Cre/lox regulated conditional rescue and inactivation with zebrafish UFlip alleles generated by CRISPR-Cas9 targeted integration

The ability to regulate gene activity spatially and temporally is essential to investigate cell-type-specific gene function during development and in postembryonic processes and disease models. The Cre/lox system has been widely used for performing cell and tissue-specific conditional analysis of gene function in zebrafish. However, simple and efficient methods for isolation of stable, Cre/lox regulated zebrafish alleles are lacking. Here, we applied our GeneWeld CRISPR-Cas9 targeted integration strategy to generate floxed alleles that provide robust conditional inactivation and rescue. A universal targeting vector, UFlip, with sites for cloning short homology arms flanking a floxed 2A-mRFP gene trap, was integrated into an intron in rbbp4 and rb1. rbbp4off and rb1off integration alleles resulted in strong mRFP expression,>99% reduction of endogenous gene expression, and recapitulated known indel loss-of-function phenotypes. Introduction of Cre led to stable inversion of the floxed cassette, loss of mRFP expression, and phenotypic rescue. rbbp4on and rb1on integration alleles did not cause phenotypes in combination with a loss-of-function mutation. Addition of Cre led to conditional inactivation by stable inversion of the cassette, gene trapping and mRFP expression, and the expected mutant phenotype. Neural progenitor Cre drivers were used for conditional inactivation and phenotypic rescue to showcase how this approach can be used in specific cell populations. Together these results validate a simplified approach for efficient isolation of Cre/lox-responsive conditional alleles in zebrafish. Our strategy provides a new toolkit for generating genetic mosaics and represents a significant advance in zebrafish genetics.


Introduction
The ability to regulate gene activity spatially and temporally is essential to investigate cell-typespecific gene function during development and in postembryonic processes and disease models. Conditional gene inactivation can be achieved with the Cre/lox system in which the bacteriophage Cre recombinase promotes site-specific recombination at compatible loxP sites engineered in a gene of interest (Sauer and Henderson, 1988). Effective application of Cre/lox in vertebrates requires the isolation of transgenic lines expressing Cre, and gene modification to introduce loxP sequences at desired locations. Homologous recombination driven site-directed genetic modification in embryonic stem cells enabled creation of the mouse Cre/lox resource that contains more than 2500 Cre drivers and lox conditional gene alleles (Bult et al., 2019). In zebrafish, the lack of embryonic stem cell technology and inefficient homologous recombination in the early embryo necessitated alternative approaches to generate Cre/lox tools (Carney and Mosimann, 2018). Tol2 transposon transgenesis (Balciunas et al., 2006;Kawakami et al., 2000) combined with defined promoters (Mosimann et al., 2011) or BAC recombineering (Förster et al., 2017) has been used to develop Cre lines, and a rich source of zebrafish cell-type-specific Cre drivers has been generated by random Tol2 transposon enhancer trap screens (Jungke et al., 2015;Marquart et al., 2015;Tabor et al., 2019;Zhong et al., 2019). Cre/lox regulated conditional control of gene function in zebrafish was previously limited to generation of knock out alleles by random Tol2 transposon insertional mutagenesis with a floxed/FRT gene trap, which can be reverted by Cre or Flip-mediated recombination (Clark et al., 2011;Ni et al., 2012;Trinh et al., 2011). Only recently have gene editing methods been developed in zebrafish to simplify Cre driver isolation Kesavan et al., 2018) and generate genuine loxregulated conditional alleles by targeted integration (Burg et al., 2018;Han et al., 2021;Hoshijima et al., 2016;Li et al., 2019;Sugimoto et al., 2017). These advances in zebrafish Cre/lox genetics are being driven by rapidly evolving methods for homology directed gene editing using CRISPR-Cas9 and TALEN endonucleases.
We and others recently demonstrated CRISPR-Cas9 targeted integration is an efficient method to isolate zebrafish proneural specific Cre and CreERT2 drivers that are expressed under the control of endogenous gene regulatory elements Kesavan et al., 2018). Efforts to introduce loxP sequences directly into the zebrafish genome by somatic gene targeting was first described using TALENs to target a double-strand break in the cdrh2 gene, followed by repair from a DNA oligonucleotide template containing homologous sequences flanking a loxP site (Bedell et al., 2012). This approach was expanded upon using CRISPR-Cas9 to create a floxed allele in two zebrafish genes by sequential loxP oligonucleotide targeting at sites flanking an exon, and demonstrated robust gene knockdown and loss of function after Cre-mediated exon excision (Burg et al., 2018). Although this approach is effective for creating a Cre/lox regulated allele, a single integration event of a floxed cassette with a linked reporter has distinct advantages, including only one generation of targeting and fluorescent genotyping of the integration allele. The first example of a germline floxed conditional allele with linked reporter used a single TALEN site downstream of an exon (Hoshijima et al., 2016). The repair template consisted of a linear cassette containing 1 kb long homology arms flanking a floxed exon plus reporter, to replace the endogenous exon by homologous recombination. A similar strategy using long homology arms was used to isolate a zebrafish conditional shha allele by integration of a modified Flex invertible gene trap (Ni et al., 2012), called Zwitch, by homologous recombination from an intact plasmid (Sugimoto et al., 2017). Stable inversion of the Zwitch cassette by Cre led to effective shha gene knockdown and defective shha signaling. More recently homologyindependent targeting strategies have been described to generate conditional knockouts with a linked reporter, or bidirectional knockin of a dual reporter gene trap cassette, providing a method for labeling conditional gene knockout cells (Han et al., 2021;Li et al., 2019). As an alternative to isolation of stable germline conditional alleles, rapid tissue-specific gene knockdown and cell labeling can be achieved by somatic targeting with a floxed Cas9-2A-GFP; U6:gRNA transposon in a Cre-specific transgenic background, allowing for simultaneous bi-allelic inactivation with GFP cell labeling (Hans et al., 2021). Together, these approaches provide effective conditional gene knockdown strategies with varying degrees of complexity in design and execution. However, there is still a need for a simple method to efficiently recover robust conditional alleles through a single gene engineering event.
To address this need for a simple method to recover zebrafish germline conditional alleles in a single generation, we applied our GeneWeld CRISPR-Cas9 knock in strategy Wierson et al., 2020) to target integration of a highly mutagenic floxed gene trap into an intron in two genes. We previously showed our GeneWeld strategy simplifies targeting vector assembly and enhances the efficiency of on target integration Wierson et al., 2020). In the present study, we built a universal vector, UFlip, which is derived from the GeneWeld pGTag and pPRISM targeted integration vectors Wierson et al., 2020), and the Tol2 Gene Breaking Transposon vectors (Clark et al., 2011;Ichino et al., 2020). The GeneWeld vectors contain sites for cloning short homology arms flanking a cargo of interest, and universal gRNA sites for efficient CRISPR-Cas9 cutting in vivo. In the UFlip vector, homology arm cloning sites flank a floxed gene trap cassette that contains a 2A-mRFP primary reporter based on the Tol2 RP2 gene trap (Clark et al., 2011;Ichino et al., 2020). The cassette also contains a heart or lens BFP secondary reporter for transgenic identification. Using 48 bp flanking short homology arms to drive integration at unique intron target gRNA sites, we recovered precise UFlip integration alleles in rbbp4 and rb1 with frequencies of 4-14%. We demonstrate that each molecular component in the floxed UFlip cassette functions as expected in vivo. Integration alleles with UFlip in the active, gene 'off' orientation show mRFP expression from the gene trap in the expected gene-specific pattern. UFlip 'off' alleles lead to >99% knock down of gene expression compared to wild-type levels and recapitulate indel mutant phenotypes, demonstrating highly effective loss of function by primary transcript splicing into the gene trap. Recombination of an 'off' allele by Cre injection or a transgenic Cre source leads loss of RFP expression, stable inversion of the UFlip cassette to the passive orientation, and robust conditional rescue of mutant phenotypes. UFlip 'on' alleles are phenotypically normal in combination with lossof-function mutations. Recombination of an 'on' allele by Cre injection or a transgenic Cre source leads to induction of RFP expression, stable cassette inversion, and robust conditional inactivation. These results demonstrate GeneWeld UFlip CRISPR targeted integration is an effective approach to generate zebrafish Cre/lox conditional alleles that can be used for cell type-specific genetic mosaic analysis. The ability to perform conditional inactivation as well as conditional rescue provides a significant advance in zebrafish genetics.

