An expanded domain of fgf3 expression in the hindbrain of zebrafish valentino mutants results in mis-patterning of the otic vesicle

Development of the inner ear requires interactions with adjacent hindbrain tissue. Many studies have shown that the hindbrain can induce otic placodes in adjacent ectoderm (Stone,1931; Yntema,1933; Harrison,1935; Waddington,1937; Jacobsen, 1963;Gallagher et al., 1996;Woo and Fraser, 1998;Groves and Bronner-Fraser,2000). Several of the relevant hindbrain signals have recently been identified (reviewed by Whitfield et al., 2002). In zebrafish, two members of the FGF family of signaling molecules, Fgf3 and Fgf8, are expressed in the anlagen of rhombomere 4 (r4) during late gastrulation, when induction of the otic placode begins(Reifers et al., 1998;Phillips et al., 2001;Maroon et al., 2002). At this time, pax8 is induced in the adjacent otic anlagen. Disruption of both fgf3 and fgf8 prevents induction of the otic placode,and conditions that expand the expression domains of these genes lead to production of supernumerary or ectopic otic vesicles(Phillips et al., 2001;Raible and Brand, 2001;Vendrell et al., 2001;Maroon et al., 2002). In addition, disruption or depletion of Fgf3 perturbs inner ear development in chick and mouse (Represa et al., 1991; Mansour et al.,1993), and misexpression of Fgf3 in chick is sufficient to induce ectopic otic vesicles (Vendrell et al., 2000). It has also been shown that chick Fgf19, which is expressed in a pattern similar to that of Fgf3(Mahmood et al., 1995),cooperates with the hindbrain factor Wnt8c to induce a range of otic placode markers in tissue culture (Ladher et al.,2000). Thus, multiple hindbrain factors are involved in otic placode induction, and FGF signaling plays an especially prominent role.

Much less is known about the role played by hindbrain signals in later stages of inner ear development. Experiments in chick embryos show that rotation of the early otic vesicle about the anteroposterior axis reorients gene expression patterns in a manner suggesting that proximity to the hindbrain influences differentiation of cells within the otic vesicle(Wu et al., 1998;Hutson et al., 1999). In zebrafish, Xenopus, chick and mouse embryos, Fgf3 continues to be expressed in the hindbrain after otic placode induction(Mahmood et al., 1995;Mahmood et al., 1996;McKay et al., 1996;Lombardo et al., 1998;Phillips et al., 2001). This raises the question of whether this factor also helps regulate subsequent development of the otic placode or otic vesicle.

Analysis of the valentino (val) mutant in zebrafish provides indirect evidence that hindbrain signals are necessary for normal development of the otic vesicle (Moens et al., 1996; Moens et al.,1998). val encodes a bZip transcription factor that is normally expressed in r5 and r6. val/val mutants produce an abnormal hindbrain in which the r5/6 anlagen fails to differentiate properly and gives rise to a single abnormal segment, rX, which shows confused segmental identity. Although the val gene is not expressed in the inner ear,val/val mutants produce otic vesicles that are small and malformed. As otic induction appears to occur normally in val/val mutants(Mendonsa and Riley, 1999), we infer that altered hindbrain patterning perturbs signals required for later aspects of otic development. Mice homozygous for a mutation in the ortholologous gene, kreisler (Mafb — Mouse Genome Informatics),also show later defects in development of the otic vesicle(Deol, 1964;Cordes and Barsh, 1994). The inner ear defects in kreisler mutants are thought to result from insufficient expression of Fgf3 in the hindbrain(McKay et al., 1996). In contrast to zebrafish, mouse Fgf3 is initially expressed at moderate levels in the hindbrain from r1 through r6. As development proceeds,expression downregulates in the anterior hindbrain but upregulates in r4(Mahmood et al., 1996). After formation of the otic placodes, Fgf3 expression also upregulates in r5 and r6. This upregulation fails to occur in kreisler mutants, possibly accounting for subsequent patterning defects in the inner ear(McKay et al., 1996).

