|Home | About | Journals | Submit | Contact Us | Français|
We present simple and efficient methods for creating heritable modifications of the zebrafish genome. Precisely modified alleles are generated by homologous recombination between the host genome and dsDNA donor molecules, stimulated by the induction of chromosomally targeted DSBs. Several kilobase-long tracts of genome sequence can be replaced. Tagging donor sequences with reporter genes that can be subsequently excised improves recovery of edited alleles by an order of magnitude and facilitates recovery of recessive and phenotypically silent conditional mutations. We generate and demonstrate functionality of: i) alleles with a single codon change, ii) an allele encoding an epitope-tagged version of an endogenous protein, iii) alleles expressing reporter proteins, and iv) a conditional allele in which an exon is flanked by recombinogenic loxP sites. Our methods make recovery of a broad range of genome editing events very practicable, significantly advancing applicability of the zebrafish for studying normal biological processes and modeling diseases.
The zebrafish has been established as a formidable experimental system for the discovery and analysis of gene functions regulating a wide spectrum of developmental and physiological processes as they occur in the intact organism (Hammerschmidt and Mullins 2002; North et al. 2009; Ceol et al. 2011; Kikuchi and Poss 2012; Wolman and Granato 2012; Barbosa et al. 2015; Ruprecht et al. 2015). The near transparency of the embryo has even permitted visualization and analysis of the dynamics of signaling molecules in intact tissues (Yu et al. 2009; Muller et al. 2012). Furthermore, the extent of functional conservation among vertebrates makes the zebrafish an outstanding system for examining the mechanistic basis of disease (Jurynec et al. 2008), assaying human gene activity (Bamford et al. 2000; Jou et al. 2013), or discovering compounds that augment or inhibit developmental processes or disease-like conditions (Peterson et al. 2004; North et al. 2007; Hagedorn et al. 2014). The sophistication with which biological processes can be probed in the zebrafish make it poised to exploit tools of genome manipulation, which could be used to generate models of human disease, study gene function in specific tissue or temporal contexts, or generate alleles that allow study of tagged protein activities under physiological conditions. Whereas early efforts have demonstrated the feasibility of genome editing in the zebrafish (Bedell et al. 2012; Hruscha et al. 2013; Zu et al. 2013; Auer et al. 2014; Irion et al. 2014; Kimura et al. 2014; Shin et al. 2014; Hisano et al. 2015), available methods are inefficient and limited in their ability to allow recovery of genome modifications, such as conditional alleles or novel recessive alleles, which do not confer an immediate phenotype (Auer and Del Bene 2014). As a result there is not yet an established procedure for the routine recovery of edited genomes. To more fully utilize the potential of the zebrafish, we have developed an easy-to-apply approach for creating and recovering genome modifications that facilitate the study of normal and disease gene function and analyses of biological processes in vivo.
The development of synthetic sequence-specific DNA nucleases (also called programmable nucleases) that are easy to generate and implement opened a new realm of possibilities for genome modification (Hsu et al. 2014; Kim and Kim 2014). Double strand breaks (DSBs) trigger host-cell repair responses that can be utilized to produce altered alleles: in the absence of a template guiding repair, the lesion is likely to be healed via the error-prone Non-Homologous End-Joining pathway; in the presence of single- or double-stranded DNA sharing homology with the targeted locus, the template can either guide repair or participate in an exchange of sequences via homologous recombination (HR). The induction of chromosomal DSBs greatly stimulates recombination between the targeted locus and available homologous DNA sequences (Jasin 1996; Paques and Haber 1999). Cleavage of chromosomal loci by synthetic nucleases can be highly efficient, with ≥50% of the genomes in a developing zebrafish embryo experiencing a targeted DSB (Dahlem et al. 2012). Thus targeted DSBs can be harnessed to stimulate genome modification through homology-directed repair of induced lesions or true HR exchange events with exogenously supplied templates.
Here we describe a comprehensive approach to accurate genome editing in the zebrafish that involves targeted HR, allowing tracts of genomic sequence to be precisely replaced with donor sequences. We also introduce methods that greatly increase the efficiency of inducing and recovering precisely edited alleles. Typically 5 - 15% of the treated animals transmit edited alleles through their germ lines. The identification and recovery of edited alleles is made extremely simple and efficient by temporarily tagging donor sequences with reporter genes whose acquisition initially marks the edited allele and which can be subsequently excised. The tools and approaches we report make it possible to generate and rapidly recover most types of desired gene modifications. We illustrate application of these methods by generating different types of edited alleles at four distinct genome sites, including designed coding changes, alleles encoding antigenically tagged versions of endogenous proteins expressed under physiological conditions, alleles that drive expression of introduced proteins under the control of endogenous promoters, and conditional ‘floxed’ alleles.
