Duplicated genes with split functions: independent roles of protocadherin15 orthologues in zebrafish hearing and vision

Sensory receptors in the ear and eye are thought to have evolved from a common ancestral ciliated cell (reviewed by Popper et al., 1992). Both receptors have highly specialized apical surfaces. Receptor cells in the ear possess an apical bundle of actin-filled stereocilia that transduce mechanical stimuli (reviewed by Hudspeth,1989), whereas the photoreceptors of the eye have an outer segment that contains membrane disks specialized for detection of light. Through genetic studies, several molecules required for both hearing and vision have been identified. In humans, mutations in these genes cause Usher syndrome (for a review, see Petit, 2001),one of the most prevalent inherited disorders(Ahmed et al., 2003b). Individuals suffering from Usher 1 syndrome are born deaf and develop visual problems in the second to fourth decade of life due to retinitis pigmentosa. Mutations in most of these genes have been shown to cause deafness in mice as well, but no phenotype comparable to retinitis pigmentosa has been reported(reviewed by Steel and Kros,2001; Ball et al.,2003; Libby et al.,2003).

As in mammals, fish sense mechanical stimuli with hair cell receptors in the inner ear and an additional sensory system unique to aquatic animals, the lateral line (Manley and Koppl,1998; Popper and Fay,1997). In recent screens for genes necessary for inner ear function in zebrafish larvae, 24 loci linked to an auditory/vestibular-specific phenotype were identified(Nicolson et al., 1998) (T.N.,unpublished). Here we show that the zebrafish auditory/vestibular mutant orbiter is affected in the gene pcdh15. Mutations in PCDH15 have been recently shown to be the cause of human Usher syndrome 1F (Ahmed et al.,2001; Alagramam et al.,2001b; Ben-Yosef et al.,2003) and deafness in Ames waltzer mice(Alagramam et al., 2001a). Furthermore, we report the characterization of a second, duplicated pcdh15 gene in zebrafish, which is expressed in the retina. Our results indicate that one gene acquired an essential function in the eye and the other one in the ear. Both genes are required for the structural integrity of the respective receptor cells in these organs.

All studies were done with larvae in the Tübingen (Tü)background. albino larvae were used for in situ hybridization. Zebrafish stocks were maintained as described(Haffter et al., 1996).

Linkage analysis was performed using WIK wild-type fish crossed with heterozygous orbiter^th263b-Tü fish. The mutation was mapped using SSLP (simple sequence length polymorphisms) and SSCP (single strand conformation polymorphisms) (Cleangels SSCPplus gels were used according to the manufacturer's instruction; ETC, Kirchentellinsfurt,Germany). For fine-mapping, single mutant larvae were tested for recombination events. To initiate a chromosomal walk, we used the marker z26404 to screen PCR-able pools of genomic zebrafish BAC (BioCat, Heidelberg, Germany) or PAC(RZPD, Berlin, Germany) libraries. All clones found were sized using pulsefield gel electrophoresis. New SSLP or SSCP markers were created using sequences from the ends of the identified BAC- or PAC-clones. After defining a critical interval, six clones spanning the contig (PAC1, BAC1, PAC10, PAC13,BAC6, BAC7) were sequenced at the Sanger Center (Welcome Trust Sanger Institute, Cambridge, UK). The sequence within the critical interval was analyzed using Genscan(http://genes.mit.edu/GENSCAN.html)and blast searches of the NCBI database(http://www.ncbi.nlm.nih.gov/BLAST).

To clone the 3′ and 5′ ends of the genes several rounds of RACE with either the Marathon or SMART RACE cDNA amplification kit (Clontech, Palo Alto, CA) were performed according to the manufacturers' instruction. The Goodfellow T51 panel was used to map pcdh15b(Geisler et al., 1999). Mapping prediction on the T51 panel map was done using the `Instant Mapping'website from The Children's Hospital Zebrafish Genome Project Initiatives(http://zfrhmaps.tch.harvard.edu/ZonRHmapper/instantMapping.html).