Results
UFlip, a universal vector to generate stable Cre/lox regulated conditional alleles by CRISPR-Cas9 targeted integration The Universal Flip (UFlip) vector was designed to be used with the GeneWeld strategy for CRIS-PR-Cas9 targeted integration driven by short homology (Wierson et al., 2020). The vector contains a cassette with a floxed gene trap plus secondary marker flanked by cloning sites for homology arms (HA) complementary to the genomic CRISPR target site, and universal gRNA sites (UgRNA) for in vivo liberation of the cassette after injection into zebrafish embryos ( Figure 1A). The gene trap was derived from the gene-breaking Tol2 transposon RP2 that has previously been shown to lead to >99% knockdown of gene expression (Clark et al., 2011;Ichino et al., 2020). RP2 was modified by insertion of the porcine teschvirus-1 2A peptide sequence in between the splice acceptor and mRFP, which enables expression of a primary fluorescent reporter from the gene trap without creating a fusion protein. The ocean pout antifreeze gene transcriptional terminator follows the mRFP cDNA. Like our recently published endogenous Cre driver lines , the primary gene trap reporter is followed by a secondary marker that drives tissue-specific expression of blue fluorescent protein (BFP) from the Xenopus gamma crystallin 1 (gcry1) lens promoter or the zebrafish myosin light chain 7 (myl7) cardiac muscle promoter, simplifying identification of an integration allele. Alternating pairs of loxP and lox2272 sites oriented head to head flanking the cassette are designed to drive two Cre recombination events that invert and lock the cassette in place, a design strategy described previously to generate floxed conditional alleles in mice (Robles-Oteiza et al., 2015;Schnütgen et al., 2003) and zebrafish (Sugimoto et al., 2017). rox sites flanking the entire cassette provide an alternative method for inversion using Dre recombinase ( Figure 1A), which we demonstrate below provides highly efficient recovery of inverted alleles through the germline. Integration of UFlip into an intron in the active, gene 'off' orientation is expected to lead to premature transcriptional termination of the primary transcript in the gene trap and splicing of upstream exons into 2A-mRFP ( Figure 1B), resulting in a loss-of-function allele. In contrast, integration of UFlip into an intron in the passive, gene 'on' orientation is not predicted to interrupt endogenous gene expression, since RNA polymerase will The vector contains a floxed rox loxP lox2272 gene trap plus secondary marker loxP lox2272 rox cassette. The cassette is flanked by cloning sites for homology arms (HA) complementary to a genomic CRISPR target site, and universal gRNA sites (UgRNA) for in vivo liberation of the targeting cassette. (B) Gene 'off' alleles are generated by integration of the UFlip cassette into an intron in the active orientation, leading to transcription termination and splicing of the primary transcript in the mRFP gene trap. (C) Gene 'on" alleles are generated by integration of the UFlip cassette into an intron in the passive orientation. This is driven by cloning the genomic 5' homology arm downstream of the UFlip cassette, and cloning the genomic 3' homology arm upstream of the UFlip cassette. Integration at the genomic CRISPR-Cas9 target site occurs in the opposite orientation. During transcription RNA polymerase reads through the integrated UFlip cassette, which is spliced out with the intron during processing of the primary transcript. (D) Cre-mediated recombination at an 'off' allele locks the cassette in the 'on' orientation. The first recombination occurs stochastically at either lox2272 or loxP sites. The diagram shows the intermediate that forms if the first recombination occurs at the lox2272 sites. (E) Cremediated recombination at an 'on' allele locks the cassette in the 'off' orientation. The first recombination occurs stochastically at either lox2272 or loxP Figure 1 continued on next page read through the cassette in the intron, which is then spliced out of the mature transcript ( Figure 1C). For clarity, the UFlip alleles described here are referred to as 'off' when the cassette was integrated in the active, gene trap orientation. Alleles are designated 'on' when UFlip was integrated in the passive orientation, that leads to removal of the UFlip cassette during splicing and allows proper functioning of the gene.
The process by which Cre recombination leads to stable inversion of the UFlip cassette is illustrated in Figure 1D, E (Schnütgen et al., 2003). Introduction of Cre leads to stochastic recombination at either the loxP or the lox2272 pairs of recombinase sites. Figure 1D, E show the intermediate results after 1 recombination event at the lox2272 sites. In the intermediate the cassette has inverted from 'off' to 'on' ( Figure 1D) or from 'on' to 'off' (Figure 1E), resulting in a pair of loxP sites flanking one of the lox2272 sites. Recombination at the head to tail oriented loxP sites removes the intervening lox2272, resulting in a cassette flanked by a single pair of incompatible loxP and lox2272 sites. This prevents further recombination and produces stable inversion of the cassette.