In our approach to genome editing in the zebrafish, DSBs were generated continually at a locus of interest during early developmental stages of a zebrafish embryo while exogenous double stranded DNA (dsDNA) donor sequences were provided as a template to guide repair of the breaks (Zu et al. 2013; Irion et al. 2014; Shin et al. 2014). Targeted cleavage of the genome was accomplished by injecting just-fertilized zygotes with a pair of mRNAs encoding a heterodimeric TALEN (Dahlem et al. 2012) or with CRISPR/Cas9 components (Charpentier and Doudna 2013; Hwang et al. 2013; Jao et al. 2013) designed to cleave a unique sequence within the genome. Nuclease activity in vivo was assessed by detection of targeted mutations in the genomes of 1 day post-fertilization (dpf) injected embryos (Dahlem et al. 2012); only TALENs or sgRNAs that induced targeted DSBs in every embryo were selected to trigger HR (Table S1 lists target sites and the mutagenesis efficiency of each TALEN). Donor template was provided by simultaneous injection of plasmid or plasmid-derived dsDNA. Donor templates consisted of novel sequences flanked by homology arms of approximately 1 kbp identical to the host sequences bordering the nuclease target site. Animals with minimal sequence heterogeneity at the targeted locus were selected as hosts (Experimental Procedures). Donor sequences were always designed so their incorporation would destroy the nuclease recognition site in the genome and introduce novel diagnostic restriction sites or primer-binding sequences. Hence successfully targeted loci could not be re-cleaved by the nuclease and could be distinguished unambiguously by PCR anchored at unique donor sequences and host genome-specific sequences distal to the homologies. Following injection of zygotes, genomic DNAs of individual embryos were always analyzed by diagnostic PCR (Table S2 lists primers). These analyses indicated virtually all founder embryos harbored correctly edited alleles. The relative abundance of edited alleles in treated F0 embryos was used as an assay to optimize editing conditions. In all cases reported here, the configuration of germ-line transmitted edited alleles was verified by complete sequence analysis of single amplicons that bridged the host sequences that lay distal to the homologies and thus extended across all recombination junctions and newly introduced sequences.
As previous studies (Dahlem et al. 2012) indicated DSB cleavage activity commences about 2 hours after injection of TALEN mRNA into zebrafish zygotes, we anticipated DSB-stimulated HR would occur in a mosaic fashion during multi-cellular stages of embryonic development. To detect such events, we developed a simple assay in which individual cells with edited alleles could be readily recognized in the somatic tissue of developing embryos. Embryos homozygous for the null goldenb1 mutation have severely reduced pigmentation at 2-3 dpf (Streisinger et al. 1981). As golden (gol) functions cell autonomously to promote pigmentation, we measured the ability to convert the golb1 allele, which encodes a premature termination product (Y151X) due a C→A mutation at the 3′ end of exon 5 (Lamason et al. 2005), to a wildtype form (Figure 1A). Recombination/repair was stimulated by injecting golb1 mutant zygotes with gol-int5 TALEN mRNA to generate a DSB within intron 5, approximately 30 bp 3′ of the b1 mutation. Donor sequences were provided by co-injection of 50pg gol(b1→WT) donor plasmid DNA harboring modified gol locus sequences that would restore WT coding in exon 5 and introduce nearby intronic changes destroying the TALEN target site and creating a diagnostic NotI restriction site. The modified donor sequences were flanked by perfect homology arms (1 kbp left arm and 1.1 kbp right arm) derived by PCR from the gol locus of the targeted genome. Whereas injection of TALEN mRNA or donor plasmid alone did not yield embryos with pigmentation at 2 dpf, nearly every embryo (95%, n = 368) with both TALEN and donor plasmid had normally pigmented cells (Figure 1B, Table 1). PCR analysis indicated virtually all targeted embryos had acquired donor sequences at the gol locus (Figure 1C).
To determine if founders transmitted WT alleles to offspring, injected F0 founder embryos were raised to adulthood and mated with golb1/b1 partners. Eight of 51 (16%) F0 founders produced wholly pigmented offspring, indicating transmission of gol+ gametes (WT gametes represented 0.6 - 23.1% of germ lines, μ = 6.6%, Table S3). Sequence analysis of the 8 transmitted golb1→WT alleles (Figure S1) indicated 7 had precisely replaced genomic sequences with donor sequences, whereas the eighth also suffered a 7 bp deletion at the TALEN cut site within intron 5. It appears all WT alleles arose through precise HR events, but one of these also experienced the kind of mutagenic event typical of Non-Homologous End-Joining repair. In sum, gene editing in which a small number of nucleotides replace the endogenous sequence via HR can be triggered to occur efficiently in zebrafish.
