The Usher gene cadherin 23 is expressed in the zebrafish brain and a subset of retinal amacrine cells

In situ hybridization

Wholemount in situ hybridization was performed essentially as described in The Zebrafish Book, using a digoxigenin-labeled probe generated using a digoxigenin labeling kit (Roche, Penzberg, Germany). Probes were synthesized from cdh23 fragments cloned into pcR2.1 TOPO (Invitrogen, Carlsbad, CA). The 5′ antisense probe consisted of nucleotides 448–1,566 from cdh23 cDNA, while the 3′ antisense probe contained nucleotides 9730–10,515. For in situ hybridization followed by immunostaining on cryosections, the temperature of incubation was reduced from 63 °C to 56 °C. Images of in situ hybridization were taken on a Zeiss AxioImager.M1 wide-field microscope using a 10× or 20× dry objective. Images were acquired with an AxioCam MRc5 color digital camera using Axiovision software, and subsequently adjusted for contrast in Photoshop CS2 (Adobe, San Jose, CA) using the levels, curves or brightness/contrast adjustment functions.

Immunohistochemistry

Zebrafish were fixed in 4% paraformaldehyde in PBS overnight, cryoprotected in increasing concentrations of fish skin gelatin/sucrose in PBS, ending in 25% gelatin, 15% sucrose, and then frozen in blocks and cut into 14 µm sections on a Leica CM180 cryostat. Immunostaining was performed using standard on-section immunohistochemistry protocols. Permeabilization of sections was achieved by including 0.02% saponin (S7900; Sigma, St. Louis, MO) in all block, wash, and antibody incubation solutions. Antibodies and dilutions were as follows: Ab5E11 (mouse, 1:500, gift from J. Fadool, Florida State University, Tallahassee, FL), calretinin (rabbit; 1:5,000; Swant [Swiss Antibodies], Marly, Switzerland), FMRF (rabbit, 1:100; Phoenix Pharmaceuticals, Burlingame, CA), glutamate decarboxylase (GAD65/67) rabbit; 1:500; Abcam, Cambridge, MA), parvalbumin (mouse; 1:500; Millipore, Billerica, MA), serotonin (rabbit; 1:250; Sigma), acetylated tubulin (mouse; 1:2000; Sigma), tyrosine hydroxylase (TH; mouse; 1:500; Immunostar, Hudson, WI), Cy3 antimouse and Cy3 antirabbit (1:1,000; Jackson Immunoresearch, West Grove, PA). Images of immunostained cryosections were acquired on either an Olympus Fluoview (Olympus America Inc., Center Valley, PA) or Zeiss LSM710 (Carl Zeiss AG, Oberkochen, Germany) confocal microscope, and processed for display using ImageJ [18] and Adobe Photoshop. ImageJ version 1.43u was used to convert files to tif format and to produce color overlays. Photoshop CS2 (Adobe) was used to adjust contrast using curves, levels and/or brightness/contrast image adjustment functions.

Eye injections

Serotonin (H7752; 100 mM; Sigma) in embryo medium was injected intraorbitally using a protocol modified from [19,20]. Briefly, glass microinjection pipettes were pulled on a Nashirige PC-10 (Nashirige International USA, Long island, NY) and filled with 100 mM serotonin (5-hydroxytryptamine, 5-HT) and 100 mM ascorbic acid in embryo medium. The tip of the glass micropipette was broken with forceps to a diameter sized to slip between the edge of the lens and retina without causing excessive damage. Approximately 50 nl was injected into the left eye of 4 days postfertilization (dpf) larvae anesthetized in Tricaine and mounted in 2% low-melt agarose; the right eye was left uninjected to control for nonspecific immunostaining. Following injection, larvae were transferred to fresh embryo medium and allowed to recover at 28 °C for 3 h, and were then fixed in 4% paraformaldehyde overnight at 4 °C. Following this, they were processed for in situ hybridization and subsequent antiserotonin immunohistochemistry.

