Endogenous zebrafish proneural Cre drivers generated by CRISPR/Cas9 short homology directed targeted integration

We previously reported efficient precision targeted integration of reporter DNA in zebrafish and human cells using CRISPR/Cas9 and short regions of homology. Here, we apply this strategy to isolate zebrafish Cre recombinase drivers whose spatial and temporal restricted expression mimics endogenous genes. A 2A-Cre recombinase transgene with 48 bp homology arms was targeted into proneural genes ascl1b, olig2 and neurod1. We observed high rates of germline transmission ranging from 10 to 100% (2/20 olig2; 1/5 neurod1; 3/3 ascl1b). The transgenic lines Tg(ascl1b-2A-Cre)is75, Tg(olig2-2A-Cre)is76, and Tg(neurod1-2A-Cre)is77 expressed functional Cre recombinase in the expected proneural cell populations. Somatic targeting of 2A-CreERT2 into neurod1 resulted in tamoxifen responsive recombination in the nervous system. The results demonstrate Cre recombinase expression is driven by the native promoter and regulatory elements of the targeted genes. This approach provides a straightforward, efficient, and cost-effective method to generate cell type specific zebrafish Cre and CreERT2 drivers, overcoming challenges associated with promoter-BAC and transposon mediated transgenics.

To streamline isolation of zebrafish Cre drivers for specific cell lineages and cell types, we applied our recently reported method for CRISPR/Cas9 short homology directed targeted integration 25 to generate Cre transgenic lines. As proof of principle we targeted a 2A-Cre cassette in frame into a coding exon in the proneural transcription factor genes ascl1, olig2 and neurod1 that define neural stem, progenitor and post-mitotic cell lineages [26][27][28][29][30][31][32] . We chose to target ascl1b and olig2 since both are differentially expressed in highly aggressive pediatric brain cancers characterized by embryonal-like, poorly differentiated tumors 33,34 , consistent with their role in progenitor cell function during neural development. We and others had shown previously ascl1b and olig2 are overexpressed in zebrafish embryonal brain tumor models, reinforcing the conservation of molecular mechanisms driving neuroectodermal tumor types 35,36 . Together with neurod1, a marker of early neural commitment and differentiation, ascl1b and olig2 are good candidates to generate Cre drivers that would be useful for functional studies in neurogenesis and brain tumor pathogenesis. The 2A-Cre and 2A-CreERT2 targeting vectors used for integration are part of our pPRISM vector series that contains cassettes with linked fluorescent secondary marker for allele tracking (Welker et al., in preparation). Recovery of 2A-Cre integration alleles was highly efficient, with frequencies ranging from 10 to 100%, similar to other cargos as reported previously 25 . As expected, the expression of functional Cre recombinase was restricted to cell populations defined by ascl1b, olig2 and neurod1. Somatic targeting of Cre-ERT2 into neurod1 shows tamoxifen inducible recombinase activity in the nervous system. Together, our results demonstrate CRISPR/Cas9 directed targeted integration is an efficient and robust method for isolating endogenous Cre drivers that reflect the endogenous promoter activity of the targeted genes.

Results and discussion
CRISPR/Cas9 mediated Knock-in of 2A-Cre into ascl1b, olig2, and neurod1. Proneural Cre drivers were generated by integration of a 2A-Cre recombinase cDNA cassette in frame into a coding exon of the zebrafish ascl1b, olig2 and neurod1 genes (Fig. 1). We used our recently published strategy for CRISPR/Cas9 precision targeted integration with short homology arms that is likely to drive integration by homology mediated end joining (HMEJ) 25 . The 2A-Cre targeting construct contains a secondary fluorescent marker cassette with the γ-crystallin (γ-cry) promoter driving EGFP expression (Fig. 1a). Efficient targeted integration after injection into embryos is driven by in vivo liberation of the homology arms flanking the cassette, which are released by Cas9 induced double strand breaks at the universal gRNA (UgRNA) sites in the vector (Fig. 1a).