Isolation of rbbp4 and rb1 UFlip conditional alleles
To test the ability of the UFlip cassette to generate zebrafish Cre-responsive conditional alleles, we chose to target integration of UFlip two genes, retinoblastoma binding protein 4 (rbbp4) and retinoblastoma 1 (rb1). We had previously isolated indel mutations in both genes that show strong loss-offunction phenotypes at the morphological and cellular level Schultz-Rogers et al., 2022;Solin et al., 2015). Intronic genomic DNA was amplified from wild-type WIK fish, cloned, sequenced, and analyzed to locate unique Cas9 gRNA sites that were shared among fish and that did not map to repetitive elements or retrotransposons. Cas9 target sites were identified in rbbp4 intron 4 and rb1 intron 6 ( Table 1) and synthetic guides ordered from Synthego (https://www.synthego. com). Efficient indel formation at the intronic gRNA sites was confirmed by co-injection of gRNA and Cas9 mRNA into single-cell embryos, followed by extraction of genomic DNA at 2 dpf. A PCR amplicon surrounding the target site was directly sequenced followed by ICE analysis (https://ice. synthego.com/#/). The rbbp4 intron 4 and rb1 intron 6 guides showed 50% and 95% indel formation at the target site, respectively (Figure 2A, B). UFlip targeting vectors were assembled with 48 bp 5' and 3' homology arms complementary to the DNA flanking the Cas9 genomic DNA double strand break site (Table 1). One cell stage WIK embryos were co-injected with targeting vector (10 pg), Cas9 mRNA (150 pg), universal gRNA (25 pg), and gene-specific gRNA (25 pg). Four to six embryos with positive expression of the primary mRFP reporter and secondary BFP reporter were selected to test for evidence of on target integration by PCR amplification of the 5' and 3' junctions (data not shown). Primary and secondary reporter positive siblings were raised to adulthood. rbbp4 intron 4 was targeted with the UFlip-2A-mRFP cassette in the active orientation to recover an 'off' allele. Embryos expressing the primary reporter mRFP in the nervous system, where rbbp4 is normally expressed, and the lens:BFP secondary marker, were raised to adulthood. Two out of seven adult F0 fish transmitted embryos with widespread mRFP and lens BFP expression ( Table 2; Table 2-source data 1). One adult transmitted an allele with precise 5' and 3' junctions at the intron target site. F1 embryos were raised to adulthood and the presence of the precise rbbp4-2A-mRFP-off, gcry1:BFP (rbbp4 off ) allele  confirmed by fin clip and sequencing ( Figure 2C). An F1 adult with the precise integration allele was outcrossed to WIK to establish an F2 family of the rbbp4 off is61 allele. We initially targeted rb1 intron 6 with the original UFlip-mRFP primary reporter construct designed to create an in-frame fusion of the polypeptide with mRFP. We recovered rb1-mRFP-off, gcry1:BFP (rb1 off ) and rb1-mRFP-on, gcry1:BFP (rb1 on ) alleles with frequencies of 6% (2/33) for rb1 off and 4% (1/25) for rb1 on ( Table 2; Table 2-source data 1). 5' and 3' junction PCR and sequence analysis of F1 embryos showed the integrations were precise, except for the 5' junction in the allele transmitted by the second rb1 off founder #26 ( Figure 2C). The 5' junction included a duplication of the 5' Homology Arm sequence that flanked a 115 bp segment of the vector backbone. The rb1 off founder #18 transmitting the precise allele was lost, therefore we moved forward with the allele from founder #26. As shown below, the 153 bp insertion on the 5' side of the rb1 off integration did not appear to affect the functionality of the allele. F1 adults were outcrossed to wildtype WIK to establish rb1 off is58 and rb1 on is57 families. We subsequently targeted the same rb1 intron six site with the updated UFlip-2A-mRFP vector, which allows expression of mRFP without creating a fusion protein with the targeted gene polypeptide. We recovered 2/8 (12.5%) founders transmitting lens:BFP +alleles with ubiquitous expression of the mRFP primary reporter ( Table 2; Table 2-source data 1). Founder F0#14 transmitted two alleles, one with precise junctions, the other with a precise 3' junction and a 5' junction with a homology arm duplication and insertion of 19 bp ( Figure 2C). F1s from this founder will be used to establish the line rb1 off is63 . For the experiments outlined below demonstrating Cre-meditated rb1 conditional rescue and inactivation, we used the established F3 generation rb1 off is58 and rb1 on is57 lines generated with the original UFlip construct, which did not show the mRFP gene trap reporter expression.
The UFlip construct contains rox sites flanking the entire cassette to allow the use of Dre recombination to recover an allele of the opposite orientation when starting with the rbbp4 off allele ( Figure 3A). To test this, we injected Dre synthetic mRNA into F1 and F2 rbbp4 off/+ embryos. Injected embryos were raised to adulthood and five F1 and six F2 adults were outcrossed to wildtype WIK. Embryos were screened for expression of the primary mRFP reporter and the linked BFP secondary marker. Three of 5 F1 adults, and 4/6 F2 adults, transmitted both the mRFP/BFP positive original 'off' allele to their progeny, as well as the mRFP negative/BFP positive Dre-inverted 'on' allele ( Table 3). The frequency of embryos inheriting the inverted allele vs. the original allele ranged from 7.7% to 43.2% (Table 3). We confirmed the rbbp4 on allele recovered by Dre recombination contained the expected 5' and 3' junctions by PCR amplification and sequencing ( Figure 3B, C). A single F2 adult was used to establish an F3 family of the rbbp4 on is62 allele. These results demonstrate efficient Dre mediated recombination at rox sites in the germline, and a simple method for isolation of an alternative conditional allele starting with one UFlip allele.
sequence of the intron 6 reverse strand gRNA. Gel image of PCR amplicons surrounding the target site from eight embryos injected with Cas9 and the gRNA (1-8), and two uninjected embryos (U). Amplicons from embryo #1 and an uninjected embryo were sequenced and analyzed with Synthego's ICE software, and indicate 95% indel efficiency at the target site. Plots show the range and percentage of indels present in the sequences. PAM sequences shown in bold and underlined. (C) 5' and 3' genomic-UFlip integration junctions were PCR amplified from F1 transgenic zebrafish fin clip genomic DNA. The PCR products were sequenced and aligned to the reference sequence expected for a precise integration at the genomic target site. Capitalized red nucleotides represent 48 bp homology arms. Lowercase green nucleotides represent random inserted sequences.

rbbp4 off allele provides robust gene knockdown and loss of function
To demonstrate that integration of the UFlip cassette in the active, 'off' orientation leads to effective gene knockdown we examined gene expression by quantitative PCR in rbbp4 off/off homozygotes. Homozygous rbbp4 off/off 3 dpf larvae show a>99% reduction in rbbp4 mRNA levels in comparison to wild type sibling larvae ( Figure 4A, B; Figure 4-source data 1). In comparison to wild-type larvae, the rbbp4 off allele is easily identifiable by primary reporter gene trap RFP expression in the retina, brain, and mesodermal tissues, and secondary marker lens specific BFP expression ( Figure 4C, D arrowhead). In trans-heterozygous combination with the established loss-of-function indel allele rbbp4 Δ4 ( Figure 4D, E), the rbbp4 off allele recapitulates the microcephaly and microphthalmia phenotypes found in rbbp4 Δ4/Δ4 homozygotes Schultz-Rogers et al., 2022). In comparison to control, anti-activated caspase 3 labeling and quantification in sectioned rbbp4 off/Δ4 2 dpf head tissue reveals extensive apoptosis throughout the midbrain (** p<0.01) and retina (** p<0.01) ( Figure 4F, G). This cellular phenotype is not significantly different from our previous description of apoptosis in rbbp4 Δ4/Δ4 homozygotes ( Figure 4F-H). Together these results demonstrate the rbbp4 off allele causes robust gene knockdown by the primary reporter RFP gene trap, induces the expected pattern of RFP expression, and recapitulates the rbbp4 strong loss of phenotype at the molecular, cellular, and morphological levels.

