Stable Gene Silencing in Zebrafish with Spatiotemporally Targetable RNA Interference

Abstract

The ability to regulate gene activity in a spatiotemporally controllable manner is vital for biological discovery that will impact disease diagnosis and treatment. While conditional gene silencing is possible in other genetic model organisms, this technology is largely unavailable in zebrafish, an important vertebrate model organism for functional gene discovery. Here, using short hairpin RNAs (shRNAs) designed in the microRNA-30 backbone, which have been shown to mimic natural microRNA primary transcripts and be more effective than simple shRNAs, we report stable RNA interference-mediated gene silencing in zebrafish employing the yeast Gal4-UAS system. Using this approach, we reveal at single-cell resolution the role of atypical protein kinase Cλ (aPKCλ) in regulating neural progenitor/stem cell division. We also show effective silencing of the one-eyed-pinhead and no-tail/brachyury genes. Furthermore, we demonstrate stable integration and germ-line transmission of the UAS-miR-shRNAs for aPKCλ, the expressivity of which is controllable by the strength and expression of Gal4. This technology shall significantly advance the utility of zebrafish for understanding fundamental vertebrate biology and for the identification and evaluation of important therapeutic targets.

STUDIES of genetic model organisms contribute tremendously to our understanding of diverse biological processes and human disorders at molecular and cellular levels (Davis 2004; Aitman et al. 2011). As a recently established animal model, zebrafish has salient features that promise to provide new insights into biology and medicine. These include rapid external development, transparent embryonic and larval stages, simpler vertebrate organ systems, and amenability for genetic and chemical screening. The utility of zebrafish, however, has been hampered by the inability to silence genes in a spatiotemporally controllable manner. Although valuable reverse genetic methods are available (Nasevicius and Ekker 2000; Wienholds et al. 2002; Doyon et al. 2008; Meng et al. 2008; Huang et al. 2011; Sander et al. 2011; Bedell et al. 2012), none offer gene silencing at any desired stages of the life cycle and in any cell types of interest.

RNA interference (RNAi) is a powerful approach to dissect gene function (Fire 2007; Mello 2007). It was originally discovered in Caenorhabditis elegans, where the observation of gene inactivation by both sense and antisense RNAs (Guo and Kemphues 1995) led to the finding of double-stranded RNA (dsRNA)-mediated gene silencing (Fire et al. 1998). In vertebrates, the effectiveness of RNAi was hindered by the dsRNA-induced interferon response causing nonspecific effects (Manche et al. 1992; Stark et al. 1998) until the finding of efficacious small interfering RNAs (siRNAs) (Elbashir et al. 2001). However, their delivery into zebrafish remains either ineffective or has nonspecific effects (Gruber et al. 2005; Zhao et al. 2008). Likewise, simple short hairpin RNAs (shRNAs) driven by RNA polymerase III promoters (Paddison et al. 2002) also appears ineffective in zebrafish (Wang et al. 2010).

Micro-RNAs (miRNAs) are endogenous ∼21- to 23-nucleotide RNAs that can regulate gene expression. Originally discovered in C. elegans (Lee et al. 1993; Reinhart et al. 2000), they were later found conserved across animals and plants (Bartel 2009; Ebert and Sharp 2012). The observation that designed miRNAs can inhibit cognate mRNA expression in human cells (Zeng et al. 2002) has paved the way for recent developments of shRNAs employing the primary miR-30 backbones (miR-shRNAs) (Dickins et al. 2005; Silva et al. 2005). This natural configuration is reported 12 times more efficient than simple hairpin designs. MiR-shRNAs are recently shown to be effective in vivo in mice (Dickins et al. 2005; Premsrirut et al. 2011; Zuber et al. 2011) and zebrafish (Dong et al. 2009; De Rienzo et al. 2012). However, the capability of germ-line transmission is only reported by one study in zebrafish (Dong et al. 2009), where functional efficacy and specificity of silencing were not quantitatively documented. Therefore, the potential of RNAi for stable and conditional gene silencing in zebrafish remains uncertain.

With the goal of achieving spatiotemporal and dosage control of gene silencing, we have exploited the miR-shRNA technology in combination with the bipartite Gal4/upstream activating sequence (UAS) system (Ptashne 1988), which offers excellent versatility for spatiotemporal control together with amplification of gene expression level. Using atypical protein kinase Cλ (apkcλ), one-eyed pinhead (oep), and no tail/brachyury (ntl) as test loci, we show that RNAi in zebrafish is effective, rapid, dosage controllable, stable with conditional capability, and scalable.