CRISPR/Cas9 components can also be used to initiate DSBs that significantly stimulate HR. Mutant embryos with normally pigmented melanophores were readily produced following co-injection of gol mutant fertilized eggs with gol(b1→WT) donor plasmid DNA, Cas9 protein, and sgRNA targeted to the gol-int5 TALEN recognition site (Figures 1A and 1B). Although it is difficult to use the two approaches to induce genomic DSBs at precisely the same site and with precisely the same efficiency, it appears DSBs created by CRISPR/Cas9 components or TALENs are similarly capable of stimulating HR events (Figures 1B and 1D).
The replacement of short stretches of sequence at the gol locus indicated it should also be possible to introduce defined coding sequences at a precise location. Given the relative dearth of proven functional antibodies against zebrafish proteins, one application of sequence insertion would be to generate a modified allele encoding an antigen-tagged version of a protein in order to track an endogenous protein expressed under native conditions and/or recover interacting DNA or protein partners. We used DSB-stimulated HR to introduce sequences encoding the V5 epitope in frame immediately downstream of the AUG translation initiation codon of the no tail (ntl) gene. ntl is an essential gene required for development of the notochord and posterior mesoderm (Halpern et al. 1993). It encodes a T-box transcription factor expressed in the primitive mesoderm and the nascent notochord (Schulte-Merker et al. 1994). Previous work indicated additional proteins likely interact with No Tail to modify its transcription function (Goering et al. 2003).
The strategy for editing the ntl locus is illustrated in Figure 2A. To stimulate HR, the ntl-ex1 TALEN was engineered to target about 45bp downstream of the ntl translation start site. Donor sequences encoded the V5 epitope fused to the normal N-terminus of No Tail; additional silent codon changes were engineered so that incorporation of the donor would render the edited ntl allele resistant to further TALEN activity. The novel sequences in the donor plasmid were flanked by perfect homology arms of 1 and 1.2 kbp.
All embryos co-injected with TALEN mRNA and circular donor plasmid harbored HR products as detected by PCR using primer pairs specific to donor and host genome sequences (Figures 2A and S2A). A V5-epitope-modified ntl allele was recovered from the germ lines of founders and established in a true-breeding line. F2 ntlV5/+ embryos expressed nuclear V5-tagged protein specifically in the ntl expression domain (Figures 2B-E). To test if the V5-tagged No Tail protein provided wildtype activity, ntlV5/+ heterozygotes were intercrossed and their progeny analyzed. All progeny, including ntlV5/V5 homozygotes, appeared morphologically normal (Figures 2F and 2G) and were viable to adulthood (not shown). Thus genome editing by HR allows for production of modified alleles in which epitope-tagged proteins, detectable by common commercially available antibodies, are expressed under conditions that truly mimic native proteins.
We investigated factors that affect HR. We found genome editing is quite sensitive to the degree to which DSBs are induced at the target locus (Jasin 1996; Paques and Haber 1999). In the absence of induced DSBs, we failed to detect recombination between genomic sequences and homologous donor DNAs (Figures 1C and S2). To test the sensitivity to DSB-induction, we generated pairs of TALENs with identical DNA binding motifs but different DNA cleavage domains: one harboring the first-generation heteromeric DD/RR version of the Fok I nuclease domain (Miller et al. 2007), the other harboring the more active derived heteromeric DDD/RRR version of the nuclease domain (Doyon et al. 2011). As illustrated by analysis of mutagenesis at the gol locus (Figure S3A), the DDD/RRR form reproducibly induced higher levels of mutations. The ability of DD/RR or DDD/RRR variants of the gol-int5 TALEN to stimulate gene editing of the golb1 mutation was assayed as described in Figure 1. Precise editing events were more frequent among the genomes of embryos injected with the DDD/RRR variant (Figure S3B), demonstrating the efficiency with which DSBs are induced to stimulate HR notably affects the frequency of gene editing events. Using the induction of targeted mutations as a proxy measurement for the efficiency with which DSBs were induced, the TALENS used in the experiments reported here induced DSBs in 23-86% of the genomes of injected F0 embryos (Table S1).