The official names of the PAC clones from RZPD are the following: PAC1:BUSMP706C1835Q2; PAC5/7: BUSMP706H0192Q2; PAC9: BUSMP706M01204Q2; PAC12:BUSMP706N24105Q2; PAC10: BUSMP706N24105Q2; PAC11: BUSMP706O192Q2; PAC13:BUSMP706J031239. See the RZPD-Database(https://www.rzpd.de)for further information. GenBank accession number of the clones sequenced by the Sanger center: BAC1: AL592289; BAC6: AL592062; BAC7: AL645689; PAC10:AL592202; PAC13: AL592204; PAC1: AL592077. GenBank accession numbers for zebrafish pcdh15a: AY772390; and pcdh15b: AY772391.

Total RNA was prepared from pools of 5-10 wild-type sibling or mutant larvae using the NucleoSpin RNAII Kit (MACHEREYNAGEL, Easton, PA). Reverse transcription reactions were performed using the Superscript reverse transcriptase (Invitrogen, Carlsbad, CA) with an oligo (dT) primer.

Whole mount in situ hybridization was performed as described(Schulte-Merker et al., 1994). A fragment containing bp 180-1800 from the extracellular domain(Fig. 2) or bp 4516-5908 from the polymorphic intracellular part were used as probes for pcdh15a. Both probes gave the same signal; the signal with the extracellular probe was stronger and appeared more specific. bp 4075-5697 from the intracellular part was used as probe for pcdh15b.

For cryosectioning, larvae were immersed with increasing concentration of fish gelatin (FG, 8-25%) (Sigma G-7765 or G-7041) and sucrose (SC 8-15%) in Ringers during a 48-hour period at room temperature. They were transferred to forms made of aluminium foil covered with 20% FG, 15% SC and quickly frozen on a metal plate bathing in liquid nitrogen. 14 μm sections were cut at-20°C using a Leica CM 3050 cryomicrotome.

For electron microscopy, larvae were processed as previously described(Seiler and Nicolson, 1999). Whole larvae were embedded in Epon and oriented such that the eye was sectioned in a transverse dorsal to ventral direction. All sections were done in the plane of the optic nerve. Images were taken from areas adjacent to the optic nerve in the center of the retina.

Four different morpholinos were used in this study and injected at the one cell stage as described (Nasevicius and Ekker, 2000). One was directed against the translation start (ATG Mo) of pcdh15b with the sequence 5′-CCACGACCTCTGGCGATCTTCATA-3′, one against the 3′ splice site of coding exon number 4 (GT Mo) with the sequence 5′-AAAAGATAATCACATCTCGGTCCAG-3′, and an unrelated control morpholino (a 4 bp mismatched oligo against fkd6;5′-CGACGAGCTCTGCCGATCTTGATA-3′). A second control morpholino was a 4 bp mismatched oligo against pcdh15b(5′-TGCGCGCCAGACAGGGTGATGAC-3′). In the ATG Mo-injected larvae (40 ng), a slightly shorter body axis of the larvae was occasionally observed. Differentiation of other organs like the heart, brain and ear appeared normal,indicating that this effect was not due to a delay in development. To ensure that the 40 ng dose had no effect on the onset of eye differentiation, we stained 14 μm sections of eyes of these larvae with anti-tyrosine hydroxylase antibody and counted the positive cells. No difference in number could be detected between uninjected (1.4 cells/section n=13) and larvae injected with 40 ng ATG Mo (1.5 cells/section n=14).

For confocal microscopy, an inversed Leica DM IRBE microscope equipped with a 100× oil lens was used. For antibody staining, the larvae were fixed with 4% formaldehyde in PBS for at least 24 hours prior to sectioning. The sections were incubated with 1% bovine serum albumin. The following antibodies were used in this study: zpr-1 (The zebrafish international resource Center, Eugene, OR, USA), rhod-1 (Ret-P1, Milan, Analytica AG/Switzerland), and Alexa488 goat anti-rabbit Ab (Molecular Probes, Eugene,OR).