Materials and Methods

Zebrafish strains

Wild-type embryos were obtained from natural spawning of AB adults. The transgenic line Tg[ubi-GFF] (Asakawa and Kawakami 2010) was a gift from K. Kawakami. The apkcλ mutant heart and soul was kindly provided by D. Stainier.

ShRNA design

ShRNA design was performed using the siRNA design tool from the following website: http://www.genscript.com/design_center.html, which ranks target candidates based on a proprietary algorithm using ΔE parameters. We further refined the list by removing the candidates that have more than three G/C in 6–11 nt of the target sequences, based on a thermodynamic study of siRNA (Ui-Tei et al. 2008).

Vector construction

The shRNA precursor structure was designed based on the miR-30e precursor in zebrafish. The miR-30e target and guide sequences were replaced with the predicted shRNA target and guide sequences. To best mimic the miR-30e precursor structure, we kept the original loop region and the flanking sequences of the miR-30e precursor and introduced two mismatched “TT” into the shRNA guide sequence between nt 12 and nt 13. We synthesized a pair of complementary oligos that include all the designed shRNA structures (guide sequence, loop sequence, target sequence, and the flanking sequences) and the annealed oligos were inserted into zebrafish miR-30e backbone in expression vectors kindly provided by the late Ting Xi Liu (Dong et al. 2009). The following 5xUAS sequence was used: cggagtactgtcctccgagcggagtactgtcctccgagcggagtactgtcctccgagcggagtactgtcctccgagcggagtactgtcctccg.

DNA and RNA injection

DNA and RNA injections were performed at the one-cell stage. For genetic mosaic analyses, RNAi constructs (UAS-TdTomato-shRNA1^apkcλ-miniTol2 or UAS-TdTomato-shRNA5^apkcλ-miniTol2; 75 pg per embryo) were co-injected with EF1α-GFF-PT2KXIG (elongation factor 1a regulatory element-driven Gal4) (50 pg per embryo) and Tol2 tranposase (10 pg per embryo). Sense strand-capped RNA was synthesized by SP6 transcription from NotI-linearized plasmid using the mMESSAGE mMACHINE system (Ambion). RNAi-resistant sense RNAs for apkcλ, ντλa ανδ οεπa, as well as RNAs encoding the oep-shRNA and ntla-shRNA, were injected at 400 pg per embryo. For time-lapse in vivo imaging, 3NLS:EGFP sense RNA (200 pg per embryo) was co-injected with EF1α-GFF-PT2KXIG DNA (50 pg per embryo) and 5xUAS-tdTomato-shRNA^apkcλ plasmid DNA (75 pg per embryo).

Heat shock of zebrafish embryos

Heat shock of zebrafish embryos was performed at the stage of 8 hours postfertilization (hpf) (75% epiboly). Fertilized embryos were collected from the cross between the UAS-TdTom-miR-shRNA1^apkcλ founder A and the hsp70-Gal4 transgenic animal, and were raised at 28.5° before heat shock. A 500-ml glass beaker filled with 100 ml egg water was preheated to 37° in a water bath right before heat shock. Around 100 embryos were transferred to the preheated egg water at 8 hpf and incubated in a 37° water bath for 1 hr. The embryos were then transferred to 28.5° egg water immediately and raised in a 28.5° incubator until analysis.

Phenotypic analyses

Phenotypic analyses were performed manually under a bright field dissection microscope. For apkcλ RNAi, four morphological phenotypes were analyzed, which are typically observed in the apkcλ mutant (heart and soul) (Horne-Badovinac et al. 2001): (1) defect in heart tube assembly, (2) patchy retinal pigmented epithelium, (3) failure of the brain ventricle to inflate, and (4) abnormal body curve. The number of embryos that are positive for each phenotype as well as the total number of all embryos was counted for both RNAi groups and control groups. Statistic analyses were performed based on positive percentage for each phenotype. For ntla and oep RNAi, the typical no-tail phenotype and one-eyed pinhead phenotype were analyzed.

