Mutations affecting development of the zebrafish ear

The vertebrate inner ear has two major functions: it is responsible for the maintenance of body balance and for auditory perception. The morphological form of the inner ear is relatively similar in all groups of jawed vertebrates (Retzius, 1881, 1884). It consists of three chambers of varying prominence, the utriculus, sacculus and lagena and three semicircular canals. The semicircular canals both begin and end in the utriculus and each is located in a separate plane, at right angles to the other two. A spherical widening, known as the ampulla, is located at one end of each semicircular canal. The walls of the inner ear are lined with epithelium which contains specialized sensory areas. These regions contain a mechanoreceptive cell type known as the hair cell. Three large and five smaller sensory areas are present in the ear of the adult zebrafish (Platt, 1993). A large sensory area, called the macula is localized in each of the ear chambers in bony fishes, forming the support for a largely inorganic aggregation of calcium salts, termed the otolith (Lim, 1980). Three smaller sensory areas (cristae) are located in the ampulae of the semicircular canals.

The sensory epithelia of the inner ear chambers respond to movements of otoliths providing information about tilt and linear acceleration, while the cristae of the semicircular canals provide information about turning movements (Hawkins and Horner, 1981). In addition to serving the role of a balance organ, the maculae of the cyprinid fishes are most likely also involved in hearing (Popper and Fay, 1993). In zebrafish and other cyprinids, the sound induced vibrations of the swim bladder are thought to be conveyed to the ear chambers by a chain of specialized bones, named Weberian ossicles (Chranilov, 1927; reviewed by Popper and Fay, 1993). In higher vertebrates the function of hearing is fulfilled by the cochlea, a structure homologous to the lagena.

Development of a structure as complex as the vertebrate inner ear undoubtedly requires an intricate network of genetic interactions. Several genes characterized at the molecular level are known to be involved in the development of the mammalian ear (reviewed by Steel and Brown, 1994). Genetic lesions in Pax-3, Int-2 and Hoxa-1 have been shown to produce inner ear defects (Epstein et al., 1991; Chisaka et al., 1992; Mansour et al., 1993). A large group of loci involved in the development of the mouse inner ear were isolated on the basis of behavioral abnormalities. Mutations in these loci produce a behavioral disorder known as the waltzing syndrome which involves hyperactivity, tendency to run in circles and jerking movements of the head. The ear defects of these mutants frequently involve malformations of the inner ear chambers and the semicircular canals (Deol, 1968, 1980). The kreisler (kr) phenotype is one of the most severe in this group and in extreme cases leads to the complete absence of any recognizable inner ear elements (Deol, 1964). Positional cloning of the kreisler locus has recently revealed that it encodes a transcription factor expressed in rhombomeres 5 and 6 of the hindbrain (Cordes and Barsh, 1994).

The development of the zebrafish ear has been described in detail (Waterman and Bell, 1984; Haddon and Lewis, 1996). The first morphological manifestation of the ear is a condensation of ectodermal cells known as the otic placode. The otic placodes first become visible around the 14 somite stage and are located on both sides of the hindbrain approximately midway between the eye and the first somite. Cells in the initially solid mass of the otic placode rearrange, forming the lumen of the otic vesicle. Two otoliths appear in the otic vesicle at approximately 19 hpf (hours postfertilization). Between 19 and 44 hpf the otic vesicle increases in size and the otoliths become larger (Fig. 1A). A series of morphological changes which lead to the formation of semicircular canals is initiated at 44 hpf with the appearance of finger-like protrusions growing into the cavity of the otic vesicle (Fig. 1B; Waterman and Bell, 1984). Protrusions growing from the opposite walls of the otocyst fuse together, forming three columns of tissue which delimit the space of individual semicircular canals. The anterior semicircular canal is the first to be formed and the lateral is the last. The process of semicircular canal formation is completed by 64 hpf (Waterman and Bell, 1984). The subdivision of the zebrafish ear into utriculus, sacculus and lagena is a relatively late event. The three chambers are not distinct at 5 days postfertilization (dpf) (Fig. 1C).

