Repression of Hedgehog signalling is required for the acquisition of dorsolateral cell fates in the zebrafish otic vesicle

The vertebrate inner ear is a complex structure with asymmetries about all three body axes. These asymmetries arise early during ear development; by otic placode and vesicle stages, several otic genes are expressed asymmetrically (reviewed by Whitfield and Hammond, 2007). Patterning defects at these early stages will have consequences for subsequent ear development. To date, a small number of signalling molecules that influence otic patterning at these early stages have been identified, of which Hedgehog (Hh) is one.

Members of the Hh family of signalling molecules act as morphogens during axial patterning of many tissues: for instance patterning the dorsoventral (DV) axis of the neural tube and the anteroposterior (AP) axis of the developing limb bud (Echelard et al., 1993; Krauss et al., 1993; Riddle et al., 1993) (reviewed by Hammerschmidt et al., 1997; Ingham and McMahon, 2001; Ingham and Placzek, 2006). Graded levels of Hh signalling give rise to different fates along the axes of these tissues. Similarly, Hh signalling from ventral midline structures is vital for otic axial patterning in all vertebrates so far examined.

In zebrafish, we have previously shown that Shha and Shhb are necessary and sufficient to specify posterior otic identity (Hammond et al., 2003). When Hh signalling is lost or severely reduced, posterior otic structures are lost and anterior structures are duplicated in their place. Similar double anterior phenotypes have been reported in Xenopus embryos overexpressing mRNA encoding the Hh inhibitor Hip (Waldman et al., 2007). Conversely, when Hh signalling is overactivated by shh or dnPKA overexpression in the zebrafish embryo, anterior otic structures are absent and posterior regions are duplicated (Hammond et al., 2003). In mouse and chick, however, manipulation of Shh activity predominantly affects otic DV and mediolateral (ML) patterning; AP effects, if present, are not obvious (Bok et al., 2005; Riccomagno et al., 2002). This apparent difference in the role of Hh in otic patterning between amniote and anamniote vertebrates is surprising, as the structure of the inner ear is similar in both groups, except for the presence of the ventrally positioned cochlea, a specialised auditory endorgan, in the amniote ear.

Subsequently, however, we have established that whereas a loss of Hh function does not affect the otic DV and ML axes in zebrafish (Hammond et al., 2003), increasing Hh levels by shh mRNA injection causes an expansion of ventromedial (VM) otic territories at the expense of dorsolateral (DL) domains. To investigate further, we analysed the otic phenotypes of a panel of lines carrying mutations in genes encoding inhibitors of the Hh pathway: ptc1, ptc2, su(fu) (sufu – ZFIN), dzip1 and hip, all of which are expressed in and around the developing otic vesicle. These lines provide a series with increased Hh signalling throughout the embryo (Koudijs et al., 2008; Koudijs et al., 2005). Ptc (Patched), the Hh receptor, represses Hh pathway activity in the absence of Hh ligand (Chen and Struhl, 1996) (reviewed by Ingham and McMahon, 2001). ptc1 is expressed in a posteroventromedial domain of the zebrafish otic vesicle and ptc2 in a wider ventral domain (Hammond et al., 2003). Hip (Hedgehog interacting protein) is a membrane-bound protein that binds to the Hh ligand and prevents it binding to the Ptc receptor (Chuang and McMahon, 1999; Ochi et al., 2006). hip is expressed in a complex pattern in the zebrafish, initially concentrated towards the anterior of the otic vesicle (Hammond and Whitfield, 2009). Dzip1 (Daz interacting protein 1) and Su(fu) (Suppressor of fused) both act within the Hh-receiving cell to regulate activity of the transcription factor Gli, which mediates the Hh response (Méthot and Basler, 2000; Sekimizu et al., 2004; Wolff et al., 2004) (reviewed by Huangfu and Anderson, 2006). Both su(fu) and dzip1 are expressed ubiquitously throughout the zebrafish embryo (Koudijs et al., 2005; Wolff et al., 2004).

The overriding otic phenotype in these lines is a ventralisation and medialisation of the ear: with increasing Hh activity, dorsolateral structures are progressively lost. In the strongest phenotype, in embryos mutant for ptc1 and ptc2, the otic vesicle is strongly medialised and ventralised as well as posteriorised, and has a stronger phenotype than that generated by shh mRNA injection (Hammond et al., 2003). Gene expression pattern changes in the otic vesicle prefigure the defects in ptc1^–/–; lep^–/– and shh mRNA-injected otic vesicles.