Time-lapse in vivo imaging

To observe the division orientation of radial glia progenitor cells, time-lapse in vivo imaging was performed on embryos injected with 3NLS:EGFP sense RNA, EF1α-GFF-PT2KXIG DNA, and 5xUAS-tdTomato-shRNA^apkcλ-miniTol2 DNA (see DNA and RNA injection for detail) at 24 hpf. The imaging was performed as previously described (Dong et al. 2012). In brief, embryos were mounted in low melting point agarose and properly placed on a temperature-controlled stage of the confocal microscope. We used a Nikon C1 spectral confocal microscope with upright objectives. Fluorescently labeled individual neural progenitor cells were imaged for 2 hr with a fixed interval of 90 sec under a ×40 water-dipping objective. The parameters of confocal imaging were determined as sufficient to capture the orientation of cell division, while reducing photobleaching during the imaging period.

Immunohistochemistry

Immunohistochemistry was performed on whole mount embryos as described previously (Guo et al. 1999). Embryos were fixed with 4% paraformaldehyde. Embryos older than 60 hpf were treated with proteinase K before incubation in primary antibodies (rabbit anti-aPKC 1:100, Santa Cruz sc-216) at 4° overnight. After four washes with phosphate-buffered saline with 0.1% Tween, 0.5% Triton X-100, and 1% DMSO, the embryos were incubated with Alexa Fluor secondary antibody (1:2000, Invitrogen) at 4° overnight. After wash and glycerol sequential treatment, the embryos were mounted and imaged using Nikon C1 confocal microscope.

Quantitative real-time RT–PCR analyses

For the analyses of knockdown of apkcλ expression by transient RNAi, the control or RNAi embryos were collected at 10 hpf and total RNA from pooled embryos (20 embryos per sample for each group) was extracted using Trizol reagent (Invitrogen) followed by treatment with Turbo DNA-free DNase (Ambion). First-strand cDNA was reverse transcribed using oligo-dT primers and Superscript reverse transcriptase (Invitrogen). Real-time PCR amplifications of apkcλ, oep, ntla, and gapdhs were carried out with SYBR Green PCR Master Mix (Applied Biosystems). Primers used were:

• apkcλ F: 5′-CCCGCACCAAGTCCGGGTAA-3′,

• apkcλ R: 5′-GCTGAGAAGAAACGGTGCACGGA-3′;

• oep F: 5′-ATGGACTTTTGCATTGCTTCCCACA-3′,

• oep R: 5′-AGTGTTCTGAGGGAGCCCGACC-3′;

• ntla F: 5′-CCCAGCCATTACTCCCACCGC-3′,

• ntla R: 5′-TGGGCCAGGGTTCCCATCCC-3′;

• gapdhs F: 5′-ACTCCACTCATGGCCGTTAC-3′,

• gapdhs R: 5′-TGAGCTGAGGCCTTCTCAAT-3′.

Droplet digital PCR analyses

For the analyses of knockdown of apkcλ expression by stable RNAi transgenesis, the sibling control or RNAi transgenic embryos were collected at 24 hpf and total RNA from single embryos was extracted using Trizol reagent (Invitrogen) followed by treatment with Turbo DNA-free DNase (Ambion). First-strand cDNA was reverse transcribed using oligo-dT primers and Superscript reverse transcriptase (Invitrogen). Droplet Digital PCR amplifications of apkcλ and gapdhs were carried out with primers and probes carrying fluorophore modifications. The sequences of the primers are the same as used in qRT–PCR analyses. Additionally, FAM-modified apkcλ probe (5′-FAM-ATGTGCTCCATGGACAATGACCAGCT-3′) and HEX-modified gapdhs probe (5′-HEX-TTCCAGTGCATGAAGCCTGCTGAGAT-3′) were used. Data were analyzed using Quantasoft version 1.2.10.0.