In a large mutagenesis screen in zebrafish we searched for recessive, zygotic loci involved in ear morphogenesis during the first 5 days of development. We identified mutations affecting several aspects of development: specification of the otic placode; growth of the otic vesicle; otolith formation; morphogenesis of the semicircular canals and late expansion of the ear structure. Here, we report initial genetic and phenotypic characterization of 20 mutations affecting early ear development.

Mutations were induced using N-ethyl-N-nitrosourea (ENU) as described previously (Solnica-Krezel et al., 1994). All mutants were initially identified in the progeny of crosses between individuals of the F₂ generation bred from the mutagenized G₀ males (Driever et al., 1996). All crosses that led to the identification of mutant phenotypes were repeated. For further analysis, heterozygous carriers of mutant alleles were outcrossed to wild-type AB (Chakrabarti et al., 1983), Tübingen (Mullins et al., 1994) or India (Knapik et al., 1996) strains. Mutant heterozygotes were identified in the progeny of outcrosses by sibling mating.

Complementation tests were performed between heterozygous carriers of mutations falling into the same phenotypic category. At least 30 embryos, often from two independent crosses were scored for the mutant phenotype. Complementation testing was not performed for m219.

Embryos were maintained at 28°C in egg water (Westerfield, 1994) with the addition of methylene blue (Sigma, Inc.) at a concentration of 1 mg/l. Staging was performed as described previously (Kimmel et al., 1995). For observations, embryos were anesthetized with a 0.02% solution of 3-aminobenzoic acid methyl ester (Sigma, Inc.) and embedded in 3% methylcellulose in egg water. Living embryos were initially observed and photographed under a dissecting microscope. Detailed observations were conducted using an Axiophot microscope and Nomarski optics (Zeiss, Inc.).

In situ hybridization was performed with digoxigenin-labeled probes on whole embryos as described previously (Oxtoby and Jowett, 1993). Expression patterns were photographed in embryos mounted in 3% methyl cellulose.

As a part of a large scale mutagenesis screen we searched for mutations affecting development of the zebrafish ear (Driever et al., 1996). The ear of the zebrafish embryo is relatively large and easily accessible to visual inspection (Fig. 1). Fig. 1D shows a schematic representation of morphological elements which are easily recognizable in the zebrafish ear at 5 dpf (compare with the photograph in 1C). The shape of semicircular canals is easy to follow at this stage in both lateral and dorsal views. In the course of screening, several aspects of ear development, including size and shape of the otic vesicle, presence and positioning of otoliths, morphology of the semicircular canals and the otic capsule were evaluated at 1, 2, 3 and 5 dpf.

Some aspects of ear development were not accessible to inspection during our screen. For example, we were unable to monitor abnormalities in the positioning of cristae. Likewise, we did not screen for defects in developmental events taking place after 5 dpf, such as formation of the lagena and its otolith.

Not all phenotypes affecting the ear were subject to further analysis. Numerous mutations resulting in a general delay of development frequently led to a somewhat reduced ear size at day 2 and did not form semicircular canals. Mutations affecting formation of cartilage also frequently involved abnormal mor-phogenesis of semicircular canals (Neuhauss et al., 1996).

Phenotypic characterization of the isolated mutations allowed us to divide them into five categories, each affecting different aspects of ear development: (1) abnormal specification of the otic placode; (2) defective early morphogenesis of the otic vesicle; (3) absence of otoliths; (4) aberrant late morphogenesis of the otic vesicle; (5) defective late expansion of the ear structure. Below we present a phenotypic analysis of mutants belonging to each of the above categories.

The mutation quadro (quo)^m271 affects ear development at the earliest stages of otic placode formation. The otic placodes of quo^m271 are frequently smaller or split into two fields. The phenotype of this mutant is variable and asymmetrical. In some individuals the otic vesicle is reduced in size, in others two incomplete otocysts are formed. Otoliths frequently develop in each of the incomplete otocysts (Fig. 2A,B). The epithelial protrusions associated with formation of semicircular canals initially form but do not reach their mature shape (arrowheads in Fig. 2B). The ear phenotype of quo^m271 is not fully penetrant. In a cross between two heterozygous animals approximately 15% (29/190) of progeny develop ear abnormalities at 1 dpf.

