Cluap1 is Essential for Ciliogenesis and Photoreceptor Maintenance in the Vertebrate Eye

Abstract

Purpose.: To identify the mutation and cell biological underpinnings of photoreceptor defects in zebrafish au5 mutants.

Methods.: Whole genome sequencing and SNP mapping were used to determine the genomic interval that harbors the au5 mutation. A candidate mutation was cloned and sequenced, and mRNA rescue used to validate that the affected gene was correctly identified. In situ hybridization, immunohistochemistry, and confocal imaging were used to determine the effects on photoreceptor development and maintenance in mutant retinae, and to determine if ciliogenesis or cilia-dependent development was affected in mutant embryos. Expression of tagged proteins and high-speed in vivo confocal imaging was used to quantify intraflagellar transport (IFT) and IFT particle localization within multiciliated cells of the Xenopus epidermis.

Results.: The au5 mutants possess a nonsense mutation in cluap1, which encodes a component of the IFT machinery. Photoreceptor defects result from degeneration of photoreceptors, and defects in ciliogenesis precede degeneration. Cilia in the olfactory pit are absent, and left-right heart positioning is aberrant, consistent with a role for cluap1 during ciliogenesis and cilia-dependent development. High-speed in vivo imaging demonstrates that cluap1 undergoes IFT and that it moves along the cilium bidirectionally, with similar localization and kinetics as IFT20, an IFT-B complex component.

Conclusions.: We identified a novel mutation in cluap1 and determined that photoreceptor maintenance is dependent on cluap1. Imaging data support a model in which cluap1 is a component of the IFT-B complex, and cilia formation requires cluap1 function. These data may provide new insights into the mechanism of photoreceptor degeneration in retinal ciliopathies.

Introduction

Photoreceptor degeneration is the major cause of human adult blindness and it results from the progressive dysfunction and eventual death of photoreceptor cells. In numerous instances, this disorder is linked to genetic mutations and more than 200 genes are implicated in photoreceptor degeneration (https://sph.uth.edu/retnet/). Approximately one-third of these known photoreceptor degeneration genes are associated with ciliary transport defects. ^1–3  

Vertebrate photoreceptor cells have a unique morphology, which consists of several distinct regions: the outer segment, the inner segment, the nuclear region, and the synapse. Although outer segments are enriched with proteins and membrane required for phototransduction, they lack ribosomes, and thus, development and maintenance of the outer segment requires delivery of cellular components from the cell body. This delivery occurs through the connecting cilium, a narrow constriction elaborated from the primary cilium and connecting the outer segment and inner segment. Defects of cilia structure or function in photoreceptors result in disruption of ciliary trafficking and ultimately lead to photoreceptor degeneration. ^2,3 Indeed, ciliopathies are an emerging class of human diseases and many are associated with progressive blindness. ^1–3  

Rapid transport and recycling of proteins and membranes in photoreceptor cells depends on active intraflagellar transport (IFT), a key transport system for movement of cargo within cilia. ⁴ Intraflagellar transport involves the bidirectional motility of a multisubunit protein complex, called the IFT particle, which consists of two major complexes: IFT-A required for retrograde movement and IFT-B required for anterograde movement. Intraflagellar transport core proteins and IFT-associated proteins are necessary for cilia assembly, maintenance, and signaling, and collectively, they are critical for survival of photoreceptors. Numerous genetic studies in mice and zebrafish have shown that defects associated with many of these IFT proteins lead to photoreceptor degeneration. ^5–10 However, several IFT components still remain to be identified and detailed cellular functions for many IFT proteins are unknown. 

Clusterin-associated protein 1 (Cluap1) has been identified as a component of the IFT machinery. Initially identified from human cancer studies, Cluap1 was reported to localize to the nucleus and is upregulated in several cancers. ^11,12 Moreover, loss-of-function mutations in cluap1 (previously known as qilin) result in ciliopathy-related phenotypes, such as kidney cysts and photoreceptor defects, in zebrafish. ^13,14 Cluap1 ^−/− mice have cilia defects, and Cluap1 protein is detected in primary cilia. ^15,16 Likewise, mutation of dyf-3, the Caenorhabditis elegans homolog of cluap1, results in morphological abnormality in sensory cilia structure, ¹⁷ and a DYF-3::GFP fusion protein undergoes IFT along sensory cilia with a similar velocity to other IFT particles, suggesting that Cluap1 is associated with the IFT complex. ¹⁸ Although Cluap1 protein has been localized to mammalian cilia, ^15,16 directed IFT of Cluap1 has never been demonstrated in a vertebrate. Moreover, although loss of cluap1 is associated with ocular anomalies, ¹³ the root cause of these defects is not defined. 

