Leapfrogging: primordial germ cell transplantation permits recovery of CRISPR/Cas9-induced mutations in essential genes

The use of CRISPR/Cas9 and TALEN programmable nucleases is revolutionizing genetic analyses and has been applied to a remarkable number of different organisms. However, the production of founder organisms carrying gene disruptions to produce mutants for loss of function (LOF) analysis has its challenges. The efficient mutagenesis of essential genes can result in lethality in the F0 generation and therefore failure to transmit mutant alleles to subsequent generations. Genetic screens in mouse and zebrafish have estimated that as many as 30% of genes are embryonic lethal (Driever et al., 1996; Haffter et al., 1996; Ayadi et al., 2012). Therefore, improvements in current genetic tools and/or manipulations to circumvent the lethality associated with mutation of essential genes would greatly accelerate progress in making mutant lines.

We and others have shown that programmable nucleases are efficient genome editing tools in the human disease model Xenopus, both in the diploid frog Xenopus tropicalis and its close allotetraploid relative Xenopus laevis (Young et al., 2011; Ishibashi et al., 2012; Lei et al., 2012; Blitz et al., 2013; Nakayama et al., 2013; Guo et al., 2014; Nakajima and Yaoita, 2015a; Wang et al., 2015). In an attempt to circumvent founder lethality, we sought to develop a method to confine targeted gene mutations to the germline, thereby ‘protecting’ somatic tissues from the deleterious effects of LOF mutations. Under such conditions, we expect that germ cells harboring specific mutations will successfully mature in healthy host animals that could transmit mutant alleles at high frequency to the F1 generation. Here, we present leapfrogging, which combines whole-embryo mutagenesis with transplantation of mutant primordial germ cells (PGCs) into wild-type sibling embryos.

Our approach was stimulated by studies in the early 1960s by Blackler and colleagues, who showed that transplantation of Xenopus posterior ventral flank from neurula or early tailbud stage embryos can confer the donor germline to recipient embryos, bolstering the idea that germ plasm-bearing cells establish the germline (Blackler, 1960; Blackler and Fischberg, 1961). We aimed to develop a more efficient transplantation procedure, combined with CRISPR/Cas9 mutagenesis, to accelerate research on identifying the functions of thousands of embryonic lethal genes. In Xenopus and other anurans, germplasm is first localized in the vegetal pole of the egg and early embryo, which is a more accessible position for both ablation and transplantation. Germline ablation has been partially or completely achieved by either vegetal UV irradiation or by cytoplasmic extrusion following pricking of the zygote's vegetal pole (Buehr and Blackler, 1970; Nieuwkoop and Sutasurya, 1979). During the early cleavages following fertilization, the germ plasm gradually coalesces into a small number of cells located near the vegetal pole (reviewed in Nieuwkoop and Sutasurya, 1979; Houston and King, 2000a). Leapfrogging combines efficient F0 embryo-wide mutagenesis with transplantation of mutation-bearing PGCs to wild-type hosts that have had their endogenous PGCs removed. We show that transplantation is readily achieved at late blastula stages when the PGCs are still in the vegetal-most domain. Leapfrogging results in efficient transmission of mutant alleles to F1 offspring, demonstrating successful transfer. We also demonstrate that embryonic lethal goosecoid (gsc) mutants can be analyzed for phenotypes in the F1 generation by intercrossing leapfrogged F0 adults. We anticipate that leapfrogging will accelerate CRISPR/Cas9- and TALEN-based genetic analyses in Xenopus and similar approaches may be adapted to a variety of organisms where programmable nucleases can be applied.

