The primary role of zebrafish nanog is in extra-embryonic tissue

The transcription factor Nanog is part of the core circuitry that regulates mammalian pluripotency (reviewed by Theunissen and Jaenisch, 2014). In vitro studies have shown that removal of Nanog triggers differentiation of mouse and human embryonic stem cells (Chambers et al., 2007; Hyslop et al., 2005; Loh et al., 2006; Mitsui et al., 2003). However, a subset of Nanog mutant mouse embryonic stem cells are able to self-renew (Chambers et al., 2007). In vivo studies have revealed that Nanog is required for inner cell mass pluripotency and epiblast development (Mitsui et al., 2003). However, in chimeras with wild-type cells, Nanog mutant cells can give rise to tissues from all germ layers (Chambers et al., 2007). Thus, mouse Nanog is involved in, but not absolutely required for, the maintenance of the pluripotent state (Carter et al., 2014; Chambers et al., 2007; Schwarz et al., 2014).

The roles of zebrafish Nanog in pluripotency and differentiation are less well-understood. Xu et al. (2012) reported that nanog was provided maternally and present in all embryonic and extra-embryonic cells. Morpholino-mediated knockdown of nanog mRNA resulted in developmental arrest prior to gastrulation. Nanog morphants displayed defects in the formation of the yolk syncytial layer (YSL), the extra-embryonic tissue that attaches the embryo to the yolk and generates Nodal and BMP signals that pattern mesendoderm (Carvalho and Heisenberg, 2010; Chen and Kimelman, 2000; D'Amico and Cooper, 2001; Hong et al., 2011; Kimmel and Law, 1985; Mizuno et al., 1996; Xu et al., 2012). Gene expression analysis in nanog morphants revealed the absence of YSL markers such as mxtx2 and the misregulation of hundreds of embryonic genes, including Nodal and its target genes. Injecting mxtx2 mRNA into YSL precursors of nanog morphants partially rescued YSL formation and the expression of Nodal and several of its target genes. Although no cell-autonomy data were shown to determine whether Nanog was required in embryonic cells, the study suggested that the primary role of Nanog is to regulate the formation of the YSL (Xu et al., 2012).

Two subsequent studies analyzed potential roles of zebrafish Nanog in embryonic cells (Lee et al., 2013; Perez-Camps et al., 2016). Lee et al. (2013) defined a set of genes expressed at the maternal-to-zygotic transition (MZT), expression of which was reduced in nanog morphants. Chromatin immunoprecipitation experiments suggested that many of these genes were direct targets of Nanog (Bogdanovic et al., 2012; Lee et al., 2013; Leichsenring et al., 2013; Xu et al., 2012). Based on the reduced expression of genes in morphants and the Nanog binding data, the study concluded that Nanog, along with Pou5f1 (now known as Pou5f3 in zebrafish) and the SoxB1 family, was involved in the first wave of zygotic transcription in embryonic cells. Subsequent reviews have interpreted these results to conclude that Nanog is directly required for zygotic genome activation in embryonic cells (Langley et al., 2014; Lee et al., 2014; Onichtchouk and Driever, 2016; Paranjpe and Veenstra, 2015), even though the majority of zygotic genes are activated in nanog morphants (Lee et al., 2013; Xu et al., 2012). Perez-Camps et al. (2016) reported that morpholino knockdown of nanog caused defects in BMP signaling and target gene expression, and suggested that Nanog acts to promote ventral cell-fate specification. Surprisingly, neither study (Lee et al., 2013; Perez-Camps et al., 2016) mentioned the extra-embryonic YSL phenotype of nanog morphants (Xu et al., 2012) or tested the postulated direct roles of Nanog in embryonic cells.

Here, we clarify the embryonic and extra-embryonic requirements for Nanog using tissue-specific rescue and chimera analysis. Our results indicate that the primary role of zebrafish nanog is YSL formation and that it is not essential for embryonic cell differentiation.