Single larvae were placed dorsal side up in the center of a petridish (35 mm diameter) containing 3% prewarmed (28°C) methylcellulose. Moving gratings were projected by a Proxima 4200 DLP projector onto a screen within the visual field of the larva, at an apparent distance 4.65 cm from the larva's right eye. Projection size on the screen was 8 cm × 6 cm,subtending a visual angle of 65.6° horizontally and 53.1° vertically. A custom-made graphics library running under Microsoft DOS, allowing full control of timing and intensity of the projection was used to create the stimuli. Mean intensity levels were adjusted by introducing neutral density filters into the light beam. The contrast C of grating stimuli was calculated from: C =(I_max -I_min)/(I_max +I_min).

Spatial frequency of the grating stimulus was varied between 0.025 and 0.14 cycles/°. All measurements were made under a mean luminance of 120 cd/m², temporal frequency of 0.4 Hz, pattern contrast of 62.5% and a grating velocity of 7.6°/second.

We measured the CSF in 4-day-old larvae injected with either an ATG Mo (40 ng n=7 or 30 ng n=7) or the control Morpholino (40 ng n=7). The experiment has been repeated with similar results in an independently injected clutch.

Electroretinograms (ERGs) were performed on larvae at 4.5 dpf as previously described (Makhankov et al.,2004). All pre-recording steps were done under red illumination to minimize bleaching of the visual pigment. Preparation and recordings were performed in a tight Faraday cage. All specimens were dark-adapted for 30 minutes prior to positioning them in the recording chamber. Each larva was placed on its side on the surface of a moist sponge with E3 medium (mM: 5 NaCl, 0.17 KCl, 0.33 CaCl₂ and 0.33 MgSO₄) and paralyzed by directly adding a droplet of the muscle relaxant Esmeron (0.8 mg/ml in larval medium; Organon Teknika, Eppelheim, Germany). The Ag/AgCl electrode system was used to record the ERG response. The recording electrode was positioned in the center of the cornea. The reference Ag/AgCl pellet was placed under the body of the larva. A 3-minute period was chosen to adapt the larvae to both dark- and light-adapted states prior to measurement. Relative background light intensitiy was 175 lux for the light-adapted state. The duration of light stimulus was 100 mseconds with an interstimulus interval of 5 seconds. Stimulus illumination was increased in 0.5 log unit steps over the range from -5 log unit to -1 log unit. Unattenuated light intensity over the subject's head was 20,000 lux (optical density, OD was equal 0 log unit), as measured by a light meter (Tekronix J17, Texas Instruments, USA). A virtual instrument (VI) under NI LabVIEW 5.1 was developed to use in all experiments. Amplified analog signals were sampled by means of NI PCI 6035E DAQ board connected to a NI BNC-2090 BNC terminal block. Sampling was done in buffered acquisition mode with a sampling rate of 250 Hz. Recording was triggered by the shutter signal. To analyze the ERG response with respect to the actual onset of the light stimulus, mechanical shutter delay was measured by means of a photodiode. Traces were normalized to the baseline by subtracting the average potential before the stimulus onset. Responses were averaged between three and five, times depending on the signal-to-noise ratio.