Stable transgenesis

Stable UAS-miR-shRNA^apckcλ transgenic lines were generated using the Tol2 transposon system as described previously (Suster et al. 2009). RNAi constructs (UAS-TdTomato-MiR-shRNA1^apkcλ-miniTol2 or UAS-TdTomato-miR-shRNA5^apkcλ-miniTol2; 50 pg per embryo) together with Tol2 transposase (10 pg per embryo) were injected into one-cell-stage embryos of AB wild-type fish. Injected embryos were raised to adulthood (G0) and screened for germ-line transmission by crossing with Tg[Ubi-GFF] transgenic fish. The germ-line transgenic founders would yield embryos with TdTomato expression, which could be easily distinguished under the epifluorescent microscope. The embryos with ubiquitous TdTomato expression were raised to adulthood (F1), which carry both UAS-^apckcλ RNAi and Ubi-GFF transgenes. Additionally, UAS-RNAi G0 germ-line founders were crossed with AB wild-type fish. All the progenies were raised to adulthood and genotyped by PCR using TdTomato-specific primers to screen for stable transgenic F1 fish. Stable transgenic F1 were obtained for UAS-TdTomato-shRNA1^apckcλ and UAS-TdTomato-shRNA5^apckcλ.

Results and Discussion

Conditional miR-shRNA expression system

We designed shRNAs employing the primary miR-30 backbones (hence called miR-shRNAs) (Figure 1A). The fluorescent reporter TdTomato conveniently marked the cells that express the miR-shRNAs. Three genes were used as the test set: (1) The atypical protein kinase C lambda (aPKCλ), for which a distinct loss-of-function phenotype has been documented in zebrafish. We chose this gene also because its disruption produces a highly specific cellular phenotype, that is, an alteration of spindle rotation in dividing radial glia progenitor cells (Horne-Badovinac et al. 2001). (2) The no tail a (ntla) and (3) one-eyed pinhead (oep) genes, for which distinct loss-of-function phenotypes can be readily assessed (Schulte-Merker et al. 1994; Zhang et al. 1998). Using the Web-based shRNA design tool (www.genescript.com) together with the filtering criteria based on thermodynamic properties (Ui-Tei et al. 2008), six shRNAs targeting the apkcλ gene (Supporting Information, Figure S1), four shRNAs each that targeted the ntla (Figure S2), and oep (Figure S3) genes were selected.

Figure 1.

Functional validation of miR-shRNAs by transient in vivo transgenesis. (A) Diagram of the conditional miR-shRNA expression system. The guide stand (bottom) is highlighted in light blue. (B) Mosaic expression of miR-shRNA5^apkcλ (left), miR-shRNA4^oep (top right), and miR-shRNA4^ntla (bottom right) causes morphological defects similar to those observed in respective mutants, which can be rescued by delivery of shRNA-resistant wild-type mRNAs. The miR-30e vector-injected embryos serve as controls. (C) Quantitative RT–PCR analysis shows a gene-specific reduction of endogenous mRNA levels by shRNAs cognate to their target genes. The miR-30e vector-injected embryos serve as controls. (D) Fluorescent immunostaining of aPKCλ (green) shows knockdown effects at the protein level by different miR-shRNAs^apkcλ. In a single miR-shRNA^apkcλ-expressing cell, as indicated by the tdTomato fluorescence (red), miR-shRNA1^apkcλ (middle panel) and miR-shRNA5^apkcλ (left panel) lead to significant knockdown of aPKCλ, whereas the ineffective miR-shRNA7^apkcλ does not show obvious knockdown effect (right panel). (E) Mosaic expression of miR-shRNA5^apkcλ causes defects in mitotic division orientation of radial glia progenitors in the developing zebrafish forebrain. Left: examples of mitotic division orientation in the miR-30e vector-injected (control) and miR-shRNA5^apkcλ-expressing radial glia progenitors. Right: Quantification of division orientation in miR-30e vector control and miR-shRNA5^apkcλ-expressing radial glia progenitors (n = 27 for control, and n = 24 for miR-shRNA5^apkcλ). The nuclei of radial glia progenitors were labeled with 3NLS:EGFP (green) and the individual radial glia progenitors expressing UAS-miR30e vector (control) or UAS-miR-shRNA5^apkcλ were highlighted by tdTomato (red).