The ability of intact plasmid circles, linearized plasmids, or purified dsDNA fragments containing only donor sequences and flanking homology arms to function as donor molecules for HR was analyzed. Under our experimental conditions, plasmid circles always proved most efficient as donor molecules (Figure 1C, Table 1, and data not shown). Although the reason for these differences is not clear, similar findings have been made in Drosophila (Carroll and Beumer 2014) and in some zebrafish studies (Irion et al. 2014). We hypothesize linear molecules may be more prone than intact circles to degradation or concatenation, either of which might affect the availability of donor sequences for HR. Noting that linear molecules generated by cleavage with the I-SceI meganuclease are excellent substrates for random insertion and tend to produce single copy transgenes in zebrafish (Grabher et al. 2004), we tested whether linear donor molecules generated by I-SceI cleavage might function well for gene editing purposes. As I-SceI enzyme cleaves its target site asymmetrically and associates in a relatively stable manner with the longer digestion product (Perrin et al. 1993), we generated a series of pKHR donor vectors (Figure S4) containing a central domain that accepts donor sequences flanked by a pair of head-to-head oriented I-SceI sites. The effect of I-SceI-digestion on the efficiency of gene editing was tested at the zebrafish kcnh6a locus.
The kcnh6a gene encodes a potassium channel expressed exclusively in the heart that is critical for normal cardiac excitability (Arnaout et al. 2007). Mutations in the human homologue KCNH2/HERG cause inherited arrhythmia disorders associated with sudden cardiac death (Splawski et al. 2000). Cardiac activity in zebrafish embryos lacking kcnh6a function can be rescued by expression of wildtype human KCNH2 RNA but not RNA encoding pathogenic KCNH2 variants associated with Long QT Syndrome (Jou et al. 2013). To further develop this bioassay, we chose to produce zebrafish lines in which fluorescent reporter proteins would be expressed from the kcnh6a locus in lieu of its normal protein product. The strategy for editing the kcnh6a locus is shown in Figure 3A. Two donor constructs were generated in pKHR4 with the purpose of introducing a fluorescent reporter protein into the kcnh6a locus. In each, either eGFP or mCherry coding sequences followed by translation and transcription termination motifs were flanked by perfect homology arms of about 1 kbp in such a manner that HR between each arm and the host genome would introduce the reporter sequence in frame just downstream of the kcnh6a AUG translation initiation codon. Recombination was stimulated with the kcnh6a-int1 TALEN designed to induce a DSB in intron 1, about 150 bp 3′ of the initiation codon. Donor DNA was injected into 1-cell zygotes with or without kcnh6a-int1 TALEN mRNA or I-SceI enzyme. Integration of reporter sequences at the kcnh6a locus of injected embryos was detected by PCR amplification with diagnostic primers (Figures 3A and 3B).
Without TALEN activity to cleave the target locus and stimulate HR, neither gfp nor mcherry sequences were detected at the kcnh6a locus. Following induction of DSBs, although intact plasmid DNA served as an adequate substrate for HR events, in experiments with two different donor plasmids, co-injection of donor plasmid DNA with I-SceI enzyme further stimulated the generation of kcnh6a locus-specific integration events (Figure 3B). As demonstrated by quantitative assessment of edited alleles in injected F0 embryos, the enhancement of genome editing events was dependent on the amount of I-SceI enzyme digestion and could be accomplished by digestion of donor DNA with I-SceI enzyme in vitro, prior to injection of zygotes (Figure 3C).
Embryos injected with both kcnh6a(eGFP) donor plasmid and I-SceI enzyme were grown to adulthood, and germ-line transmitted alleles were recovered from 2 of 14 (14%) F0 founders. The resulting kcnh6aGFP alleles were amplified from the genomes of F1 embryos and sequence analysis confirmed that donor sequences had been perfectly integrated as expected from HR.
kcnh6aGFP/+ heterozygous embryos were fully viable and expressed the GFP reporter protein exclusively in the heart (Figures 3D and 3E). F2 embryos homozygous for the knock-in allele were phenotypically indistinguishable from previously described kcnh6a null mutants (Arnaout et al. 2007), exhibiting cardiac edema (Figures 3F and 3G) and hearts with contractile defects (Supplemental Movies 1 and 2).
Reporter sequences can be readily introduced at additional loci for uses such as the in vivo lineage analysis of cells that fail to express a gene regulating developmental fate. The TALEN and homology arm sequences used to mediate introduction of V5 epitope sequences just downstream of the ntl AUG initiation codon (Figure 2A) were also used to introduce eGFP coding sequences followed by translation and transcription termination motifs (Figure 4A). Virtually all injected embryos had evidence of HR at the ntl locus (Figure S2B). Embryos carrying the ntlGFP allele expressed GFP appropriately in the notochord of developing embryos (Figures 4B and 4C). As expected, the knock-in allele destroyed ntl function, as the ntlGFP allele failed to complement the null ntlb195 mutation (Figures 4D and 4E).
