Maternally provided gene products regulate the earliest events of embryonic life, including formation of the oocyte that will develop into an egg, and eventually into an embryo. Forward genetic screens have provided invaluable insights into the molecular regulation of embryonic development, including the essential contributions of some genes whose products must be provided to the transcriptionally silent early embryo for normal embryogenesis, called maternal-effect genes. However, other maternal-effect genes are not accessible due to their essential zygotic functions during embryonic development. Identifying these regulators is essential to fill the large gaps in our understanding of the mechanisms and molecular pathways contributing to fertility and to maternally regulated developmental processes. To identify these maternal factors, it is necessary to bypass the earlier requirement for these genes so that their potential later functions can be investigated. Here, we report reverse genetic systems to identify genes with essential roles in zebrafish reproductive and maternal-effect processes. As proof of principle and to assess the efficiency and robustness of mutagenesis, we used these transgenic systems to disrupt two genes with known maternal-effect functions: kif5ba and bucky ball.
Maternally provided gene products regulate the earliest events of embryonic life, including formation of the oocyte, egg and eventually the embryo. Disruption of oocyte development or early embryogenesis causes congenital anomalies and are apparent in 2-5% of human births according to the National Institute of Child Health and Human Development. Chromosomal aneuploidy underlies some birth defects, but the genetic basis for others remains poorly defined (Ambartsumyan and Clark, 2008; Hassold et al., 2007). The most devastating mutations disrupt early embryogenesis when the diverse cell types necessary to build an embryo and the basic body plan are forming. Mutations occurring in later development are detectable in the children born with the consequent congenital deviations. In contrast, mutations in genes essential for processes occurring before implantation or during gastrulation result in embryonic lethality (Ambartsumyan and Clark, 2008; Hassold et al., 2007). In mammals, embryonic development occurs in utero, so mutations disrupting essential regulators of early embryogenesis often go undetected due to arrest in utero and miscarriage (Hassold et al., 2007; Hassold and Hunt, 2007; Zhao et al., 2006). Consequently, our understanding of the molecular and genetic regulation of this extremely sensitive developmental period remains incomplete.
In fish and humans, the earliest developmental events are regulated by maternally supplied gene products because the early zygote is transcriptionally silent (Abrams et al., 2020; Marlow, 2010; Sato et al., 2019; Vastenhouw et al., 2019; Wu and Vastenhouw, 2020; Zhao et al., 2006). These maternally supplied genes are known as maternal-effect genes because the embryo relies on gene function supplied by its mother. Mutations disrupting these genes in the mother affect the progeny regardless of the genotype of the embryo. Where studied, the basic aspects of oocyte development are remarkably conserved among animals, including regulation of meiotic initiation and arrest (Marlow, 2010). A number of maternal-effect genes have been discovered in the mouse; however, the precise contribution of maternal-effect genes is often masked by in utero arrest of the embryos (Marlow, 2010; Wu and Dean, 2020; Zhao et al., 2006). That is, although the gene is essential, the phenotype detected is lack of or underrepresentation of genotypically mutant progeny (if any) of mutant mothers, which hinders determining the specific cause of embryonic arrest. The external fertilization and development of zebrafish, and the large numbers of progeny produced weekly allow the recovery and examination of every egg produced, and the determination of the cellular and molecular basis of the developmental disruption.
Forward genetic screens have provided invaluable insights into the molecular regulation of embryonic development, including contributions of some maternal-effect genes (Dosch et al., 2004; Pelegri et al., 2004; Pelegri and Mullins, 2004; Wagner et al., 2004). However, maternal-effect genes with additional essential zygotic functions during embryogenesis are missed because the mutants do not reach reproductive maturity. Indeed, although the zebrafish is an excellent genetic system, traditional mutagenesis strategies and modern reverse genetic approaches alone have not permitted straightforward identification of maternal functions of zygotic lethals (Doyon et al., 2008; Foley et al., 2009a,b; Lawson and Wolfe, 2011; Moens et al., 2008; Sander et al., 2011a,b; Sood et al., 2006). To identify these factors, one needs to bypass the early zygotic requirement of the gene so that potential reproductive or maternal functions can be investigated. Methods to circumvent zygotic lethal phenotypes in the zebrafish were pioneered in the Schier and Raz labs (Ciruna et al., 2002). Their germline replacement approach takes advantage of early separation of somatic and germline lineages to generate animals with normal somatic composition and mutant germlines through host germline ablation and transplantation (replacement) with mutant germ cells (Ciruna et al., 2002). This strategy allows the animal to survive to produce mutant gametes, which can be examined for reproductive and maternal-effect phenotypes. Although this approach has been applied to examine the function of specific genes (Bennett et al., 2007; Borovina et al., 2010; Ciruna et al., 2002, 2006; Williams et al., 2018), thus far, no systematic germline replacement screen of zebrafish zygotic lethal mutations has been attempted because this approach is challenging and inefficient.