Transient in vivo RNAi is highly potent in knocking down gene activity

For functional validation of shRNAs, we employed a transient in vivo transgenesis method. Although the mosaic expression is an inevitable feature associated with transient transgenesis, this can be alleviated by coexpression of the Tol2 transposase (Li et al. 2010) and selection of high expressers with the visible reporter TdTomato (Figure S4). DNA plasmids carrying the Pef-1a-Gal4 (elongation factor 1a promoter-driven Gal4) and UAS-miR-shRNA together with the transposase were co-injected into one-cell-stage zebrafish embryos. Three of six miR-shRNAs targeting apkcλ, and three of four miR-shRNAs targeting either oep or ntla genes led to gene-specific phenotypes (with variable penetrance) that could be rescued by delivery of shRNA-resistant wild-type mRNAs for the respective genes (Figure 1B, Table S1, Table S2, and Table S3). Quantitative RT–PCR analysis using total RNAs from 10 hpf shRNA-expressing embryos showed a gene-specific reduction of endogenous mRNAs cognate to the shRNA target genes, and codelivery of two shRNAs (1 + 5) targeting apkcλ further decreased the endogenous apkcλ level (Figure 1C). Immunostaining of the aPKC protein (recognizing both λ and θ isoforms) showed at single-cell levels that the two most effective shRNAs, apkcλ-shRNA5 and apkcλ-shRNA1, diminished the apically localized aPKC protein in radial glia progenitors, whereas the ineffective apkcλ-shRNA7 did not (Figure 1D).

Transient in vivo RNAi is a powerful tool for genetic mosaic analysis

A useful application of transient RNAi is in vivo genetic mosaic analysis at single-cell resolution. Using time-lapse imaging, we further analyzed the miR-shRNA-expressing individual radial glia progenitor cells for the mitotic spindle rotation phenotype associated with the loss of apkcλ gene activity (Horne-Badovinac et al. 2001). While most control progenitors (>90%) displayed a division angle between 60° and 90° and none between 0° and 30° (Figure 1E and File S1), a significant portion of the apkcλ-shRNA5-expressing progenitors had a division angle between either 0° and 30° or 30° and 60° (Figure 1E and File S2), a ratio that was similar to what has been observed in the apkcλ mutant (heart and soul, has) retina (Horne-Badovinac et al. 2001). Together, these results validate the efficacy and specificity of miR-shRNAs in vivo and demonstrate the saliency of transient RNAi for in vivo genetic mosaic analysis at single-cell resolution.

Stable gene silencing with UAS-miR-shRNA^apkcλ transgenic lines

Next we established transgenic lines that carry the UAS-TdTom-miR-shRNA1^apkcλ or UAS-TdTom-miR-shRNA5^apkcλ transgene using the Tol2 transposon system. G₀ germ-line transgenic founders were identified as those that produced red fluorescent progeny when crossed with the Ubiquitous-Gal4 (Ubi-GFF) transgenic animal (Asakawa and Kawakami 2010). One of 40 UAS-TdTom-miR-shRNA1^apkcλ and 3 of 50 UAS-TdTom-miR-shRNA5^apkcλ founders transmitted the transgenes to germ line (Figure 2A, Table S4, and Table S5). We noted that among the F₁’s derived from the UAS-TdTom-miR-shRNA1^apkcλ founder (A) and Tg[Ubi-GFF] cross, ∼30% of the TdTom-positive embryos displayed a heart defect with cardiac edema that was strictly correlated with the expression of TdTom (Figure 2B, middle panels compared to top panels). This heart defect was rescued by the delivery of shRNA-resistant apkcλ mRNA (Figure 2, B, bottom panels, and C). Furthermore, to test for the spatiotemporal effect of RNAi, we crossed the UAS-TdTom-miR-shRNA1^apkcλ founder (A) with Tg[hsp-gal4], and subjected the resulting embryos to a transient heat shock. The data showed an shRNA-dependent cardiac defect upon heat shock (Figure 2D). Together, these results demonstrate the stable integration and germ-line transmission of RNAi in zebrafish and uncover a cardiac impairment by UAS-TdTom-miR-shRNA1^apkcλ that is temporally regulatable.

Figure 2.

Functional validation of miR-shRNAs by stable in vivo transgenesis. (A) Schematic shows the stable transgenesis of miR-shRNAs^apkcλ in zebrafish using the Tol2 transposon system. (B) F₁ embryos derived from a cross between the UAS-TdTom-miR-shRNA1^apkcλ founder A and Ubi-GFF transgenic line. TdTom-negative F₁ embryos are normal (top panels). In contrast, ∼30% of the TdTom-positive F₁ embryos display a heart defect with cardiac edema (middle panels). This heart defect is rescued by the delivery of shRNA-resistant apkcλ mRNA (bottom panels). (C) Quantification of heart defect and rescue in B. (D) Embryos (60 hpf) derived from a cross between the UAS-TdTom-miR-shRNA1^apkcλ founder A and hsp70-Gal4 transgenic animal. Heat shock was performed at 8 hpf at 37 °C for 1 hr. The TdTom-positive embryo (right) displayed a heart defect with cardiac edema, whereas the heat-shocked control sibling (left) did not. (E) Droplet digital PCR analyses of endogenous apkcλ mRNA levels in stable UAS-TdTom-miR-shRNAs^apkcλ transgenic lines at different conditions of Gal4 induction. The representative images of 48-hpf embryos for each condition were shown on top of the bar graph. *P < 0.05, **P < 0.01, ***P < 0.001, vs. control sibling.