One goal of genome editing in the zebrafish is to recapitulate and analyze the functions of human disease alleles. Many of these editing events, such as those that produce recessive alleles, will be difficult to recognize phenotypically in the F1 generation. With current genome editing methods in the zebrafish, it would be difficult to routinely recover silent sequence modifications, as detection of these would require labor-intensive DNA screening of the genomes of thousands of F1 individuals. Given that non-homologous stretches of sequences encoding reporter proteins can be integrated efficiently into the genome via HR, we reasoned that donor sequence-linked reporter genes could be used to tag the acquisition of donor sequences. To test this concept, we determined whether acquisition of a reporter gene could be used to identify genomes that had acquired linked sequence modifications at the gol locus.
The donor plasmid used to correct the golb1 null mutation was modified so it harbored a reporter gene within the intron of donor sequences, 3′ of the TALEN recognition site (Figure 5A). The 1.8 kbp α-crystallin::Venus (CV) reporter gene expressing a green fluorescent protein under the control of the α-crystallin lens promoter (Hesselson et al. 2009) was bordered by FRT recombination sites that could mediate excision of the reporter gene upon expression of FLP recombinase.
Homozygous golb1 zygotes were injected with a combination of gol-int5 TALEN RNA and gol(b1→WT;CV) donor plasmid, resulting in normal-appearing embryos, 84% of which exhibited pigmented tissue (Table 1). Analysis of genomic DNA indicated each F0 embryo had some genomes in which reporter sequences were integrated into the gol locus as expected (Figure S2C). F0 embryos were raised to adulthood and mated with golb1 mutant partners to examine their germ lines for the ability to transmit an expressed CV transgene and a WT gol allele. Of 39 F0 adults screened, 13 (33%) produced progeny expressing the fluorescent reporter in the lens. Among the 13 F0’s that transmitted the CV transgene, 4 (31%) produced offspring that both expressed the reporter and were fully pigmented (Figures 5B and 5C). Analysis of genomic DNA sequences indicated the embryos that expressed both the reporter and pigment had precisely replaced mutant gol sequences with a single copy of the donor sequences.
Following recovery of the reporter-tagged genome editing event, the FRT-bordered reporter gene can be readily excised (Boniface et al. 2009). Zygotes carrying the edited golb1→WT;CV allele were injected with FLP recombinase mRNA. Resulting 2 dpf embryos lost GFP expression but still exhibited normal pigmentation (Figures 5D and 5E). Consistent with the phenotypic analyses, genomic DNA from each injected embryo experienced precise loss of reporter sequences from edited alleles (Figure 5F).
In sum, the reporter gene can be used to simplify and streamline the recovery of genomes that have been modified by HR and enrich for those that have acquired linked donor sequences of interest. Once an edited allele is recovered, the reporter gene can be removed efficiently using sequence-specific recombination tools.
The experiments at gol indicated acquisition of a linked reporter gene could be used to identify accompanying silent editing events. We tested the utility of the linked reporter for recovering one type of phenotypically silent editing event, generation of a conditional ‘floxed’ allele, which would enable tissue-specific and temporal analyses of gene function that are not available in the zebrafish. The strategy for introducing loxP sites flanking the 203 bp exon 6 of kcnh6a is presented in Figure 6A. HR was stimulated with DSBs induced in intron 6 by the kcnh6a-int6 TALEN. Donor sequences were assembled in the pKHR5 backbone (Figure S4), which provides a loxP site immediately 3′ to the FRT-flanked CV reporter gene. Right arm homology was derived from genomic sequences extending about 1 kbp 3′ of the TALEN site. The left arm extended about 1.5 kbp upstream from the TALEN site, spanning the entirety of exon 6. The left arm was constructed by overlapping PCR to introduce a loxP site into intron 5, 0.4 kbp 5′ of the TALEN site, with an additional 1.1 kbp host sequences extending distal to the loxP site. Incorporation of all novel donor sequences would yield a floxed allele of the configuration 5′ -loxP-exon6-FRT-CV reporter-FRT-loxP-3′.
Zygotes were injected with TALEN mRNA, kcnh6a(loxP) donor plasmid, and I-SceI enzyme. PCR analysis diagnostic of locus-specific integration events indicated every injected embryo had evidence of HR (not shown), and thus F0 embryos were raised to adulthood. Six independently segregating CV+ alleles were recovered among the offspring of 5 of the 43 (12%) F0 adults that were screened. Individual F1 progeny that expressed the reporter were analyzed for the presence of left and right donor-host genome junctions and for the retention of backbone vector sequences. These preliminary analyses revealed two patterns of donor sequence integration (Figure S5): i) kcnh6a integrants with expected donor/target locus junctions; and ii) random integrants with vector backbone and donor sequences but lacking target locus junctions. Sequence analysis of CV reporter-expressing F1 individuals indicated edited kcnh6a alleles arose from precise exchange of sequences, incorporating both flanking loxP sites and the CV reporter gene.