rbbp4 on allele maintains rbbp4 function
In the rbbp4 on allele, the UFlip cassette is predicted to be read through by RNA polymerase and then removed from the primary transcript when intron 4 is spliced out ( Figure 4I). To test whether the rbbp4 on allele alters normal rbbp4 mRNA expression levels RT-qPCR was performed on 3 dpf rbbp4 on/+ heterozygous and rbbp4 on/on homozygous larvae. rbbp4 on/+ heterozygotes showed ~25% reduction in wild-type rbbp4 mRNA levels, while rbbp4 on/on homozygotes showed ~40% reduction ( Figure 4J, Figure 4-source data 1). This suggests the integrated UFlip cassette may impact splicing rate, stability, or other processing of the primary transcript. However, neither rbbp4 on/+ heterozygotes nor rbbp4 on/Δ4 trans-heterozygous show a gross phenotype in 5 dpf larvae ( Figure 4K and L) or evidence of apoptosis in the developing midbrain and retina at 2 dpf ( Figure 4M and N). Therefore, the level of processed, mature rbbp4 transcript is enough to maintain wild type levels of rbbp4 activity. Experimental group 1: F1 embryos from the F0 founder rbbp4 off/+ adult crossed to WIK were injected with 15 pg Dre mRNA. Inversion was confirmed by PCR on genomic DNA isolated from -injected embryos . 5 sibling injected embryos were raised to adulthood and were outcrossed to WIK. F2 embryos were screened for Primary (RFP+) and Secondary (BFP+) marker expression. Embryos inheriting the original rbbp4 off allele show expression of both markers. Embryos inheriting an allele that was inverted from "off" to "on" by Dre recombination show expression of the secondary marker (BFP+) but lose expression of the primary marker due to the inverted orientation of the cassette. Experimental group 2: The same experiment was performed with an adult F1 rbbp4 off/+ female crossed to WIK. F2 Dre mRNA injected embryos were raised to adulthood, and 7 adults screened for transmission of the inverted rbbp4 on allele to the F3 generation. Cre/lox recombination of rbbp4 off leads to stable inversion and conditional gene rescue To test the ability of the UFlip cassette to undergo stable inversion and provide conditional rescue, we first used Cre mRNA injection. A total of 12.5 pg Cre synthetic mRNA was injected into embryos from a cross between rbbp4 off/+ and rbbp4 Δ4/+ heterozygous adults ( Figure 5A-G). In comparison to uninjected rbbp4 off/Δ4 larvae ( Figure 5B), 5 dpf rbbp4 off/Δ4 Cre injected larvae show normal morphology and an absence of RFP expression ( Figure 5C). Confocal live imaging of 2 dpf in injected rbbp4 off/Δ4 confirmed loss of RFP and rescue of the cell death phenotype in the retina ( Figure 5-figure supplement 1A-L). Stable inversion of the rbbp4 off cassette to the 'on' orientation was confirmed by PCR and sequence analysis, which showed the expected sequences at the 5' and 3' junctions after inversion at the lox sites ( Figure 5-figure supplement 1M-P). Activated caspase 3 labeling in 2 dpf sectioned head tissue showed a nearly complete absence of apoptosis in the midbrain (** p<0.01) and retina (**** p<0.0001) compared to controls ( Figure 5D-F, Figure 5-source data 1). Genomic DNA quantitative PCR of the original 5' and 3' rbbp4 on allele junction fragments in the Cre injected embryos indicated a 93% reduction ( Figure 5G, Figure 5-source data 1). Together these data demonstrate Cre mediated recombination at the incompatible pairs of loxP and lox2272 sites in the rbbp4 off allele leads to efficient, stable inversion of the cassette to the 'on' orientation, which results in robust conditional rescue of the rbbp4 loss-of-function phenotype.
To test whether cell type-specific Cre sources could be used to rescue rbbp4 activity in distinct cell populations ( Figure 5H-K), we used our previously published endogenous proneural Cre driver lines ascl1b-2A-Cre, gcry1:GFP is75 and neurod1-2A-Cre, gcry1:GFP is77 . 2 dpf larvae from rbbp4 off/+ crossed with ascl1b-2A-Cre/+; rbbp4 Δ4/+ were sectioned and head tissue labeled to detect activated caspase 3. In comparison to control rbbp4 off/Δ4 embryos, ascl1b-2A-Cre; rbbp4 off/Δ4 embryos appeared to have reduced caspase labeling in the midbrain, where Cre is expressed ( Figure 5H The online version of this article includes the following source data and figure supplement(s) for figure 4: Source data 1. Source data for rbbp4 RT-quantitative PCR analysis in wildtype, heterozygous and homozygous embryos from rbbp4 off/+ and rbbp4 on/+ incrosses.   Cre/lox recombination of rbbp4 on leads to efficient conditional gene inactivation and recapitulates morphological and cellular loss-offunction phenotypes To assess whether Cre recombination of the rbbp4 on allele to the 'off' orientation would lead to conditional inactivation and induction of expression of the gene trap mRFP, we first used 12.5 pg Cre mRNA injection into embryos from a cross between rbbp4 on/+ and rbbp4 Δ4/+ heterozygous adults ( Figure 6). In comparison to uninjected rbbp4 on/Δ4 larvae which appear morphologically normal ( Figure 6B), 5 dpf rbbp4 on/Δ4 Cre mRNA-injected larvae display the rbbp4 microcephaly and microphthalmia lossof-function phenotype, and induction of RFP expression ( Figure 6C). Sectioned tissue from 2 dpf embryos showed a significant increase in activated caspase 3 labeling throughout the midbrain (* p<0.05) and retina (* p<0.05) ( Figure 6D Proneural Cre/lox rbbp4 on conditional inactivation reveals cell-typespecific requirement for rbbp4 in ascl1b progenitor survival We conducted the same experimental approach as described above using the endogenous proneural Cre driver lines ascl1b-2A-Cre, gcry1:GFP is75 and neurod1-2A-Cre, gcry1:GFP is77 , but in combination with the rbbp4 on allele to test for cell-type-specific conditional inactivation ( Figure 6H-M). Embryos from rbbp4 on/+ + crossed with ascl1b-2A-Cre/+; rbbp4 Δ4/+ adults were examined for induction of gene trap RFP expression and assessed for Cre-mediated allele inversion by 5' and 3' junction genomic qPCR ( Figure 6-figure supplement 2; Figure 6-figure supplement 2-source data 1). Three dpf ascl1b-2A-Cre/+; rbbp4 on/Δ4 larvae show RFP expression throughout the brain and neural tube and a 20% inversion rate ( Figure 6-figure supplement 2 B, B', D). Although the larvae appeared morphologically normal, compared to controls the sectioned tissue from 2 dpf embryos showed a significant increase in activated caspase 3 labeling in the midbrain tectum and dorsal thalamic region (** p<0.01) ( Figure 6H-I and L; Figure 6-source data 1). Activated caspase 3 labeling was not detected in the retina (n.s. p=0.08543), consistent with the lack of retinal expression of ascl1b and Cre ( Figure 6H"-I" and L). Conditional inactivation of rbbp4 on with neurod1-2A-Cre also lead to induction of RFP expression in both the brain and the retina, albeit at lower levels, and 20% cassette inversion efficiency ( Figure 6-     suggest Cre-mediated conditional inactivation of rbbp4 in the ascl1b neural progenitor population leads to apoptosis. Although rbbp4 is expressed in the neurod1 population as shown by induction of RFP, the absence of apoptosis indicates rbbp4 may not be required for survival in committed progenitors or newborn neurons. These results support two critical conclusions: First, conditional inactivation of rbbp4 on with ubiquitous Cre leads to efficient stable inversion of the cassette from 'on' to 'off', and replicates the rbbp4 loss-of-function phenotype. Second, rbbp4 on can be inactivated in a cell-typespecific manner that reveals a potential cell-specific function for rbbp4 in ascl1b-neural progenitor cell survival.

