees. In addition, since Cas9-induced mutations are stochastic, gel band patterns between animals that were edited in a similar manner will be variable.

Figure 5. scGESTALT barcode editing.

scGESTALT barcode zebrafish were crossed to zebrafish that express heat shock-inducible Cas9 and U6-driven sgRNAs 5–9. Resulting embryos were injected with Cas9 protein and sgRNAs 1–4 at the one-cell stage. Embryos were heat shocked at 30 hpf to induce transgenic Cas9 for a late round of editing. Double transgenic (scGESTALT⁺, hsp:Cas9⁺; lanes 2–8, n = 7 embryos) and single transgenic (scGESTALT⁺, hsp:Cas9⁻; lanes 9–12, n = 4 embryos) were identified by screening for GFP expression. The gel shows PCR results of amplifying the scGESTALT barcode (unedited = ~300 bp). Large smear patterns (120–250bp) are observed in early and late edited embryos (lanes 2–8), whereas embryos that were only mutated at sites 1–4 display less editing (lanes 9–12. The band at ~200 bp in lane 12 likely represents large deletion(s) between sites 1–4 that occurred early in development and was inherited by most cells. Note that samples with such dominant large deletions should not be used for downstream experiments and analyses as they are likely to have low barcode diversity). Sample in lane 11 was likely not efficiently injected. Lane 1 represents a control embryo, which was injected with Cas9 protein only (no sgRNAs 1–4, n = 1 embryo) and was not heat shocked. As expected, the barcode is not edited in this case. This procedure was approved by the HU/FAS Committee on the Use of Animals in Research & Teaching under Protocol No. 25–08.

BioAnalyzer electropherograms of successful scGESTALT libraries (Step 101).

The profiles of amplified lineage barcode libraries are shown in Figure 6. A successful library typically ranges in size between 500–800bp with an average of 600–650 bp. Some larger fragments may be observed, but are generally not problematic for sequencing. Alternatively, the ratio of AMPure beads at steps 95 and 99 can be increased (e.g. to 0.8× or 1×) to remove those fragments. If fragments are considerably smaller than this range, it may indicate that the barcode contains large deletions. Refer to the previous section to determine how to avoid using animals with undesirable barcode mutation patterns.

Figure 6. BioAnalyzer electropherograms of scGESTALT sequencing libraries.

Traces were obtained from 23–25 dpf juvenile zebrafish brains (n = 1 animal). The single-cell emulsion (step 82) from one inDrops collection was split into 4 fractions of ~3,000 cells each, and the resulting libraries were indexed uniquely. The average size of the libraries is 600–650bp. Peaks at 35 and 10380 bp represent gel migration markers. This procedure was approved by the HU/FAS Committee on the Use of Animals in Research & Teaching under Protocol No. 25–08.

Detailed discussion of results from inDrops transcriptome libraries have been described before⁷³. It is worth noting that zebrafish neural cells have low RNA content, with estimates in the 1–2 picogram range²³. This is still sufficient for scRNA-seq profiling and BioAnalyzer electropherograms display similar profiles as previously described for inDrops libraries from other cell types⁷³.

Clustering analyses and lineage trees (Steps 108–109).

The digital gene expression matrix output (rows contain gene counts and columns represent single cells) from the inDrops processing pipeline is used to identify cell types using any relevant scRNA-seq computational tool. We will use Seurat¹⁸ as an example for the rest of this section. First, the matrix is column-normalized and log-transformed (see tutorial at https://satijalab.org/seurat/). Next, low quality single-cell libraries are filtered out from the dataset. The metrics for what constitutes ‘low quality’ will vary between cell types, tissues, dissociation protocols and sequencing depth. For zebrafish brain data, we filtered cells with fewer than 500 expressed genes or greater than 9% mitochondrial content⁷². Potential cell doublets or multiplets should also be filtered out (e.g. by identifying cells with high numbers of gene counts and unique molecular identifiers that are outliers of a normal distribution). Next, highly variable genes are identified and used for principal component analysis. The top significant principal components are used for dimensionality reduction and for grouping all cells into distinct clusters (using a Louvain modularity algorithm). Clusters are expected to be supported by cells from multiple biological replicates, otherwise it may indicate technical artefacts in the data. The resulting clusters can be compared to each other to identify differentially expressed genes, which may serve as gene markers for individual cell types. Clusters can be classified using the identified gene markers and validated with in situ hybridizations.

To generate scGESTALT lineage trees, the inDrops index sequences are used to match transcriptomes and lineage barcodes for the same cells. Cell lineage trees are generated using maximum parsimony (adapted from phylogenetics) based on patterns of shared edits (for more details see refs^66,72). To represent early and late editing, the tree is anchored with edits that occur at sites 1–4 of the barcode array (early editing from injections at the 1-cell stage), and is then extended with additional edits accumulated at sites 5–9 (late editing from heat shock-induced Cas9 activity). A clade on the tree represents all lineage barcodes that share at least one common edit, and sub-clades that branch from the original clade contain increasingly restricted subsets of barcodes that contain the previous edit(s) as well as additional shared edits (Fig. 1, Fig. 7). Individual cells (identified by their inDrops index) containing each of the recovered barcodes is connected to the tips of the tree (i.e. the terminal lineage barcode sequence). Cells connected to the same terminal node are clonal (i.e. contain the same lineage barcode). Cell annotations (e.g. cell type and spatial/regional origin) can be color coded and added to the tree to explore various lineage relationships, such as to determine the relationship between cells belonging to the same cluster or cells from different regions of the tissue.