In sum, to recover conditional alleles of kcnh6a, the offspring of 43 F0 individuals were initially screened for GFP expression in the lens, and five founders that transmitted the CV reporter were identified. To identify precisely edited conditional kcnh6a alleles, only the genomes of reporter-expressing F1 offspring were screened. Three independently induced conditional kcnh6a alleles were recovered, representing one-half of all the CV+ alleles transmitted by F0 founders. To propagate the conditional allele, only CV+ progeny carrying kcnh6aloxP were raised to adulthood. Hence recovery of the conditional kcnh6a allele was streamlined dramatically by virtue of expression of the linked reporter gene.
To test the functionality of the conditional kcnh6aloxP allele, kcnh6aloxP/+ heterozygotes were intercrossed and the development of their progeny was examined in the absence or presence of Cre recombinase. As the loxP sites resided in introns flanking both the α-crys::Venus reporter gene and exon 6, we anticipated intercross progeny carrying the conditional allele to be viable and express GFP in the lens. Approximately three-fourths (78/112 = 0.70) of the offspring expressed the reporter, and all of these appeared morphogically normal at 2 dpf (Figures 6B, 6C, and 6F). Genotyping of intercross progeny revealed that the conditional kcnh6aloxP allele segregated as a fully viable allele (Figure 6G). In contrast, following injection of intercross zygotes with 50pg cre mRNA, none (0/49) of the embryos expressed the lens reporter gene and approximately one-quarter (12/49 = 0.24) displayed cardiac edema and a non-beating ventricle (Figures 6D-F), characteristics typical of loss of kcnh6a function. Analysis of the genomic DNA of mutants indicated efficient excision of all sequences between the loxP sites such that we could not detect intact conditional alleles in the genomes of injected embryos (Figure 6G). These results indicate the conditional allele is fully functional: the floxed allele has WT activity and the loxP-flanked sequences are excised efficiently in the presence of Cre recombinase, producing a non-functional allele.
Here we demonstrate that precise genome editing stimulated by targeted induction of HR can be accomplished with high efficiency in the zebrafish. Genome modification by HR allows for the generation of a large spectrum of designed changes to the genome. We have created modified alleles that i) harbor single codon alterations, ii) express an antigen-tagged version of an endogenous product, or iii) express GFP from a targeted locus under control of the endogenous promoter. In addition, we used HR to introduce loxP sites flanking an essential exon and thus produced a fully functional conditional kcnh6a allele, which can be readily converted from WT to mutant function in the presence of the Cre recombinase. These types of genome modifications will allow analysis in the zebrafish of disease-associated variants, candidate enhancer sequences, trafficking and interactions of proteins, or tissue-specific utilization of genes. It appears most loci can be modified by the methods presented here: in this work four different sequences were targeted to produce seven different alleles.
Almost all genome editing events produced via HR, with dsDNA used as donor molecules, result in precisely altered alleles in the zebrafish. This is in stark contrast to the use of ssDNA oligonucleotides as donor templates for genome editing in the zebrafish, where a large fraction of the modified alleles are accompanied by unintended mutations (Bedell et al. 2012; Hruscha et al. 2013; Auer and Del Bene 2014).
Work in both yeast and mammalian cells established the significant impact of DSBs on the stimulation of HR between a targeted host locus and exogenously supplied homologous sequences (Jasin 1996; Paques and Haber 1999). Our work with TALENs carrying FokI domains with differing activities indicates the efficiency with which targeted DSBs are induced has a pronounced effect on the efficiency of HR. In addition, the structure of the donor template DNA affects the efficiency of HR in the zebrafish embryo. Notably, we find linear dsDNA bordered by a pair of head-to-head oriented I-SceI recognition sites recently cleaved by the I-SceI meganuclease functions effectively as a source of donor template for HR events that lead to genome editing. We have developed a series of vectors that should be useful for generating I-SceI-cleaved template for genome modification. In our hands, digestion of donor molecules generated in these vectors with I-SceI stimulated the production of edited alleles at multiple loci (data not shown), but we cannot assert this will be true for all editing events. In all, induction of DSBs at the target locus to stimulate HR, combined with the use of I-SceI-digested donor DNA molecules, leads to the routine and efficient production of precisely edited alleles.