sgRNAs were designed to CRISPR/Cas9 sites in exon 1 of ascl1b and exon 2 of olig2 and neurod1 ( Table 1). The sgRNA mutagenesis efficiency was tested by co-injection with Cas9 mRNA into 1-cell zebrafish embryos, followed by PCR amplification of the targeted exon and analysis of heteroduplex formation in the PCR product by gel electrophoresis. After confirmation of efficient mutagenesis, 48 bp of sequence on either side of the genomic target Cas9 was used to design and clone 5′ and 3′ homology arms into the 2A-Cre targeting vector. Zebrafish embryos were injected at the 1-cell stage with the genomic target site sgRNA, the UgRNA, Cas9 mRNA, and the 48 bp homology arm 2A-Cre targeting vector. At 3 days post-fertilization (dpf), larvae were screened for expression of the lens specific γcry:EGFP secondary marker (Fig. 1b), and positive larvae selected to raise to adulthood to test for germline transmission. Targeting of ascl1b, olig2 and neurod1 resulted in 58%, 48% and 37%, respectively, of F0 injected embryos showing EGFP expression in the lens (Table 1), demonstrating the presence of the injected targeting vector. Evidence of somatic targeted integration at the genomic site was confirmed by junction fragment PCR on injected embryo.
To test whether somatic integration of the 2A-Cre cassette led to expression of functional Cre recombinase in the expected neural cell populations, we injected the genomic sgRNA, UgRNA, Cas9 mRNA, and the donor vector into embryos from the transgenic ubi:Switch floxed reporter line Tg(ubi:loxP-EGFP-loxP-mCherry) (Mosimann et al., 2011). Injected F0 mosaic Tg(ubi:loxP-EGFP-loxP-mCherry) embryos showed a switch from EGFP to mCherry expression in cells throughout the brain and retina, while EGFP expression remained in cells outside of the central nervous system (Supplementary Fig. S1). These results suggested that Cre expression was controlled by the endogenous regulatory elements of the targeted proneural genes, leading to recombination in the specific neural progenitor populations defined by those genes. Furthermore, the results suggest that the expressed Cre has recombinase activity that leads to recombination at loxP sites and excision of the floxed EGFP cassette. All neural progeny descended from the neural progenitor populations inherit the recombination event and express mCherry. The observed recombination in F0 injected animals indicated on-target integration of the 2A-Cre cassette was relatively efficient. Together, these results suggested targeted integration would be an effective method to generate endogenous Cre drivers that can promote spatially restricted cell-type specific Cre-mediated recombination.
Stable germline 2A-Cre knock-in alleles were established from adult F0s outcrossed to wildtype WIK to identify individuals transmitting the γcry:EGFP secondary marker to their progeny. We first measured the frequency of germline transmission of the γcry:EGFP secondary marker and found rates for ascl1b, olig2 and neurod1 of 100%, 10% and 20%, respectively (Table 1). F1 embryos positive for EGFP expression were tested for on-target precise integration at the genomic target site by PCR amplification and sequencing of the 5′ and 3′ genome/cassette junctions (Fig. 1, Supplementary Fig. S2). We found F0 transmission of precise integration alleles through the germline was relatively high, with frequencies at ascl1b, olig2 and neurod1 of 33% (1/3), 5% (1/20) and 20% (1/5), respectively (Table 1). These results at 3 independent loci demonstrate 2A-Cre in-frame targeted integration alleles can be efficiently recovered after screening a minimum of 20 F0 adults.   