To examine the relationship between hindbrain and otic vesicle development in zebrafish, we have examined patterning of these tissues in wild-type andval/val mutant embryos. We find that val/val mutants produce excess and ectopic hair cells at virtually any position in the epithelium juxtaposed to the hindbrain. Expression of the anterior otic markersnkx5.1 (hmx3 — Zebrafish Information Network) andpax5 is also seen ectopically throughout this region of the otic vesicle. Conversely, expression of the posterior marker zp23(pou23 — Zebrafish Information Network) is ablated inval/val embryos. Analysis of hindbrain patterning shows thatfgf3 is misexpressed in the rX region of val/val mutants. Disruption of fgf3 function by injection of an antisense morpholino oligomer blocks formation of ectopic hair cells and suppresses AP patterning defects in the otic vesicle of val/val mutants. By contrast,fgf8 is expressed normally in val/val embryos, and loss offgf8 does not suppress the inner ear defects caused by theval mutation. These data indicate that the expanded domain offgf3 plays a crucial role in the etiology of inner ear defects inval/val mutants and suggest that Fgf3 secreted by r4 normally specifies anterior fates, suppresses posterior fates and stimulates hair cell formation in the anterior of the otic vesicle.

Wild-type zebrafish embryos were derived from the AB line (Eugene, OR). Mutations used in this study were valentino(val^b337) and acerebellar(ace^ti282a). Both of mutations were induced with ENU and are thought to be functional null alleles(Moens et al., 1996;Moens et al., 1998; Brand et al., 1998). Embryos were developed at 28.5°C in water containing 0.008%Instant Ocean salts. Embryonic ages are expressed as hours post-fertilization(h).

Live val/val homozygotes were reliably identified after 19 h by the small size and round shape of the otic vesicle. In addition, fixedval/val embryos stained for pax2.1, pax5 or zp23showed characteristic changes in posterior hindbrain patterning. At earlier stages, val/val mutants were identified by loss of krox20(egr2 — Zebrafish Information Network) staining in rhombomere 5(Moens et al., 1996). Liveace/ace (fgf8/fgf8 — Zebrafish Information Network)mutants were readily identified after 24 h by the absence of a midbrain-hindbrain border and enlarged optic tectum(Brand et al., 1996). In addition, ace/ace specimens that were fixed and stained forpax2.1 or pax5 showed no staining in the midbrain-hindbrain border. At earlier stages (14 h), ace/ace mutants were identified by loss of fgf3 expression in the midbrain-hindbrain border.

Embryos were fixed in MEMFA (0.1 M MOPS at 7.4, 2 mM EGTA, 1 mM MgSO₄, 3.7% formaldehyde) and stained as previously described(Riley et al., 1999). Primary antibodies used in this study were: polyclonal antibody directed against mouse Pax2 (Berkeley Antibody Company, 1:100 dilution), which also recognizes zebrafish pax2.1 (Riley et al.,1999); Monoclonal antibody directed against acetylated tubulin(Sigma T-6793, 1:100), which binds hair cell kinocilia(Haddon and Lewis, 1996). Secondary antibodies were Alexa 546 goat anti-rabbit IgG (Molecular Probes A-11010, 1:50) or Alexa 488 goat anti-mouse IgG (Molecular Probes A-11001,1:50).

Whole-mount in situ hybridization was performed as described(Stachel et al., 1993) using riboprobes for fgf3 (Kiefer et al., 1996a), fgf8(Reifers et al., 1998),dlA (Appel and Eisen,1998; Haddon et al.,1998b), pax5 (Pfeffer et al., 1998), dlx3 and msxc(Ekker et al., 1992),nkx5.1 (Adamska et al.,2000), otx1 (Li et al., 1994), and zp23(Hauptmann and Gerster, 2000). Two-color in situ hybridization was performed essentially as described by Jowett (Jowett, 1996) with minor modifications (Phillips et al.,2001).

fgf3-specific morpholino oligomer obtained from Gene Tools was diluted in Danieaux solution [58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO₄,0.6 mM Ca(NO₃)₂, 5.0 mM HEPES, pH 7.6] to a concentration of 5 μg/μl as previously described(Nasevicius and Ekker, 2000;Phillips et al., 2001). Approximately 1 nl (5 ng fgf3-MO) was injected into the yolk cell at the one-to two-cell stage.