Our data demonstrate that, in addition to a requirement for Hh signalling for AP otic patterning, inhibition of Hh signalling is crucial for the correct development of dorsolateral structures in the zebrafish inner ear. Otic vesicle patterning is very sensitive to small increases in Hh signalling; Hh pathway activity must therefore be tightly regulated for correct inner ear development. In addition, we show that the effects of derepression of Hh signalling on the zebrafish ear are likely to be mediated directly. Our data indicate that a requirement for inhibition of Hh signalling during zebrafish and amniote inner ear patterning is at least partially conserved.

Wild-type zebrafish strains were AB, Tup Longfin (TL) or WIK. Mutant lines were dre^tm146d (su(fu)^–/–), igu^ts294 (dzip1^–/–), lep^tj222 (ptc2^–/–), ptc1^hu1602 (ptc1^–/–) and uki^hu418B (hip^–/–) (Brand et al., 1996; Heisenberg et al., 1996; Karlstrom et al., 1996; Koudijs et al., 2008; Koudijs et al., 2005; Piotrowski et al., 1996; van Eeden et al., 1996; Whitfield et al., 1996). All are recessive loss-of-function alleles. Embryonic stages are given as hours post-fertilisation (hpf) at 28.5°C or as somite stages (S) (Kimmel et al., 1995; Westerfield, 1995).

Whole-mount in situ hybridisation was carried out as described (Hammond et al., 2003). Probes used were dlx3b, eya1, fgf8, fst1 (fsta – ZFIN), hmx3, msxc, otx1, pax2a, pax5, ptc1, ptc2 (Hammond et al., 2003), hip, pcna (Koudijs et al., 2005), tbx1 (Piotrowski et al., 2003), foxi1 (Solomon et al., 2004) and raldh3 (aldh1a3 – ZFIN) (Pittlik et al., 2008).

Genomic DNA was prepared as described (Westerfield, 1995). Primers were: dre, F 5′-TTCTGCTGTCAGAGGGTTTC-3′, R 5′-CAACTGCATTACTGGCAATC-3′; lep, F 5′-CACATTAAGATGGAAACCTG-3′, R 5′-TATCGAGCCTTTATTTAGCC-3′; ptc1, F 5′-GCATATATGTGTGGAGCTATTCTC-3′, R 5′-GAGCTGTGATCTTCAGCAAC-3′; uki, F 5′-GGGAGGCCTGGCTTTAG-3′, R 5′-CCATGGTTAATAGCCTTTGC-3′ (Koudijs et al., 2008; Koudijs et al., 2005). Sequencing was carried out at the Genetics Core Facility, University of Sheffield, using an ABI 3730 capillary sequencer. PCR primers were used for sequencing, except for lep, where the primer 5′-GTGGTGGTGTTTAACTTTGC-3′ was used.

Staining was carried out as described (Haddon and Lewis, 1996).

Embryos were fixed overnight in 4% paraformaldehyde (PFA) at 4°C and stored in methanol. They were washed in PBTw (1×PBS/0.1% Tween), treated with 10 μg/ml proteinase K (1 hour), washed in PBTw, blocked in 10% serum/PBTw (1 hour) and hybridised overnight in 1:500 mouse anti-Collagen type II antibody (II-II6B3 monoclonal, DSHB) at 4°C. After washing in PBTw, anti-mouse IgG-HRP (Sigma; 1:200) was applied before staining with a DAB detection kit (Vector Laboratories).

Embryos were fixed overnight in 4% PFA at 4°C, and stained as described (Schilling et al., 1996).

RNA injection was carried out as described (Hammond et al., 2003).

ptc1^–/–; lep^–/– double-mutant embryos were sorted from siblings at 13-14S based on somite phenotype (Koudijs et al., 2008). Ten to 15 embryos were treated in each well of a 12-well culture dish in 2 ml of embryo medium containing 0.25-50 μM cyclopamine/1% ethanol (Calbiochem) or 1% ethanol alone.

Acridine Orange treatment was carried out as described (Abbas and Whitfield, 2009).

Microscopy was carried out as described (Hammond et al., 2003).