Specification of the ear rudiment is detectable by expression of the dlx-3 gene before the morphological appearance of the otic placode (Ekker et al., 1992). To investigate whether quo^m271 affects specification of the otic placode, we analyzed dlx-3 expression by in situ hybridization at the 10 somite stage (14 hpf). At this stage dlx-3 is expressed in two ellipsoid patches of cells located symmetrically one on each side of the wild-type hindbrain (arrowhead in Fig. 2C). The dlx-3 expression domain is split into two parts in a fraction of embryos originating from a cross between two quo^m271 heterozygotes (arrowheads in Fig. 2D). Such expression pattern is never seen in the progeny of wild-type individuals and indicates that the ear patterning defect in quo^m271 mutants is present at the onset of the otic placode formation.

The ear defect of quo^m271 mutants is also evident in the expression pattern of msx-C. In the wild type at 60 hpf msx-C is expressed in three distinct patches in the ventral portion of the otic vesicle (Ekker et al., 1992; arrows in Fig. 2E). In quo^m271 mutants, the msx-C expression domains are frequently fused or missing (Fig. 2F). In order to assess whether the ear phenotype of quo^m271 is associated with a hindbrain patterning defect, we performed in situ hybridization with the krox-20 probe on 10 somite stage embryos. In wild-type embryos the krox-20 gene is expressed in two distinct domains which correspond to rhombomeres 3 and 5 (Oxtoby and Jowett, 1993). We did not detect any obvious differences between the expression pattern of krox-20 in wild-type and quo^m271animals at this stage (not shown).

Seven mutations in four loci were found to affect early development of the otocyst (Table 1, Fig. 3). The abnormal shape of the otic vesicle is detectable in these mutants by 24 hpf. The wild-type otic vesicle is ellipsoid at 30 hpf and its long axis is parallel to the anteroposterior body axis (Fig. 1A). In contrast, the otic vesicles of mind bomb (mib)^m178, pandora (pan)^m313, and flachland (fll)^m517 are frequently smaller and more round (Fig. 3A,C,D). The otocyst of snakehead (snk)^m273 is narrower in its dorsoventral dimension than its wild-type counterpart (Fig. 3B). All four loci affect development of the otoliths. The otoliths of snk^m273 are barely visible. In mib^m178 and pan^m313 the otoliths are often missing or only one otolith can be found. Frequently only one otolith is present in fll^m517.

Defects of the semicircular canals are variable among mutants in this group. This aspect of ear development is most severely affected in pan^m313 and snk^m273 which rarely initiate formation of the semicircular canals. The otic vesicle of snk^m273 appears abnormally large relative to other organs at 2 dpf (Fig. 3F). In mib^m178 protrusions form in the otic vesicles but these do not develop the normal pattern of columns (arrowhead in Fig. 3E). Their presence suggests that the reduced ear size in this mutant is not simply due to a developmental arrest. The semicircular canals are relatively unaffected in fll^m517 and develop normally in at least some mutant animals (not shown). In order to investigate by additional means whether the patterning of the otic vesicle in fll^m517 is normal, we performed in situ hybridization with the msx-C probe on the mutant embryos at 60 hpf. The msx-C expression pattern in fll^m517 is indistinguishable from wild-type pattern at this stage (not shown).

The brain morphology of mutants in this category is abnormal. The midbrain ventricle appears to be abnormally large in mib^m178. The mib locus is associated with a particularly intriguing phenotype; supernumerary primary neurons are generated in the mib mutants (Schier et al., 1996). The remaining mutations result in a reduction in the size of the brain ventricles, which is most pronounced in snakehead (Schier et al., 1996).

Six mutations isolated in our screen result in a lack of otolith formation (Table 1; Fig. 4). In empty ear (emp)^m107, straight edge (sed)^m327, haos (hos)^m751 and m219 the gross morphology of the otic vesicles and later of the semicircular canals is normal (Fig. 4A,E). In contrast, helter skelter(hek)^m504 and znikam(znk)^m574 produce changes in ear morphology. At 36 hpf the otocyst of hek^m504 animals is larger than its wild-type counterpart (Fig. 4C compare with B). In some hek^m504 mutant animals the positioning of the protrusions forming the semicircular canals appears to be slightly abnormal at 2 dpf (not shown). The abnormalities of the semicircular canals are no longer apparent at 5 dpf. The morphological defect in znk^m574 becomes most obvious around 72 hpf and has varying prominence. In the most affected mutant animals the otic capsule and the semicircular canals are barely visible at this stage.