Cre in a pattern that recapitulates the targeted gene, we used whole mount in situ hybridization to examine Cre transcript localization. Adult F2 or F3 heterozygotes were outcrossed to wild type WIK and 3 dpf sibling larvae were hybridized with a probe complementary to Cre or to the target gene mRNA. Each of the lines showed that the pattern of Cre expression was identical to the endogenous gene mRNA (Fig. 2). ascl1b-2A-Cre/+ larvae hybridized with ascl1b or Cre probes showed the same pattern of signal in the forebrain, along the midbrain ventricle and midbrain-hindbrain border in the optic tectum, and on the dorsal surface of the hindbrain (Fig. 2a-d), as previously reported 37 . olig2-2A-Cre/+ larvae showed Cre expression matched olig2 in wild type larvae ( Fig. 2e-h) 38 , and was restricted to the posterior half of the cerebellum, a subset of cells in the neural retina, and the ventral forebrain. Like the endogenous neurod1 in wild type (Fig. 2i,j) 39,40 , Cre expression in neurod1-2A-Cre/+ was observed in the forebrain, adjacent to the midline in the midbrain and hindbrain, and absent from the peripheral midbrain layer of the optic tectum, consistent with neurod1 expression in committed progenitors and newborn post-mitotic neurons. It was highly expressed in the cerebellum and in the inner and outer neural layers of the retina (Fig. 2i-l). For all three lines, the consistency between the Cre and endogenous gene expression patterns suggested that both were controlled by the same regulatory elements, providing gene specific spatial and temporal Cre expression in neural progenitor populations defined by the proneural transcription factors.  (Fig. 3a). The offspring were imaged at 3 dpf and showed expression had switched from EGFP to mCherry in neurons in the developing forebrain, midbrain, hindbrain and retinas of the double heterozygous 2A-Cre; ubi:switch larvae ( Fig. 3b-d). ascl1b-2A-Cre led to near complete switching of EGFP to mCherry along the anterior-posterior axis of CNS, from the olfactory placode extending to the neural tube, with the exception of the neural retina (Fig. 3b). This is consistent with the early expression of ascl1b in the midbrain and hindbrain regions of the neural plate in 4-10 somite stage embryos, which broadens to cells in the anterior telencephalon, diencephalon, midbrain tegmentum and hindbrain by 25 somites/22 h post fertilization, but is completely absent from the developing retina 41,42 . Little switching occurred in cells outside of the CNS. These results indicated Cre expressed by the ascl1b-2A-Cre allele led to efficient recombination at the ubi:Switch transgene loxP sites and excision of the floxed EGFP cassette.
Similar to ascl1b-2A-Cre, the olig2-2A-Cre transgenic line led to recombination and switching from EGFP to mCherry expression throughout the CNS, including the retina (Fig. 3c). The pattern was consistent with early expression of olig2 in the neural plate presumptive ventral diencephalon at 8 hpf 43 ; forebrain dorsal thalamus, subpallium, posterior ventral thalamus and posterior tuberculum at 24-30 hpf 44 ; and along the forebrain ventricles, retina, midbrain/hindbrain border and cerebellum at 48 hpf 45,46 . The pattern of switching induced by the neurod1-2A-Cre line extended throughout neurons in the forebrain, midbrain, hindbrain, retina, and the PNS (Fig. 3d), as expected, given neurod1 is expressed in committed neural progenitors and early post mitotic neurons throughout the nervous system 47 . In single transgenic ubi:Switch larvae switching does not occur and only EGFP expression is detected (Fig. 3e).