One drawback of dead end-mediated germline replacement is that few females are produced (Ciruna et al., 2006, 2002). This clearly impedes studies of oocyte development or of maternal-effect functions. This male bias is in part due to insufficient numbers of donor PGCs to support female-specific gonadogenesis. Consequently, dead end morphant embryos become sterile males (Ciruna et al., 2002; Siegfried and Nusslein-Volhard, 2008; Slanchev et al., 2005; Weidinger et al., 2003). Additional evidence suggests that signals from oocytes support female gonadogenesis (Bertho et al., 2021; Cao et al., 2019; Dranow et al., 2016, 2013; Hartung et al., 2014; Kaufman et al., 2018; Rodriguez-Mari et al., 2010; Rodriguez-Mari and Postlethwait, 2011, 2011; Romano et al., 2020; Wu et al., 2020). Specifically, diminished oocyte numbers or stages result in masculinization and female to male sex reversal (Kaufman et al., 2018; Rodriguez-Mari et al., 2010, 2011; Rodriguez-Mari and Postlethwait, 2011; Romano et al., 2020; Wu et al., 2020). To circumvent this problem, we used a transgenic mutagenesis approach to generate mosaic gonads in which germ cells carrying mutagenic cassettes and potential mutations are marked with fluorescent reporters.
Here, we report reverse genetic systems to identify genes with reproductive and maternal-effect functions. This approach will be particularly useful for genes whose maternal-effect functions are masked by earlier zygotic roles in embryogenesis. This transgenic approach selectively mutates the germline and thus allows the animal to survive to produce mutant gametes, which can be examined for reproductive or maternal-effect phenotypes. As proof of concept, we used this system to disrupt two genes with known reproductive and maternal-effect phenotypes, kinesin 1 (kif5ba) and bucky ball (buc). The potential to examine the function of every gene in its genome makes the zebrafish an extremely powerful vertebrate system to unravel molecular and genetic control of developmental processes and of adult physiology and disease. Complete phenotypic characterization of the zebrafish phenome will significantly improve our understanding of processes that are difficult to access in mammals, in particular maternal-effect processes.
RESULTS AND DISCUSSION
Vector based system to generate germline and maternal-effect mutants
Traditional mutagenesis strategies and modern reverse genetic approaches alone have only provided limited access to zebrafish maternal-effect genes (Doyon et al., 2008; Foley et al., 2009a,b; Lawson and Wolfe, 2011; Moens et al., 2008; Sander et al., 2011a,b; Sood et al., 2006). To access these genes, we developed a Gateway plasmid-based system for germline specific mutagenesis based on previous work (Ablain et al., 2015; Kwan et al., 2007; Villefranc et al., 2007; Walhout et al., 2000). CRISPR/Cas9 mutagenesis is now a standard method in zebrafish and other organisms, and biallelic conversion events have been widely observed in mitotic cells (Ablain et al., 2015; Auer et al., 2014a,b; Barrangou, 2013; Blackburn et al., 2013; Hruscha et al., 2013; Hwang et al., 2013a,b). However, the effectiveness of CRISPR/Cas9 in meiotic cells, when there are four copies of each chromosome and distinct checkpoints and repair pathways, is unknown.