In the least affected individuals, the ear morphology is fairly normal. Fig. 4D shows a moderately severe form of this phenotype at 3 dpf. The overall size of the ear is reduced and the columns delimiting the semicircular canals are not as distinct as in the wild type. Interestingly, these morphological abnormalities are no longer obvious by 5 dpf.

The absence of otoliths is not fully penetrant in sed^m327, hos^m751and znk^m574. In some sed^m327 and znk^m574 mutant animals a single, small otolith is present in one of the normal positions. In some outcrosses, hos^m751 mutants develop supernumerary, mispositioned otoliths (arrowheads in Fig. 4E). At 60 hpf the expression pattern of msx-C in znk^m574 appears to be somewhat abnormal. The two anterior expression domains are positioned closer to each other than in the wild type (not shown). In emp^m107, hek^m504 and hos^m751 the msx-C expression closely resembles the wild-type pattern (not shown).

Two loci affect otic vesicle morphogenesis at later stages of development than the group II mutations (Table 1; Fig. 5). The little richard (lit) phenotype is first distinguishable shortly before the onset of semicircular canal formation. Of the two alleles recovered for this locus, lit^m433 produces a stronger phenotype (Fig. 5B, compare with A). During late day 2 of development, growth of the otic vesicle in lit^m433 becomes slower and the formation of semicircular canals is not initiated. At 3 dpf the ear of these mutants is very small and undifferentiated in comparison to the wild type (Fig. 5B,E - compare with A,D). Two small otoliths are positioned next to each other. The reduced size of the otic vesicle is accompanied by an overall shortening of the posterior head region. The distance between the pectoral fin (arrowheads in Fig. 5A,B) and the eye is reduced in the mutant. The msx-C expression at 60 hpf is absent (Fig. 5H, compare with G). Lack of msx-C expression and absence of epithelial protrusions suggest that late patterning events are not initiated in the otic vesicle of lit^m433 embryos.

The golas (gos) mutant embryos combine defective ear morphology and abnormal pattern of pigment cells. Two alleles are available at this locus: gos^m241 and gos^m618. gos^m618 is stronger and its phenotype is first detectable around 36 hpf. The epithelial protrusions start developing in these mutants but have abnormal shape and fail to form mature columns. At 3 dpf the otic capsule of gos^m241 is slightly smaller than that of the wild type (Fig. 5A,C). The ear reduction is much stronger in gos^m618 (Fig. 5F, compare with the wild type in D). In the most extreme cases the otic capsule surrounding two small otoliths is barely visible (not shown). Melanocytes are not observed in gos^m618 and are restricted to the dorsal midline in gos^m241.

The two alleles isolated at the antytalent (ant) locus affect expansion of the ear (Table 1; Fig. 6). ant^m426 appears to produce a somewhat stronger phenotype than ant^m317. Both alleles cause a slight reduction of ear size during day 4 of development. The semicircular canals and otoliths appear to have normal positioning and morphology at this stage. At 5 dpf the otic capsule and the semicircular canals of the mutant embryos are reduced and not easily recognizable in the lateral view (Fig. 6A; compare with the wild type in Fig. 1C). Also in the dorsal view, the otic capsule is flattened as compared to the wild type (Fig. 6C, compare with B). Two normal size otoliths persist at least until 5 dpf.

A number of ear mutations isolated in our screen have not been analyzed here. These mutations fall into four subgroups. 13 mutations result in the presence of only one otolith (m96, m150, m198, m234, m269, m392, m490, m539, m548, m554, m675, m690, m720). Limited complementation analysis within this category revealed that at least five of them: m150, m198, m234, m269 and m392 affect the same locus. The second subgroup includes mutations resulting in a reduction of otolith size (m251, m389, m428, m431). A group of 4 mutations (m286, m288, m293, m300) affect otolith size as well as eye size and pigmentation (Malicki et al., 1996). The fourth group of 17 mutations lead to abnormal growth and differentiation of cartilage (Neuhauss et al., 1996). These mutations also affect development of epithelial protrusions in the otocyst and consequently formation of the semicircular canals.