Donor embryos were labelled with 5% rhodamine-dextran/3% biotin-dextran (Molecular Probes) as described (Piotrowski et al., 2003). Embryos were cooled to 21.5°C overnight to obtain embryos at the correct stage. ptc1^–/–; lep^–/– embryos were identified before embedding based on somite morphology (Koudijs et al., 2008). Embryos were embedded dorsal side up in 1% low melting point agarose in Ringer’s solution. Otic vesicles were extirpated from the left-hand side of donor and host embryos using glass microelectrodes and fine-gauge hypodermic needles. Labelled donor otic vesicles were transplanted into unlabelled hosts at 18-19S. Host embryos were cultured overnight in Ringer’s solution, and then transferred to embryo medium. Analysis of fluorescence and semicircular canal projection formation was carried out at 3 dpf. To detect biotin-dextran, embryos were fixed overnight in 4% PFA, washed, permeabilised and quenched as in Kane and Kishimoto (Kane and Kishimoto, 2002), and labelled using an ABC kit (Vector Laboratories) followed by a Cy3-tyramide kit (Perkin-Elmer).

Several genes encoding inhibitors of the Hh signalling pathway are expressed in and around the developing otic vesicle: these include su(fu), dzip1, ptc1, ptc2 and hip (Hammond et al., 2003; Hammond and Whitfield, 2009; Koudijs et al., 2005; Wolff et al., 2004). We examined otic patterning in embryos carrying homozygous mutations in these genes, both individually and in combination. These were uki (hip^–/–), lep (ptc2^–/–), dre (su(fu)^–/–) (Koudijs et al., 2005), ptc1^–/– (Koudijs et al., 2008) and igu (dzip1^–/–) (Sekimizu et al., 2004; Wolff et al., 2004). Hh signalling activity is upregulated in these mutants, ranging from a modest increase in dre to maximal Hh pathway overactivation in ptc1^–/–; lep^–/– double mutants, based on ptc1 and gli1 expression levels and phenotypic severity (Koudijs et al., 2008; Koudijs et al., 2005). In igu homozygotes, low-level Hh signalling is expanded, and high-level Hh signalling is reduced (Sekimizu et al., 2004; Wolff et al., 2004).

At 4 days post-fertilisation (dpf), inner ear phenotypes ranged from normal (ptc1^–/–), through mild (dre, lep), moderate (igu, uki) and substantial (ptc1^+/–; lep^–/–) disruption of dorsolateral structures, to severely ventralised and medialised (ptc1^–/–; lep^–/–) (Fig. 1; Table 1). As the severity of the ear phenotype increased, the dorsolateral septum, endolymphatic duct (a dorsomedial structure) and cristae (lateral structures) were progressively lost and the semicircular canal projections developed increasingly abnormally. All these structures were absent from ptc1^–/–; lep^–/– otic vesicles. We describe how each structure within the ear was affected in detail below.

The dorsolateral septum (dls) (Fig. 1A, dls) separates the anterior and posterior semicircular canals. It was normal in ptc1^–/– mutant ears (Fig. 1B), present but malformed and often positioned incorrectly in lep (Fig. 1D) and absent from dre, igu, uki, ptc1^+/–; lep^–/– and ptc1^–/–; lep^–/– ears (Fig. 1C,E-H). This did not result from increased cell death: we treated uki homozygotes and siblings with Acridine Orange between 42 and 74 hpf, when the dls normally forms (Haddon and Lewis, 1996), and observed no alteration in cell death in the dorsal region of the otic vesicle (data not shown).

In wild-type otic vesicles, the three cristae (anterior, lateral and posterior) develop ventrolaterally, as revealed by msxc expression (Fig. 2A) and by staining with FITC-phalloidin (Fig. 2D). Cristae in dre embryos were indistinguishable from wild type (Fig. 2U). In igu, lep, uki and ptc1^+/–; lep^–/– embryos, the lateral crista was lost or reduced in 29, 45, 77 and 100% of cases, respectively (Fig. 2B-H,U). In a number of uki mutants we also observed an additional small domain of msxc expression, possibly representing an incomplete separation of the anterior macula and lateral crista domains (arrowhead, Fig. 2H). Hair cells developed in both this region and the reduced lateral crista (arrowhead, Fig. 2E). No cristae formed in ptc1^–/–; lep^–/– double mutants (see Fig. 4F).

The endolymphatic duct (ED) forms dorsally from an outpocketing of the otic epithelium and is strongly marked by foxi1 expression (Abbas and Whitfield, 2009). In dre, lep and igu the ED was not significantly different in length between mutants and stage-matched 3 dpf siblings (t-test: P=0.39, n=8; P=0.95, n=18; P=0.65, n=21, respectively), whereas in uki^–/– and ptc1^+/–; lep^–/– embryos, ED length was significantly reduced (P=0.0009, n=15 and P=0.0052, n=8, respectively) (Fig. 2K,L). No ED was present in ptc1^–/–; lep^–/– embryos (Fig. 2I,J).