Wild-type val was ligated into pCS2 expression vector by Andrew Waskiewicz (Cecilia Moens' laboratory) and was kindly provided as a gift. RNA was synthesized in vitro and ∼1 ng of RNA was injected into the yolk of cleaving embryos at the one- to four-cell stage.

val/val mutants produce small otic vesicles with shortened anteroposterior axes, but relatively normal dorsoventral axes. This gives the mutant ear a characteristic circular shape that is very distinct from the ovoid shape of the wild-type ear. This is thought to arise secondarily from abnormal development of the hindbrain(Moens et al., 1998), signals from which are required for normal ear development. To test this idea directly, we characterized early patterning of the otic vesicle and hindbrain in val/val mutants. In val/val mutants, the size, number and distribution of otoliths in the inner ear vary considerably(Fig. 1A,B). In wild-type embryos, otoliths form only at the anterior and posterior ends of the otic vesicle where they attach to the kinocilia of tether cells(Fig. 1C)(Riley et al., 1997). Tether cells are the first hair cells to form and occur in pairs at both ends of the nascent otic vesicle where they facilitate localized accretion of otolith material. The supernumerary and ectopic otoliths observed in val/valembryos were each associated with pairs of tether cells, as seen in live embryos under DIC optics (not shown). Visualizing tether cells by their expression of deltaA (dla — Zebrafish Information Network) (Haddon et al.,1998a; Riley et al.,1999) confirms that val/val mutants produce excess and ectopic tether cells (Fig. 1D). In both wild-type and val/val embryos, tether cells acquire the morphology of mature hair cells by 22 h(Riley et al., 1997) (data not shown) and can be visualized by nuclear staining with anti-Pax2 antibody. This antibody was originally directed against mouse Pax2 but also binds zebrafishpax2.1 (pax2a — Zebrafish Information Network), which is preferentially expressed in maturing hair cells(Riley et al., 1999). Because of the unusual positions of some hair cells in val/val mutants, their cell type identity was confirmed in some specimens by staining with anti-acetylated tubulin, which labels hair cell kinocilia(Haddon and Lewis, 1996). This confirmed the presence of excess and ectopic hair cells at 24 h inval/val mutants (Fig. 1F). val/val mutants continue to show greater numbers of hair cells than wild-type embryos through at least 33 h(Fig. 2;Table 1). In addition, ectopic patches of hair cells continue to develop between the anterior and posterior maculae in most val/val mutants(Fig. 1G). However, the spatial distribution of hair cells varies widely from one specimen to the next(Fig. 1G,I-K). In general, hair cells can emerge at any position along the ventromedial surface of the otic vesicle in val/val mutants, unlike wild-type embryos in which hair cells are restricted to the anterior (utricular) and posterior (saccular)maculae. These data suggest that the signal(s) that normally regulate the location and number of hair cells are misregulated in val/valmutants, an interpretation further supported by analysis of FGF expression in the hindbrain (see below).

We next examined expression of various otic markers to further characterize altered patterning in val/val embryos. Expression of pax5 is first detectable in the inner ear at 17.5-18.0 h(Pfeffer et al., 1998). This expression domain is normally restricted to the anterior part of the otic vesicle adjacent to r4 and is maintained through at least 30 h(Fig. 3A,C). Inval/val embryos, pax5 expression extends along the entire length of the medial wall of the otic vesicle(Fig. 3B,D). Another anterior marker, nkx5.1, is also expressed throughout the medial wall of the otic vesicle in val/val mutants(Fig. 3F). By contrast,zp23 is normally expressed in posterior medial cells adjacent to r5 and r6 in the wild type but is not detectably expressed in val/valembryos (Fig. 3G,H). Otic patterning is not globally perturbed, however. Mutant embryos show a normal pattern of dlx3 expression in the dorsomedial epithelium(Fig. 4F). Similarly,otx1 is expressed normally in ventral and lateral cells ofval/val mutants (Fig. 4A-D). Based on studies in mouse, the dorsal and lateral domains of dlx3 (dlx3b — Zebrafish Information Network) andotx1 probably help regulate development of the semicircular canals and sensory cristae (Depew at al.,1999; Krauss and Lufkin,1999; Morsli et al.,1999; Mazan et al.,2001). It has previously been reported that formation of semicircular canals is totally disrupted in val/val mutants(Moens et al., 1998). However,we find that this is a highly variable phenotype, ranging from grossly abnormal morphogenesis to nearly normal patterning at day 3(Fig. 4G-I). Morphology typically becomes increasingly aberrant with time, possibly resulting from improper regulation of endolymph, as seen in kreisler mutant mice(Deol, 1964;Brigande et al., 2000) (see Discussion). Regardless of whether semicircular canals develop properly, all three sensory cristae are produced and express msxc (data not shown). Thus, some aspects of axial patterning are relatively normal inval/val embryos at early stages, and the only consistent defect is that medial cells abutting the hindbrain all show anterior character. This is consistent with the hypothesis that factors locally expressed in the hindbrain regulate anterposterior fates in the medial wall of the otic vesicle, and that such factors are misregulated in the rX region of val/val mutants. Such misexpression could also explain the abnormal pattern of hair cells produced in val/val mutants.