The interpretation of morpholino experiments can be complicated by potential partial loss-of-function phenotypes and the short half-life of morpholinos. To avoid these confounding effects in our studies, we generated nanog mutants using transcription activator-like effector nucleases (TALENs) (Carroll, 2014). We isolated an allele containing a 7 bp deletion predicted to cause a frameshift and premature termination codon before the homeodomain required for DNA binding (Fig. 1A). The mutant nanog mRNA was not detectable at sphere stage [4 hours post-fertilization (hpf)], presumably owing to nonsense-mediated decay (Fig. 1B). Homozygous zygotic nanog (Znanog) mutant embryos showed no phenotypic defects and could be raised to fertile adults. By contrast, maternal-zygotic nanog mutants (MZnanog) and maternal-only nanog mutants (Mnanog) arrested at sphere stage, did not undergo normal epiboly (Fig. 1C), and died when embryonic tissue detached from the yolk. The defects observed in MZnanog embryos were rescued by injection of nanog mRNA at the 1-cell stage (Fig. 1D). Rescued embryos could be raised to fertile adults, establishing that the observed phenotype is solely due to disruption of the nanog gene. The nanog mutant phenotype strongly resembles the previously reported nanog morphant phenotype (Xu et al., 2012). As Mnanog and MZnanog mutants had very similar or identical phenotypes, whereas Znanog mutants were viable, we conclude that maternal, but not zygotic, Nanog has essential roles during embryogenesis.

YSL development and the expression of YSL markers are impaired in nanog morphants (Xu et al., 2012). Several lines of evidence indicate that YSL formation is similarly impacted in MZnanog mutants. First, yolk injection of SYTOX Green dye indicated that marginal blastomere nuclei form a normal syncytium in MZnanog at sphere stage (Fig. 2A). However, by 6.5 hpf (shield stage in wild type) these yolk syncytial nuclei are aberrantly clustered together and display an enlarged morphology in MZnanog, reminiscent of the phenotypes caused by microtubule defects in the YSL (Takesono et al., 2012). We next investigated the F-actin band, a network of actin filaments that forms in the YSL during gastrulation and is postulated to promote blastopore closure (Cheng et al., 2004; Köppen et al., 2006; Wilkins et al., 2008). We stained for actin using phalloidin and found that the F-actin band was absent in MZnanog at 8 hpf (Fig. 2B), consistent with the phenotypic arrest in the mutants before epiboly (Fig. 1C). Interestingly, MZnanog embryos still form the shield structure and occasionally exhibit dorsal cell internalization and marginal constriction, even in the absence of epiboly or the F-actin band. These results suggest that YSL formation is initiated in MZnanog mutants but the YSL fails to differentiate properly and maintain the attachment of embryo to yolk.

We next investigated gene expression in the YSL. MZnanog embryos lacked expression of the key YSL determinant mxtx2 at sphere stage (Fig. 2C). Expression of other YSL marker genes, such as slc26a1, gata3 and hnf4a (Xu et al., 2012), was decreased or absent as determined by RT-qPCR (Fig. 2D). RNAseq experiments showed that the expression of many genes expressed in the YSL (Xu et al., 2012) was reduced in MZnanog mutants at 6.5 hpf (Fig. 2E). Expression of many of these genes was not completely eliminated, probably owing to their co-expression in embryonic cells. Compared with zygotically expressed housekeeping genes (Fig. 2F), YSL genes were significantly downregulated in MZnanog embryos (P=7.5×10⁻⁶; Fig. 2G). These results confirm and extend previous morphant studies that concluded that Nanog is required for YSL development (Xu et al., 2012).

nanog morphants were reported to display a widespread reduction in zygotic gene expression (Lee et al., 2013). To determine whether similar defects are found in MZnanog mutants, we used in situ hybridization, RT-qPCR and RNAseq. Using in situ hybridization and RT-qPCR, we demonstrated that the expression of several early zygotic genes was decreased in MZnanog at sphere stage (Fig. 3A,B). RNAseq experiments confirmed that the expression of many of the previously defined early zygotic genes (Lee et al., 2013) was reduced in MZnanog embryos at sphere stage (Fig. 3C). Compared with maternally provided mRNAs, which are largely unaffected (Fig. 3D), the levels of early zygotic RNAs are significantly reduced in MZnanog mutants (P=0.00307; Fig. 3E). The expression of 79/251 early zygotic genes was reduced >2-fold in MZnanog embryos. Many of these genes are part of developmental signaling pathways (Table S1). Of the 79 genes reduced in MZnanog mutants, 68 were also significantly decreased in the nanog morphants (Lee et al., 2013). These results indicate that nanog mutants and morphants have similar gene expression defects.