We first examined the efficacy of extirpating germ plasm-bearing cells by simple removal of vegetal explants. Late blastula (stage 9) embryos were dissected between ∼5 and 6 h post-fertilization (hpf), before the vegetal movements of gastrulation begin at 6.5 hpf (Fig. 1A). Vegetal explants and the embryo remainders (‘carcasses’) were subjected to whole-mount in situ hybridization to visualize expression of deleted in azoospermia like (dazl) mRNA (Houston et al., 1998; Houston and King, 2000b; Sekizaki et al., 2004), a marker for PGC localization (Fig. 1B). All vegetal explants contained numerous dazl expression foci (Fig. 1C), suggesting that their transplantation would efficiently shuttle donor PGCs to recipient embryos. Most carcasses (90%; 18/20) showed nearly complete removal of dazl signal (compare Fig. 1B and D, left). A small fraction of carcasses (2/20) showed a faint dazl signal (Fig. 1D, right). We conclude that our procedure can effectively remove the majority of PGCs from embryos.

Since we wished to combine CRISPR/Cas9 mutagenesis with PGC transplantation, we sought an indirect assay for determining the efficacy of mutagenesis in the PGC-containing transplanted tissues, by using the remaining embryo carcass as a proxy for these cells. However, since the diffusibility of Cas9 ribonucleoprotein complexes in the early cytoplasm might be limited, it remained possible that the carcass might not reflect mutagenesis of PGCs in the vegetal pole. Therefore, we first assessed whether mutagenesis in the carcass is a reasonable approximation of the efficacy of mutagenesis within the PGC transplant. We injected Cas9-sgRNA complexes that target the gene tyrosinase (tyr), which results in the non-lethal albino phenotype when biallelically mutated, into the animal pole at the one-cell stage. Late blastula stage 9 embryos were then dissected, subdividing the embryo into PGC explant, animal cap and the remaining endomesodermal tissues (Fig. S1). The extent of mutagenesis was assessed in these explants by direct sequencing of PCR amplicons (DSP; Nakayama et al., 2014) containing the targeted region in tyr. Sequencing traces for populations of amplicons show a mixing of peaks beginning in the vicinity of the cleavage site, providing a rough measure of mutational efficacy. We found that all three dissected domains show similar DSP traces, suggesting that animal pole injections result in efficient mutagenesis in the vegetal-most PGC-containing explants. Therefore, we routinely use DSP on carcasses to verify the efficacy of mutagenesis in transplant-bearing animals before raising them to adulthood.

Of an original 30 tyr ‘leapfrogged’ embryos, 17 successfully passed through metamorphosis to froglet stages, and the first 10 to reach sexual maturity were assayed for germline transmission of tyr mutant alleles. F0 animals were crossed with animals from a homozygous albino (tyr^−/−) population that we previously established. Since the albino phenotype is only observed in homozygous tyr-deficient (null) offspring, scoring of the F1 animals for this phenotype effectively assayed the rate of mutant alleles transmitted by the gametes of leapfrogged animals. Both male and female animals were test crossed in this manner (Fig. 2A,B) and the results from these 10 crosses (with over 3500 offspring scored) are displayed in Table 1. Six of the animals bearing leapfrog transplants, representing both sexes, showed a remarkable rate of 100% germline transmission of mutant alleles (e.g. Fig. 2C). Three of the four remaining test crosses resulted in no albino embryos whereas the fourth had 41% transmission of mutant alleles. There are several possible explanations for the cases where low tyr transmission rates were observed. First, mutagenesis in some F0 donor embryos might have been very low. Since we confirm the efficacy of mutagenesis using a sequencing-based assay, this possibility is unlikely. Second, in some cases, the removal of the endogenous wild-type PGCs may have been insufficient, prior to transplantation of mutated PGCs. Lastly, the PGC transplant tissue may have been largely or completely ‘expelled’ from the embryo as a result of insufficient healing. Despite these possibilities, the high frequency of tyr^−/− embryos generated demonstrates that leapfrogging results in efficient transfer of mutant PGCs into somatically wild-type animals.