Fgf3 and Fgf8 are both expressed in the r4 anlagen during late gastrulation and cooperate to induce the otic placode(Phillips et al., 2001). We hypothesized that persistent expression of one or both of these factors in r4 plays a later role in patterning the otic placode and vesicle. In both wild-type and val/val embryos, fgf8 is expressed at high levels in r4 at 12 h (Fig. 5A,B) but is downregulated by 14 h (not shown). This argues against a role for Fgf8 in the etiology of the inner ear phenotype inval/val embryos. By contrast, fgf3 expression shows a consistent difference between val/val and wild-type embryos. In the wild type, hindbrain expression of fgf3 is restricted to r4 and is maintained through at least 18 h when the otic vesicle forms(Fig. 5C,E, and data not shown). In val/val mutants, fgf3 shows similar developmental timing but is expressed in an expanded domain extending from r4 through rX(Fig. 5D,F). Within rX, the level of expression falls off gradually towards the posterior such that there is no clear posterior limit of expression. Ectopic expression of fgf3in val/val embryos is first detectable at 10 h, corresponding to the time when val normally begins to function in the r5/6 anlagen (data not shown). Initially, ectopic expression of fgf3 in rX is much weaker than in r4. Expression in rX subsequently increases to a level similar to that seen in r4 by 12 h (Fig. 5D). These data suggest that expansion of the domain offgf3 in the hindbrain could play a role in misexpression of AP markers and production of ectopic hair cells in the inner ear.

The above data also suggest that val normally functions, directly or indirectly, to exclude fgf3 expression from r5/6. To explore this more fully, we examined the effects of val mis-expression by injecting val RNA into wild-type embryos. In more than half (55/98)of val-injected embryos, hindbrain expression of fgf3 was dramatically reduced or ablated (Fig. 6A,B). Similar effects were seen at 10, 12 and 14 h (data not shown). At 24 h, otic vesicles were usually small (15/64) or totally ablated(36/64) (Fig. 6C,D). Disruptingfgf3 by itself impairs, but does not ablate, otic tissue(Phillips et al., 2001;Vendrell et al., 2001;Maroon et al., 2002). This indicates that val mis-expression affects other processes in addition to fgf3 expression. Indeed, ubiquitous mis-expression of valfrequently caused truncation of the trunk and tail (46/64,Fig. 6C) and could therefore impair mesendodermal signals on which otic development relies (reviewed byWhitfield et al., 2002). However, even among embryos with normal axial development, about half showed partial loss of fgf3 expression (5/10) and impaired otic development (18/34). In many of these cases, these defects were limited to one side of the embryo(Fig. 6E,F), possibly resulting from variation in the amount of RNA inherited by early cleavage stage blastomeres. In contrast to fgf3, expression of fgf8 was relatively normal in most (82/85) val-injected embryos, even those with axial truncations (Fig. 6H). These data support the hypothesis that valspecifically represses fgf3 expression in the hindbrain. This is in sharp contrast to the function of the mouse homolog kreisler, which is required to activate high level expression of Fgf3 in r5 and r6(McKay et al., 1996). Such species differences may have been important for evolutionary changes in inner ear structure and function (see Discussion).