In this study, we have characterized the zebrafish au5 mutant, ²³ which harbors a novel mutation in cluap1. We show that cluap1 mutants possess normal development of photoreceptors, but this is followed by progressive degeneration associated with defective cilia. In addition, using high-speed live imaging, we show that Cluap1 travels together with IFT particles, demonstrating that Cluap1 is an IFT particle subunit in vertebrates. These data may provide new insights into the mechanism of photoreceptor degeneration in retinal ciliopathies. ^3,19  

Materials and Methods

Animals

Zebrafish and Xenopus laevis were maintained as described. ^20–22 All protocols used within this study were approved by the Institutional Animal Care and Use Committee of The University of Texas at Austin and conform to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The allele used in this study: cluap1^au5 . ²³  

Mutant Cloning

au5 heterozygous carriers (AB background) were outcrossed with wild-type TU fish to generate a mapping line. Heterozygous AB/TU carriers of the au5 mutation were identified and then incrossed to generate homozygous mutant embryos. Fifty mutant zebrafish embryos from four parental pairs were collected, genomic DNA was isolated (DNeasy Blood & Tissue Kit; Qiagen, Valencia, CA, USA), and 1 μg was used for Illumina sequencing at the University of Texas Genomic Sequencing and Analysis Facility. On an Illumina HiSequation 2000 machine (Illumina, San Diego, CA, USA), 150 million paired-end 100-bp sequences were generated for an average genome coverage of 19X. To identify putative mutations, sequencing reads were analyzed using the BSFseq mapping pipeline on MegaMapper (Megason Lab, Harvard Medical School, Boston, MA, USA). ²⁴ Candidate single nucleotide polymorphisms (SNPs) were confirmed by cDNA sequencing. 

mRNA Synthesis and Injection

Wild-type cluap1 open reading frame was cloned from zebrafish cDNA and inserted into pCS10R and pCS10R-GFP vectors for mRNA synthesis. Capped, poly-adenylated cluap1 and cluap1-GFP were synthesized via a mMESSAGE mMACHINE SP6 Transcription Kit (Life Technologies, Grand Island, NY, USA). For mRNA rescue experiments, 100 pg cluap1 mRNA was injected into one-cell stage zebrafish embryos. For Xenopus experiments, 200 pg cluap1-GFP, 150 pg RFP-IFT20, and 150 pg GFP-IFT20 ²⁵ were injected into two ventral cells of four-cell stage embryos. 

Immunofluorescence and Whole-Mount In Situ Hybridization

Immunostaining in zebrafish cryosections was performed as described. ²⁶ All antibodies were used at 1:200 dilutions. Primary antibodies were mouse monoclonal anti-acetylated tubulin (T7451; Sigma-Aldrich Corp., St. Louis, MO, USA), anti-zpr1 (ZIRC), and anti-zpr3 (ZIRC). Secondary antibodies were Cy2 or Cy3-conjuated goat anti-mouse (Jackson, West Grove, PA, USA). Phalloidin (Alexa Fluor 488 or 555, Invitrogen, Grand Island, NY, USA) and Sytox-Green were applied to sections to stain F-actin and DNA, respectively. In situ hybridizations were performed as described. ²⁷ For cmlc2 probe synthesis, a partial cDNA fragment of cmlc2 was cloned by RT-PCR using primers, 5′-tgtatttaggaggctctgggtgtc-3′ and 5′-ctgctgatgtgaatgttgaactgg-3′. 

Imaging and Image Analysis

Confocal imaging of cryosections was performed on either a Zeiss LSM5 Pascal or LSM 700 microscope (LSM 5LIVE; Zeiss, Thornwood, NY, USA). High-speed in vivo imaging of IFT was performed as described ²⁵ on a Zeiss LSM 5LIVE microscope. Intraflagellar transport velocities were measured using LSM 5LIVE duoScan software. Whole embryo images were captured on a Leica MZ16FA stereomicroscope (Leica, Buffalo Grove, IL, USA), mounted with a DFC480 digital camera. 