Mutant orbiter zebrafish larvae have a vestibular defect and cannot sense acoustic-vibrational stimuli(Granato et al., 1996; Nicolson et al., 1998). Extracellular microphonic potentials of orbiter^th263bmechanosensory hair cells are absent, suggesting that the primary events in the mechanotransduction process are affected(Nicolson et al., 1998). To identify the orbiter gene, we undertook a positional cloning approach. By testing 2951 mutant larvae for recombination events using polymorphic CA repeat markers, we mapped orbiter to chromosome 13,north of the marker z26404 (4 recombinations in 2951 larvae) and south of the marker z25253 (99 recombinations in 2780 larvae). Using z26404, we started a chromosomal walk and constructed a contig comprising several BAC and PAC clones that spanned the critical interval(Fig. 1A). To identify the affected gene, three PAC and three BAC contig clones were sequenced. Analysis of these sequences revealed that these clones contain exons present in the first half of a gene that is homologous to human and mouse protocadherin 15(PCDH15). The entire contig contained bp 1-3225 of pcdh15;bp 1352-3225 was within the critical interval. No other genes were present on the clones. We performed RACE to clone the full-length zebrafish pcdh15 gene, which we designated pcdh15a (see below). Sequence analysis revealed that it shares high homology with human and mouse protocadherin 15. The Pcdh15 protein belongs to a class of atypical cadherins:the protein is composed of 11 cadherin repeats, a single transmembrane domain,and an intracellular domain with no similarity to other proteins(Fig. 1B).

The five orbiter alleles can be grouped in two phenotypic classes. One class causes strong defects such as no response to acoustic vibrational stimuli and a vestibular defect, no uptake of the styryl dye FM1-43 into hair cells, and no detectable microphonic potentials [microphonics only measured in th263b (Nicolson et al.,1998)]. This class includes the th263b, t25708,and t288 alleles. The other class is a weak, hypomorphic class with occasional response to acoustic vibrational stimuli but a fully penetrant vestibular defect, weak uptake of FM1-43 (data not shown)(Nicolson et al., 1998; Seiler and Nicolson, 1999),and reduced microphonic potentials [tc256e(Nicolson et al., 1998)]. The weaker alleles include tc256e and t23281.

Sequencing of pcdh15a cDNA from the mutants revealed that it is the gene affected in orbiter mutants. For the alleles of the strong class, we found that the mutation in t25708 leads to a valine to glutamine exchange (V142Q) in close proximity to the conserved cadherin calcium-binding motif (with the consensus DXNDN) of the first cadherin domain. The mutation in th263b causes a truncation of the protein in the third cadherin repeat (R360X), presumably resulting in a nonfunctional protein. t288 causes a leucine to proline (L386P) exchange, adjacent to the calcium binding domain of the third cadherin domain (here DENNQ instead of DXNDN) (Fig. 1B).

Sequencing of the alleles of the hypomorphic class revealed that tc256e causes an isoleucine to asparagine (I419N) exchange in the linker between the third and fourth cadherin domain. t23281 causes an asparagine to lysine (N1143K) exchange in the tenth cadherin domain, changing the calcium-binding motive from DENDH to DEKDH(Fig. 1B). The weaker phenotype seen in homozygous t23281 animals suggests that this cadherin domain is less critical for Pcdh15 function.

In addition to the pcdh15 gene mutated in orbiter, we identified a second ortholog of pcdh15. We designated the gene mutated in orbiter as pcdh15a, and the other gene pcdh15b. pcdh15b is located on chromosome 12 between the markers dkk1 and z1473 (T51 RH panel). The predicted Pcdh15b protein has the same domain structure and shares high amino acid similarity (70%) with the extracellular domain of zebrafish Pcdh15a, and mouse and human PCDH15(Fig. 1C). The intracellular domain is less similar to the intracellular domains of Pcdh15a and mouse and human protocadherin 15 (17-33%, Fig. 1C), but the intracellular tail is divergent among all species(Fig. 1C). Nevertheless, all Pcdh15 intracellular domains share the feature of being proline- and serine-rich, and there are short stretches of high similarity. Pcdh15b has a putative PDZ class I binding motif at its C-terminus [consensus X-S/T-X-L(Sheng and Sala, 2001)]. This motif is also conserved in human and mouse PCDH15, but is not present in zebrafish Pcdh15a, which is shorter at the C-terminus.

Phylogenetic analysis based on a ClustalW alignment revealed that the two zebrafish proteins are more related to each other than to the mammalian orthologs (Fig. 1D). This suggests that they were duplicated after the split of the mammalian and fish lineage, probably during the suggested genome duplication event in ray-finned fish (Prince and Pickett,2002; Taylor et al.,2003).