Briefly, we generated mutagenesis cassettes that include selectable markers [tissue-specific expression of fluorescent proteins (FPs)] and that express target guide RNAs ubiquitously (U6 promoter) and Cas9 from germline promoters [bucky ball (female meiotic cells) (Heim et al., 2014) and ziwi (all germ cells) (Leu and Draper, 2010); Fig. 1A,B, Fig. S1 and S2]. Using this approach, the germline is marked by GFP when the germline promoter is activated, and the cardiac myosin light chain promoter drives GFP in the heart to allow for earlier selection (Fig. 1A,C, Table S1). To generate transgenic animals, these cassettes, along with transposase RNA, were injected into embryos to generate stable lines by Tol2-mediated transgene integration (Kawakami, 2005, 2007; Kawakami et al., 2004, 1998). We anticipated mutations would be induced later by the buc mutagenesis cassette rather than using the ziwi mutagenesis cassette because the buc promoter is activated later in more advanced female germ cells compared with the ziwi promoter, which is expressed early in mitotic germ cells (Heim et al., 2014; Leu and Draper, 2010) (Fig. 2A). Hereafter, we refer to constructs and lines expressing cas9 from the buc promoter as OMS for ovary mutagenesis system and those from the ziwi promoter as GMS for germline mutagenesis system. Because buc is activated only in females, founder males can be used to propagate the transgenes and to generate mutant alleles in subsequent generations. This will be valuable for mutations that cause female sterile phenotypes (oocytes arrest and no eggs are produced) or if the maternal-effect phenotypes are nonviable, e.g. buc mutants. Significantly, even if oocyte arrest occurs, histological assays can be used to examine affected gametes because the transgenic oocytes (OMS and GMS) or sperm (GMS) are marked with fluorescent reporters (Figs 1 and 2C-F). By sequencing the targeted region in marked oocytes or eggs, mutations induced in the germline can be identified.
Validation of OMS and GMS mutagenesis systems
Here, we report cassettes and recovered transgenic (Tg+) founders targeting two genes, kif5ba and buc (see Tables S1 and S2) both of which have known maternal-effect functions. Founders were identified based on GFP expression in their hearts and, in the case of females, transmission of GFP to their progeny. Although GFP should be a proxy for Cas9 because both proteins are produced from the same transcript (Fig. 1A), we confirmed that maternal Cas9 expression, like GFP, persists in embryos (Fig. 2B). Analysis of GFP-positive progeny of founders and F1 parents indicates that ubiquitous expression of guide RNAs is not toxic to germ cells (Fig. 2, Fig. S3). Similarly, germline Cas9 driven by buc (OMS) or ziwi (GMS) promoters is not toxic as fertile adults were recovered (Fig. 2C-E).
Phenocopy of Mkif5ba with mutagenesis vectors
Zebrafish kinesin I genes kif5ba and kif5bb function redundantly to promote craniofacial morphogenesis (Santos-Ledo et al., 2017). Double mutants lacking both fail to undergo proper jaw morphogenesis and are inviable (Santos-Ledo et al., 2017). Zygotic mutants disrupting kif5ba alone are viable; however, embryos lacking maternal kif5ba (Mkif5ba) fail to properly localize axis and germline determinants and consequently have dorsal-ventral patterning defects and lack primordial germ cells (PGCs) (Fig. 3A) (Campbell et al., 2015). As a proof of concept, we used the vector mutagenesis systems to disrupt kif5ba. We cloned a previously validated guide targeting the motor domain of kif5ba (Campbell et al., 2015) into the GMS cassette (hereafter called Tg:GMS:kif5ba) (see Tables S1 and S2). The restriction enzyme MboI cuts the wild-type allele; however, when mutated, MboI no longer cuts the mutant allele (Campbell et al., 2015). We used this assay to analyze the mutagenesis frequency in six GFP+ F1 progeny of a Tg:GMS:kif5ba male founder and their F2 progeny (Fig. 3B, Fig. S3A,B). Analysis of genomic DNA from somatic tissue (fin) revealed that, as expected, if no germline mutations were induced, five of the six F1-progeny had homozygous wild-type somatic tissues (Fig. 3B). However, one female (female 3) was heterozygous, indicating de novo mutation of kif5ba occurred in her father's sperm (Fig. 3B). Next, we examined the progeny from pairwise intercrosses of Tg:GMS:kif5ba F1s to screen for germline mutations (Fig. 3B). Normally, a cross between two homozygous wild-type fish yields only homozygous wild-type progeny, whereas a cross between a homozygous wild-type fish and a heterozygote yields half homozygous wild-type and half heterozygous progeny. Instead, we found deviations from these expected genotypes in the progeny of Tg:GMS:kif5ba carriers, indicative of CRISPR/Cas9-mediated germline mutation (Fig. 3C). Sequencing of F2 progeny confirmed that new mutations were induced in the germline of Tg:GMS:kif5ba F1s. Both in frame and deleterious mutations were recovered (Fig. 3D).