Because our strategy of using a reporter gene nested within donor sequences allows for the extremely efficient identification and recovery of genomes with modifications, it dramatically changes the numbers and types of genome alterations that can be practically isolated on a routine basis. Using any of the previously reported approaches to genome editing in zebrafish (Auer and Del Bene 2014), it is necessary to screen the genomes of approximately 500 - 5,000 F1 individuals to identify carriers of sequence modifications that do not produce a phenotype in the F1 generation. This is because current approaches to genome editing produce founders whose germ lines are genetically mosaic. Pooling the data from all the heritable edited alleles recovered in this study, we find that an edited allele is transmitted by 6.3% of the F0 gametes on average (median = 5.4%) (Table S3). Thus only a few percent of the gametes of a small subset of potential founders carry a modified allele of interest. To identify an altered allele, it is often necessary to screen by DNA analysis up to 100 offspring from each of 20-100 founders. Further, to recover the edited allele from a transmitting founder once identified, it has been necessary to raise large numbers of F1 progeny and then genotype these as adults.
The approach presented here improves both the production of edited alleles in the germ lines of founders and the efficiency with which the alleles are recognized. By providing the α-crys::Venus gene within donor sequences, animals with genomes that have acquired donor sequences can be unambiguously recognized by virtue of expression of GFP in the lens. In our experience, the presence of the CV reporter gene, within an intron and in opposite transcriptional orientation to the host gene, does not have a detectable effect on expression of the endogenous gene. Moreover, if needed, the reporter can be subsequently excised. As the unintegrated donor plasmid bearing the reporter gene is capable of driving GFP expression in the injected embryo, this expression cannot be used to identify F0 embryos with integration events. However, use of the linked reporter significantly enhances the recovery of desired genome editing events: In the two examples described in this study, 30-50% of the F1 individuals expressing the reporter gene had precisely edited alleles that arose via HR. This finding is wholly consistent with expectations, given our finding that about 10% of F0s transmit a bona fide edited allele and about 10% of F0s transmit donor sequences that have integrated elsewhere in the genome (as determined by PCR analysis, data not shown). Thus one need only raise a handful of F1 offspring that express the donor-linked CV reporter gene, anticipating that a large fraction of these will be heterozygous for the modified allele. In rare cases where non-targeted integration of donor sequences may occur with high frequency, we have found use of the donor vectors that additionally carry the cmlc2::mCherry reporter outside of the homology regions (Figure S4) further streamlines recovery of correct editing events. When these vectors are used, the simultaneous expression of both reporter genes indicates germ lines that must have acquired donor sequences by means other than precise HR. Germ lines that only harbor the α-crys::Venus reporter will be enriched for edited alleles that arose via HR.
We demonstrate how HR can be used to generate several highly useful types of alleles in the zebrafish. First, we show at the gol locus that it is simple to recover and propagate coding sequence changes by virtue of the expression of a linked reporter that was acquired along with the sequence change of interest. Second, at ntl we demonstrate that in-frame introduction of epitope-coding sequences can be used to tag endogenous proteins that are produced under physiological conditions and maintain WT function. Third, at kcnh6a we show single HR-mediated gene-editing events can produce fully functional, conditional floxed alleles, whose loxP sites can be induced to recombine efficiently in the presence of Cre recombinase. We envision zebrafish with engineered conditional alleles could be used routinely as bioassays to distinguish functional from pathogenic variants of human disease-associated genes. Conditional mutations also allow study of the tissue- and temporal-specific utilization of genes (Ni et al. 2012). Finally, the ability to manipulate conditional alleles will have a great impact on studies of regeneration or cancer in the zebrafish, in which it is important to preserve normal gene function during early stages of animal development and growth.
Zebrafish Danio rerio were maintained in accordance with approved institutional protocols at the University of Utah. Wildtype zebrafish were from the Tübingen (Tü) or AB strains. Embryos and adults were maintained under standard conditions (Westerfield 2000).
Potential TALEN target sites were identified using the TALEN Targeter program at https://tale-nt.cac.cornell.edu/node/add/talen-old. TALEN targets are listed in Table S1. The RVD repeat array of each TALEN monomer was assembled using Golden Gate assembly (Dahlem et al. 2012). The RVD repeat arrays for left and right TALEN monomers were cloned into the pCS2TAL3DDD (Addgene 48637) and pCS2TAL3RRR (Addgene 48636) vectors, respectively, which contain FokI domain sequences (Doyon et al. 2011). TALEN mRNA was generated by transcription in vitro of pCS2-TALEN plasmids linearized with NotI (mMESSAGE mMACHINE SP6 kit, Life Technologies). The gol-int5 sgRNA target site (Figure 1A) was identified using the CRISPR design program at http://crispr.mit.edu. gol-int5 sgRNA was prepared by transcription in vitro essentially as described (Gagnon et al. 2014).