rb1 off allele provides robust gene knockdown and loss of function
To validate the UFlip approach for generating conditional alleles at a second locus, we characterized the activity of the rb1 off is58 and rb1 on is57 alleles (Figure 7). RT-qPCR of homozygous rb1 off 3 dpf larvae with primers spanning rb1 exons 6-8 showed >99% knockdown of rb1 mRNA levels in comparison to wild-type sibling larvae ( Figure 7A, B; Figure 7-source data 1). To simplify identification of trans heterozygous rb1 mutant larvae, we isolated an rb1 exon 2 CRISPR integration allele, rb1-stop-myl7:GFP is59 (rb1-stop) using a targeting vector design similar to our recently published method to isolate endogenous Cre lines in proneural genes . The stop-myl:GFP cassette contains three translation termination codons, in each of the three reading frames, followed by a transcription terminator plus a myl7:GFP secondary marker (Figure 7-figure supplement 1A, B). The rb1 stop allele is larval lethal in combination with the rb1 Δ7 indel mutation and homozygous rb1 stop/stop 5 dpf larvae show the same level of increased numbers of pH3 labeled cells throughout the midbrain (n.s. p=0.2951) and retina (n.s. P=0.1534) (Figure 7-figure supplement 1C; Figure 7-figure supplement 1-source data 1) as rb1 Δ7/Δ7 homozygotes . Trans-heterozygous rb1 off/stop larvae from a cross between rb1 off/+ and rb1 stop/+ heterozygotes are easily distinguished from siblings by the expression of lens BFP by the rb1 off allele ( Figure 7D arrowhead) and heart GFP by the rb1 stop allele ( Figure 7D asterisk). rb1 off/stop 5 dpf larvae appear morphologically normal compared to wild type, with the exception of the lack of a swim bladder ( Figure 7C, D), similar to rb1 Δ7/stop larvae ( Figure 7E). In contrast to wild type ( Figure 7F), 5 dpf sectioned head tissue shows pH3-labeled cells throughout the rb1 off/stop midbrain (*** p<0.001) and retina (**** p<0.0001), like rb1 Δ7/stop transheterozygotes ( Figure 7F-H; Figure 7- figure supplement 2A; Figure 7-source data 1) and rb1 stop/ stop and rb1 Δ7/Δ7 homozygotes (Figure 7-figure supplement 1C). Together these results demonstrate the rb1 off allele provides robust gene knockdown and recapitulates the zebrafish rb1 loss-of-function morphological and cellular phenotypes.

Cre/lox recombination of rb1 off leads to stable inversion and conditional gene rescue
To test the rb1 off and rb1 on alleles for Cre-mediated conditional rescue and inactivation, we used the same approach as outlined above for the conditional rbbp4 off and rbbp4 on alleles. First, 12.5 pg Cre mRNA injection was used to test for recombinase mediated inversion and conditional gene rescue in embryos from a cross between rb1 off /+and rb1 stop /+adults ( Figure 8A-G). Stable inversion of the rb1 off cassette to the 'on' orientation was confirmed by PCR and sequence analysis, which showed the expected sequences at the 5' and 3' junctions after inversion at the lox sites ( Figure 8-figure  supplement 1). In comparison to uninjected rb1 off / stop larvae ( Figure 8B), 5 dpf rb1 off / stop Cre injected larvae showed a significant reduction in the number of pH3-labeled cells in the midbrain (** p<0.01) and retina (** p<0.01) ( Figure 8C, F; Figure 8-source data 1). Genomic DNA quantitative PCR of the original 5' and 3' rb1 off allele junction fragments in the Cre injected embryos indicated a>78% inversion rate ( Figure 8G; Figure 8-source data 1). Together these data demonstrate Cre mediated recombination of the rb1 off allele leads to efficient, stable inversion of the cassette to the 'on' orientation and conditional rescue of the rb1 loss-of-function phenotype. neurod1-specific rb1 off conditional rescue demonstrates rb1 is required to suppress neurod1 neural progenitor cell cycle entry The proneural neurod1-2A-Cre, gcry1:GFP is77 driver was used to test for conditional gene rescue of rb1 in the neurod1 progenitor population. Five dpf larvae from adult rb1 off/+ crossed with neurod1-2A-Cre/+; rb1 stop/+ was labeled with antibodies to Cre and pH3 ( Figure 8D-I). In comparison to control rb1 off/stop larvae, neurod1-2A-Cre/+; rb1 off/stop larvae showed a significant reduction in pH3 positive cells in the midbrain (* p<0.05) and retina (** p<0.01) where Cre is expressed ( Figure 8D-H; Figure 8source data 1). Rescue of the phenotype correlated with >34% Cre mediated inversion of rb1 off to the 'on' orientation ( Figure 8I; Figure 8-source data 1). The results demonstrate cell-type-specific conditional rescue and a requirement for rb1 to suppress cell cycle entry in neurod1 progenitors. and 24. (K, L) Gross phenotype of wildtype +/+ (K), heterozygous rb1 on/+ (L) and transheterozygous rb1 on/stop (M) 5 dpf larvae. Arrowhead in L, M points to rb1 off allele gcry1:BFP secondary reporter expression in lens. Asterisk in M marks the rb1 stop allele mly7:GFP secondary reporter expression in heart. pH3 and HuC/D labeling of sectioned head tissue from 5 dpf +/+ (N), heterozygous rb1 on/+ (O) and transheterozygous rb1 on/stop (P).OT, optic tectum; R, retina; Th, thalamic region. Error bars represent mean ± s.e.m. Scale bars: 200 μm (C-E, K-M), 50 μm (F-H, N-P).
The online version of this article includes the following source data and figure supplement(s) for figure 7: Source data 1. Source data for rb1 RT-quantitative PCR analysis in wildtype, heterozygous and homozygous embryos from rb1 off/+ and rbbp4 on/+ incrosses.      Cre/lox recombination of rb1 on leads to efficient conditional gene inactivation and recapitulates cellular rb1 loss-of-function phenotype Conditional rescue using the rb1 on allele was first tested by injection of 12.5 pg Cre mRNA into embryos from a cross between heterozygous rb1 on/+ and rb1 stop/+ adults to induce inversion from the passive to the active orientation ( Figure 9A). At 3 dpf rb1 on/stop uninjected and injected larvae appear morphologically normal like control rb1 on/+ larvae (Figure 9-figure supplement 1 A, B). However, in comparison to uninjected control sectioned head tissue ( Figure 9B), Cre injected rb1 on/stop showed increased levels of pH3 labeling throughout the brain and retina ( Figure 9C), similar to the loss-offunction phenotype of rb1 Δ7/stop and rb1 off/stop ( Figure 7D and E). Quantification showed a significant increase in the number of pH3-positive cells in injected larvae in both the midbrain (**** p<0.0001) and retina (** p<0.01) ( Figure 9F; Figure 9-source data 1). Induction of the mutant phenotype correlated with >80% inversion rate by Cre recombination (Figure 9G; Figure 9-source data 1). Stable inversion of the UFlip cassette in Cre injected embryos was confirmed by PCR junction analysis and sequencing of the junction amplicons (Figure 9-figure supplement 1 C-F). These results demonstrate Cre recombination leads to stable inversion of the UFlip cassette and conditional gene inactivation which replicates the rb1 loss-of-function phenotype at the cellular level.