Detection of splice variants

Total RNA was isolated from wild-type zebrafish eyes or enucleated larvae (remaining tissue following the removal of eyes) by homogenization of the larvae in TRIZOL reagent (Invitrogen, Life Technologies, Grand Island, NY) using a 25 gauge needle, followed by RNeasy purification (Qiagen, Valencia, CA). Reverse transcription was performed with the Superscript III First Strand cDNA synthesis kit (Invitrogen) using oligo(dT) primers. PCR products were subcloned into the pCRII-TOPO vector (Invitrogen) and sequenced. The splice variants were confirmed at least once with another independent PCR. Primer sets were as follows (written 5′ to 3′). Nested primers to amplify the region from extracellular cadherin repeat, or ectodomain 14 (EC14) to the 3′ UTR were as follows. External primers: 5′-TGT GGA CCG CTA CCT GCT TAA AGT-3′ 5′-TGA GGC ATT CAG AGT CCA CAC ACA-3′, internal primers: 5′-TCA GTG AGA ATG TGG GTG GTG GAA-3′ 5′-GTG TCG CTG CCT TTG TTT CTG TGT-3′. Primers to amplify the cytoplasmic domain: 5′-CTG AAG GCA GTA GTT GCA GGC T-3′ 5′-TCA TAA CTC TGT GAT CTC TAA CGG AC-3).

Optokinetic response

As a test for visual function, the OKR of 5 dpf homozygous larvae of both alleles, as well as their wild-type siblings, was measured as previously described [21]. In brief, larvae were embedded in 3% prewarmed methylcellulose, aligned to lay dorsal side up and placed in the center of a white paper drum (d=9 cm). Vertical black-and-white sine-wave gratings of varying spatial frequency and contrast were projected onto the inside of the drum [22]. The grating pattern was rotated around the larva with an angular velocity of 7.5 °/s. To minimize the frequency of saccades, the direction of motion was altered with a frequency of 1/3 Hz.

Eye movements were recorded at 5 frames per second by a CCD camera attached to a dissecting microscope. Custom-made software based on NI LabView 7.1 and NI-IMAQ 3.7 (National Instruments, Austin, TX) was used to extract the orientation of each eye in real time. This software is now licensed to TSE (Homburg, Germany), under the name VisioTracker. Eye velocity was calculated as the first derivative of the orientation with respect to time. Saccades were filtered out from raw measurements of eye velocities by applying a fixed threshold of 20 °/s. The resulting curves of slow-phase velocities were smoothed by a running average with a sliding window of seven frames [23]. Finally, eye velocities were averaged over frames with identical stimulus properties, as well as for the two eyes of each larva. Graphs were generated using R 2.9.2.

Zebrafish were cared for according to the guidelines set forth by the Institute for Laboratory Animal Research.

Expression of cdh23 transcripts in zebrafish eye

To initiate our study of the role of cdh23 in zebrafish visual function, we characterized its expression in the developing zebrafish eye by reverse transcriptase (RT)–PCR (Figure 1). To describe the splice isoforms present specifically in zebrafish eyes, we performed RT–PCR on dissected eyes isolated from zebrafish at various stages of development. Using nested primers directed against EC14 and the 3′UTR (Figure 1A), we found that, in contrast to the case in mice, zebrafish eyes express the full-length isoform at all developmental stages examined: larval (7 dpf), juvenile (6 weeks postfertilization [wpf]), adult (6 months postfertilization [mpf]), and end-of-life (3 years postfertilization [ypf]; Figure 1B). Alternative splicing of cdh23 has been studied extensively in the mouse, and the presence of shorter cdh23 splice variants has been implicated as one explanation for the lack of visual phenotype in mouse cdh23 mutants [10]. Using primers similar to those used in our study, Söllner et al. [16] amplified two additional splice variants from whole larvae: a short transmembrane (TM) isoform that is missing EC15–25 (and is also missing exon 68), and a soluble isoform that skips from EC17 to a short sequence in exon 69, the final exon. We therefore investigated the possibility that eye-specific variants are present in the zebrafish. We were able to amplify the short TM isoform from 7 dpf and 6 mpf eyes, and the soluble isoform from 6 wpf eyes (Figure 1C). We were unable to amplify either shorter variant from 3 ypf eyes. Using primers specific to the region encoding the cytoplasmic domain (Figure 1A), both +exon68 and -exon68 variants of cdh23 were amplified from zebrafish eyes and enucleated larvae at 7 dpf (not shown) and 6 wpf (Figure 1D). Only the +exon68 form was amplified with the EC14–3′ UTR primer pairs described above. We do not know the biologic relevance of these two shorter variants, although both would be affected by the cdh23^tc317e mutation. However, it is possible that there are additional splice variants present that we did not detect.