To determine the spatial and temporal expression of pcdh15a and pcdh15b, we performed mRNA in situ hybridization. At the four-somite stage (11 hours post fertilization, hpf), pcdh15a is expressed in the anterior neuroectoderm and in the first forming somites(Fig. 2A). At 24 hpf, pcdh15a is expressed in the eye(Fig. 2B) and hatching gland(out of focus). Expression is also detectable in the newly developed mechanosensory hair cells (Fig. 2B,C). Expression of pcdh15a in hair cells was visible in all later stages investigated (Fig. 2D-F′). At day 3, additional expression in regions of the brain is present, together with expression at the lateral borders of the epiphysis (Fig. 2D). At day 4,expression in the brain persists and is detectable in the hair cells of neuromasts (Fig. 2E,F′). Within the neuroepithelia of the ear and lateral line, levels of pcdh15a mRNA are highest in the hair cells(Fig. 2F,F′).

pcdh15b is weakly expressed in the whole embryo at 24 hpf (data not shown). At 48 hpf, strong expression in the epiphysis is visible but not in the eye (data not shown). At day 2.5, the gene is expressed at the ventral margin of the eye (Fig. 3A),where the first photoreceptors differentiate. Following the wave of differentiation of photoreceptors (Larison and Bremiller, 1990; Raymond et al., 1995), pcdh15b is expressed in the proximal regions of the eye around day 3 and restricted to the lateral margin at day 5(Fig. 3B,C). In thick sections,expression is evident in photoreceptors(Fig. 3E,F). In addition, pcdh15b is expressed in the brain and center of the epiphysis at day 3 (Fig. 3D) and 5. Weak expression was detected in the ear and neuromasts (data not shown). No signal was detected with the corresponding sense probes.

To analyze the hair bundles in orbiter mutants in more detail, we stained the actin-filled stereocilia of the hair bundles with fluorescently-labeled phalloidin (Fig. 4). We observed a moderate degree of splaying in the hair bundles in the semicircular canals in the strong allele th263b(Fig. 4B). Not all stereocilia were split apart, and some bundles appeared normal. Hair bundles were less affected in larvae carrying the weak allele tc256e(Fig. 4C). The defect in th263b is much weaker in comparison to other splayed-bundle mutants like mariner (Fig. 4D).

In Ames waltzer/Pcdh15 mutant mice, the planar polarity of the hair bundle of cochlear hair cells was sporadically disturbed, with hair cells rotated at 90° (Alagramam et al.,2001a; Raphael et al.,2001). By staining the actin in the bundle and the cuticular plate with fluorescently-labeled phalloidin, we could visualize the polarity of the hair bundles in zebrafish neuromasts (NM), because the kinocilium and its insertion point are not stained with phalloidin. The kinocilium is located at either the caudal or rostal margin of the hair cell (n=12 NM, three specimens) (Fig. 4E-F). No difference in polarity could be detected in th263b (n=10 NM,5 spec) (Fig. 4E′-F′; arrows indicate orientation each bundle with arrowhead pointed towards the kinocilium; see Fig. S1 in the supplementary material for more examples). The number of hair cells (HC) in neuromasts was reduced (12.5 HC/NM in tc263b; 17HC/NM in WT; counting NM above and around the ear) (Fig. 4F). A similar reduction was seen in mariner neuromasts (13,25 HC/NM, n=12, 3 spec). It is not clear why the hair bundles in Ames waltzer mice have a polarity defect, whereas mutant orbiter hair bundles do not.