neurod1-specific rb1 on conditional inactivation leads to neural progenitor cell cycle entry
The proneural neurod1-2A-Cre, gcry1:GFP is77 driver was used to test for conditional inactivation of rb1 in the neurod1 progenitor population. 3 dpf larvae from adult rb1 on/+ crossed with neurod1-2A-Cre/+; rb1 stop/+ was labeled with antibodies to Cre and pH3 ( Figure 9D-I). In comparison to control rb1 on/stop larvae ( Figure 9D), neurod1-2A-Cre/+; rb1 on/stop larvae showed a significant increase in pH3 positive cells in the midbrain (*** p<0.001) and retina (* p<0.05) where Cre is expressed ( Figure 9E and H; Figure 9-source data 1). Induction of the mutant phenotype correlated with >25% inversion of rb1 on to the 'off' orientation ( Figure 9I; Figure 9-source data 1). The results are consistent with the rb1 off conditional rescue experiment, and demonstrate a requirement for rb1 in suppressing cell cycle entry in the neurod1 progenitor cell population in the developing midbrain and retina.
Overall, these analyses show conditional alleles generated by targeted integration of the floxed UFlip cassette provide robust control of gene activity. Combined with cell-type-specific Cre drivers, the alleles demonstrate cell-type-specific requirements for gene activity through both conditional inactivation and rescue.