To clarify the embryonic and extra-embryonic requirements for Nanog, we performed tissue-specific rescue experiments. We first attempted rescue by injection of nanog mRNA into the yolk at the 64-cell stage, or directly into the YSL at the 512-cell stage, but observed little to no rescue. We hypothesized that nanog expression is required early during embryogenesis, and that delayed expression by mRNA injection at later stages is insufficient to rescue the phenotype. We therefore performed YSL rescue experiments by depositing mRNA in the precursor cells of the YSL by vegetal yolk injection at the 4-cell stage (Xu et al., 2012). Co-injection of GFP mRNA was used to verify that expression was predominantly in the YSL, although we cannot exclude the possibility that low levels of injected mRNA are inherited and translated in embryonic cells (Fig. S1A,B). Similar to 1-cell nanog injection, 4-cell yolk injection of nanog mRNA was sufficient to rescue the epiboly defects and yolk detachment in most MZnanog mutants (Fig. 4A). Additionally, expression of mxtx2 and ndr2 was restored in 4-cell yolk-injected MZnanog embryos (Fig. S1C), indicating that YSL function was rescued. At 24 hpf, the majority of mutants displayed axis rescue, tail elongation, and differentiation of various cell types (Fig. 4B,C). In contrast, injection of nanog mRNA into one of 16 cells did not result in any rescue of MZnanog embryos (Fig. 4A,C). We next attempted to rescue MZnanog by mxtx2 mRNA injection (Xu et al., 2012). Four-cell yolk injection of mxtx2 mRNA rescued the epiboly defect and yolk detachment in MZnanog mutants (Fig. 4D,E), and some rescued embryos developed differentiated tissues, including somites, notochord, eyes, brain, melanocytes, otoliths and blood. Generally, rescue by nanog mRNA was more pronounced than rescue by mxtx2 mRNA. These results suggest that the primary role of Nanog is in YSL development, and that YSL defects are a major cause of the embryonic phenotypes observed in MZnanog mutants.

As an additional test to determine whether nanog acts primarily in the YSL, we performed whole blastoderm transplants (Holloway et al., 2009). We transplanted MZnanog blastoderm tissue onto wild-type embryos in which yolk cell was left intact but most of the blastoderm tissue was removed (Fig. 5A). Transplants were performed at the 256- to 512-cell stage, and donor and host tissues were labeled with different fluorescent dyes to track their contributions to the resulting chimera. Shortly after transplantation, we verified that the YSL was exclusively contributed by wild type and that embryonic cells were predominantly contributed by MZnanog tissue (Fig. 5B). Strikingly, these chimeras (n=9) underwent epiboly similar to wild-type embryos or wild-type-to-wild-type controls (n=10) (Fig. 5C,E). In the reciprocal experiment, we transplanted wild-type blastoderm tissue onto MZnanog embryos in which yolk cell was left intact but most of the blastoderm tissue was removed. These chimeras arrested during epiboly (n=5) and had a phenotype similar to MZnanog mutants (Fig. 5D). These experiments cannot rule out an additional role for Nanog in marginal cells but support the idea that Nanog is primarily required in the YSL.