To determine the function of Pcdh15b, we designed two morpholinos: one against the ATG translational start site (ATG Mo), and one against the GT splice site of the fourth coding exon (GT Mo). The ATG Mo is predicted to inhibit the initiation of translation of pcdh15b message, whereas the GT Mo should prevent the correct splicing of exon 4. Because the ATG Mo inhibits protein synthesis of Pcdh15b, its efficiency cannot be evaluated at the mRNA expression level. However, we could test the efficiency of the GT Mo by amplifying cDNA with primers flanking the targeted exon. RT-PCR analysis confirmed that the GT Mo indeed caused aberrant splicing of the targeted exon. Wild-type transcript levels are reduced by approximately 60% on day 3, and reduced by 40% on day 4 (Fig. 5D). The results of the histological analysis and electroretinograms (see below) suggest that the ATG Mo was more effective than the GT Mo. Morpholino-injected animals appeared to develop normally and overall morphology was normal (see Fig. S2 in the supplementary material). No obvious defects in balance or the acoustic startle reflex or hair bundle morphology were observed in pcdh15b morpholino-injected (morphant)larvae (data not shown). In addition, we examined uptake of the vital dye,FM1-43, an indicator of mechanotransduction in zebrafish hair cells (Seiler et al., 1999). Uptake was unaffected in pcdh15b morphants (see Fig. S3 in the supplementary material).

To test whether morphant larvae have a visual defect, we measured the optokinetic response (OKR). The OKR is a reflexive movement of the eye when stimulated with a moving pattern. The velocity of the eye movement depends on the density and contrast of the pattern. We measured the gain (eye velocity/stimulus velocity) of 4-day-old larvae injected with 40 ng of an unrelated control morpholino (control Mo) and larvae injected with 30 ng or 40 ng ATG Mo during stimulation with increasing spatial frequency (i.e. decreasing stripe width of the moving grate)(Fig. 5A). Comparison of the pcdh15b morphants with control injected larvae (40 ng, n=7)revealed a significant difference in sensitivity (ANOVA with repeated measures, d.f.=2, F=44.1, a<0.001). The comparison of responses in larvae injected with 40 ng or 30 ng pcdh15b ATG Mo is significant as well (a<0.001), indicating an obvious dosage effect. The experiment was repeated, obtaining similar results in an independently injected clutch of eggs (data not shown, n=7, 40 ng ATG Mo, 31% reduction in contrast sensitivity). In contrast, no OKR defect was detectable in orbiter^th263b mutant larvae when compared to wild-type larvae (data not shown).

To determine whether injection of pcdh15b morpholinos results in physiological defects that are specific for retinal function, the electroretinogram (ERG) b-wave was measured in 4.5-day-old larvae. Larvae injected with 40 ng of the ATG Mo (n=6) or 16 ng of the GT Mo(n=6) were compared with uninjected larvae (n=5) and larvae injected with the control Mo (n=5). As shown in Fig. 5B and C, the b-wave increases in amplitude with increasing stimulus intensity in all experimental groups. Injection of both morpholinos resulted in a significant reduction in b-wave amplitude. (3-way ANOVA d.f.=3, F=11.7, a<0.001). Although both morpholinos are targeted against different sequences, they cause a comparable reduction in b-wave amplitude, indicating that this effect is specific for the knockdown. Comparison of the pcdh15b morphants with control-injected and uninjected larvae is significant for 40 ng ATG MO(a<0.001) and 16 ng GT MO (a<0.001; Scheffe test). The weaker b-wave amplitude reduction between the GT Mo compared to the ATG Mo injected larvae was not significant, though. Representative traces from individual specimens are shown in Fig. 5E,F. Due to a very low amplitude and differing width, we were not able to evaluate the a-wave data conclusively. Nevertheless, our behavioral and physiological analyses indicate a clear defect in retinal responses in pcdh15b morphants.