In zebrafish, the anterior (ap), posterior (pp) and ventral (vp) (horizontal) semicircular canal pillars form by fusion of epithelial projection tissue by 3 dpf. In ptc1^–/– and dre embryos, all three pillars formed normally (Fig. 1A-C). In lep, igu and uki, the ap and pp were often thin and spindly, whereas the vp was enlarged and dysmorphic (Fig. 1D-F). In ptc1^+/–; lep^–/– embryos, all pillars were very reduced (Fig. 1G). In ptc1^–/–; lep^–/– double-mutant embryos, no canal projections formed (Fig. 1H).

The dysmorphic vp in lep, igu and uki showed several gene-expression abnormalities (arrows, Fig. 2M-T). It expressed raldh3 and Collagen type II at high levels, and stained positively with Alcian Blue, markers that are all absent from the wild-type vp. All three markers were also occasionally seen in the malformed ap and pp of lep, uki, igu and ptc1^+/–; lep^–/– embryos (data not shown). Expression of otx1 in the dysmorphic vp was relatively normal, however (arrows, Fig. 2N,R). To test whether the vp in lep, uki and igu was enlarged as a result of overproliferation, we examined expression of proliferating cell nuclear antigen (pcna) mRNA in this tissue. However, pcna expression levels did not appear to be increased at 60 hpf, when the ventral canal projection formed (data not shown).

The anterior and posterior maculae were grossly normal in the Hh inhibitor mutants (see, for example, uki: Fig. 2D,E), with the exception of ptc1^+/–; lep^–/– embryos, in which the anterior macula was positioned slightly medial to normal (Fig. 2F), igu^–/–, in which the posterior macula was often small and badly formed (data not shown) (Hammond et al., 2003), and ptc1^–/–; lep^–/–, which had a single medial macula with double posterior character (Fig. 4 and see below).

To investigate whether the DL otic defects in the Hh inhibitor mutants are due to patterning changes at otic vesicle stages, we examined expression of tbx1 (a marker of lateral otic epithelium), pax2a (medial), dlx3b (dorsal) and eya1 (ventral), between 25S and 30 hpf (Fig. 3). In uki, igu, lep and dre, there were no obvious alterations in otic expression of these markers (data not shown). In ptc1^–/–; lep^–/– otic vesicles, however, medial and ventral otic territories, marked by pax2a and eya1, were expanded relative to ear size at the expense of lateral and dorsal domains, marked by tbx1 and dlx3b. Injection of 5 nl of 50 μg/ml shh RNA into wild-type (WIK) embryos produced a similar, but variable, effect (Fig. 3).

To test whether the expansion of VM otic fate at the expense of DL identity could be a result of increased cell death in DL otic domains, we stained embryos with Acridine Orange to highlight apoptotic tissue. However, we saw no increase in cell death in DL regions of ptc1^–/–; lep^–/– otic vesicles (16S to 29 hpf) compared with age-matched siblings (data not shown). At 16S, the cavitating otic vesicle of ptc1^–/–; lep^–/– embryos was similar in size to that of wild-type siblings but by 29 hpf was significantly reduced (see Fig. S1 in the supplementary material). Cell death was, however, increased anteroventrally to the otic vesicle (see Fig. S2 in the supplementary material). This might include the forming statoacoustic ganglion. Interestingly, in zebrafish embryos in which Hh signalling is absent (smo, con), the otic expression patterns of pax2a, eya1 and dlx3b are not altered (Hammond et al., 2003). Therefore, although excessive Hh signalling can perturb DV and ML otic identity, Hh does not appear to be required for correct DV and ML patterning of the zebrafish otic vesicle.

In addition to the DV and ML otic defects described above, ptc1^–/–; lep^–/– double mutants exhibited a striking posterior otic duplication concomitant with a loss of anterior otic structures. This was similar, but not identical, to the otic phenotype in embryos overexpressing shh or dnPKA mRNA, in which Hh signalling is upregulated throughout the embryo (Hammond et al., 2003). In ptc1^–/–; lep^–/– homozygotes the otic vesicle was small and round, with no cristae or semicircular canal projections (Fig. 1H; Fig. 4B,D). Two separate macular domains of hair cells developed at the anterior and posterior poles of the vesicle, as in wild-type embryos, but both domains were positioned medially, similar to the posterior domain in wild-type embryos (Fig. 4A-D). Later, these fused to form a single medial macula on the VM wall (Fig. 4F), overlaid by a single dumbbell-shaped otolith, which originated as two separate medial otoliths that later fused (Fig. 1H,I).