Results

au⁵ Mutants Possess Photoreceptor Defects

au⁵ mutants were identified in a forward genetic screen for embryos possessing morphological defects in eye development, and were identified based on a photoreceptor phenotype. ²³ au⁵ mutation is recessive, fully penetrant, and embryonic lethal by 10 to 12 days post fertilization (dpf). au⁵ mutant embryos do not inflate their swim bladder. Mutant embryos present with a ventral curve to their body axis (Figs. 1A, 1B), which appears at approximately 1.5 dpf. Because of this curvature, mutants are easily distinguished from phenotypically wild-type siblings by 2 dpf. Mutants are also slightly microphthalmic at 3 dpf (Fig. 2D) and 5 dpf (Fig. 1B). Analyzing retinal development at 7 dpf in au⁵ mutants revealed a complete absence of photoreceptors in the central retina (Figs. 1C, 1D). 

Figure 1

 

au⁵ mutants possess photoreceptor defects. Whole embryo views of wild-type sibling (A) and au⁵ mutant (B) at 7 dpf. au⁵ mutants possess a ventral body curvature. (A′, B′) High-magnification zoom of (A) and (B) showing mild microphthalmia in au⁵ mutants. (C, D) Cross-section views of wild-type (C) and au⁵ (D) retinas at 7 dpf stained with phalloidin (red) and Sytox-Green (green). au⁵ mutants possess photoreceptor defects when compared with their wild-type siblings. (C′) and (D′) are high-magnification views of the photoreceptor layer from the boxed in regions of (C) and (D). Photoreceptor layer is marked by a black bracket in (C′). Scale bars: 50 μm (C, D), 20 μm (C′, D′).

Figure 2

 

au⁵ mutants possess a premature stop codon in cluap1. (A, B) Sequencing results of cluap1 from phenotypically wild-type siblings (A) and au⁵ mutants (B). au⁵ mutants possess a G-to-A transition mutation in cluap1 (arrow) that generates a premature stop codon at amino acid 41 (W41X). (C) Schematic of wild-type cluap1 protein and cluap1^au5. Yellow and red boxes represent the predicted NN-CH domains of cluap1, a divergent N-terminal calponin homology (CH)-like domain (6 amino acids (a.a.)–69 a.a.) and coiled-coil box domain (202 a.a.–307 a.a.), respectively. (D) Examples of normal and au⁵ mutant phenotype at 3 dpf. (E) Graph depicting proportion of embryos with phenotypes shown in (D). The au5 mutant phenotype is rescued by injection of cluap1 mRNA. Uninjected embryos of wild-type (n = 50), cluap1 mRNA injected embryos of wild-type (n = 47), uninjected embryos of au5 (n = 45) and cluap1 mRNA injected embryos of au5 (n = 40).

au⁵ Mutant Phenotypes Result From a Nonsense Mutation in cluap1

To identify the gene mutated in au⁵ , we used next-generation sequencing and SNP mapping, generating 19X average genome-wide sequencing coverage. Using the MegaMapper analysis pipeline, ²⁴ a 6-megabase window was identified on chromosome 24 that was predicted to harbor the mutation (Supplementary Fig. S1). Within this window, a G-to-A transition mutation at nucleotide 123 of the cluap1 coding sequence was identified (Figs. 2A, 2B). This mutation generates a premature stop codon at amino acid 41 (Trp41Stop) of cluap1, truncating the protein by approximately 90% (Fig. 2C). 

To confirm that this is the causative mutation in au⁵ mutants, we performed an mRNA rescue experiment by injecting full-length wild-type cluap1 into au⁵ embryos. Overexpression of cluap1 did not affect development of wild-type embryos, and the mutant phenotype was rescued in all injected embryos (Figs. 2D, 2E). Thus, we conclude that au⁵ possesses a mutation in cluap1 and hereafter refer to this mutant as cluap1^au5 . 