Expression pattern of cdh23 in the developing zebrafish brain and retina

We next characterized the expression pattern of cdh23 in developing larvae by in situ hybridization (Figure 2). The first detectable staining was observed approximately 18–24 h postfertilization (hpf) in two paired nuclei deep in the diencephalon (Figure 2A,B), and concurrently in the first developing inner ear hair cells (data not shown). By 48–54 hpf, additional nuclei appeared near the olfactory bulb and in the hindbrain (Figure 2C,E). In contrast to the expression pattern observed for pcdh15b [17], we did not detect the cdh23 transcript in retinal photoreceptors. Instead, a very small subset of amacrine cells was positive for cdh23 mRNA starting at 70 hpf (Figure 2F). Approximately 30 amacrine cells/retina were labeled, as counted in transverse 14 µm serial cryosections through the entire retina (average of six retinas, data not shown). Labeling with either 5′ or 3′ antisense probes produced identical results; staining patterns were indistinguishable in cdh23^tc317e siblings (mixed wild-type and heterozygous genotypes) and homozygous mutants (shown for the retina in Appendix 1). The images in all other figures show results using the 5′ antisense probe, which exhibited more robust labeling. No staining was observed using a sense control probe (Figure 2D).

We generated serial cryosections following in situ hybridization of whole larvae to examine the brain structures expressing cdh23 in more detail (Figure 2G-L). Expression of cdh23 in the brain was localized to the subpallium, eminentia thalami, ventral thalamus, sparse cells in dorsal thalamus, and scattered cells throughout hindbrain. Of the brain nuclei labeled, none were obvious visual centers, although the subpallium and thalamic nuclei have multiple inputs and outputs, and could therefore participate in visual signal processing [26,27]. The cdh23-labeled hindbrain neurons were not concentrated in a single nucleus, but instead were scattered throughout the hindbrain. Specific brain nuclei were identified by consulting a zebrafish brain atlas [28] and by performing on-section immunocytochemistry following in situ hybridization for the following antigens, whose localization has previously been described: acetylated tubulin, TH, FMRF, and calretinin (data not shown). To determine whether or not this expression pattern was maintained in older fish, we performed in situ hybridization on transverse vibratome sections from 6 wpf brains and retinas, as well as 6 mpf retinas. In the 6 wpf brain, the cdh23-positive nuclei evident in the larval stages were still labeled (Figure 2M and data not shown); additional cell populations were also labeled in the optic tectum and subthalamic area. As in the larval retina, a sparse subpopulation of amacrine cells, but not photoreceptors, were cdh23 positive in both 6 wpf and 6 mpf retinas (Figure 2N,O).

Identification of the subtype of cdh23-expressing amacrine cells

To further explore the identity of cdh23-positive amacrine cells, we combined in situ hybridization for cdh23 mRNA with immunohistochemistry for various amacrine cell markers. When determining colocalization of an amacrine cell marker with cdh23 mRNA labeling, this technique gave better single-cell resolution compared to two-color in situ hybridization. Attempts to develop a reliable antibody to zebrafish Cdh23 were unsuccessful; therefore, a simple double immunostain was not feasible. By reducing the stringency of our whole mount in situ hybridization and washing conditions, we were able to preserve immunogenicity of several antigens. Using this approach, we determined that the cdh23-positive cells were indeed amacrine cells as shown not only by morphology and position, but also by colocalization with the amacrine cell–specific antigen 5E11 [29] (Figure 3A). Most amacrine cells are either GABAergic or glycinergic [30]. Immunostaining for the GABAergic cell marker glutamate decarboxylase 65/67 (GAD65/67) showed that cdh23-positive amacrine cells were invariably GABAergic (Figure 3B). Many different GABAergic amacrine cell subclasses have been further defined based on the presence of a second molecular marker. We next examined colocalization of cdh23 with antigens known to define subtypes of GABAergic amacrine cells.

Two GABAergic amacrine populations in zebrafish are known to be present at a density very similar to that which we observed for cdh23, and therefore warranted investigation. First, a serotonin transporter (SERT)–positive amacrine population has been described in zebrafish [31]. While it has not been extensively characterized in zebrafish, this cell may be similar to the A17 amacrine described in other organisms [32–35]. The A17, or indolamine-accumulating amacrine cells, express a SERT, though these cells do not accumulate serotonin (5-HT) under normal conditions [36]. By exploiting the presence of this SERT, the cell type can be labeled by allowing the uptake of exogenously applied 5-HT, followed by immunostaining with an anti-5-HT antibody [37]. Using a protocol modified for zebrafish larvae, we performed intraocular injection of 5-HT in live larvae, then processed the larvae for in situ hybridization and anti-5-HT immunostaining. None of the cdh23-positive cells colabeled with anti-5-HT (Figure 3C). A second sparse GABAergic amacrine population described in zebrafish is the dopaminergic interplexiform cell [38–40]. As was the case for the 5-HT-positive population, cdh23 mRNA did not localize to cells labeled by an antibody to TH, a marker for dopaminergic amacrine and interplexiform cells. We also compared cdh23 in situ labeling to a pax6 promoter-driven GFP transgenic line [41]. The cells labeled by GFP in this line are of at least two types: the parvalbumin-positive type and the ChAT-positive type. No colocalization of GFP and cdh23 in situ label was observed (Appendix 2). In addition, cdh23-positive amacrine cells did not colabel with antibodies to endogenous calretinin (Appendix 2) or parvalbumin (data not shown). We were therefore unable to positively identify a secondary marker for the GABAergic, cdh23-positive amacrine cell population.