Using albinism to assay for gametes carrying tyr alleles provided a rapid and easy high-throughput assay for germline transmission of mutant alleles, but these experiments do not demonstrate that leapfrogging permits the recovery of mutations in an essential gene. Therefore we applied leapfrogging to a gene that displays a well-known embryonic lethal mutant phenotype, gsc, as a test case. gsc encodes the Goosecoid homeodomain transcription factor, which was identified in Xenopus based on its early gastrula stage expression in Spemann's organizer (Blumberg et al., 1991; Cho et al., 1991). Morpholino antisense oligonucleotide-mediated inhibition of gsc mRNA translation in Xenopus severely reduces development of the anterior head (Sander et al., 2007). We synthesized an sgRNA targeting a sequence within the homeobox, near the splice donor site in exon 2 (Fig. 3A). Since this site is just upstream of the coding sequence for the VWFKNRR motif of the DNA recognition helix encoded by exon 3, we expected that any mutation at the target site, including single in-frame codon deletions or insertions, would disrupt proper folding of the DNA-binding domain, thereby resulting in null alleles.

Preliminary testing of the gsc sgRNA in F0 embryos showed varying degrees of loss of anterior head tissue, including cyclopia (Fig. 3B,C; data not shown) and was accompanied by high lethality. A large and reproducible population decline occurs in the second week of development (Fig. 3D). We hypothesize that these tadpoles die from starvation as a result of defects in mouth and/or pharyngeal structures or from cardiac defects (Yamada et al., 1995; Rivera-Pérez et al., 1995; Filosa et al., 1997). Interestingly, even apparently phenotypically wild-type F0 animals (presumably less mutagenized) that survive this initial lethality show reduced overall body size as froglets compared with their uninjected siblings and continue to expire as immature adults (data not shown), illustrating the challenges in raising such F0 founder animals using conventional breeding schemes.

Since mutagenesis in F0 embryos appeared to be efficient from DSP assays, we created F0 animals bearing PGC transplants from gsc CRISPR/Cas9-injected siblings. When individuals from both sexes reached sexual maturity, an intercross of leapfrogged animals was performed. Eighty-five embryos were produced from the first mating and embryos were allowed to develop to early tailbud stage for morphological and molecular analyses. We found that 73% of embryos developed with moderate to severe anterior head truncations while 27% appeared phenotypically wild type. Nine morphologically wild-type embryos and 15 presumptive mutants were fixed at early tailbud stage 29/30 and these were subjected to whole-mount in situ hybridization with a probe cocktail to detect otx2, egr2 (formerly krox20) and hoxb9 (Papalopulu et al., 1991; Lamb et al., 1993; Godsave et al., 1994; Blitz and Cho, 1995; Pannese et al., 1995). Mutant embryos show loss of the anterior portion of the otx2 expression domain without effects on more posterior neural expression (Fig. 4A-C; Fig. S2) in a pattern virtually identical to that previously observed in gsc morpholino knockdown experiments in X. laevis (Sander et al., 2007).

A second mating using the same pair produced 370 embryos, which were grown to tailbud stage 40 to assess the extent of anterior truncation morphologically (Fig. 4D-F). This second mating produced a similar ratio of 68% mutant to 32% wild type. Interestingly, of the 252 embryos showing head truncations, 135 were cyclopic (Fig. 4E) whereas 117 were headless (Fig. 4F). Since our mutagenesis strategy targeted the DNA-binding domain to ensure LOF, the most parsimonious explanation is that the LOF effect on anterior development is variably penetrant.

To correlate the phenotype with mutations at the gsc target site, we subsequently genotyped both phenotypically wild-type and mutant embryos. Nearly all (15/16) phenotypically wild-type embryos were heterozygotes, with one being homozygous wild type (Fig. S3). In contrast, 100% of phenotypically mutant embryos carried biallelic gsc mutations with indels around the target site (Fig. S3). Because all of the observed mutations disrupt the coding sequence of the Gsc homeodomain (Figs S3 and S4), and all but one remove the DNA recognition helix entirely, we anticipate these would abrogate DNA binding, thereby resulting in null alleles. We conclude that there is a 100% correspondence between biallelic gsc LOF genotypes and mutant phenotypes.