It has been proposed that Nanog is directly required in embryonic cells (Lee et al., 2013; Perez-Camps et al., 2016), but this assumption has not been tested rigorously. For example, mesendodermal marker gene expression is lost in nanog morphants (Lee et al., 2013; Xu et al., 2012) and MZnanog mutants (Fig. 3B, Table S1). Similarly, miR-430 expression and activity are abrogated in nanog morphants (Lee et al., 2013). Such defects may be caused directly by absence of Nanog in embryonic cells or indirectly by lack of Nanog in the YSL (Fig. S1C) (Xu et al., 2012). In support of the latter model, MZnanog embryos injected with Nodal (ndr1) mRNA can express mesendodermal marker genes at similar levels as in wild-type embryos (Fig. S2). This results reveals that Nanog is not required for Nodal-induced cell-type specification. However, supporting a direct role for Nanog in embryonic cells, we found that a GFP reporter normally silenced by miR-430 is active in MZnanog embryos and in transplanted MZnanog cells (Fig. S3). This result supports a cell-autonomous requirement for Nanog in miR-430 activity.

To test directly the role of Nanog in embryonic cell proliferation and differentiation, we co-transplanted fluorescently labeled MZnanog mutant and wild-type cells into wild-type hosts at sphere stage, and tracked their contributions to the host embryos (Fig. 6A). By 30 hpf, transplanted wild-type and MZnanog mutant cells had proliferated to a similar extent in the host embryo and contributed to many tissues, including brain (ectodermal derivative), muscle (mesodermal derivative) and hatching gland (axial mesodermal derivative) (Fig. 6B). To further test whether cells lacking Nanog can differentiate and contribute to tissues from all three germ layers, we used transgenic markers to follow donor cell differentiation (Fig. 6C), including fli1a:GFP (vasculature, a derivative of the mesodermal germ layer), actc1b:GFP (trunk muscle, mesodermal germ layer), isl1:GFP (trigeminal sensory neurons, ectodermal germ layer) and sox17:GFP (gastrointestinal tract, endodermal germ layer) (Higashijima et al., 1997; Lawson and Weinstein, 2002; Sagasti et al., 2005; Sakaguchi et al., 2006) (Fig. S4). We transplanted cells from transgenic Mnanog embryos into wild-type hosts and found that donor cells proliferated in the host embryo and activated each of the four marker transgenes (Fig. 6D). We additionally found that germ cells were specified in MZnanog embryos, and MZnanog germ cells migrated correctly to the host gonad when transplanted into wild-type host embryos (Fig. S5).

Finally, we tested whether transplanted cells lacking Nanog can make long-term contributions to the adult host. We turned to GESTALT (genome editing of synthetic target arrays for lineage tracing), a long-term lineage-tracing method that uses CRISPR-Cas9 genome editing to introduce permanent mutations in a DNA barcode (McKenna et al., 2016). We injected Cas9 protein and sgRNAs into Mnanog; GESTALT transgenic embryos to edit the barcode and transplanted cells into wild-type hosts (Fig. 7A). Chimeric embryos were grown to adulthood, and intestine, heart, eyes and brain tissues were isolated. To detect the contributions of Mnanog; GESTALT cells, GESTALT barcodes were amplified and sequenced from each sample. Across all organs, we found hundreds of barcodes containing different combinations of edits, demonstrating that Mnanog cells can contribute to adult organs (Fig. 7B). These results reveal that zebrafish Nanog is not required for the proliferation, differentiation or survival of embryonic cells.

Our results support the conclusion that the primary role for zebrafish Nanog is in the specification of the YSL (Xu et al., 2012) and that it has no absolutely essential autonomous functions in embryonic cells.