Cluap1 is has been previously associated with ciliogenesis in mice and fish. ^13,15,16 As further evidence that the identified mutation in cluap1 was causative for the au⁵ phenotypes, we examined ciliogenesis and cilia-related developmental defects. The olfactory epithelium of zebrafish is normally ciliated, and we found that olfactory epithelial cells lacked cilia in cluap1^au5 mutants (Figs. 3A, 3B). Moreover, although Cluap1 has been shown to be required for nodal ciliogenesis and left/right asymmetry in mice, ¹⁶ its role in in the zebrafish left/right patterning is not known. We performed whole-mount in situ hybridization using cardiac myosin light chain 2 (cmlc2) as a marker for the position of the heart tube. ²⁸ At 28 hours post fertilization, the heart tube is positioned to the left side of wild-type embryos (Figs. 3C, 3D). However, in embryos derived from cluap1^au5 heterozygous incrosses, heart tube position was randomized, with 13% of embryos possessing cmlc2 expression in the middle of the heart field, and 4.3% of embryos on the right side (Figs. 3C, 3D). Given that 25% of embryos from such crosses are cluap1^au5 mutants, these data demonstrate that left-right asymmetry was defective in cluap1^au5 mutants. Taken together, these data demonstrate that cluap1^au5 mutants display cilia defects, consistent with a role for cluap1 in ciliogenesis. ^13,15,16  

Figure 3

 

cluap1^au5 mutants possess defects in ciliogenesis and cilia-dependent developmental processes. (A, B) Olfactory pit cilia marked by acetylated alpha tubulin antibody at 3 dpf. When compared to phenotypically wild-type siblings (A, A′), no cilia are detected in the olfactory pits of cluap1^au5 mutants (B′). Scale bars: 10 μm. (C) Heart tube position determined by cmlc2 expression 28 hpf. Dorsal views. (D) Quantification of the proportion of embryos with left, middle/none, and right heart tube position. The heart of all wild-type embryos (n = 52) is on the left side. In clutches of cluap1^au5 embryos (n = 46), heart position is defective, with 13.0 % in the middle and 4.34 % on the right.

Progressive Photoreceptor Degeneration in cluap1^au5 Mutants

A previous study of cluap1 in zebrafish revealed an absence of photoreceptors at 5 dpf, but the cause of this defect has not been explored. ¹³ Because human ciliopathies are frequently associated with retinal degeneration, ^2,29,30 we hypothesized that the photoreceptor defects observed in cluap1^au5 mutants might be the result of degeneration, rather than defects in specification or differentiation. 

To test this hypothesis, we analyzed photoreceptor cell formation at the early stages of retinal development. In zebrafish, differentiation of photoreceptors initiates at approximately 2 dpf, with the photoreceptor layer becoming morphologically distinct by 3 dpf, at which point photoreceptors have formed obvious outer segments and become functional. ³¹ At 3 dpf, the retina in cluap1^au5 mutants was well formed, and an outer nuclear layer was present (Figs. 4A, 4B). However, using zpr3 and zpr1 as immunohistochemical markers for rods and red-green cones, respectively, revealed that whereas these were present in the cluap1^au5 mutant retina, there were fewer photoreceptors expressing these proteins, and those that did possessed markedly shorter outer segments (Figs. 4C–F). By 5 dpf, the photoreceptor layer in cluap1^au5 mutants was almost completely absent from the central retina, and photoreceptors were present only in the peripheral retina, near the ciliary marginal zone, where photoreceptors are continually generated (Figs. 4G–L). These data suggest that in cluap1^au5 mutants, photoreceptor cells are impaired in their ability to form outer segments, and that they ultimately degenerate. 

Figure 4

 

Photoreceptor defects in cluap1^au5 mutants result from degeneration. Transverse sections of zebrafish retinae immunostained for Zpr-1 (a green-red cone cell marker, red) or Zpr-3 (a rod cell marker, red) and Sytox-Green (DNA, green) at 3 dpf (A–F) and 5 dpf (G–L). Dashed boxes in (C–F) and (I–L) indicate regions from which high-magnification views of the photoreceptor layers are presented in (C′–F′) and (I′–L′), respectively. Photoreceptor cells expressing Zpr-1 and Zpr-3 are detected at 3 dpf in cluap1^au5 mutants, but by 5 dpf, these have degenerated in the central retina. Scale bars: 50 μm (A–L), 20 μm (C′–L′).

Ciliogenesis Defects Precede Photoreceptor Defects in cluap1^au5 Mutants

Previous studies have demonstrated that Cluap1 is required for cilia assembly and maintenance, ^13,15,16 but the development of connecting cilia has yet to be investigated directly in cluap1-deficient photoreceptors in any animal. Because we observed that photoreceptor degeneration was progressive over time, we were able to examine connecting cilia directly in cluap1^au5 mutants at 3 dpf. Using acetylated α-tubulin immunostaining, cilia were easily detected in photoreceptor cells of wild-type siblings (Fig. 5A, arrowheads). In contrast, although photoreceptor cells are present at 3 dpf in cluap1^au5 mutants (Fig. 5B), no cilia were detected in these photoreceptor cells (Fig. 5B). Similarly, no cilia were detected at 7 dpf in cluap1^au5 mutants, despite obvious cilia on the photoreceptors of wild-type siblings (Figs. 5C, 5D). Thus, loss of cilia in precedes photoreceptor degeneration in cluap1^au5 mutants. 