To determine whether a greater impairment of apkcλ gene activity might be achievable if the strength of the Gal4 driver was increased, we used KalTA4, an optimized version of Gal4 for zebrafish (Distel et al. 2009). The KalTA4 mRNA was injected into one-cell stage-F₂ embryos from Tg[UAS-TdTom-miR-shRNA1^apkcλ A; Ubi-GFF] fish. While the KalTA4 mRNA-injected, nontransgenic siblings were mostly normal, KalTA4 mRNA-injected, TdTom-positive transgenic embryos showed an apkcλ mutant-like morphology (Figure 2E, 4 and Table S6). Similarly, EF1a-GFF DNA-injected, TdTom-positive transgenic embryos also showed an apkcλ mutant-like phenotype (Figure 2E, 5 and Table S7). To assess the endogenous apkcλ mRNA level, we employed the Droplet Digital (dd) PCR technology (Figure S5), which measures target (e.g., apkcλ mRNA) and control (e.g., gapdh mRNA) molecules in the same biological reaction for superb precision and sensitivity (Hua et al. 2010). The results showed a clear correlation between the phenotypic severity and the apkcλ mRNA knockdown level (Figure 2E). Together, these results uncover a dosage-dependent knockdown of apkcλ gene activity through stable UAS-miR-shRNA transgenesis and controllable expression of Gal4.

Conclusions

We show that RNAi employing the miR-shRNA and Gal4/UAS system is a promising technology for conditional gene silencing in zebrafish. Compared to previous reports (Dong et al. 2009; De Rienzo et al. 2012), our bipartite Gal4/UAS-based design offers greater versatility and synergizes with the ongoing development of Gal4 enhancer trap lines in zebrafish.

Two platforms are established that can be used either independently or in combination to serve different research needs. The first platform is the in vivo transient RNAi. We demonstrate that this method leads to robust and specific phenotypes associated with three different genes, apkcλ, oep, and ntla. We also show that it is a powerful tool for genetic mosaic analysis at single-cell resolution. The ability to detect significant and dosage-dependent decrease of target mRNA expression also makes the in vivo transient RNAi a high throughput screening tool for identifying effective miR-shRNAs. The second platform is the in vivo stable RNAi. We show that the miR-shRNA transgenes can be stably inherited in subsequent generations and leads to significant gene knockdown that is temporally regulatable and dependent on the strength and expression of the Gal4 driver.

However, it is worth pointing out that the stable UAS-shRNA^apkcλ transgene, which is likely present at single or few copies in the genome, is insufficient to produce the severe mutant phenotype observable in transient RNAi, when crossed with stable Gal4 lines including several tissue-specific Gal4 lines tested. This is likely to be a general concern associated with stable RNAi, and is possibly due to the difference in shRNA transgene copy numbers between transient and stable RNAi. Increasing the UAS-shRNA copy numbers through homozygosing the transgene or crossing two different UAS-shRNA lines that potentially quadruples the copy numbers did not produce an obvious difference in our hands, suggesting a great many more copy numbers are necessary to produce a transient RNAi-like effect in stable lines. It may not be feasible to incorporate tens or hundreds of miR-shRNA copies in the transgene for concerns of cloning and transgene stability. Instead, the efficacy of stable RNAi transgenic systems may be further enhanced by incorporating the dual Tet-regulatory system (Knopf et al. 2010) recently developed in zebrafish for further amplifying the UAS transgene expressivity. Taken together, this study demonstrates the first prototype of stable and conditional RNAi-mediated gene silencing employing the Gal4/UAS system in zebrafish that has the potential to transform the utility of this vertebrate for uncovering novel biology and modeling human diseases.