Four lines of evidence support an essential role of Nanog in YSL formation. First, both MZnanog mutants (this study) and nanog morphants (Xu et al., 2012) lack expression of mxtx2, the master regulator of YSL maturation, as well as expression of several other YSL markers. Second, absence of Nanog blocks formation of the F-actin band within the YSL (Xu et al., 2012) and leads to epiboly defects and embryo detachment (Xu et al., 2012; this study). Interestingly, MZnanog embryos occasionally exhibit marginal constriction, even in the absence of epiboly or the F-actin band. Given the importance of the actin ring for marginal constriction after 50% epiboly (Behrndt et al., 2012; Holloway et al., 2009; Köppen et al., 2006; Popgeorgiev et al., 2011; Wilkins et al., 2008), MZnanog embryos seem to constrict off the yolk by a fundamentally different mechanism. Third, expression of nanog mRNA (this study) or mxtx2 mRNA (Xu et al., 2012; this study) predominantly in YSL precursors partially rescues the YSL gene expression, epiboly and embryonic patterning defects caused by the loss of nanog. Fourth, MZnanog mutant yolk and YSL do not support the epiboly movements of wild-type embryonic cells upon blastoderm transplants. These observations strengthen and extend the previously reached conclusion (Xu et al., 2012) that Nanog is essential for proper YSL development.

Four lines of evidence indicate that Nanog is not directly required in embryonic cells. First, nanog (this study) or mxtx2 (Xu et al., 2012; this study) expression predominantly in YSL precursors is sufficient to rescue many aspects of the phenotype, including epiboly, embryo attachment and embryonic cell differentiation. Second, in blastoderm transplants, MZnanog mutant cells undergo epiboly when combined with a wild-type yolk and YSL (this study). Third, MZnanog embryos injected with Nodal (ndr1) mRNA can express mesendodermal marker genes to levels similar to wild-type embryos (this study). Fourth, and most unexpectedly, embryonic cells that lack Nanog and are transplanted into wild-type hosts proliferate and differentiate into derivatives of all germ layers and into germ cells (this study). Strikingly, GESTALT-mediated long-term lineage tracing revealed that these cells survive to adulthood in a wild-type host. Together, these observations indicate that Nanog has no absolutely essential autonomous role in embryonic cell differentiation and survival.

Although our results cast doubt on the previously held assumption that Nanog is directly required in embryonic cells for zygotic gene activation (Lee et al., 2013), we did find that MZnanog embryos exhibit embryonic gene expression defects during MZT and that MZnanog mutant cells display reduced miR-430 activity. Seventy-nine genes expressed at sphere stage displayed a 2-fold or greater reduction of transcript levels in MZnanog mutants. Many of these genes encode components of developmental signaling pathways. The gene set reduced in nanog mutants broadly overlapped with genes downregulated in nanog morphants (Lee et al., 2013). Thus, previously observed gene regulation defects (Lee et al., 2013; Xu et al., 2012) were not simply a consequence of delayed development in morphants. Moreover, maternally provided nanog is present throughout the embryo (Fig. 1) (Xu et al., 2012), and Nanog binds promoters of many developmental regulators in the early embryo (Bogdanovic et al., 2012; Lee et al., 2013; Leichsenring et al., 2013; Xu et al., 2012). In agreement with our study, Veil et al. (2018) found that MZnanog cells can contribute to all categorized lineages. By quantification of cell contributions, they concluded that MZnanog cells show reduced potential for contribution to the host. It is not clear whether this reduced potential is due to proliferation or differentiation defects, or an increase in cell death. Although our study shows that Nanog is not strictly required in embryonic cell differentiation, these observations suggest a non-essential contribution of Nanog to the expression of some genes during MZT and in potentiating cell differentiation.

How can the seemingly contradictory observations on the roles of Nanog be reconciled (Lee et al., 2013; Perez-Camps et al., 2016; Xu et al., 2012; Veil et al., 2018; this study)? We suggest the following model. Nanog is primarily required for the proper development of the YSL and the activation of the YSL gene expression program. These genes include patterning signals that in turn regulate gene expression in embryonic cells. Nanog is not required for zygotic genome activation and cellular differentiation but binds in conjunction with other pluripotency factors to cis-regulatory regions of embryonic genes, including miR-430, and contributes to their transcription. In this scenario, zebrafish Nanog primarily acts extra-embryonically, and its embryonic requirements might be redundant with other factors and are similar to those in mammals (Carter et al., 2014; Chambers et al., 2007; Schwarz et al., 2014), i.e. involved in but not essential for the acquisition and maintenance of pluripotency and differentiation.