Figure 5

 

Ciliogenesis defects precede photoreceptor degeneration in cluap1^au5 mutant photoreceptor cells. Transverse sections from wild-type and cluap1^au5 mutant retinae immunostained with an acetylated tubulin antibody (red) and phalloidin (green) to detect cilia and F-actin in photoreceptor outer segments, respectively. (A–B′) Photoreceptors of 3 dpf and (C, D) 7 dpf. Cilia are lacking in cluap1^au5 mutant photoreceptors (B, D) when compared with phenotypically wild-type siblings (A, C). Yellow arrowheads indicate cilia in the photoreceptors. Scale bars: 10 μm.

Cluap1 Undergoes IFT in a Vertebrate

In C. elegans , Cluap1/DYF-3 is required for normal IFT, and Cluap1 itself undergoes bidirectional transport with similar velocity as other IFT-B complex particles. ¹⁸ However, the transport mechanisms of IFT particles in C. elegans differ from those in vertebrates, ^32,33 and vertebrate Cluap1 has been variably reported to localize either along axonemes ¹⁵ or to the tips and base of cilia. ¹⁶ To directly analyze the localization and dynamics of Cluap1 in a vertebrate system, we turned to the muticiliated cells of the Xenopus epidermis, which provide an excellent system in which to examine IFT by live imaging in a vertebrate. ^25,34–36  

We expressed Cluap1-GFP and observed that it localized to both the basal body and the axoneme in a pattern similar to IFT20, a subunit of the IFT-B complex (Fig. 6). Moreover, high-speed in vivo imaging showed that Cluap1-GFP moved bidirectionally along the axoneme (Fig. 7B), and its velocity in both the anterograde and retrograde direction was similar to that of IFT20 (Fig. 7C). Finally, from high-speed, two-color imaging, we could observe that Cluap1-GFP and RFP-IFT20 colocalized and were cotransported in the same particles (Figs. 7D–F). These data support a model in which vertebrate Cluap1 undergoes IFT as part of the IFT-B complex in vertebrate cilia. ^13,15,37  

Figure 6

 

Cluap1-GFP localizes to the axoneme and basal body of cilia of Xenopus multiciliated cells. (A) Cluap1-GFP is detected at the basal body and axoneme of cilia emanating from multiciliated cells on the surface of Xenopus embryo (stage 25). (B–B″) High-magnification view of dashed box in (A). Cluap1-GFP colocalizes with RFP-IFT20 (Cluap1-GFP, green; RFP-IFT20, red). Scale bars: 10 μm.

Figure 7

 

Cluap1-GFP undergoes IFT with a similar velocity to IFT-20. (A) Still image from a time-lapse video of Cluap1-GFP in Xenopus multiciliated cell (stage 25). Scale bar: 5 μm. (B) High-magnification view of boxed region in (A). Time-lapse images of Cluap1-GFP movement along the axoneme demonstrate that Cluap1-GFP undergoes both anterograde and retrograde IFT. (C) Graph depicting the velocities of Cluap1-GFP and GFP-IFT20. Cluap1 moves with a similar velocity to IFT20 in both the anterograde and retrograde directions. Mean velocities ± SD are 1.076 ± 0.28 μm/s (n = 46 from 8 cells) for Cluap1-GFP anterograde, 1.23 ± 0.30 μm/s (n = 68 from 8 cells) for Cluap1-GFP retrograde, 1.05 ± 0.20 μm/s (n = 27 from 9 cells) for GFP-IFT20 anterograde and 1.19 ± 0.29 μm/s (n = 53 from 9 cells) for GFP-IFT20 retrograde. (D, E) Still images of time-lapse movies showing Cluap1-GFP (D) and RFP-IFT20 (E) localization in cilia. Scale bar: 10 μm. (F) High-magnification view of boxed regions in (D, E). Time-lapse images of Cluap1-GFP and RFP-IFT20 movement along the axoneme support that Cluap1-GFP and RFP-IFT20 are transported in the same IFT particle.