In summary, leapfrogging mitigates founder lethality when targeting essential genes and a high percentage of homozygous mutant F1 animals can be generated for subsequent characterization of non-mosaic LOF phenotypes.

Methods for creating mutant lines are needed that can mitigate lethality in F0 animals when mutating essential genes while retaining efficient germline transmission. Several potential solutions to this problem have been published. In rodents, the introduction of programmable nucleases, performed either in vivo or in vitro, into adult spermatogonial or oogonial stem cells has been accomplished (Fanslow et al., 2014; Chapman et al., 2015; Sato et al., 2015; Takahashi et al., 2015; Wu et al., 2015). These mutated germ cell precursors variably contribute to the germline while endogenous unmutated germ cells remain. Significant technology development, especially to target both sexes, would be required to successfully apply these methods to new systems. A second approach is to enrich Cas9 or TALEN mRNAs in the germ plasm using 3′ UTRs derived from germ plasm-localized mRNAs such as nanos1 or ddx25 (Moreno-Mateos et al., 2015; Nakajima and Yaoita, 2015b). Since successful partitioning to the germ plasm requires careful titration to minimize targeting of somatic nuclei, the frequency of germline transmission is highly variable. The use of 3′ UTRs to drive Cas9 mRNA into germ cells has not been demonstrated in Xenopus, but has been used in zebrafish (Moreno-Mateos et al., 2015). However, we speculate that the use of highly efficient doses of Cas9-sgRNA complexes or efficient doses of Cas9 mRNA plus sgRNA (see Nakayama et al., 2014) might have an additional advantage over titrated low doses of Cas9-3′UTR fusion mRNAs. The former approach is expected to yield mutations at earlier stages of development (Bhattacharya et al., 2015), resulting in a smaller diversity of alleles being transmitted in the F0 leapfrogged germline. In our gsc leapfrogging experiment, we recovered 10 different mutant alleles from the 37 F1 embryos analyzed (Fig. S3). The most parsimonious explanation for the allelic combinations found is that the number of mutant alleles carried by each parent is probably between 4 and 8 (see Fig. S3C). It would be valuable to directly compare the efficacy of each approach in the future by creating F0 lines.

While we used CRISPR/Cas9 to mutagenize PGCs, we believe that TALENs can also be successfully applied to generate efficient mutagenesis using leapfrogging. The use of PGCs allows for a single technology to produce both male and female F0 animals with a high percentage of their donor-derived gametes bearing mutant alleles. Since the recipient of these modified PGCs is somatically wild type, transplant-bearing animals are viable and only carry mutations in germ cells. Intercrossing of leapfrogged animals is capable of producing a high frequency of mutants. Our experiments with tyr suggest that germline transmission rates can frequently be 100%. This high efficiency of germline transmission is especially advantageous if one needs large amounts of biological material (i.e. egg or embryo extracts and ChIP-seq, where genotyping of hundreds to thousands of individual embryos would be prohibitive), from mutant embryos. This is a major strength of the Xenopus system. Since leapfrogging produces non-mosaic phenotypic animals in the F1 generation, this method circumvents a full generation of time-consuming and laborious screening (e.g. test crosses of numerous F1 animals to find those showing germline transmission) required with the F2 analysis of the more standard breeding regimen.