TL/AB strain zebrafish (Danio rerio) were used in this study. All vertebrate animal work was performed at the facilities of Harvard University, Faculty of Arts & Sciences (HU/FAS). The HU/FAS animal care and use program maintains full Association for Assessment and Accreditation of Laboratory Animal Care International accreditation, is assured with the Office of Laboratory Animal Welfare (A3593-01), and is currently registered with the United States Department of Agriculture. This study was approved by the Harvard University/Faculty of Arts & Sciences Standing Committee on the Use of Animals in Research & Teaching under Protocol 25–08.

A TALEN pair targeting the first exon of nanog (TALEN L: TCCCGAATCTGAGCTGGC; TALEN R: TGTGACCCCGCCGGAGTGT) was generated using the TALE Toolbox (Sanjana et al., 2012). Wild-type embryos were injected at the 1-cell stage with 450 pg TALEN pair mRNA. Mutations were verified in injected embryos, and from clutches of outcrossed putative founder adults, by PCR from genomic DNA, followed by T7 Endonuclease I assay (NEB). Mutations within individual founder fish were identified by cloning PCR products followed by Sanger sequencing. An allele containing a 7 bp deletion was isolated and used for all further experiments (a166). These fish were genotyped using a PCR strategy – two allele-specific primers, in combination with a constant primer, separately identify wild-type and mutant alleles. Allele description, and primer and vector sequences can be found in Table S2.

Total RNA was isolated from embryos using EZNA Total RNA kits (Omega Biotek). cDNA was generated using iScript cDNA Synthesis kit (Bio-Rad). Open reading frames for nanog and mxtx2 were amplified by PCR from embryonic cDNA and cloned into the pCS2 vector for generation of mRNA and probes. Open reading frames for vox and bmp2b were amplified by PCR and cloned using the Strataclone kit (Agilent) for generation of probes. Plasmids encoding GFP, DsRed, ndr1, mxtx2, nanog, GFP-3xPT-miR-430 (Giraldez et al., 2005) and GFP-nanos1 (Köprunner et al., 2001) mRNAs were transcribed using mMessage mMachine kits (Thermo Fisher). Vector sequences can be found in Table S2. Antisense probes for in situ hybridization were transcribed using the DIG RNA labeling kit (Roche). All RNAs were purified using EZNA Total RNA kits (Omega Biotek). In situ hybridization was performed as previously described (Thisse and Thisse, 2008); stained embryos were cleared and imaged with a Zeiss Axio Imager.Z1 microscope.

Total RNA and cDNA were generated as above. qPCR was conducted using iTaq (Bio-Rad) on a CFX96 (Bio-Rad). Primer sequences are listed in Table S2.

Total RNA was isolated from MZnanog and wild-type embryos at 4 hpf (sphere stage) and 6.5 hpf (shield stage) (n=40 embryos each condition) following a previously published protocol (Pauli et al., 2012). RNA quality was confirmed by Bioanalyzer (Agilent). RNAseq libraries were generated using the TruSeq RNA Library Prep Kit v2 (Illumina) and sequenced on a HiSeq 2500, generating single-end 51 bp reads. Reads were aligned for each sample using TopHat v2.0.13 (Trapnell et al., 2009) with the following command ‘tophat -o <output directory> -p 16 --no-novel-juncs -G <gene table> <Bowtie2 genome index> <fastq reads>’. Transcript abundance and differential expression were determined using Cufflinks v2.2.1 (Trapnell et al., 2012) with the following command for each developmental stage ‘cuffdiff -p 16 -b <genome.fa -u -L <labels> -o <output directory> <gene table> <wild-type aligned reads .bam file> <mutant aligned reads .bam file>’. Differential gene expression plots were generated in Rstudio. Raw and processed data are available at Gene Expression Omnibus under accession number GSE89245.