To test the role of Fgf3 in otic vesicle patterning, embryos were injected with fgf3-MO, an antisense oligomer that specifically inhibits translation offgf3 mRNA (Nasevicius and Ekker,2000; Phillips et al.,2001; Maroon et al.,2002). Injection of fgf3-MO into wild-type embryos results in a range of defects with varying degrees of severity(Phillips et al., 2001). The size of otic vesicle is usually reduced, and about half (42/86) of Fgf3-depleted wild-type embryos show little or no pax5 expression in the inner ear (Fig. 7A). Expression of nkx5.1 is also reduced or ablated in the otic vesicle and vestibulo-acoustic ganglion in about half (30/62) of injected wild-type embryos (data not shown). By contrast, expression of zp23 often expands anteriorly in the otic vesicle to include medial cells adjacent to r4(21/32 embryos, Fig. 7D). Hair cell production is reduced by up to 70% in severely affected embryos(Fig. 7G;Table 1, note range of data). Injection of fgf3-MO into val/val mutants leads to further reduction in the size of otic vesicle. Expression of pax5 is strongly reduced in most cases: In one experiment, 37% (26/71) showed pax5 expression limited to the anterior of the otic vesicle(Fig. 7B) and 38% (27/71)showed no detectable expression (Fig. 7C). Similarly, nkx5.1 is strongly reduced or ablated in about half (16/30) of injected val/val embryos(Fig. 7F). Most (12/15)val/val embryos injected with fgf3-MO express zp23 in the otic vesicle, including tissue adjacent to r4(Fig. 7E). Hair cell production is reduced to a level comparable with that seen in Fgf3-depleted wild-type embryos (Table 1). In addition,depletion of Fgf3 prevents formation of ectopic hair cells in the majority(19/25) of val/val embryos (Fig. 7H,I). Thus, Fgf3-depletion prevents formation of excess and ectopic hair cells as well as misexpression of AP markers in val/valmutants. As the hindbrain is the only periotic tissue known to expressfgf3 at this time, we infer that the expanded domain of fgf3in val/val mutants is crucial for generation of the above inner ear defects.

Although expression of fgf8 did not appear to correlate with changes in inner ear patterning in val/val mutants, we sought to characterize patterning defects in ace/ace mutants and examine genetic interactions between ace and val. Defects inace/ace embryos are less variable than in embryos injected with fgf3-MO (Phillips et al.,2001). The otic vesicle in ace/ace mutants is reduced in size but usually retains an oval shape at 24 h. Hair cell production is reduced by more than half in the majority of ace/ace mutants(Table 1), and more than a third (7/19) of specimens produce no posterior hair cells at all(Fig. 8E). In ace/ace;val/val double mutants, the size of otic vesicle is further reduced and the number of hair cells is comparable with that in ace/acesingle mutants (Fig. 8F;Table 1). Hair cells often form adjacent to r4 and/or rX in ace/ace; val/val double mutants and are usually located in a more medial position than are hair cells inace/ace mutants (Fig. 8F). In addition, pax5 is expressed along the full length of the anteroposterior axis of the ear(Fig. 8D). Expression ofnkx5.1 is also expanded in acelace-val/val double mutants,while zp23 is not expressed (data not shown). Thus, the acemutation strongly perturbs inner ear patterning, but loss of fgf8function does not suppress the patterning defects associated with theval mutation. This is probably because expression of fgf3 is expanded in the hindbrain of ace/ace; val/val double mutants as inval/val mutants (Fig. 8B). Together, these data indicate that val andace affect different developmental pathways, and that the early patterning defects seen in the val/val mutant ear are not caused by mis-regulation of fgf8 expression.

Development of the first hair cells is normally restricted to regions of the otic placode directly adjacent to r4 and r6(Fig. 1), suggesting that signals emitted by those rhombomeres specify the equivalence groups from which hair cells emerge. Data presented here suggest that Fgf3 is an important r4-derived factor that regulates formation of anterior hair cells, as well as expression of various AP markers in the ear. In val/val embryos,fgf3 is expressed ectopically in rX(Fig. 5), and ectopic hair cells form within the adjacent otic vesicle(Fig. 1). Expression ofnkx5.1 and pax5, which are normally restricted to the anterior region of the placode next to r4, expand posteriorly inval/val mutants to include all cells abutting the hindbrain(Fig. 3). The posterior markerzp23 is not expressed in the otic vesicle in val/valmutants. Depletion of Fgf3 suppresses all of the above patterning defects in the val/val mutant ear. Moreover, in many Fgf3-depleted embryos,anterior otic markers are totally ablated and zp23 expression expands anteriorly to include cells adjacent to r4.