Having confirmed induction of germline mutations, we examined the germline marker nanos3 to determine whether germ cells were present in the GFP+ progeny of Tg:GMS:kif5ba F1 females (Fig. 3, Fig. S4). As expected for mosaic loss of maternal kif5ba function, a fraction of the progeny from each Tg:GMS:kif5ba F1 female lacked germ cells expressing nanos3 (Fig. 3E, Fig. S4). The penetrance of phenotypic embryos was nonmendelian and varied from female to female (ranging from 21-91%), with the highest frequency of phenotypic progeny from the Tg:GMS:kif5ba F1 mother that was already heterozygous for a mutation at the kif5ba locus. In addition, maternal kif5ba promotes dorso-ventral (DV) patterning by promoting the parallel vegetal microtubule array that mediates asymmetric distribution of dorsal factors (Campbell et al., 2015). To determine whether GMS-induced alleles recapitulated this Mkif5ba phenotype, we examined the phenotype of embryos from Tg:GMS:kif5ba F1 females at day 1 (d1). As expected, dorso-ventral phenotypes ranging from mild to severe dorsalization and axis duplication (V1 to V5 based on Kishimoto et al., 1997) were observed (Fig. 3F-J) among the green heart+ progeny – the heart was not scoreable in embryos with duplicated axes (Fig. 3F, Fig. S3).
To confirm that mutations were induced in the germline and not the soma, we injected kif5ba guide RNA and Cas9 protein into kif5bbe6/e6 homozygous mutants to determine whether the resulting somatic cell ‘crispants’ phenocopied the zygotic craniofacial defects observed in kif5ba;kif5bb double mutants at d5 (Santos-Ledo et al., 2017). As expected, jaw extension was compromised in ‘crispants’ (Fig. 3K,M). Next, we injected the gms:kif5ba plasmid system into kif5bb homozygous mutants to determine whether jaw defects were observed, which would be expected if there were leakage or somatic cell mutations. We observed no jaw defects in green heart-positive (transgenic+) or in their transgene negative siblings on d5 (Fig. 3L,M). Moreover, the transgenic+ animals were viable to adulthood, further indicating that somatic mutations were not induced because kif5ba;kif5bb double and compound mutant (mutant;heterozygote) fish are not viable (Santos-Ledo et al., 2017). Based on these results, we conclude that GMS-induced mutations in kif5ba can effectively phenocopy traditional maternal-effect loss of function kif5ba phenotypes, and that this system can be used to bypass somatic-lethal mutations.
Phenocopy of buc
Loss of buc results in failure to establish the animal-vegetal axis (Bontems et al., 2009; Dosch et al., 2004; Heim et al., 2014; Marlow and Mullins, 2008). To test the OMS and GMS systems at another locus, we generated guide RNAs targeting exon 4 of the buc gene and confirmed that the guides were mutagenic in transient assays (Fig. 4A, Tables S1 and S2). Mutagenic guides were cloned into the mutagenesis vectors and the resulting OMS:buc and OMS:buc plasmids were sequenced (Fig. 4A, Tables S1 and S2). We recovered one female and four male founders for OMS:buc (Fig. 4B,C, Fig. S3C). The single female OMS:buc founder produced embryos with either wild-type animal-vegetal polarity (n=22; 91.6%) or lacking polarity (n=2; 8.4%) (Fig. 4B,C, Fig. S3C). Thus, confirming the GMS system can be used to generate and assess maternal-effect phenotypes in just one generation, a significant advantage over traditional screens for maternal-effect functions in which phenotypes are detectable after four generations (Dosch et al., 2004; Wagner et al., 2004) and diploidized haploid screens (Pelegri et al., 2004; Pelegri and Mullins, 2004; Pelegri and Schulte-Merker, 1999).
To detect de novo mutations, T7 endonuclease assays were performed using genomic DNA from somatic tissues of the F1 progeny of two founder males. These analyses revealed 29% and 20% mutagenesis frequencies (Fig. 4D), indicating that expressing Cas9 from the ziwi promoter in the presence of guide RNA targeting buc induced germline mutations in founder males. Sequencing of genomic DNA confirmed the mutations detected by the T7 assays and revealed that all four F1 progeny of founder male E2 (three males and one female) carried the same mutation and that four female offspring from founder male E3 had a different mutant allele of buc (Fig. 4D, Fig. S5). Next, we examined the F1 females for buc phenotypes, specifically no animal-vegetal polarity and multiple micropyles – a somatic cell fate that is expanded in buc mutants (Heim et al., 2014; Marlow and Mullins, 2008). As expected for buc mutation, five F1 females produced progeny with buc phenotypes ranging in penetrance from 6-53% (Fig. 4E,I).