It is useful to note that leapfrogging also offers a new method to study the function of maternally expressed genes. Recent studies have shown that maternal RNAs are present until late gastrulation (Owens et al., 2016) and maternally deposited proteins may also persist until at least tailbud stage 33 (Peshkin et al., 2015). This indicates that maternal gene products might play a much larger role in embryonic development than previously recognized, and that a great deal of biology in the early embryo may be refractory to F0 analysis using genome editing, therefore requiring genetic crosses. A successful approach to LOF analysis of maternal gene products utilizes the host transfer method (Heasman et al., 1991; Olson et al., 2012). Stabilized antisense oligonucleotides act in conjunction with endogenous RNaseH in oocytes to target and degrade maternal RNAs during in vitro oocyte maturation. Manipulated oocytes are surgically reintroduced into adult females that provide the jelly coat needed for fertilization. While this approach has been successful in revealing the functions of maternal gene products, it can be technically challenging and often produces incomplete knockdowns. In summary, leapfrogging provides a genetic method for F1 analysis of maternal gene knockouts. Additionally, only a leapfrogged female is required for these analyses as the maternal products are provided in the eggs. Fertilization with wild-type sperm would permit assessment of loss of maternal gene function. We note that in cases where the targeted gene is required for the production of germ cells, leapfrogging might be difficult or impossible to achieve, whereas the host transfer method may be more advantageous as it allows for knockdowns after oogenesis is complete.

To achieve a high percentage of mutant (LOF) alleles in the germline of F0 animals by leapfrogging, there are a number of important considerations. First, it is essential that a high percentage of genomes in the donor PGCs bear mutations in the targeted gene. A number of recommendations for efficient CRISPR/Cas9-mediated mutagenesis in Xenopus tropicalis are published (Nakayama et al., 2014). We find that the most critical determinants for efficient CRISPR/Cas9 mutagenesis in Xenopus are injection of optimized quantities of Cas9-sgRNA complexes, combined with the selection of efficient sgRNAs. Careful dose optimization and preliminary testing of a handful of sgRNA target sites is recommended prior to transplantation experiments.

Second, careful determination of the location of sgRNA target sites within a gene is essential. A common strategy is to target a site near the 5′ end of the ORF, with the expectation that indels will result in frameshift mutations to block protein translation. While this might be a good strategy for making mutants via the standard three-generation breeding scheme, this approach is poorly suited for leapfrogging. Two-thirds of the alleles from repair of double-strand breaks within an ORF are expected to result in frameshifts, but one-third of the alleles will be in-frame indels (usually small deletions). This expectation has been experimentally verified on a large scale in zebrafish embryos, where 68% of indels were found to result in frameshifts and 32% were in-frame deletions, with >75% being smaller than 12 bp deletions (Varshney et al., 2015). Therefore, a significant number of mutant alleles are expected to have either wild-type levels of activity or be hypomorphic, and thus not complete LOF. F1 animals resulting from such F0 intercrosses would be ineffective for displaying mutant phenotypes. For leapfrogging to be maximally successful, all mutant alleles need to be strong LOF mutations. We recommend selecting sgRNA target sites within the coding regions of recognizable protein folding domains so that even in-frame indels will display defects. In the case of gsc, by targeting within the DNA-binding domain, in-frame deletions will result in misfolding and loss of DNA binding capability. Therefore, careful target choice results in ∼100% of indels being LOF mutants. This principle of targeting folded protein domains can be applied to other classes of protein-coding genes, an approach validated by a recent study (Shi et al., 2015). Numerous sites were targeted across a number of chromatin regulatory genes to scan for optimal loss of protein function and a higher proportion of null mutations were found when targeting folded domains.

Third, for successful leapfrogging, efficient removal of PGCs from the recipient embryo must also accompany transplantation of well-mutagenized donor PGCs. When we examined our PGC extirpation performed at blastula stages, a minority of dazl-stained foci were found (Fig. 1). These remaining wild-type PGCs might contribute to the variable rates of transmission of mutant alleles that we see. An alternative interpretation is that the deep cells positive for residual dazl expression in the vegetal yolk mass (Fig. 1B) may not represent future PGCs. It has been suggested that microRNAs contribute to the clearing of germ plasm transcripts from somatic cells that do not contribute to the PGCs (Koebernick et al., 2010; Yang et al., 2015). In the future it would be interesting to deplete PGCs in host embryos to determine whether this results in increased frequency of mutant alleles in the ‘leapfrogged’ germline. This can be achieved by morpholino knockdowns of ddx25 (also known as deadsouth), nanos1 or dnd1 (also known as dead end), all of which have been shown to deplete PGCs (Horvay et al., 2006; Lai et al., 2012; Yamaguchi et al., 2013). This approach may permit the use of smaller tissue fragments for transplantation, thus further minimizing the likelihood of causing physical damage caused by transplantation, and also reducing endodermal carryover.