The fact that any hair cells are produced at all in Fgf3-depleted embryos indicates that additional hair cell-inducing factors must be present.fgf8 is clearly required for normal hair cell formation and could partially compensate for loss of fgf3(Reifers et al., 1998;Phillips et al., 2001). However, several observations indicate that the role of fgf8 is distinct from that of fgf3. First, periotic expression offgf8 declines sharply just before the placode forms at 14 h, thereby limiting its ability to influence later otic patterning. Second, expression patterns of nkx5.1, pax5 and zp23 are not altered inace/ace embryos (Fig. 8C, and data not shown), indicating that AP patterning is relatively normal. Third, loss of fgf8 inhibits hair cell formation but does not prevent formation of ectopic hair cells in val/valmutants. The latter are dependent on fgf3 instead. Thus, in contrast to fgf3, there is little evidence to suggest that the r4 domain offgf8 regulates regional patterning in the otic placode. Instead,fgf8 may play a more general role in stimulating hair cell competence during the process of placode induction.

Paradoxically, anterior hair cells are not as severely impaired inace/ace mutants as are posterior hair cells. Posterior hair cells are totally ablated in about 1/3 of ace/ace mutants. This is difficult to explain based solely on the expression domain of fgf8, but may reflect changes in the dimensions of the otic placode. In ace/acemutants, the otic placode is often reduced to a domain juxtaposed to r4 and r5 only. Thus, secretion of Fgf3 from r4 may be sufficient to induce some anterior hair cells in the absence of Fgf8, whereas cells in the posterior otic placode may lie too far from r6 to benefit from inductive factors possibly secreted from there. No clear candidates for r6-specific inducers are known, but the Fgf-inducible genes erm, pea3 and sprouty4are expressed in r6 (Fürthauer et al., 2001; Raible and Brand,2001; Roehl and Nüsslein-Volhard, 2001) (S.-J. K., B. T. P., R. H. and B. B. R., unpublished), suggesting that at least one as yet unidentified Fgf homolog is expressed there.

The reason for expanded expression of fgf3 in val/valmutants is not clear, but there are several possibilities. First, this could result from mis-specification of segment identity in the rX territory. Several other genes normally expressed in adjacent segments, including hoxb1in r4 and hoxb4 in r7, eventually come to be expressed in rX(Prince et al., 1998). However, these changes do not occur until 20 somites (19 h). By contrast,expression of fgf3 in rX is first detected at 10 h inval/val mutants, corresponding to the time when val normally begins to function (Moens et al.,1998). This raises the alternative possibility that Val protein normally acts to transcriptionally repress fgf3. In support of this,mis-expression of val inhibits r4-expression of fgf3, but not fgf8 (Fig. 6). Direct support for transcriptional regulation by Val will require analysis of the promoter/enhancer regions of fgf3.

In sharp contrast to val function in zebrafish, mouse kreisler is required, directly or indirectly, for upregulation of Fgf3 in r5 and r6(McKay et al., 1996). This difference is notable because so many other aspects of early hindbrain and ear development are conserved between these species. The high degree of sequence identity leaves little doubt that the zebrafish genes are orthologous to kreisler and Fgf3 (Kiefer et al.,1996a; Moens et al.,1998). There are, however, differences in the N- and C- terminal regions of Fgf3 in zebrafish and mouse. These regions are thought to be important for mediating the characteristic receptor binding preferences and signaling properties of Fgf3. Nevertheless, these functional properties are actually very similar between the fish and mouse proteins(Kiefer et al., 1996b). This,combined with the broad similarities in their expression patterns and involvement in early otic development, strengthen the notion that the fish and mouse fgf3 genes are indeed orthologs. Because zebrafish often has multiple homologs of specific tetrapod genes, it is possible that a secondfgf3 gene might be present in the zebrafish genome that shows an expression pattern more like the mouse gene. If so, it will be important to address its function as well. However, we have shown that the knownfgf3 ortholog plays an essential role in the etiology of the ear phenotype in val/val embryos, as key aspects of the phenotype are suppressed by injecting fgf3-MO. Morpholino oligomers are highly gene-specific in their effects, and even though they do not totally eliminate gene function,they generate phenotypes that are indistinguishable from those caused by known null mutations (Nasevicius and Ekker,2000; Phillips et al.,2001; Raible and Brand,2001; Maroon et al.,2002). On balance, it appears that the general role of Fgf3 in otic development has been conserved in mouse and fish but that differential regulation in the hindbrain represents a real difference between these species.