Here we report identification of 20 mutations in 13 loci affecting several consecutive stages of ear development. The earliest phenotype, produced by a mutation in the locus quadro affects the very first step of ear morphogenesis – formation of the otic placode. In contrast, the latest onset phenotype – lack of ear expansion in the antytalent mutants – becomes apparent only after formation of the semicircular canals. The phenotypes of other mutants become noticeable at the intermediate stages and affect shaping of the otocyst, otolith formation and morphogenesis of the semicircular canals. The phenotypes of quadro and antytalent mutants appear to be opposites in yet another way: antytalent mutations affect growth of the ear but not its pattern, while quo^m271 results in abnormal patterning of a normally sized ear. In many other mutants the ear patterning defects are accompanied by a reduction of ear size.

Early embryological experiments suggested that pattern formation in the developing ear depends on continued interaction of the otic vesicle with the hindbrain. In amphibian embryos, surgical replacement of the hindbrain with midbrain or spinal cord tissues results in the abnormal development of the otocyst which frequently does not form semicircular canals (Detwiler and Van Dyke, 1950, 1951). A close association of otic vesicle morphogenesis with brain development is also suggested by mouse mutants. In kreisler embryos the morphological appearance of rhombomeres is abnormal and the expression patterns of Hoxb-1; Krox-20 and Fgf-3 do not obey their wild-type expression boundaries in the hindbrain (Frohman et al., 1993). The developing otic vesicle is shifted to a more lateral location, the inner ear chambers, semicircular canals and the endolymphatic duct do not develop properly (Deol, 1964). The kreisler gene is expressed in the hindbrain but not in the otic vesicle suggesting that ear abnormalities in this mutant are secondary to the neural tube defect (Cordes and Barsh, 1994). Likewise, in Hoxa-1 knockout mice abnormal hindbrain morphology is associated with malformations of the inner ear compartments and the endolymphatic duct (Chisaka et al., 1992).

Several mutations isolated in our screen involve both ear and brain defects. This association may reflect disruption of proper signaling events between the brain and the developing ear. Alternatively, the affected loci may function independently in the neural tube and the otic vesicle. Brain patterning was analyzed using molecular markers in two of the Group II mutants (Table 1) (Schier et al., 1996). The expression patterns of krox-20 and pax-2 in fll^m517 and rtk-1 in snk^m273 do not display any obvious abnormalities.

The association of a strong ear patterning defect with a loss of pigmentation in golas is intriguing. Apart from these two abnormalities the golas mutants closely resemble the wild type. The pigmentation phenotype suggests a neural crest defect. One possible explanation for the association of these two phenotypes could be that mutations in the golas locus result in the absence of the neural crest-derived cells necessary for normal ear morphogenesis. Fate mapping experiments conducted in quail-chick chimeras indicate that neural crest-derived cells contribute to the formation of the otic capsule (Le Lievre, 1978; Noden, 1983). Their role in inner ear morphogenesis is, however, not clear. Neural crest ablation experiments conducted in medaka (Oryzias latipes) did not result in any obvious abnormalities of ear development (Langille and Hall, 1988). Mouse pigmentation mutants frequently display hearing abnormalities but do not have defects in early ear morphogenesis (Deol, 1970, 1980). In the mouse, the neural crest-derived melanoblasts and melanocytes are found in the vicinity of the developing ear. Their absence in the mutant piebald lethal (s^l) is not associated with gross defects of ear morphogenesis (Pavan and Tilghman, 1994).