Lastly, we recommend producing a number of transplant-bearing animals, both males and females, so that these sufficient numbers are available to survive to sexual maturity. In our various transplantation experiments, 50-75% of leapfrogged embryos survive to adulthood. Further investigation into the size of transplants required for successful leapfrogging should be tested as smaller transplants might improve survival rates. Modification of other conditions for transplantation might increase our success further. For example, recent testing of conditions for successful PGC transplantation in Xenopus laevis revealed that a higher ‘strength’ embryo culture solution [1× Marc's modified Ringer solution (MMR)] is required for efficient healing (our unpublished observations) than we used in the current study for Xenopus tropicalis.

Combining programmable nuclease-mediated genome editing with transplantation of PGCs should be applicable to many other animals. PGC transplantation has been performed in a number of species and therefore the only barrier to using leapfrogging is an efficient method for delivery and mutagenesis by programmable nucleases. It is likely that transplantation of PGCs can be performed successfully in other animals that (like Xenopus) have a maternally localized germ plasm, such as the sturgeon (Saito et al., 2014), which is a fish of economic importance. Teleost fish such as zebrafish, medaka, goby and others also have maternally deposited germ plasm, although it is initially not confined to a single location (Yoon et al., 1997; Miyake et al., 2006; Herpin et al., 2007). However, PGCs have been both directly transplanted between zebrafish embryos and also after growth in cell culture (Ciruna et al., 2002; Kawakami et al., 2010), making leapfrogging feasible in this model system. In other animals, the PGCs are not maternally derived but instead are specified by an inductive mechanism. PGC transplantation has also been demonstrated in several of these species. For example, transplantation of ventral marginal zone between early gastrula stage embryos in caudate amphibians (salamanders and newts) such as the Mexican axolotl Amblystoma mexicanum results in transfer of the germline between animals (Nieuwkoop, 1947; Smith, 1964; Chatfield et al., 2014). Leapfrogging might also be viable in birds. PGCs have been transplanted from cultured early chick extraembryonic tissue and can contribute to the germline in recipients (van de Lavoir et al., 2006; Nakamura et al., 2013). We envision combining this technology with programmable nucleases (Park et al., 2014; Véron et al., 2015) to make leapfrogging possible in this and a number of other species.

Vegetal explants were prepared from X. tropicalis embryos between early stage 9 and the first appearance of bottle cells at stage 10. These developmental stages, using Nieuwkoop and Faber staging for Xenopus laevis (Nieuwkoop and Faber, 1967), are roughly 5-6.5 hpf when embryos are cultured at 24-25°C (Owens et al., 2016). Explants were prepared using eyebrow hair knives and hair loops and dissections were performed in agarose-coated 60 mm plastic dishes in 0.3× MMR (Sive et al., 2000)containing gentamycin. Embryos were first dechorionated with Dumond watchmaker's forceps and explants measured ∼0.4-0.45 mm in width and ∼0.25-0.3 mm in depth, which corresponds to approximately 1/2 to 2/3 the distance from the vegetal pole to the blastocoel floor. These dimensions were used to maximize extirpation of germ plasm-bearing cells from recipient embryos. Vegetal tissue was first dissected from recipient embryos, which were set aside while graft donor embryos were dissected. Once donor tissues were isolated, they were quickly placed into the recipient's open vegetal wound with the interior surface of the graft being placed into the wound. Gentle pressure was applied to place the graft securely and the recipient embryo was moved to a well cut into the agarose to allow for healing while the next transplantation was being performed. Donor embryo ‘carcasses’ were similarly moved to wells. Typically ∼12 transplants can be accomplished during the 90 min time window afforded. Once grafts had healed into place, embryos were carefully placed, vegetal pole up, into individual agarose-coated wells in 12-well plates containing 1/9× MMR plus gentamycin. Donor carcasses were also moved to individual wells in the 12-well plates and care was taken to keep track of carcasses and the corresponding embryos receiving grafts. All embryos were subsequently cultured overnight in a 25°C incubator. Carcasses were typically homogenized in proteinase K-containing lysis buffer the following day to provide material for DSP analysis (Nakayama et al., 2014) to verify the success of CRISPR/Cas9 mutagenesis in DNA of the corresponding grafts.