YSL nuclei were labeled by injection of 1 nl of 0.5 mM SYTOX Green solution (Thermo Fisher) into the yolk cell at the 128-cell stage, as previously described (D'Amico and Cooper, 2001). F-actin was labeled using Alexa-647-phalloidin (Thermo Fisher), as previously described (Wilkins et al., 2008). Labeled embryos were mounted in 1% low melt agarose and imaged with either a Zeiss Pascal or Zeiss 880 inverted confocal microscope. YSL expression was performed through injection of mRNAs at the 4-cell stage into the vegetal yolk (Xu et al., 2012). Embryos with expression restricted to the YSL were scored and sorted at sphere stage using fluorescence from co-injected GFP mRNA. Phenotypes for all injected embryos were scored during gastrulation and at 24-30 hpf. At 24-30 hpf, classification used the following category definitions (shown in Fig. 4B): Class I, no rescue, ball of necrotic cells or exploded; Class II, axis rescue; Class III, axis rescue and tail extension; Class IV, similar to wild type.

Blastoderm transplants were performed as previously described with minor modifications (Holloway et al., 2009). Dechorionated wild-type and MZnanog embryos were injected with 250 pg of either Dextran-Alexa-488 or Dextran-Alexa-546 (Thermo Fisher). Embryos (512-cell stage) were placed in an agarose-coated Petri dish containing 1× Ringer's solution supplemented with 1.6% cleared egg whites, which aids blastoderm adherence. A pulled glass knife was used to remove the blastoderm from the yolk cell of one embryo. This blastoderm was placed onto the yolk cell of a recipient embryo using a flamed glass knife. Pressure was applied for a few seconds to adhere the donor blastoderm onto the recipient yolk cell. Embryos were allowed to recover for approximately 10 min before transfer to 1/3× Ringer's solution, and subsequent monitoring and imaging.

Non-transplanted embryos were anesthetized when necessary, mounted in 3% methylcellulose, and imaged with a Leica MZ 16 F microscope. For transplantation, donor embryos were injected, as appropriate, with either GFP mRNA or DsRed mRNA (50 pg each). Approximately 20-40 cells from a donor embryo were transplanted at sphere stage to a host embryo. Transplant host embryos were screened for fluorescence at 30 hpf with a Leica MZ 16 F microscope, then mounted in 1% low melt agarose and imaged with a Zeiss Pascal confocal microscope.

Mnanog; GESTALT donor embryos were generated from crosses of homozygous nanog^−/− female fish to homozygous Tg(ubb:DsRed-barcodev7,myl7:EGFP) male fish (McKenna et al., 2016). These donor embryos were injected with Cas9 protein (NEB) and sgRNAs targeting the v7 barcode (sequence in Table S2) (Gagnon et al., 2014). Transplantation was conducted as described above. Host embryos were grown to adulthood. Intestine, heart, eyes and brain were dissected from euthanized adults and frozen on dry ice. Adult organ gDNA was prepared using the Qiagen DNeasy kit. Sequencing libraries were prepared from each sample using two rounds of PCR to amplify barcodes and attach sample indexes and sequencing adapters (primer sequences in Table S2). Libraries were pooled and subjected to paired-end sequencing on an Illumina MiSeq, using a 500-cycle kit. After demultiplexing, reads were aligned and parsed to determine edited barcode sequences as described (McKenna et al., 2016). To filter sequencing noise, rare barcodes (present at <0.5% abundance per sample) were removed. The number of distinct edited barcodes (containing different combinations of CRISPR-Cas9 edits) present in each sample was counted. Each edited barcode was considered to be derived from the descendants of distinct clones of transplanted cells.

We thank members of the Zon lab for the Tg(fli1a:GFP) line and helpful discussions; Hazel Sive for a blastoderm transplantation protocol posted on her website; Adam Carte, Jeff Farrell, Jessica Leslie, Nathan Lord, Megan Norris, Andrea Pauli, Michal Rabani, Katherine Rogers and Maxwell Shafer for technical assistance; Marina Veil and Daria Onichtchouk for discussions and sharing their manuscript before publication; Miler Lee and Antonio Giraldez for helpful discussions; and the Harvard Bauer Core Facility, the Biopolymers Facility at Harvard Medical School and the Harvard Center for Biological Imaging.