While these amacrine cells were of great interest considering the role of cdh23 in vision, we were also struck by the intense labeling in the deep diencephalic paired nuclei. Robust labeling was detectable both very early in development, as well as in adult brain. Overall, these cells comprised the most intensely labeled brain region labeled by cdh23 RNA probes. Double immunofluorescence/in situ hybridization experiments showed that a subset of these diencephalic, cdh23-positive cells were also positive for TH (Appendix 2). Of the three clusters of dopaminergic diencephalic neurons present at this age [42], it was those of the ventral thalamus (ventral tegmentum) that partially colocalized with neurons in the more caudal regions of the cdh23-positive deep diencephalic nuclei. The fibers of these ventral tegmental dopaminergic neurons can be followed ventrocaudally, where they join a dense network of TH-positive fibers that project in four different directions [42]. The more rostral cdh23-positive cells in this region, those clustered in a striped pattern, were situated just caudal and dorsal to the anterior commissure, but ventral to the calretinin-positive pretectal cluster [43], and were not TH-positive (data not shown).

To identify the projection pattern of cdh23-positive cells in the retina and brain, we tested multiple Cdh23 antisera, including several new antibodies that we generated against the zebrafish Cdh23 sequence. Unfortunately, each Cdh23 antibody produced nonspecific staining using multiple staining protocols in whole mount and cryosectioned brain and retinal tissue. To find an alternative means by which to identify the projection pattern of cdh23-expressing cells, we generated reporter constructs containing cdh23 upstream regulatory elements driving GFP. We found that a 6 kb fragment (encompassing the entire 5′ UTR and some surrounding intronic sequence) drove GFP expression in hair cells of the ear and neuromasts (data not shown). By adding the intronic sequence (1 kb, 3 kb or 5 kb) directly upstream of the start codon to the 6 kb 5′ UTR construct, expression in the eye and brain was also observed. The GFP expression pattern, however, was the same for all three constructs: Labeling in the brain was too low to visualize neurites, and in the retina, mainly glycinergic rather than GABAergic amacrine cells expressed GFP (data not shown). Further efforts will be necessary to characterize the cdh23 promoter in the brain and retina.

Retinal morphology in cdh23 siblings and mutants

Retinal morphology was examined at 4 dpf and 17 dpf for two different alleles of cdh23 mutants. For homozygous larvae carrying either allele, the ability to feed properly is severely compromised, and mutant larvae typically die by 8–9 dpf. This early death significantly limits the developmental stage at which we are able to examine retinal degeneration. With excessive feeding and care, however, mutant larvae were kept alive until 17 dpf, at which point similar-sized siblings and mutants were fixed and their retinas were examined. Cell density and health, examined by both differential interference contrast microscopy and nuclear labeling with DAPI (Figure 4), were identical in siblings and mutants of both alleles at 4 and 17 dpf (Figure 4A,B shows cdh23^tc317e at 4 dpf, Figure 4C,D shows cdh23^tj264a at 17 dpf). In addition, the layering of GABAergic fibers in the inner plexiform layer was indistinguishable between siblings and mutants (Figure 4, right panels). We predict that the defects caused by these two alleles likely do not contribute to gross retinal degeneration or disorganization, but may instead contribute to subtle defects that would require more careful analysis of the cdh23-positive amacrine cell population in isolation.

OKR is normal in cdh23 mutants

To investigate whether visual function was affected in cdh23 mutants, we tested the OKR, varying either the contrast or spatial frequency of the grating or stripes. At 5 dpf, the OKR of homozygous larvae of both alleles was indistinguishable from wild-type responses. We found a reduction neither in contrast sensitivity (Figure 5A) nor in visual acuity (Figure 5B). This result is consistent with an earlier study in which vestibular function was reported to be absent in homozygous cdh23^tc317e (also referred to as cdh23^1619ag) larvae, whereas the OKR occurred under bright light conditions [44]. While we were not able to detect a reduction of the OKR in mutant cdh23 larvae at this stage of development, testing this particular reflex may not be sufficient to observe functional defects in either the amacrine cells themselves, or upstream in the visual processing centers of the brain.