We confirmed that the single ptc1^–/–; lep^–/– macula had a double posterior character by examining sensory hair cell polarity patterns (Fig. 4G-J; see Fig. S3 in the supplementary material). Hair cell polarity patterns are stereotypical for each sensory patch, and can be mapped using anti-acetylated tubulin antibody to mark the kinocilia and FITC-phalloidin to mark the hair bundle (Haddon et al., 1999). We mapped polarity in four ptc1^–/–; lep^–/– maculae: hair bundles pointed away from a midline in both the anterior and posterior halves of the maculae with a central region of confused polarity (Fig. 4G,H; see Fig. S3 in the supplementary material). A pattern of hair bundle polarities pointing away from a midline is characteristic of the zebrafish posterior macula (Fig. 4J); this is never seen in the anterior macula (Fig. 4I) (Haddon et al., 1999). We therefore conclude that the ptc1^–/–; lep^–/– macula has double-posterior identity. Note, however, that the shape of the macula was almost triangular, differing from the double-posterior ‘bow tie’- or ‘butterfly’-shaped maculae in the ears of embryos injected with shh or dnPKA mRNA (Fig. 4) (Hammond et al., 2003).

We also examined expression of a panel of otic AP markers in ptc1^–/–; lep^–/– homozygotes (Fig. 5). Expression of hmx3, fgf8 and pax5 was reduced or absent from the anterior region of the otic vesicle (Fig. 5A-F). otx1 expression was lost or reduced to a small ventrolateral domain, consistent with the absence of tissue between the duplicated halves of the posterior macula (Fig. 5G-J) (Hammond and Whitfield, 2006). Curiously, however, fst1, a posterior otic marker, was not expressed in ptc1^–/–; lep^–/– otic vesicles (n=16), despite being upregulated around the otic vesicle (Fig. 5K,L). This differs from the phenotype in shh RNA-injected embryos, in which fst1 was upregulated at the anterior of 5/22 otic vesicles (Hammond et al., 2003), and from dre, lep, igu and uki (see below). The otic vesicles of ptc1^–/–; lep^–/– embryos were also significantly smaller than those of shh RNA-injected embryos by 4 dpf (Fig. 1H,I). These data, together with the triangular-shaped macula, suggest that the ptc1^–/–; lep^–/– otic phenotype is more severe than that of our shh- or dnPKA-injected embryos. To confirm this, we applied low doses of cyclopamine, a potent inhibitor of the Hh signalling pathway, to ptc1^–/–; lep^–/– homozygotes from 15S, in order to reduce Hh signalling activity slightly. Using 0.25 μM cyclopamine, the triangular ptc1^–/–; lep^–/– macula was rescued to a ‘bow tie’ shape in 3/8 cases (see Fig. S4 in the supplementary material).

In dre, lep, igu and uki, there were no morphologically obvious otic AP patterning defects, and expression of hmx5, pax5, fgf8 and otx1 in the ear was normal (data not shown). However, fst1 was ectopically expressed at the anterior of dre, lep, igu and uki otic vesicles. This ectopic expression was variable and was most extreme in uki and least severe in dre (Fig. 6). A small domain of fst1 expression was also seen at the anterior of some dre, uki, lep and igu siblings (presumed heterozygotes; for example, Fig. 6B). This was not seen in wild-type (TL) embryos (0/26). Note, however, that expression of fst1 was upregulated in other regions of the embryo in the Hh inhibitor mutants (data not shown). Taken together, these data indicate that Hh inhibitory activity is required for AP patterning of the otic vesicle as well as for DV and ML patterning.

To investigate whether levels of Hh pathway activity in the ear region correspond to the severity of the otic phenotype in the Hh-inhibitor mutants, we used ptc1 and ptc2 expression levels as a readout of Hh signalling (Concordet et al., 1996; Goodrich et al., 1996). ptc1 and ptc2 RNA levels in the embryo, including the ear region, generally correspond to the severity of otic phenotype. Expression of ptc1 and ptc2 is lost or severely reduced in smo and con mutants, in which Hh signalling is absent or severely downregulated, and is upregulated in shh RNA-injected embryos and ptc1^–/–; lep^–/– double mutants (Hammond et al., 2003) (see Fig. S5 in the supplementary material; data not shown). In our hands, ptc1 and ptc2 levels were not substantially raised in dre, lep and uki mutants, including in the otic vesicle region (data not shown); however, Koudijs et al. (Koudijs et al., 2005) report a slight increase in ptc1 levels in lep and uki mutants. This corresponds to the slightly more severe otic phenotypes in these mutants compared with dre.