Higher resolution analyses of recombinase switching in the ascl1b-2A-Cre, olig2-2A-Cre, and neurod1-2A-Cre 3 dpf larval midbrain and hindbrain were consistent with cell-specific expression of Cre recombinase (Fig. 4). F3 ascl1b-2A-Cre, F2 olig2-2A-Cre, and F2 neurod1-2A-Cre adults were crossed to the ubi:Switch line and RFP expression was examined in the 3 dpf larval dorsal tectum, cerebellum and hindbrain. In ascl1b-2A-Cre; ubi:Switch nearly all cells throughout the tectum (Fig. 4a-a′′), including cells lining the ventricle, expressed mCherry, consistent with the high levels of ascl1b and Cre transcripts detected in the ventricular zone of the midbrain by in situ hybridization (Fig. 3a-d). A subset of cells in the cerebellum had switched from EGFP to mCherry expression (Fig. 4a′′,b′′), which may be from the small number of cells expressing ascl1b in the cerebellum (Fig. 3a′,b′,c′), or the result of earlier recombination in progenitors during embryonic brain development. mCherry expression was detected throughout the hindbrain, as expected given the earlier widespread expression of ascl1b in stem/progenitors in the neural plate and embryonic brain regions, and the expression of ascl1b and Cre transcripts detected by in situ hybridization in a subset of cells lining the ventricles and located laterally in the dorsal 3 dpf hindbrain (Fig. 3a-d). A switch from EGFP to mCherry expression wasn't detected in sibling single transgenic ubi:Switch larvae (Fig. 4c-c′′,d-d′′). In olig2-2A-Cre; ubi:Switch the switch from EGFP to mCherry was detected in a subset of cells at the posterior region of the tectum (Fig. 4e-e′′), in the cerebellum (Fig. 4e,f arrowheads), in cells lining the hindbrain ventricle (Fig. 4f-f′′ V arrow), and in cells scattered throughout the hindbrain (Fig. 4e-e′′,f-f′′). In contrast, in neurod1-2A-Cre; ubi:Switch larval brains mCherry wasn't detected in cells lining the ventricle of the tectum or hindbrain (Fig. 4g,h, arrows). mCherry was present in cells throughout the cerebellum (Fig. 4g′,h′ arrowheads) and the midbrain and hindbrain parenchyma, as expected given neurod1 expression is induced after committed progenitors migrate away from the ventricle and transition to the neuronal state ( Fig. 4g-g′′,h-h′′). The mCherry expression was consistent with the strong pattern of in situ hybridization signal for neurod1 and Cre in the 3dpf midbrain and hindbrain (Fig. 3i-l) 4 . To test whether somatic integration of CreERT2 would lead to expression of tamoxifen responsive Cre recombinase, the 2A-CreERT2 cassette was targeted to the neurod1 exon 2 sgRNA site in nacre; ubi:Switch embryos that lack melanin pigment (Fig. 5a). At 6hpf injected embryos were treated with 5 µm 4-hydroxytamoxifen (4-OHT) in ethanol or ethanol alone and placed in a light-blocking container. On day 2 and day 3 the embryos were placed in fresh solution. At 3dpf, mCherry expression was detected in 17% (4/23) of untreated injected embryos (Fig. 5b-b′′) and 11% (2/19) of mock treated injected embryos (Fig. 5c-c′′) (Table 2), however the amount of switching was reduced compared to treated embryos. Recombination in the absence of treatment (Fig. 5b-b′′) or tamoxifen (Fig. 5c-c′′) may be caused by a stress response from injection, that alters hsp70 ability to tether CreERT2 at the cell membrane and in the cytoplasm. 45% (10/22) of injected embryos treated with 4-hydroxytamoxifen showed widespread mosaic mCherry expression throughout the brain and spinal cord (Fig. 5d-d′′,e-e′′; Table 2). Some examples of off targeting were detected in the gut (Fig. 5b′′ arrowhead) and skin (Fig. 5d′′ arrows). As expected, uninjected control embryos did not show mCherry expression (0/24) ( Table 2). neurod1-2A-CreERT2genomic DNA junction fragment analysis showed evidence of integration that correlates with switching in targeted embryos (Supplementary Fig. S3). Together, these results show targeted integration of Cre-ERT2 leads to endogenous expression of tamoxifen inducible Cre recombinase.