Considering the above differences in hindbrain signaling, one might expect the ear phenotypes in val/val and Mafb/Mafb mutants to be quite different. Instead, the phenotypes appear strikingly similar. InMafb/Mafb embryos, as in val/val embryos, development of the otic vesicle is highly variable and defects can be seen in virtually all regions of the labyrinth (Deol,1964). In Mafb/Mafb mutants, formation of the wall of the otic capsule is often incomplete, with large gaps through with membranous epithelia protrude, and morphology of the labyrinth is usually grossly abnormal. Such global disruption may be related to buildup of excess fluid pressure due to failure of the endolymphatic duct to form in many or mostMafb/Mafb mutants (Deol,1964; Brigande et al.,2000). Whether a similar problem occurs in val/valmutants is not clear. The existence of an endolymphatic duct in zebrafish has only recently been documented (Bever and Fekete, 2002), but it does not begin to form until around day 8. Most val/val mutants die before this time, and they often begin to show defects in morphogenesis (e.g. of the semicircular canals) by 72 h(Fig. 4, and data not shown). Although these early defects cannot be explained by the absence of an endolymphatic duct, mutant ears often appear swollen and distended by day 3,suggesting a buildup of endolymphatic pressure. It is possible that cellular functions normally required to maintain a proper fluid balance in the early vesicle are mis-regulated in val/val mutants. Thus, hydrops may be an important contributing factor to the defects in both Mafb/Mafb andval/val mutants.

Another similarity between Mafb/Mafb and val/val mutants is that they both form ectopic patches of hair cells. However, this phenotype has a completely different etiology in the two species. In tetrapod vertebrates, sensory epithelia do not begin to differentiate until after the various chambers of the labyrinth begin to form. Thus, formation of ectopic hair cells in Mafb/Mafb mutants probably reflects the general disorganization of, and chaotic protrusions from, the labyrinth(Deol, 1964). By contrast,sensory epithelia in zebrafish begin to differentiate much earlier. Macular equivalence groups are already specified at 14 h when the placode first forms(Haddon et al., 1998a;Whitfield et al., 2002), and the first hair cells (visualized by the presence of kinocilia) are evident as soon as the lumen of the vesicle forms at 18.5 h(Riley et al., 1997). Thus,formation of ectopic hair cells in val/val mutants reflects an early defect in cell fate specification rather than a later defect in morphogenesis. It is noteworthy that there have been no detailed molecular studies of otic development in Mafb/Mafb mutants, so a direct comparison of early pattern formation is not yet possible.

It is interesting to consider that the altered pattern of fgf3expression in the val/val mutant hindbrain closely resembles the normal pattern of Fgf3 expression in chick and mouse embryos(Mahmood, 1995; Mahmood, 1996; McKay et al., 1996). Analysis of val/val mutants suggests that misexpression of fgf3 in rX leads to development of excess and ectopic hair cells in the otic vesicle. It is possible that evolutionary changes that led to normal expression of Fgf3 in r5/6 in amniotes were crucial for evolution of the cochlea, which has no known counterpart in anamniote vertebrates (Lewis et al.,1985). In the mouse, development of the cochlea requires FGF signaling at early otic vesicle stages(Pirvola et al., 2000). The FGF receptor isoform FGFR-2(IIIb) is expressed in the otic epithelium juxtaposed to the hindbrain. Targeted disruption of this isoform leads to severe dysgenesis of the cochlea. Cochlear development is also impaired inFgf3-null and Mafb/Mafb mutant mice(Deol, 1964;Mansour et al., 1993). InXenopus, Fgf3 expression shows a pattern intermediate between that of zebrafish and amniotes: The frog gene is initially expressed in r3 through r5 and only later becomes restricted to r4(Lombardo et al., 1998). Although amphibians do not possess a cochlea, they do show modifications of the posterior otic vesicle that give rise to the basilar and amphibian papillae, auditory organs not found in fish (reviewed byLewis et al., 1985). Thus,expression of fgf3 in more posterior regions of the hindbrain correlates with elaborations of the inner ear that may have been essential for enhancing auditory function in terrestrial environments.

This work was supported by National Institutes of Health, NIDCD grant 5 R01 DC03806. We thank Andrew Waskiewicz and Cecilia Moens for their kindness and generosity in providing the pCS2-val expression vector, and Arne Lekven for helpful discussion of the data. We also thank Peter Pfeffer, Eva Bober and Giselbert Hauptmann for providing pax5, nkx5.1 and zp23cDNAs.