Discussion

Ciliopathies are human genetic disorders caused by mutations in genes regulating cilia formation or function. ³⁸ In many cases, ciliopathies involve a wide range of clinical features, and these include defects in the retina, as well as other organ systems. Syndromic ciliopathies that present with photoreceptor involvement include Alstrom syndrome (MIM 203800), Bardet-Biedl syndrome (MIM 209900), Joubert syndrome (MIM 213300), Meckel-Gruber syndrome (MIM 249000), Senior-Loken syndrome (MIM 266900), Usher syndrome (MIM 276901), and nephronophthisis. Nonsyndromic retinal ciliopathies include retinitis pigmentosa (MIM 268000), cone-rod dystrophy (MIM 120970), and Leber's congenital amaurosis (MIM 204000). ^2,39 Vertebrate animal models have become powerful tools through which the relationships between the clinical features observed in ciliopathy patients, and the genetic and molecular underpinnings of disease can be elucidated. ^{2,5,9,40–46}  

In this study, we identify a zebrafish model that may provide novel insights into the genetic basis of ciliopathic blindness. We report on a novel mutation in the zebrafish cluap1 gene, and describe ciliogenesis defects in cluap1^au5 mutants that are similar to those observed in other regions of the zebrafish embryo, ¹³ as well as in Cluap1 knockout mice ^15,16 and C. elegans dyf-3 mutants. ¹⁷ Photoreceptor defects have been reported in cluap1 mutant zebrafish, but no further analysis of this phenotype has been reported. ¹³ Here, we show directly for the first time that ciliogenesis is defective in the photoreceptors of cluap1^au5 mutants, and we show that these defects precede a progressive degeneration of photoreceptors. Photoreceptor degeneration resulting from loss of cluap1 function proceeds in a central to peripheral pattern, and this is consistent with the central retinal photoreceptors being the oldest and first affected, whereas those more peripheral are younger, having differentiated later. Details like these on the etiology of the retinal pathology in cluap1^au5 mutants are significant for understanding the pathogenesis of ciliopathic eye defects. For example, although most ciliopathies involve cilia defects in photoreceptors, ^2,39 a new ciliopathy has recently been described with ciliogenesis defects in retinal precursors and not in photoreceptors. ^47,48  

Interestingly, this new ciliopathy results from mutation in intestinal cell kinase, which in turn controls the localization of IFT proteins, ⁴⁷ and our data provide new details on the link between IFT and cluap1. Cluap1 mutants in a variety of systems possess phenotypes that resemble those in mutants for other IFT-B components, and these include defects in ciliogenesis and defective Shh signaling. ^15,49–51 In mice, Cluap1 binds to IFT88, an IFT-B component, and not IFT140, an IFT-A component, ³⁷ and a recent proteomic screen identified Cluap1 as an integral component of the IFT-B complex. ⁵² Moreover, according to sequence profile-to-profile analysis and structure analysis, Cluap1 shares overall protein architecture with some IFT-B components. ⁵³ Finally, Dyn-3::GFP undergoes IFT in C. elegans , ¹⁸ but there are key differences in the mechanisms driving IFT in C. elegans and vertebrates (e.g., Refs. 32 and 33), and the previous reports of Cluap1 localization in vertebrate cilia have not been consistent. ^15,16 Our finding here, using high-speed in vivo imaging, reveals for the first time that Cluap1-GFP undergoes bidirectional IFT in vertebrate cilia and colocalizes with IFT20, an IFT-B complex protein. These data provide direct support for a role for Cluap1 in the IFT-B complex in vertebrates, and help to elucidate the etiology of ciliopathic photoreceptor degeneration. 

Supplementary Materials

pdf S2 

pdf S3 

mov S1 

Acknowledgments

We thank members of the Gross and Wallingford laboratories for comments on this study, and Sallie Ford for technical assistance. 

Supported by National Institutes of Health/National Eye Institute Grant R21-EY22770 (JMG). JBW is an early career scientist of the Howard Hughes Medical Institute and work in his laboratory is funded by grants from the National Institute of General Medical Sciences and the National Heart, Lung, and Blood Institute. Zebrafish antisera were obtained from Zebrafish International Resource Center, which is supported by National Institutes of Health–National Center for Research Resources Grant P40 RR012546. 

Disclosure: C. Lee, None; J.B. Wallingford, None; J.M. Gross, None