Synchronous embryos were obtained by in vitro fertilization, dejellied at 10 min post-fertilization by mild agitation in 3% cysteine (pH 7.6-7.8) and transferred to agarose-coated plates containing 1/9× MMR. Prior to injections, a cocktail of Cas9 protein (PNA Bio) and tyr sgRNA (Blitz et al., 2013) was prepared by preincubating the sgRNA at 60°C for 5 min. This was quick-cooled on ice and 1 μl Cas9 (1 μg/μl in 20 mM HEPES pH 7.5, 150 mM KCl and 1% sucrose) was added. The cocktail was mixed by gentle tapping and then incubated at 37°C for 10 min. One-cell stage embryos were microinjected into a single site at the animal pole with 4 nl of Cas9/sgRNA cocktail, receiving a total of 1 ng Cas9 and 250 pg tyr sgRNA. Embryos were moved to agarose-coated plates and cultured in 1/9× MMR at 25°C until early stage 9 (∼4.5 hpf). Embryos were dissected between ∼5 hpf and ∼6.5 hpf (stage 10).

To create a template for transcribing gsc targeting sgRNA [target site: CCTCAGAGAGGAAAAAGTAGagg, with the protospacer adjacent motif (PAM) in lower case], we designed two overlapping oligonucleotides as follows. A ‘top strand’ 62 nt oligo, 5′-TAATACGACTCACTATAGG-[CCTCAGAGAGGAAAAAGTAG]GTTTTAGAGCTAGAAATAGCAAG-3′, was designed containing a T7 RNA promoter (underlined), followed by the target sequence (in brackets and without the PAM) and 23 nt of the guide backbone. A second universal bottom strand oligo of 80 nt in length, 5′-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3′ with 23 nt overlap to the top oligo was used in the template assembly reaction as previously outlined (Nakayama et al., 2014). Following synthesis, the reaction was phenol/chloroform extracted, precipitated and resuspended in DEPC-treated H₂O. Guide RNA synthesis utilized 40-50 ng template in a 20 μl T7 Megascript (ThermoFisher) in vitro transcription reaction overnight at 37°C. The reaction was treated with Turbo DNaseI (ThermoFisher) and then phenol/chloroform extracted, precipitated with ammonium acetate and isopropanol, and resuspended in DEPC-treated H₂O. For efficient mutagenesis, we used 1 ng Cas9 protein (PNA Bio CP-01) and 1.2 ng of gsc sgRNA/embryo, which was precomplexed as described above. To perform survival comparisons between gsc and tyr sgRNA-injected embryos we used 1.2 ng tyr sgRNA/embryo.

Embryo lysis and DSP assays were performed as described (Nakayama et al., 2014) using proofreading Pfx Platinum DNA polymerase (Invitrogen). Oligonucleotides for tyr amplications and sequencing were previously described (Blitz et al., 2013). For genotyping of gsc mutants, we used a combination of DSP assays and sequencing of individual cloned PCR products. Oligos for gsc target region amplification and sequencing were 5′-CCACACACATAAAGCTCCACAT-3′ and 5′-ACACATTTGGGCCCTGGGTA-3′. Following PCR amplification, amplicons were cloned using a Zero Blunt Topo Cloning Kit (Invitrogen). Individual clones in pCRBluntII-TOPO were sequenced at Genewiz.