However, in ptc1^–/– mutants – which have normal ears – ptc1 and ptc2 levels were raised throughout the embryo (see Fig. S5 in the supplementary material). Other tissues in the ptc1^–/– mutant, including the somites, are more severely affected than in dre, lep and uki (Koudijs et al., 2008). The fact that there was no ear phenotype in ptc1^–/– mutants suggests that although ptc1 has an important role in many tissues, it is less important than ptc2 in the otic vesicle. Increased ptc2 levels in ptc1^–/– mutants seem to be able to compensate for the absence of ptc1 function in the otic vesicle.

To avoid complications arising from transcriptional feedback acting on ptc1 and ptc2 levels, and to confirm that levels of Hh signalling correspond to severity of the otic phenotype in the Hh inhibitor mutants, we applied graded concentrations of the Hh inhibitor cyclopamine to ptc1^–/–; lep^–/– double mutants. This would reduce Hh signalling levels throughout the embryo downstream of the non-functional Ptc receptors, and gradually rescue otic development. We applied cyclopamine at 15-16S to ptc1^–/–; lep^–/– double-mutant embryos, and analysed the ear phenotype at 88-90 hpf (Table 2). Using 50 μM cyclopamine, we observed complete or near-complete rescue of the otic phenotype in 25/30 ptc1^–/–; lep^–/– embryos. At 40 μM, as well as igu- or uki-like ears (with no dls and abnormal semicircular canal projections), we saw both a dre-like ear phenotype (no dls) and a lep-like ear phenotype (dls present but abnormal, with abnormal canal projections) as well as individuals with a very thin dls and grossly normal canal projections. Between 5 μM and 20 μM cyclopamine, all ears resembled those of ptc1^+/–; lep^–/– mutants, with three spindly canal pillars; at lower concentrations, only the anterior and posterior canal pillars formed; lower still, only the posterior pillar formed. At the lowest concentrations, the vesicle was slightly larger than in ptc1^–/–; lep^–/– double mutants, and the macula shape was sometimes partially rescued (see above). Overall, the severity of otic phenotype appeared to correlate with the level of Hh pathway activity within the embryo.

To narrow the time interval during which inhibition of Hh signalling is required for otic patterning, we applied 50 μM cyclopamine to ptc1^–/–; lep^–/– embryos from time points between 15S and 48 hpf. We analysed ear and sensory patch morphology at 80 hpf using differential interference contrast (DIC) microscopy and phalloidin staining, respectively. When cyclopamine was applied from 17S, complete rescue of the otic phenotype occurred in 4/6 cases, suggesting that inhibition of Hh signalling is not required before this time (Table 3). Application after 48 hpf had no effect. Cyclopamine applied between these time points led to partial rescue (see Fig. S6 in the supplementary material). This contrasts with a requirement for Hh signalling before 15S for correct AP otic patterning (see Fig. S7 in the supplementary material).

To investigate whether the ptc1^–/–; lep^–/– mutant otic phenotype results from the direct action of a gain of Hh signalling activity in the otic epithelium, or is caused indirectly by defects in surrounding tissues, we transplanted ptc1^–/–; lep^–/– or control wild-type (AB) otic vesicles into wild-type (AB) hosts. Donor embryos were labelled with rhodamine- and biotin-dextran at the one- to two-cell stage. At 18S the host otic vesicle was, as far as possible, extirpated and replaced with a donor otic vesicle (Fig. 7A). As described above, treatment with cyclopamine from 18S can almost entirely rescue the otic phenotype of ptc1^–/–; lep^–/– embryos; if the effects of a gain of Hh activity are indirect, surrounding tissue should be able to rescue inner ear development from this stage. At 68-70 hpf ears were assayed for rescue of semicircular canal projection development: semicircular canal projection tissue never developed in control ptc1^–/–; lep^–/– embryos (Fig. 7).