Discussion
In this study, we used GeneWeld CRISPR-Cas9 targeting to isolate Cre/lox regulated conditional alleles with linked reporters by integration of a floxed gene trap cassette, UFlip, at a single gRNA site in an intron. Gene 'off' and gene 'on' germline UFlip alleles were recovered for two genes, rbbp4 and rb1. Our results showed the UFlip cassette led to >99% gene knockdown at the molecular level, did not disrupt gene activity in the passive orientation, allowed stable Cre/lox mediated inversion for conditional inactivation or rescue, and recapitulated expected loss-of-function phenotypes at the morphological and cellular levels. We further demonstrated using proneural Cre drivers that conditional gene inactivation or rescue can be restricted to distinct neural progenitor cell populations. These studies revealed cell-type-specific requirements for gene activity that had previously only been inferred by global loss-of-function homozygous mutant analysis. Our approach is simple in design, (G) Genomic DNA qPCR quantification of rb1 off original orientation DNA 5' and 3' junctions in control rb1 off/stop (n=3) and Cre injected rb1 off/stop (n=3). (H) Quantification of pH3-positive cells in rb1 off/stop (n=3) and neurod1-2A-Cre; rb1 off/stop (n=3) midbrain (** p<0.01) and retina (* p<0.05). (I) Genomic DNA qPCR quantification of rb1 off original orientation DNA 5' and 3' junctions in control rb1 off (n=3) and neurod1-2A-Cre; rb1 off/stop (n=3). Error bars represent mean ± s.e.m. with two-tailed t-test. Scale bars: 50 μm (B -E), 10 μm (B' -E").
The online version of this article includes the following source data and figure supplement(s) for figure 8: Source data 1. Source data for quantification of pH3 labeling and rb1 off inversion after Cre injection and neurod1-2A-Cre.  We assembled UFlip vectors with 48 bp homology arms to generate integrations in rbbp4 and rb1 in both the 'off' and 'on' orientations. Of the four potential targeted integration alleles, we recovered three precise alleles at a frequency of 4-14% of founders screened. This frequency is lower than our previous reports using GeneWeld to knock in fluorescent reporters (22-100%) (Wierson et al., 2020) or 2A-Cre (10-100%) Wierson et al., 2020). In those studies, injected embryos showing expression of the fluorescent tag, or Cre-mediated switching of a floxed reporter, were selected and raised to adulthood to screen for germline transmission. In this report, some lines were generated with vectors that only enabled embryo selection based on expression of the secondary reporter, which does not reflect on target integration. In support of this, only 7 and 8 adults needed to be screened to recover the rbbp4 off and rb1 off 2A-mRFP UFlip alleles, respectively, which showed mRFP expression in the injected embryos. This is in comparison to 25 adults that were screened to recover the original rb1 on mRFP UFlip alleles. Moreover, starting with an allele in one orientation, it was straightforward to use Dre recombination at rox sites flanking the UFlip cassette to recover an allele in the opposite orientation. Dre/rox recombination has been shown to be highly efficient in somatic tissue (Park and Leach, 2013). By injection of Dre mRNA into rbbp4 off embryos we readily recovered the rbbp4 on allele in the next generation from 7/11 adults, demonstrating efficient Dre recombinase activity in the zebrafish germline. Previously reported methods for recovery of intron targeted floxed conditional alleles varied widely in frequency, from 12% for homology directed integration from a linear template with long 1-1.5 kb homology arms (Hoshijima et al., 2016), 53% by homologous recombination from a circular plasmid with 1 kb long homology arms (Sugimoto et al., 2017), to 56-60% for integration by Non-Homologous End Joining after preselection of a primary reporter (Han et al., 2021;Li et al., 2019). The simplicity of UFlip vector construction and ability to recover precise integrations, in either orientation, underscore the power of our approach for isolation of stable Cre/lox regulated conditional alleles. Similar to our previous findings when isolating endogenous Cre drivers , the UFlip founders transmitted both precise and imprecise integration events. In some cases, the imprecise event was a simple homology arm duplication or short insertion at either the 5' or 3' junction, which would be predicted not to affect gene function since the integrated cassette is located in an intron. As predicted, the rb1 off allele, from which we were able to establish a line, had a 5' homology arm duplication separated by a 115 bp insertion. This did not affect the ability of the cassette to efficiently invert in response to Cre recombinase. When inverted into the passive orientation the cassette allowed effective conditional rescue, indicating the additional sequences in the intron flanking the cassette didn't disrupt splicing or rb1 gene activity.
For effective conditional gene inactivation, it is essential that the method leads to robust knockdown of wildtype mRNA expression and recapitulates loss-of-function phenotypes for genetic analysis. Our UFlip conditional construct is based on the Tol2 transposon RP2 gene trap, which contains a transcriptional terminator that effectively blocks transcription readthrough and reduces transcript levels by >99% (Clark et al., 2011;Ichino et al., 2020). The efficacy of gene knockdown by the UFlip gene trap was demonstrated by quantitative RT-qPCR and revealed >99% reduction in mRNA levels. Cre/lox-mediated inversion of the UFlip cassette from the passive to active orientation was highly efficient after Cre mRNA injection or exposure to a transgenic source of Cre, generating expected phenotypes in combination with loss-of-function mutations. The recently published Cre-regulated CRISPR indel targeting study provides a rapid method for conditional mutagenesis and phenotypic knockout in embryos (Hans et al., 2021). In contrast, a significant advantage of our stable conditional UFlip retina (** p<0.01). (G) Genomic DNA qPCR quantification of rb1 on original orientation DNA 5' and 3' junctions in control rb1 on/stop (n=3) and Cre injected rb1 on/stop (n=3). (H) Quantification of pH3-positive cells in control rb1 on/stop (n=3) and neurod1-2A-Cre; rb1 on/stop (n=3) midbrain (*** p<0.001) and retina (* p<0.05). (I) Genomic DNA qPCR quantification of rb1 on original orientation DNA 5' and 3' junctions in control rb1 on/stop (n=3) control and neurod1-2A-Cre; rb1 on/stop (n=3). Error bars represent mean ± s.e.m. with two-tailed t-test. Scale bars: 50 μm (B -E), 20 μm (B' -E").
The online version of this article includes the following source data and figure supplement(s) for figure 9: Source data 1. Source data for quantification of pH3 labeling and rb1 on inversion after Cre injection and neurod1-2A-Cre.  alleles rests on the strength of transcriptional termination by the UFlip cassette in the active orientation. In addition, Cre induction of conditional inactivation leads to mRFP fluorescent cell labeling, as described in other recent studies (Han et al., 2021;Hans et al., 2021;Li et al., 2019), which is critical for lineage tracing and capturing mutant cells for downstream genomics applications.
We further validated effective UFlip conditional gene inactivation at the cellular level. We previously demonstrated that rb1 is required to prevent neural progenitor cell cycle re-entry in the developing zebrafish brain and rbbp4 is necessary for neural progenitor survival and differentiation . Using our ascl1b-2A-Cre and neurod1-2A-Cre drivers in combination with rbbp4 on and rb1 on conditional alleles, we demonstrated the requirement for each gene could be examined specifically in the ascl1b or neurod1 neural progenitor populations. Rb1 is required to prevent cell cycle re-entry in ascl1b and neurod1 neural progenitors. Rbbp4 appears to be required for neural progenitor survival only in ascl1b progenitors. Further analysis with tamoxifen regulated CreERT2 drivers will refine our understanding of the requirement for these genes in neural progenitor survival and proliferation. These conditional genetic tools provide a foundation for future mechanistic studies investigating the interaction of rb1 and rbbp4 specifically in neural progenitor cell cycle regulation and differentiation but can also be used with other cell-type-specific Cre drivers. Our results illustrate the robustness of the UFlip conditional alleles for cell type-specific gene knockout analysis and their potential application in different developmental or disease model contexts.
In summary, building on our previous streamlined GeneWeld approach for CRISPR-Cas short homology directed targeted integration, we describe an effective method for generating zebrafish floxed conditional alleles. We present a thorough molecular and phenotypic validation of the robustness of the UFlip floxed gene trap cassette for effective knockdown of gene expression and Cre/lox regulated conditional inactivation and rescue. UFlip alleles can be combined with any available cell or tissue-specific Cre driver for spatial and temporal gene knock out, enabling the investigation of gene function in specific developmental and post embryonic stages or disease models. The UFlip vector can be adapted to introduce a variety of gene modification cassettes, such as alternative exons and complete cDNAs. In contrast to Tol2 transposon binary UAS overexpression systems, targeted integration leads to control by endogenous gene regulatory elements and should increase phenotype consistency. The UFlip integration approach provides a powerful platform for generating new Cre/lox genetic tools to modify gene function at endogenous loci in zebrafish. Maintained in the lab of M. McGrail  Genetic reagent (Danio rerio) rb1Δ7

Contact for reagent and resource sharing
Further information and requests for resources and reagents should be directed to Maura McGrail ( mmcgrail@ iastate. edu).