Next, we generated stable transgenic lines for the OMS:buc alleles (Fig. S3). We recovered several male founders and one female founder whose progeny had wild-type animal-vegetal axes (Fig. 4F-G′) or lacked polarity (Fig. 4H,H′). As expected, the OMS system, which is expressed in meiotic cells when there are four copies of each chromosome that must be mutated, was less efficient than the GMS system, which is expressed earlier in mitotic germ cells which have only two copies of each chromosome.
Comparison of the OMS and GMS systems at two loci indicates that these systems can achieve disruption of gene function specifically in the germline at frequencies that match or exceed those of zygotic recessive alleles. The OMS system appears to be less efficient, possibly due to timing of expression (mitotic versus meiotic) and/or levels of expression from the promoters. Nonetheless, it may be suitable for genes transcribed in later oocytes, or for mutations that cause sterility using the GMS system. Detection of in-frame mutations indicates that phenotypic manifestation is an underrepresentation of mutagenesis efficiency, which encompasses both deleterious and non-deleterious mutations. Nondisruptive mutations are potentially limiting because altered target sites cannot be further mutated. Here, only a single guide RNA was used; however, including multiple guides arrayed in tandem may yield large disruptive deletions. Although the frequency of phenotype detection varies for both GMS and OMS mutagenesis, this approach represents a significant advance in the tools available to study maternal-effect genes.
MATERIALS AND METHODS
Mutant fish strains were generated using Crispr-Cas9 mutagenesis with modifications (see the plasmids list in Table S1) to the plasmid backbone published previously (Ablain et al., 2015). All procedures and experimental protocols were performed in accordance with NIH guidelines and were approved by the Einstein (protocol #20140502) and Icahn School of Medicine at Mount Sinai Institutional (ISMMS) Animal Care and Use Committees (IACUC #2017-0114).
All primers are listed in Table S2.
OMS and GMS mutagenesis plasmids
OMS and GMS plasmids (Table S1) were created using the tissue-specific promoter system described by Ablain et al. (2015) (Fig. 1, see supplementary Materials and Methods). In brief, we digested a p3E_polyA_U6:gRNA (Fig. 1B; Ablain et al., 2015) using BseRI enzyme and then inserted annealed gene-specific gRNA targeting kif5ba and buc. For the gRNAs, previously validated gRNA sequences were used to target kif5ba (Campbell et al., 2015), and to target buc, new gRNAs were tested for mutagenic activity (Table S2). Gateway recombination reactions were then used to generate expression constructs with Cas9 driven by the bucky ball (Heim et al., 2014) or ziwi (Leu and Draper, 2010) promoter. Recombination order was confirmed by sequencing using the ziwi promoter or buc promoter, and the cas9 and U6 promoter primers (Table S2).
Stable transgenic lines
To generate stable OMS and GMS transgenic lines, Tol2 Transposase RNA was transcribed from pCS2FA-transposase (Kwan et al., 2007), and combined with OMS or GMS vector circular DNA (25 ng/μl each). Embryos were injected with 1 nl of the plasmid/transposase solution at the one-cell stage. Embryos with GFP-positive hearts were selected at day 2 (d2) and raised to generate founders.
The previously validated gRNAs targeting kif5ba (Campbell et al., 2015) and Cas9 or the gms:kif5ba mutagenesis vector (as described above) were injected into kif5bbe6/e6 mutants. Larvae were scored for craniofacial morphology and swim bladder inflation at d5.
Mutation detection and genotyping
Genomic DNA was extracted from adult fins using standard procedures (Meeker et al., 2007). The genomic region surrounding the kif5ba target sequence was amplified using the primers 5′-GGAGTGCACCATTAAAGTCATGTG-3′ and 5′-GTCGGTGTCAAATATTGAGGTC-3′. The genomic region surrounding the buc target sequence was amplified using the primers 5′-TGCAGTATCCTGGCTATGTGAT-3′ and 5′-ACCACATCAGGGGTAGAAGAGA-3′ (Table S2). Products were then digested with T7 endonuclease and visualized on a gel to identify restriction patterns indicative of induced mutations.
Sequencing new alleles
Genomic DNA was extracted from adult fins using standard procedures (Meeker et al., 2007). The genomic region surrounding the kif5ba target sequence was amplified using the primers 5′-GGAGTGCACCATTAAAGTCATGTG-3′ and 5′-GTCGGTGTCAAATATTGAGGTC-3′. The genomic region surrounding the buc target sequence was amplified using the primers 5′-TGCAGTATCCTGGCTATGTGAT-3′ and 5′-ACCACATCAGGGGTAGAAGAGA-3′ (Table S2). After 35 cycles of PCR at 59°C and 57°C, for kif5ba and buc, respectively, PCR fragments were directly TA cloned into pCR4-TOPO vector (K457502, Invitrogen). After transformation, mini-prep DNA was prepared using Qiagen kits, and each plasmid was sequenced using universal primers on the vector. In this way, both the wild-type and any mutant alleles were detected.