To determine if the visual defect is due to a morphological defect in the retina, we examined cryosections of morpholino-injected larvae using immunohistochemistry. In larvae injected with 30 ng (n=5) or 40 ng(n=7, 2 independent experiments) of the ATG Mo, we observed a dose-dependent phenotype in the retinal photoreceptor cell layer(Fig. 6). The zpr-1antibody labels double cone photoreceptor cell bodies and outer segments(Larison and Bremiller, 1990). In 4-day old morphants, the general morphology of the photoreceptors labeled with zpr-1 antibody appeared disorganized. Labeled cells appeared wider, shorter, and deformed (Fig. 6A-F). Effects were also evident using an anti-rhodopsin(rhod-1) antibody that labels outer segments. Outer segments of the rod photoreceptors appeared shorter and less organized than wild-type outer segments (Fig. 6G-I). The effects on both rods and cones seen with the antibodies were stronger in larvae injected with 40 ng (Fig. 6F,I) than in larvae injected with 30 ng ATG Mo(Fig. 6E,H).

To examine the morphological defect at the ultrastructural level, we performed transmission electron microscopy (TEM) with thin sections of 4-day-old larvae. We compared uninjected (n=2) larvae to larvae injected with either 40 ng ATG Mo (n=3), 16 ng GT Mo (n=2)or 40 ng control (n=2) or 4 bp mismatch (n=3) control Mo. In the ATG Mo-injected larvae, the layer of photoreceptor nuclei appeared more compacted than in uninjected larvae (Fig. 7A,B). The outer segments of the photoreceptors were often tilted sideways and clustered together. In addition, the number of pigment cell granules positioned between the outer segments was greatly reduced(Fig. 7C). This effect was still present in 7-day-old larvae, though the outer segments were more clustered and less tilted (Fig. 7E,F). A similar but weaker defect was visible in the GT Mo-injected larvae (data not shown). A lack of interdigitation of the outer segments and therefore the absence of the protective ensheathing by the pigmented epithelium is consistent with the poorer performance of the pcdh15b morphants under bright-adapting conditions. Defects were not detectable in the control Mo-injected larvae or in larvae injected with a pcdh15b 4 bp mismatch morpholino(Fig. 1A′). In addition,we also examined orbiter^th263b retinas using TEM, and did not observe any obvious defects (data not shown).

Here we show that protocadherin 15, responsible for retinitis pigmentosa and congenital deafness in humans(Ahmed et al., 2001; Alagramam et al., 2001b), is duplicated in zebrafish. In larvae, Pcdh15a is specifically required in the ear, whereas Pcdh15b is required in the developing eye. The expression patterns of the two pcdh15 paralogs provides insight into the evolution of these two genes. After gene duplication, one of the paralogous genes is often lost from the genome due to its redundant function(Prince and Pickett, 2002). If duplicated genes acquire non-redundant functions, then both are likely to be retained (Force et al., 1999). In the case of the pcdh15 zebrafish paralogues, the high divergence of the intracellular domains may account for an adaptation or specialization of function of these two genes in the eye and ear. Although pcdh15aand pcdh15b are required in different sensory organs, both pcdh15 genes retained a function in maintaining the structural integrity of the specialized apical surfaces of the respective sensory receptor cells.

In a recent study, PCDH15 protein was shown to be localized at the stereociliary bundle of hair cells in mammals(Ahmed et al., 2003a). This result is consistent with the bundle phenotype we detected in orbitermutants, suggesting that orbiter/pcdh15a encodes a component of the stereociliary bundle. The large extracellular domain of Pcdh15a, along with the localization data from mammals, suggests that Pcdh15a may be part of the outer membrane calyx or a component of the extracellular links spanning neighboring stereocilia. Although the bundle phenotype is mild in orbiter mutants, microphonic potentials are not detectable in larvae carrying the strong allele. This phenotype raises the possibility that Pcdh15a plays a functional role in hair bundles.