Wild-type otic vesicles transplanted into wild-type hosts resulted in abnormal ears, but these contained reasonably well-developed semicircular canal projection tissue in all (6/6) cases (examples are shown in Fig. 7E,G). In two cases, both host and donor tissue was present in the otic vesicles, suggesting incomplete extirpation of the host otic rudiment; host and donor tissue fused, forming a single vesicle in both cases (Fig. 7E). When ptc1^–/–; lep^–/– tissue was transplanted into wild-type hosts, small otic vesicles formed (n=16). In contrast to the wild type-to-wild type transplants, if a substantial amount of wild-type otic tissue was present together with the transplanted mutant tissue (7/16 cases), two separate vesicles formed. In most cases, the ptc1^–/–; lep^–/– otic tissue formed a small canal-free otic vesicle (13/16) (Fig. 7F). However, in 3/16 cases, rudimentary semicircular canal projection tissue was formed (Fig. 7H). To investigate further, we labelled the biotin-dextran in the donor tissue using Cy3-tyramide and examined the ears using confocal microscopy. In all three cases, the canal projection tissue appeared to consist entirely of wild-type cells protruding through the ptc1^–/–; lep^–/– tissue (Fig. 7I,J). We were unable to assay for any rescue of sensory patch development in these abnormal ears. Nevertheless, the lack of semicircular canal projection development from ptc1^–/–; lep^–/– tissue suggests that the effect of a gain of Hh signalling on otic patterning is predominantly direct.

We have shown that the developing zebrafish ear is exquisitely sensitive to a loss of function of antagonists of the Hh pathway. Even a small increase in Hh pathway activity leads to DV and ML otic patterning defects: we see a spectrum of subtle DL defects in dre (su(fu)^–/–), lep (ptc2^–/–), igu (dzip1^–/–) and uki (hip^–/–) embryos, in which Hh signalling is only slightly raised (Koudijs et al., 2005; Wolff et al., 2004). When Hh pathway activity is maximally upregulated, either in ptc1^–/–; lep^–/– double mutants or when shh RNA is overexpressed, the otic vesicle is severely medialised and ventralised. By contrast, the effect of Hh signalling on otic AP patterning is only morphologically obvious at the extremes of Hh pathway activity, in ptc1^–/–; lep^–/– double mutants or cyclopamine-treated embryos (this work) or in shh- or dnPKA-injected embryos or severe Hh loss of function mutants (Hammond et al., 2003).

Evidence from amniotes suggests that, here too, tight regulation of Hh pathway activity is required for correct inner ear development: strictly regulated expression of Gli activator and repressor forms is required for the formation of auditory (ventral) and vestibular (dorsal) areas of the mouse inner ear, respectively, and a precise balance between Shh and Wnt signalling is required for cochlea development (Bok et al., 2007b; Riccomagno et al., 2005). Hh pathway inhibitors could provide the tight regulation of Hh activity required. In mouse otic vesicles, ptc1 is expressed in a graded fashion, highest ventromedially (Bok et al., 2007b); the expression of dzip1, su(fu) and hip has not yet been described in amniote ears.

Consistent with a graded response to levels of Hh pathway activity throughout the embryo, the severity of the otic defects in the Hh inhibitor mutants form a phenotypic series, corresponding roughly to the severity of the general embryonic phenotype (Koudijs et al., 2008; Koudijs et al., 2005). There are, however, several idiosyncrasies, suggesting that each inhibitor may have a slightly different role. This was not unexpected, as the expression patterns of ptc1, ptc2, dzip1, su(fu) and hip in and around the developing ear differ from one another (Hammond et al., 2003; Hammond and Whitfield, 2009).

Firstly, the dre (su(fu)^–/–) and lep (ptc2^–/–) otic phenotypes do not fit together into a smooth series of phenotypic severity: dre has no dls but normal semicircular canal pillars, whereas the dls in lep is present but malformed, and there are additional canal pillar abnormalities. This could result from expression pattern differences: su(fu) is expressed ubiquitously, whereas ptc2 is expressed in a broad ventral otic domain (Hammond et al., 2003; Hammond and Whitfield, 2009). Interestingly, however, both the dre and lep phenotypes, as well as intermediate phenotypes, can be produced by treating ptc1^–/–; lep^–/– embryos with 40 μM cyclopamine. This may reflect slight spatial and temporal differences in the uptake and effect of cyclopamine between individual embryos.

Secondly, although several indicators (ptc1 and ptc2 expression levels and somite phenotype) suggest that the overall phenotype of ptc1^–/– mutants is more severe than that of dre, lep, igu and uki mutants, ptc1^–/– embryos have normal ears, whereas dre, lep, igu and uki have dorsolateral otic patterning defects. This suggests that although Ptc1 has an important role elsewhere in the embryo, Ptc2 is likely to be the primary Hh receptor in the zebrafish ear. Feedback control on the transcription of ptc genes means that ptc2 levels are raised in ptc1^–/– mutants, and this appears to be able to compensate for the lack of ptc1 function in the ear. The reverse, however, is not true: ptc1 is only minimally upregulated in lep^–/– mutants, and is unable to compensate entirely for the absence of ptc2 function in the ear.