Fifty transgenic (identified by GFP expression) or non-transgenic (GFP negative) eggs were pooled together at 2 hpf. Samples were flash frozen and stored at −80°C. Samples were resuspended in 2×sample buffer with DTT at 1 µl/embryo or larvae. Samples were homogenized with a motorized pestle, centrifuged for 1 min, and incubated and boiled for 5 min prior to loading. 10 μl per sample was loaded in a 4-12% SDS-PAGE gel and proteins were transferred to PVDF membranes. Membranes were blocked in 5% milk in PBS for 1 h at room temperature. Anti-CRISPR-Cas9 antibody (Abcam, ab204448) was used at 1:1000 and incubated overnight. Membranes were washed for 3×5 min in TBS-Tween and then for 2×5 min in TBS. Rabbit-HRP secondary antibody was diluted 1:5000 and incubation was for 1 h at room temperature. Membranes were washed for 3×5 min in TBS-Tween and then for 2×5 min in TBS. Proteins were detected with ECL-Plus and chemiluminescence was imaged using a BioRad imager.
In situ hybridization
For in situ hybridization, embryos at the specified stages, were fixed in 4% paraformaldehyde overnight at 4°C. In situ hybridization was performed according to Thisse et al. (2004), except hybridization was performed at 65°C. In addition, maleic acid buffer [100 mM maleic acid (pH 8), 150 mM NaCl] was substituted for PBS during the antibody incubations, and BM Purple was used to visualize the RNA probes (Roche, 1442074).
Immunostaining and imaging
For whole-mount immunofluorescence stained shield, 30 hpf embryos or ovaries, tissues were fixed in 3.7% paraformaldehyde overnight at 4°C. The following day the samples were washed in PBS, dehydrated by washing in methanol, and then stored at −20°C. To visualize germ cells, chicken anti-GFP antibody (Invitrogen, A10262) was used at a 1:500 dilution. Secondary antibodies Alexafluor488 or Alexafluor-Cy3 (Molecular Probes) were diluted 1:500. Samples were mounted in Vectashield with DAPI and images were acquired using a Zeiss Axio Observer inverted microscope equipped with Apotome II and a CCD camera, a Zeiss Zoom dissecting scope equipped with Apotome II. Image processing was performed in Zenpro (Zeiss), ImageJ/FIJI, Adobe Photoshop and Adobe Illustrator.
Eggs from transgenic mothers, day 10 (d10) trunks, and gonads dissected at d31-d34 from the specified transgenic genotypes were placed in RNA later and stored at −80°C until use. RNA later was removed and Trizol (Life Technologies) was added. RNA was extracted using the RNAeasy mini KIT (Qiagen) and the SuperScript IV VILO kit (Thermo Fisher) was used for cDNA preparation. RT-PCR was performed using the primers in Table S2. gRNAs were amplified using the target specific forward guide primer and the universal reverse primer. PCR products were resolved using a 1.5% Ultrapure agarose (Invitrogen) gel and visualized using a BioRad gel imager.
We thank members of the Marlow lab for helpful discussions, our animal care staff for fish care (Einstein and CCMS at ISMMS) and the Microscopy CoRE at Icahn School of Mount Sinai.
Conceptualization: O.K., F.L.M.; Methodology: S.B., O.K., F.L.M.; Validation: D.D., F.L.M.; Formal analysis: S.B., O.K., K.L., A.S.-L., F.L.M.; Investigation: S.B., O.K., K.L., A.S.-L., D.D., F.L.M.; Data curation: S.B., O.K., K.L., A.S.-L., D.D., F.L.M.; Writing - original draft: F.L.M.; Writing - review & editing: S.B., O.K., K.L., A.S.-L., D.D., F.L.M.; Visualization: S.B., F.L.M.; Supervision: F.L.M.; Project administration: F.L.M.; Funding acquisition: F.L.M.
This work was supported by the National Institutes of Health (R21-HD091456 to F.L.M.). O.K. was supported by the National Institutes of Health (T32-GM007288) and by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (F30HD082903). Deposited in PMC for release after 12 months.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.198010
The authors declare no competing or financial interests.