As individuals from WT strains carry polymorphisms, a small breeding population of WT F0 animals with minimal sequence heterogeneity at the targeted locus was used for each experiment. Briefly, once the presence of the predicted nuclease target site was confirmed in founder genomes, HRMA (Dahlem et al. 2012) and/or sequence analysis was used to identify founders with little sequence heterogeneity surrounding the target site. To make homology arms, amplicons were generated from selected genomes and cloned into donor vectors. Modified sequences were introduced using oligonucleotides and overlapping PCR.
To generate DSBs, 50 pg left and 50 pg right TALEN mRNA or 250pg gol-int5 sgRNA and 600pg Cas9 protein (PNA BIO) were injected into the cytoplasm of zygotes. Following injection of nuclease components with 50 pg donor DNA, on average 68% of embryos (range 59-78%) appeared normal at 1-2 dpf.
For circular plasmid injection, plasmids were purified by phenol/chloroform extraction followed by ethanol precipitation. For linearized plasmid injection, plasmids were digested with ScaI and purified by phenol/chloroform extraction and ethanol precipitation. For injection of purified donor fragments without vector backbone, plasmids were digested with XbaI and XhoI and the insert fragments were isolated following agarose gel electrophoresis and purified by phenol/chloroform extraction and ethanol precipitation. Aliquots of I-SceI enzyme (NEB) were stored at −80°C. For co-injection with I-SceI, 0.25mU or 1mU I-SceI enzyme was mixed with 50pg donor DNA and 100pg total TALEN mRNA in 0.5× I-SceI buffer (NEB). To pre-digest donor DNA with I-SceI enzyme, 500ng donor DNA was incubated in 5ul of 1× I-SceI buffer with 2.5U or 1U I-SceI enzyme at 37°C for 1h, and subsequently mixed with 5ul of 100ng/ul each of left and right TALEN mRNA. All injection samples were maintained on ice until injection of fertilized eggs with approximately 1nl.
To prepare genomic DNA, individual 2 or 7 dpf embryos were incubated in 30 ul or 70 ul, respectively, 50 mM NaOH at 95°C, 20 min. After cooling to 4°C, 1/10 volume of 1 M Tris-HCl (pH8.0) was added to neutralize. PCR primers are listed in Table S2. PCR amplification and subsequent imaging were performed under non-saturating conditions. Quantitative PCR (qPCR) was performed to measure the effect of digestion of donor DNA with I-SceI on the efficiency of gene editing. A 10ul qPCR reaction contained: 1ul embryonic genomic DNA, 0.5× KAPA SYBR® FAST qPCR Kit Master Mix (KAPA Biosystem), 0.5× KAPA2G Fast HotStart PCR Kit (KAPA Biosystem), and 500nM each forward and reverse primers (see Table S2 and Figure 3). The amplification reaction consisted of initial denaturation at 95°C, 3 min followed by 40-45 cycles (95°C, 10 s - 66°C, 15 s - 72°C, 45 s). The reactions were monitored in Eco Real-Time PCR System (Illumina) and data were analyzed using EcoStudy Software (Illumina).
Embryos were fixed (0.4% Triton X-100, 4% Paraformaldehyde in PBS) 2 hours at room temperature, washed, blocked, and incubated overnight at 4°C in 1:250 V5 Epitope Tag Antibody (cat #MA515253, InVitrogen) or 1:250 anti-GFP antibody (Millipore). Signal was developed following tyramide amplification using the TSA Kit #2 (Molecular Probes/Life Technologies). Whole-mount in situ hybridization was performed according to standard procedures (Westerfield 2000).
We thank Janelle Evans and Haley Wilson for help with genotyping, Hannah Grunwald and Sonia Grunwald for creating kcnh6a TALENs, Matt Hockin, Kristen Kwan, and Christian Mosimann for reagents, the U of U Core Zebrafish facility for superb care and maintenance of the animals. Thanks to H. Joseph Yost and Brad Cairns for support of these projects and Brad, Rich Dorsky, Rod Stewart, and Ellen Wilson for critical feedback. Work reported here utilized U of U Cores for sequencing, HRMA, and oligonucleotide synthesis. These studies were supported by grants to D.J.G. from the University of Utah and the National Institutes of Health (5R21HD073847, 1R21OD018323, and 1R01HD081950) and by a subaward from 5P30CA042014. The authors declare no conflicts of interest. This work is dedicated to the memory of George Streisinger.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Sequences for the donor vectors pKHR 4, 5, 7, and 8 (see Figure S4) have been deposited in GenBank (KU144822 - KU144825).
K.H., M.J.J., and D.J.G. each contributed to the design and conduct of experiments and collectively wrote the manuscript.