In summary, in this study we applied our CRISPR/Cas9 short homology directed targeted integration strategy 25 to generate a set of zebrafish proneural Cre drivers that express functional Cre recombinase in the pattern of the endogenous targeted ascl1b, olig2, and neurod1 genes. Somatic targeting of 2A-CreERT2 into neurod1 www.nature.com/scientificreports/ in ubi:Switch embryos showed switching in the nervous system after treatment with tamoxifen, suggesting our approach can be used to generate endogenous CreERT2 lines that express inducible Cre recombinase activity. We were able to efficiently recover precise in-frame 2A-Cre recombinase germline integration alleles after screening a minimum of only 20 adult F0 fish. The frequencies of allele recovery were similar to our previous report describing efficient knock in of 2A-RFP and 2A-EGFP reporter cassettes at eight zebrafish loci 25 . Although ascl1b, olig2, and neurod1 do not show haploinsufficiency, for other loci it may be beneficial to maintain the wild type endogenous gene expression in the targeted allele. Previous work showed homozygous otx2:CreERT2 embryos generated by CreERT2 integration upstream of the zebrafish otx2 initiating ATG were morphologically normal, but endogenous otx2 expression was reduced 24 . In contrast to our method, the frequency of on target integration using 1000 bp of 5′ homology was 3%, and the targeted cassette did not contain a secondary marker for rapid visual genotyping. Nagy et al. (2019) inserted Cre before the translation termination codon in the mouse megakaryocyte/platelet-specific Gp1ba gene, which led to a decrease in GPIbα protein expression and multiple platelet defects 48 . These studies suggest the potential for targeted integration to negatively impact endogenous gene expression will be specific to gene and location, and may require testing alternative design strategies. Our efficient and simplified method for generating zebrafish transgenic Cre driver lines by CRISPR targeted integration can be readily applied to other loci, expanding opportunities to develop new recombinase genetic tools for investigation of specific cell types, developmental stages, or disease states. CRISPR sgRNAs target sites in the coding sequence were selected and tested for efficient indel formation by co-injection of 25 pg sgRNA plus 300 pg Cas9 mRNA into 1 cell stage wild type WIK zebrafish embryos, followed by PCR amplification of the targeted exon and analysis of heteroduplex formation by gel electrophoresis. 48 bp 5′ and 3′ homology arms were designed and cloned into the pPRISM-2A-Cre vector as described 25 . sgRNA and homology arm oligonucleotide sequences are listed in Supplementary Table S1. For targeted integration the injection mix contained 25 pg of genomic sgRNA, 25 pg of UgRNA, 10 pg of targeting vector, and 300 pg Cas9 mRNA. Wild type WIK zebrafish embryos at the 1-cell stage were injected with 2 nl of the injection mix and Isolation of stable integration alleles and genome/vector junction analysis. To identify adult F0

Ethics declarations and approval for animal experiments.
founders, adult fish were outcrossed to wild type WIK and at least 75 embryos were screened for lens-specific EGFP expression. EGFP positive F1 embryos were selected for genome/vector junction analysis. Genomic DNA was extracted by digestion of single embryos in 50 mM NaOH at 95 °C for 30 min and neutralized by addition of 1/10th volume 1 M Tris-HCl pH 8.0. Both 5′ and 3′ junctions were amplified by PCR using the primers listed in Supplementary Table S2, TOPO-TA cloned and sequenced. EGFP positive F1 embryos from transmitting founders were raised to adulthood and confirmed by fin clip. F1 adult animals with precise integration events were outcrossed to wild type WIK and single F2 adults were used to establish independent transgenic lines.
In situ hybridization and live embryo imaging. Embryos were obtained from ascl1b-2A-Cre/+, olig2-2A-Cre+ and neurod1-2A-Cre/+ outcrossed to wild type WIK, and placed in embryo media with 0.003% 1-phenyl 2-thiourea (PTU) at 24 h post fertilization to block pigment production. Whole mount in situ hybridization was performed as described previously 52 . 3 dpf larvae were fixed overnight at 4 °C in 4% paraformaldehyde or 4% paraformaldehyde/4% sucrose in PBS. neurod1, olig2 and ascl1b cDNAs were cloned by RT-PCR using total RNA isolated from 3 dpf embryos and SuperScript III (Invitrogen). Primers for reverse transcription and PCR are listed in Supplementary Table S2. Digoxigenin-labeled probes were generated from linear plasmid DNA using DIG RNA Labeling Mix (Roche) and hybridized probes were detected with anti-digoxigenin antibody (Anti-Digoxigenin-AP, Roche) and NBT/BCIP (Roche). Larvae were imaged on a Zeiss Axioskop 2 and photographed with a Cannon Rebel T3 camera. Larvae for live imaging were placed in embryo media with 0.003% 1-phenyl 2-thiourea (PTU) at 24 h post fertilization to block pigment production. At 2 or 3 dpf the PTU treated embryos were anesthetized in 160ug/ml tricaine methane sulfonate and mounted in 1.2% low-melting agarose/160ug/ml tricaine methane sulfonate. Images were captured on a Zeiss LSM 700 or Zeiss LSM 800 laser scanning confocal microscope.