A role for ptc1 in inner-ear patterning is, however, revealed in combination with lep: the otic phenotype of ptc1^+/–; lep^–/– fish is more severe than the lep phenotype, and the ptc1^–/–; lep^–/– phenotype is yet more severe. Similar redundancy between Ptc1 and Ptc2 (which are closely related proteins) has been reported in the development of other organs: for example, the somites (Koudijs et al., 2008).

Our transplant experiments suggest that surrounding wild-type tissue cannot effect a rescue of semicircular canal projection development in mutant ptc1^–/–; lep^–/– otic tissue. This suggests that a loss of Hh inhibition acts directly on the inner ear, as the transplants were performed at 18-19S, from when near-complete cyclopamine-mediated rescue of the ptc1^–/–; lep^–/– otic phenotype is possible (Fig. 7; Table 3). A direct effect of increased Hh signalling in the ear is feasible, as all components of the Hh signalling pathway so far examined, apart from Hh itself, are expressed in the developing ear, and both ptc1 and hip expression levels in the otic epithelium are responsive to Hh pathway activity (Hammond et al., 2003; Hammond and Whitfield, 2009). It is unclear, however, when and to what extent donor tissue integrates sufficiently to receive otic patterning signals from the host. Indeed, when ptc1^–/–; lep^–/– and wild-type otic tissue are both present, small separate otic vesicles form (Fig. 7F).

Previously, we and others have reported that Hh signalling in zebrafish and Xenopus is required for otic AP patterning (Hammond et al., 2003; Waldman et al., 2007). In mouse and chick, however, Hh predominantly affects otic DV patterning (Bok et al., 2005; Liu et al., 2002; Riccomagno et al., 2002; Riccomagno et al., 2005). This was puzzling, as the source of Hh in all these species appears to be the ventral midline tissues (Bok et al., 2005; Hammond et al., 2003), and the relative positioning of the otic vesicle to the midline sources of Hh is similar in both zebrafish and amniotes. Our data now suggest, however, that differences in the role of Hh signalling between amniotes and fish are less extreme than these studies indicated. In zebrafish, as in amniotes (Bok et al., 2007b), derepression of Hh signalling affects specification of dorsal otic domains. There thus appears to be a very similar requirement to keep Hh signalling repressed for the correct patterning of dorsal otic epithelium in both zebrafish and amniotes.

In ventral regions of the ear, there do seem at first sight to be substantial differences in the requirement for Hh signalling between zebrafish and amniotes. When sensory structures are examined in ventral parts of the ear, Hh signalling affects the fish otic AP axis but the amniote DV axis. In many ways, however, patterning of sensory structures along the AP axis of the zebrafish ear is equivalent to patterning along the DV axis of the amniote ear, both functionally and structurally. Auditory reception in zebrafish is performed by posterior otic structures, the saccular and lagenar maculae, whereas the major auditory endorgan in amniotes is the ventrally positioned cochlea. Like the saccule and lagena in the fish ear, the amniote cochlear duct and chick basilar papilla arise from a posteroventral region of the otic vesicle (Bok et al., 2007a; Oh et al., 1996; Riccomagno et al., 2002). In the mouse, although all sensory precursors arise from an anteroventral domain in the otocyst, the relative positions of the sensory chambers in the adult ear (utricule-saccule-cochlea) are equivalent to the AP arrangement (utricule-saccule-lagena) in zebrafish ears (Fig. 8). We conclude that both the requirement for Hh signalling in ventral regions of the otic vesicle, and the need to keep Hh signalling repressed for correct patterning of dorsolateral otic structures, have features that are conserved between zebrafish and amniote vertebrates.

We thank Laina Murphy and Joanne Spencer for technical assistance, Lisa van Hateren, Claire Allen and the aquarium staff for expert zebrafish care, and many members of the zebrafish community for mutants and probes. This work was funded by grants from the BBSRC (BB/E015875/1) to T.T.W. and the Wellcome Trust (C23207/A8066) to F.v.E. The MRC CDBG zebrafish aquaria and imaging facilities were supported by the MRC (G0400100, G0700091), with additional support from the EU FP6 (ZF-MODELS) and the Wellcome Trust (GR077544AIA). Deposited in PMC for release after 6 months.