To create digoxigenin-labeled probes, cDNA/EST and genomic sequence information for X. tropicalis was retrieved from Xenbase (www.xenbase.org; RRID:SCR_003280; Karpinka et al., 2015). Oligonucleotides were designed to PCR amplify an 877 bp fragment of dazl from genomic DNA with a bacteriophage T7 promoter added to the 5′ end of the ‘reverse’ strand as follows: forward strand, 5′-GGACGATAGTGTGCACCAATTCA-3′; reverse, 5′-GCAGCTAATACGACTCACTATAGGACCACAGATTGCCCAGTGCT-3′. The T7 promoter is underlined and a 5 bp 5′ extension (Nakayama et al., 2014) was added to enhance in vitro transcription during riboprobe synthesis. This dazl DNA template was amplified from X. tropicalis liver genomic DNA using a touchdown strategy using Pfx polymerase (Invitrogen) with the following conditions: 94°C for 5 min, followed by 13 cycles of 94°C for 20 s, 65°C annealing for 20 s, 68°C extension for 1 min, with each cycle's annealing step decreasing the temperature by 0.5°C. This segment was followed by 30 cycles of 94°C for 20 s, 58°C annealing for 20 s, 68°C extension for 1 min, followed by a 5 min extension at 68°C. Similarly, templates were prepared by genomic PCR for otx2, egr2 and hoxb9. PCR oligos were as follows. otx2: forward, 5′-CAGCAACAGCAGCAGCAGAA-3′; reverse 5′-GCAGCTAATACGACTCACTATAGTTGCCAGATCCAGGGGAAAA-3′. egr2: forward, 5′-GCGATCGCTGGATTTCTCCT-3′; reverse, 5′-GCAGCTAATACGACTCACTATAGGCACTTGTGCCCAAGCATTC-3′. hoxb9: forward, 5′-AACCCCTCAGCCAActggtta-3′; reverse, 5′-GCAGCTAATACGACTCACTATAGAAAGCGAGGGCGTTTCTTGT-3′. Probe lengths were 1.6 kb (otx2), 1.5 kb (egr2) and 0.7 kb (hoxb9), respectively.

Whole-mount in situ hybridization was carried out according to Harland (1991) with modifications (Blitz and Cho, 1995). In addition, the protease permeabilization step used 2.5 μg proteinase K (Roche)/ml PTw (PBS containing 0.1% Tween 20) for 5 min. Hybridization steps were performed at 65°C and post-hybridization RNase digestion employed 1 μg RNaseA/ml and no RNase T1. Pigment was bleached post-staining according to Mayor et al. (1995). dazl-stained embryos were photographed after clearing in Murray's solution (2 benzyl benzoate: 1 benyzl alcohol). otx2, egr2 and hoxb9 triple-stained embryos were photographed in methanol.

Sexually mature animals were primed with 10 units of human chorionic gonadotropin (HCG; Chorulon) within a few days prior to boosting, which employed 100 units of HCG. Frogs were placed in 1/9× MMR and allowed to amplex. Frogs were removed the following day, gentamycin was added to the 1/9× MMR and embryos were kept at 25°C until hatching. Tadpoles were moved to clean 1/9× MMR with gentamycin until scoring for albinism after stage 41 or as described in the text for gsc phenotypes. Institutional IACUC guidelines were followed for all animal care and experimentation.

Two important studies were recently published that demonstrate the viability of leapfrogging-like strategies in chicken (Oishi et al., 2016; Dimitrov et al., 2016).

The authors would like to thank Matt Guille and the European Xenopus Resource Center (RRID: SCR_007164) for an albino female, Takuya Nakayama, Arul Subramanian and Aaron Zorn for many valuable discussions. I.L.B. dedicates this paper to Dr Herman E. Brockman, an inspirational teacher, mentor, and tireless advocate for molecular genetics.