The expression patterns of Pax-2 and dlx-3 suggest that the otic placode of the zebrafish embryo is specified by 11-12 hours of development (Krauss et al., 1991; Ekker et al., 1992). The dlx-3 expression in quadro^m271 indicates that the ear defect in this mutant is already present by 14 hpf. It is interesting that quadro^m271 also affects development of the branchial arches. In the cephalic region of the zebrafish neural keel, neural crest migration is initiated at 15-16 hpf (Schilling and Kimmel, 1994). The crest cells destined to populate the mandibular, hyoid and the 1st gill arch migrate in the vicinity of the otic placode (Schilling and Kimmel, 1994). Inappropriate positional information in the otic placode and possibly the surrounding tissues may result in misguided migration of neural crest cells and consequently in defective morphogenesis of the branchial arches. In another scenario, defects of the otic placode and misdirected migration of the neural crest could be a consequence of inappropriate hindbrain patterning. Expression patterns of krox-20, pax-2 and hlx-1 in quo^m271 mutants suggest however that the overall brain pattern in this mutant is normal (this work and Neuhauss et al., 1996).

Lack of proper otolith formation in group III mutants (Table 1) may have diverse causes. One possibility is absence or abnormal development of the sensory epithelium which participates in otolith deposition. Another possible cause could be a biochemical defect preventing the uptake or deposition of organic or inorganic materials necessary for otolith formation. Lack of properly formed otoliths has been observed in several mouse mutants, for example: pallid (pa), muted (mu), tilted head (thd), and lethal milk (lm) (Green, 1990). Some of them have been characterized in detail. The pallid and muted phenotypes resemble a group of human diseases known as platelet storage pool deficiencies (Novak et al., 1984). In addition to the otolith phenotype, mutations in these loci produce changes in the subcellular structure of platelets, melanocytes and kidney cells. pallid has been recently shown to affect the structure and expression of the gene encoding protein 4.2, a major component of the erythrocyte membrane (White et al., 1992).

Several limitations apply to our genetic analysis of ear development. Loci involved in ear morphogenesis may not have been identified as such due to their pleiotropic function during early embryonic development. This consideration is particularly relevant to relatively late developmental events, for example, morphogenesis of semicircular canals. The function of other loci may not have been revealed due to genetic redundancy. The design of our screen precluded isolation of fully penetrant dominant phenotypes involving lethality or maternal effect loci. Only one allele was recovered for more than half of the complemented loci (Table 1) indicating that many genes involved in the ear development remain unidentified.

Several features of the zebrafish embryo will facilitate effective phenotypic and genetic analysis of identified mutants. A number of sophisticated embryological manipulations are possible in zebrafish. Mosaic animals are easily generated by blastomere transplantations (Ho and Kane, 1990). Individual cells can be labeled and their fate followed in living embryos (Kimmel and Warga, 1986; Schilling and Kimmel, 1994) Heterotopic and heterochronic transplantations are possible at various stages of development (Eisen, 1991; Ho and Kimmel, 1993). These experimental tools will enable a detailed analysis of the mutant phenotypes identified in our screen. One would like to know for example whether the abnormal ear size and shape in golas and several other mutants is an intrinsic property of the otic vesicle or results from abnormal interactions with the surrounding tissues. This question can be efficiently addressed by otic placode transplantations. Similarly, it would be interesting to trace the pattern of neural crest migration in the mutant quadro^m271. Molecular characterization of the affected genetic loci is an obvious step in the future analysis of the identified mutations. Identification of candidate genes and if this is not possible, positional cloning will lead towards this goal. The first maps of the zebrafish genome are already available (Postlethwait et al., 1994; Knapik et al., 1996) and raise hopes for quick progress in this area.

We thank Colleen Boggs, Jane Belak, Ioannis Batjakas, Heather Goldsboro, Lisa Anderson, Snorri Gunnarson and Pamela Cohen for technical help during the various stages of the screen. We are also grateful to Julian Lewis, Tanya Whitfield and Catherine Haddon for sharing unpublished results and Eliza Mountcastle-Shah for critical reading of the manuscript. This work was supported in part by NIH RO1-HD29761 and a sponsored research agreement with Bristol Myers-Squibb (to W. D.). Further support in form of fellowships came from HFSP and the Fullbright Program (to Z. R.), EMBO and Swiss National Fond (to A. S.), Helen Hay Whitney Foundation (to D. L. S. and D. Y. S.), and the Damon Runyon - Walter Winchell Cancer Research Fund (to J. M.).

Mutations in the genes colourless (Kelsh et al., 1996) and golas (this paper) fail to complement each other and thus are to be considered allelic. Future references to this gene should use the name colourless.