Figure 7. Zebrafish brain lineage tree generated using scGESTALT.

An example of a reconstructed lineage tree from a single juvenile zebrafish brain. 376 edited barcodes were recovered from single cells using inDrops. A cell lineage tree was generated from the barcodes based on shared edits using a maximum parsimony approach. Black nodes represent early barcode edits (Cas9 and sgRNA injection at 1-cell stage, Step 35); red nodes represent late edits (heat shock-induced Cas9 transgene expression, Step 38). Dashed lines join single cells to terminal nodes (represent the final edited barcode sequence) on the tree. Distinct cell types (identified from simultaneous transcriptome capture and cell clustering analyses) are color coded as indicated in the legend. The edited barcode for each cell is shown as a white bar with deletions (red) and insertions (blue). Examples of clades and subclades are indicated on the tree. A clade on the tree represents all lineage barcodes that share at least one common edit, and sub-clades that branch from the original clade contain increasingly restricted subsets of barcodes that contain the previous edit(s) as well as additional shared edits. Adapted with permission from ref⁷². This procedure was approved by the HU/FAS Committee on the Use of Animals in Research & Teaching under Protocol No. 25–08.

Supplementary Material

SupData

Supplementary Data. Raw data used for calculation of scGESTALT lineage barcode copy by qPCR. The number of integrations of the scGESTALT barcode in transgenic F1 is determined by qPCR. Genomic DNA from 13 animals (gDNA 33–50) was used in the assay. Genomic DNA from an animal with no barcode integration (control zero), an animal with known single copy integration (control single), and an animal with known double copy integration (control double) are included as standards. For each genomic DNA sample, qPCR reactions are set up in triplicate using primers for an ultraconserved control region that is known to be 2 copies, and for the DsRed region to determine copy number of the scGESTALT lineage barcode. The Ct number for each reaction is shown. The average Ct for each region is calculated (Ctrl Ct and DsRed Ct). The delta Ct and delta delta Ct are calculated and used to determine copy number (refer to Procedure Step 25 for details). Animals with a calculated copy number of less than or equal to 1 (but greater than or equal to 0.6) are considered to be single integrants.

SupFig

Supplementary Figure 1. Papain oxygenation setup. Left and middle panels, 95%O₂: 5%CO₂ gas tank fitted with a gas regulator, Tygon E-3603 tubing and a 5 ml serological pipette. Right panel, Oxygenation of papain/DNase mix in Neurobasal media (small vial to the right) is performed by bubbling 95%O₂: 5%CO₂ gas through tubing attached to a sterile 5 ml serological pipette for 2 min (Procedure Step 52). EBSS buffer (large vial to the left, used for resuspending ovumucoid, Procedure Step 53) and Neurobasal media (Procedure Step 50) are oxygenated in a similar manner.

Supplementary Figure 2. Zebrafish brain dissection. Top panels, Anesthetized fish is transferred to a sylgard dish covered with Neurobasal media and MESAB (left). The fish is pinned just posterior of the head, in the middle of the trunk and near the tail using 3 insect pins (right, asterisks mark pin positions). Bottom panels, The jaw, eyes, heart and gut tissues are removed. The skin on top of the head is pierced and peeled back to expose the brain (left, circle marks the exposed brain). Gently scoop the brain out taking care not to lose part of the hindbrain in the process (right, whole brain is encircled).

SupMov

Supplementary Movie. Papain/DNase mix oxygenation. Demonstration of the proper method for oxygenation of the papain/DNase mix.

SupSoft1

Supplementary Software 1. scGESTALT analysis pipeline scripts. This is the master pipeline for processing scGESTALT reads and generating lineage trees (Fig. 1 and Fig. 7, relevant for Step 107).

SupSoft2

Supplementary Software 2. R script (scGestaltPrepFunc) for pre-processing of inDrops scGestalt data. The script will format the data for input to the scGestalt analysis pipeline (relevant for Step 106).

SupSoft3

Supplementary Software 3. R script pipeline (Transcriptome-scGestaltMatchPipe) for matching transcriptome and scGESTALT lineage barcodes for profiled single cells (relevant for Step 109).

SupSoft4

Supplementary Software 4. R script (MatchPipeFunc) containing the code of the R functions called by the scGestaltMatchPipe pipeline (relevant for Step 109).

SupTable

Supplementary Table 1. Sequences of scGESTALT CRISPR target sites.

Supplementary Table 2. InDrops library multiplexing primer sequences.