Floxed UFlip vector for CRISPR-targeted integration to generate conditional alleles
The UFlip version 1.0 floxed mRFP gene trap vector for targeted integration ( Figure 1A) was designed to be compatible with our previously published GeneWeld CRISPR-Cas9 targeted integration strategy (Wierson et al., 2020). The UFlip cassette was assembled in the pPRISM parent vector  which has Universal UgRNA sequences on and BfuAI and BspQI type II restriction enzyme sites for cloning 5' and 3' homology arms on either side of the cassette, as previously described (Wierson et al., 2020). In the UFlip vector, the homology arm cloning sites flank head to head oriented rox sites that sit outside alternating pairs of head to head loxP and lox2272 sites for stable inversion after Cre mediated recombination (Robles-Oteiza et al., 2015;Schnütgen et al., 2003). Internal to the rox and lox sequences is the RP2 gene trap (Clark et al., 2011), consisting of a gene trap Splice Acceptor-mRFP-ocean pout (Zoarces americanus) antifreeze gene transcriptional termination and polyadenylation sequence (Gibbs and Schmale, 2000), followed by a tissue-specific lens gcry1:BFP or heart myl7:BFP secondary reporter. To assemble UFlip an intermediate vector was created by adding EagI and ClaI restriction sites to the pPRISM backbone by PCR. Six pairs of oligos, each containing the sequence of one of the rox, loxP or lox2272 sites, with complementary overhangs were ligated into the EagI and ClaI digested backbone. All three reading frames of the RP2 gene trap plus secondary reporter was directionally cloned into AvrII and SacI restriction sites located inside the pairs of rox/loxP/lox2272 sites. The UFlip version 2.0 vector was generated by mRFP primary reporter in the first version of the UFlipv1.0 vector was subsequently replaced with the porcine teschovirus-1 polyprotein 2 A peptide fused to mRFP. A UFlip version with a gene trap primary reporter containing the 2A-KalTA4 cDNA (Distel et al., 2009) for signal amplification using with a UAS:reporter is also available.
Intronic sgRNA target site selection, UFlip homology arm design and targeted integration To identify CRISPR gRNA sites in introns, intronic sequences from four adult female and four adult make WIK fish were amplified from fin clip DNA using the proofreading enzyme KOD and sequenced to identify non-repetitive sequences that were shared among adults in the population. Gene specific and vector Universal UgRNA synthetic gRNAs with 2'-O-Methyl at first and last bases, 3' phosphorothioate bonds between first 3 and last 2 bases, were ordered from Synthego . gRNA efficiency was determined by co-injection of 25 pg gRNA plus 300 pg Cas9 mRNA into one-cell stage embryos, followed by PCR amplification of the targeted intron and analysis of heteroduplex formation by gel electrophoresis. Amplicons were Sanger sequenced and the sequences analyzed for indel efficiency using Synthego's Inference of CRISPR Edits (ICE) analysis software (https://ice.synthego.com/#/). The UFlip targeting vectors were built following the GeneWeld protocol . 48 bp 5' and 3' homology arms were designed to sequences flanking the genome intronic CRISPR-Cas9 target site. The homology arms were assembled by annealing complementary oligonucleotides with appropriate overhangs for cloning into the BfuAI and BspQI type IIS restriction enzyme sites that flank the UFlip cassette. To generate an active, gene 'off' allele, the 5' and 3' homology arms were cloned into the UFlip BfuAI and BspQI sites, respectively. To integrate the UFlip cassette in the passive, gene 'on' orientation, the position of the cloned 5' and 3' homology arms was reversed, with the 3' homology arm cloned into the BfuAI site upstream of the cassette, and the 5' homology arm cloned in the BspQI site downstream of the cassette. gRNA and homology arm oligonucleotide sequences are listed in Table 1.
For synthesis of Cas9 mRNA the expression vector pT3TS-nCas9n (Addgene #46757) (Jao et al., 2013) was linearized with Xba I (New England Biolabs, R0145S). One μg linearized vector was purified with the PureYield Plasmid Miniprep System (Promega, A1223) and used as template for in vitro synthesis of capped mRNA with the Ambion mMessage Machine T3 Transcription Kit (Thermo Fisher, AM1348). in vitro synthesized mRNA was purified with the RNA Clean and Concentrator Kit RCC (Zymo, R1013).
To target CRISPR-Cas9 driven integration of the UFlip cassette into intronic sites, a 2 nl volume containing 25 pg of genomic gRNA, 25 pg of UgRNA, 10 pg of UFlip targeting vector, and 300 pg Cas9 mRNA was co-injected into one-cell stage embryos from crosses between sequence validated adult fish. Larvae were screened at 3 dpf for expression of the lens gcry1:BFP or heart myl7:BFP secondary marker. Three to four BFP-positive embryos were selected for genomic DNA isolation and confirmation of on target integration by PCR sequence analysis of 5' and 3' junctions. BFP-positive sibling embryos were raised to adulthood. Primers, gRNA and homology arm oligonucleotide sequences are listed in Table 1 and Table 4.

Isolation of stable zebrafish rbbp4 and rb1 UFlip and stop-PRISM integration alleles
To identify founder fish transmitting a UFlip allele, adults were outcrossed to wild-type WIK and embryos were screened for expression of the UFlip secondary reporter expressing BFP in the lens (gcry1:BFP) or heart (myl7:BFP). To identify on target integration alleles, genomic DNA was extracted from individual BFP-positive embryos by digestion in 50 mM NaOH at 95 °C for 30 min and neutralization by addition of 1/10 th volume 1 M Tris-HCl pH 8.0. Genomic DNA/UFlip cassette 5' and 3' junctions were amplified by PCR with gene specific and UFlip primers listed in Table 4, followed by direct Sanger sequencing. BFP-positive sibling embryos from a founder that was transmitting a precise UFlip integration allele were raised to adulthood. F1 adult animals were fin clipped and the genotype of individuals confirmed by PCR and sequencing. Confirmed F1 adults were outcrossed to wild-type WIK, and F2 adults again were confirmed for the presence of a precise UFlip integration allele. Individual confirmed F2 adults were outcrossed to WIK to establish independent transgenic lines.
The rb1-stop-PRISM-myl7:GFP allele was generated with a pPRISM (PRecise Integration with Secondary Marker) GeneWeld targeted integration vector containing a cassette with splice acceptor followed by three copies of TGA, TAA, TAG, the ocean pout (Zoarces americanus) antifreeze gene transcriptional termination and polyadenylation sequence (Gibbs and Schmale, 2000), and a myl7:eGFP-βactin polyadenylation secondary reporter. Previously described 5' and 3' homology arms complementary to the rb1 exon 2 CRISPR target site Wierson et al., 2020) were cloned into the BfuAI and BspQI type IIS restriction enzyme sites flanking the stop-PRISM-myl7:eGFP cassette. 25 pg of rb1 exon 2 genomic gRNA, 25 pg of UgRNA, 10 pg of rb1stop-PRISM targeting vector, and 300 pg Cas9 mRNA were coinjected into one-cell stage embryos in a volume of 2 nl, and adult founders screened for transmission of the rb1-stop-PRISM-myl7:GFP allele. Precise 5' and 3' junctions were confirmed in heart myl7:eGFP expressing F2 fin clipped adults. field imaging were performed on a Zeiss SteREO Discovery V12 microscope equipped with an X-Cite 120 W Metal Halide lamp (Excilitas Technologies, X-Cite 120Q). Images were captured with a Cannon Rebel T3 camera using EOS Utility software (Cannon). Bright field and fluorescence images were merged in Photoshop (Adobe). Live imaging was also performed on a Leica M165 FC stereomicroscope equipped with EL 6000 light source and K5 passive cooled sCMOS camera using Leica LAS X Multi Channel Acquisition software. Live imaging of larval retina was performed on a Zeiss LSM800 laser scanning confocal microscope.

Quantification and statistical analyses
Quantification of proliferation and apoptosis was performed on three sections of immunolabeled head tissue for each individual, from three biological replicates of zebrafish embryos or larvae. Prism (GraphPad) software was used for two-tailed unpaired Student's t-test with mean ± s.e.m. statistical analyses and production of bar graphs.

Additional files
Supplementary files • Transparent reporting form • Source data 1. Gel image files.

Data availability
All data generated or analyzed during this study are included in the manuscript and supporting files. Source data files have been provided for Table 2 Figure 9. DNA constructs reported in this study have been deposited at Addgene in the Jeffrey Essner lab list (Plasmid IDs: 173886, 173887, 173888, 173889, 173890, 173891, 180007, 180008, 180009, 180010, 180011, 18012). DNA constructs and transgenic zebrafish lines are available on request to M. McGrail.