In the pcdh15b morphants, the tilted outer segment phenotype suggests at least two possible functions of pcdh15b during differentiation. First, Pcdh15b may be necessary for adhesive contact between the outer segment of photoreceptors and the retinal pigment epithelium. Interestingly, when the retinal pigment epithelium is absent in the zebrafish retina, the outer segments are tilted sideways in a similar way (C. Seiler and A. Rojas-Munoz, unpublished). This phenotype suggests that the outer segments are actively intercalated into the pigmented epithelium. This paradigm requires trans interactions of at least two cadherins expressed in photoreceptors and the retinal pigment epithelium, but we could not detect pcdh15b in the pigment epithelium. Another possibility is that Pcdh15b is required for stabilization of photoreceptor outer segments. Such contacts may help to align the outer segments, making interdigitation with the retinal pigment epithelium possible. The localization pattern of PCDH15 in photoreceptor outer segments in the human and monkey retina(Ahmed et al., 2003a) appears to be consistent with these notions. PCDH15 protein was also detected in other layers of the fetal and adult human retina, especially the synaptic layers,and may play a role in synaptogenesis as well(Alagramam et al., 2001b).

To date, the only mouse mutant in an Usher gene reported to have a morphological defect in the eye is the shaker1/Myo7a mutant. In shaker1 mutants, the melanosomes of the retinal pigment epithelium do not migrate between photoreceptors, suggesting that myosin VIIA is needed for transport of these pigmented structures into the apical processes of retinal pigment cells (Liu et al.,1998). Slight reductions in ERG responses have also been detected in myosin VIIA mutant mice (Libby and Steel, 2001). Moreover, myosin VIIA has been shown to be expressed and required in the retinal pigment epithelium for uptake of outer segment disks (Gibbs et al., 2003). These results suggest that myosin VIIA also participates in photoreceptor and retinal pigment epithelium interactions.

In human Usher syndrome, retinitis pigmentosa develops in the second to third decade of life. This late defect in humans contrasts with the very early onset of photoreceptor malformation in zebrafish pcdh15b morphants. It is worth noting that in zebrafish, the eye develops much faster than in mammals (O'Brien et al., 2003; Sharma et al., 2003). Within 2.5 days, the zebrafish eye develops from a undifferentiated neuroephithelium to a layered and differentiated functional organ(Easter and Malicki, 2002). Furthermore, a fully populated photoreceptor cell layer develops within 30 hours (Raymond et al.,1995).

Thus, the contact and migration of outer segments into the retinal pigment epithelium happens much faster in zebrafish, and proper function of Pcdh15 appears critical at this timepoint. In various mouse models for retinal degeneration, the photoreceptor cell layer is initially disorganized and shows severe degeneration later in life [for example rd8/Crb1(Mehalow et al., 2003) rd6/Mfrp (Hawes et al.,2000; Kameya et al.,2002)]. Due to the limited efficiency of antisense morpholinos, we were not able to investigate juvenile stages to look for degeneration. However, photoreceptor interactions with the retinal pigment epithelium have been shown to be important for the outer segment disc phagocytosis, visual pigment renewal, and nutrient supply(Boulton and Dayhaw-Barker,2001; Pacione et al.,2003). Moreover, progressive death of photoreceptors was observed in the vitiligo/Mitf mouse mutant in which retinal pigment epithelium initially failed to interdigitate with photoreceptors(Sidman et al., 1996), and in mutant vestigial outer segments zebrafish, which exhibit a similar phenotype to pcdh15b morphants(Manzoor-Ali et al., 2003). Thus, impaired contact between the retina pigment epithelium and the outer segments may cause degeneration at later stages.

The phenotypes observed in either mutant orbiter/pcdh15azebrafish or pcdh15b morphants indicate a clear requirement for Pcdh15 proteins in hearing and vision in zebrafish. The sensory receptors in either the ear or the eye rely upon these protocadherins for structural integrity. Our study establishes a role for Pcdh15 in maintaining the structural integrity of the photoreceptor outer segment, and highlights the usefulness of an alternative animal model in studying the function of Usher genes.

We are grateful to the Sanger Institute for sequencing our contig clones. We thank Brigitte Sailer for her technical assistance with the TEM analysis. We would also like to thank P. Mayinger, R. Walker, and S. Jesusthasan for helpful comments on the manuscript.