Changes in retinoic acid signaling alter otic patterning

The vertebrate inner ear is the sensory organ that provides auditory and vestibular functions. It develops from a transient ectodermal thickening, the otic placode, visible on either side of the developing hindbrain. Depending on the species, the placode invaginates or cavitates to form the otic vesicle,also known as the otocyst, an epithelial structure with sharply defined borders. Subsequently, the otocyst gives rise to all structures of the inner ear including the membranous labyrinth and neurons of the statoacoustic ganglion (Noden and van de Water,1986; Couly et al.,1993; Fritzsch et al.,1997; Whitfield et al.,2002; Barald and Kelley,2004).

Previous studies support the primacy of fibroblast growth factors (Fgfs) in the induction of the otic placode. Various Fgf family members from various sources regulate otic induction in different species, with only hindbrain-derived Fgf3 playing a conserved role (reviewed in Fritzsch et al., 1997; Torres and Giráldez,1998; Whitfield et al.,2002; Brown et al.,2003). In zebrafish, Fgf3 and Fgf8 have been implicated to have overlapping functions; loss of both fgf3 and fgf8 together results in near or total ablation of otic tissue(Phillips et al., 2001; Maroon et al., 2002; Léger and Brand, 2002). In mouse, Fgf3 and Fgf10 act as redundant signals during otic induction(Wright and Mansour, 2003; Alvarez et al., 2003): Fgf3 is expressed in the hindbrain abutting the preotic domain, whereas Fgf10 is expressed in the mesoderm beneath it, and loss of both Fgf3 and Fgf10 results in the complete ablation of otic development(Wright and Mansour, 2003; Alvarez et al., 2003). Furthermore, Fgf8 has been shown to play a crucial role upstream of the Fgf signaling cascade required for otic induction in this species(Ladher et al., 2005). In chick, Fgf3, Fgf8 and Fgf19 and, in amphibians, Fgf2 and Fgf3, have been implicated in otic induction (Mahmood et al., 1995; Ladher et al.,2000; Ladher et al.,2005; Song and Slack,1994; Lombardo et al.,1998).

Recently, studies have shown that Pax2a and Pax8 are the main effectors downstream of Fgf-signaling and that cells need to express Foxi1 and Dlx3b transcription factors to be competent to respond to Fgf signaling in zebrafish(Hans et al., 2004; Solomon et al., 2004). The expression of foxi1 is restricted to bilateral domains, including the preotic domain at late gastrula stages, and disruption of foxi1expression leads to severe defects in otic placode formation and highly variable ear phenotypes (Thisse et al.,2005; Solomon et al.,2003; Nissen et al.,2003). The homeobox gene, dlx3b, is co-expressed with dlx4b in late gastrula stage embryos in a stripe corresponding to cells of the future neural plate border, and knockdown of dlx3b and dlx4b together causes a severe loss of otic and olfactory tissues(Akimenko et al., 1994; Ekker et al., 1992; Ellies et al., 1997; Kudoh et al., 2001; Solomon and Fritz, 2002; Liu et al., 2003). Loss of both Dlx3b and Foxi1 ablates all indications of otic induction even in the presence of a fully functional or over-activated Fgf signaling pathway(Hans et al., 2004; Solomon et al., 2004; Hans et al., 2007).

Retinoic acid (RA), a derivative of vitamin A, is required for proper embryonic development. Embryos deficient in RA signaling show defects in the circulatory system, limbs, trunk and hematopoietic system (reviewed by Maden, 2002). RA plays a crucial role in hindbrain patterning and rhombomere (r) identity, and the hindbrain is known to regulate otic development (reviewed by Gavalas and Krumlauf, 2000; Romand, 2003). In amniote embryos, RA is required in a concentration- and time-dependent manner for the development of the posterior hindbrain, particularly r5-r7, and compromised RA signaling leads to an expansion of anterior hindbrain at the expense of posterior hindbrain (Dupé et al.,1999; Dupé and Lumsden,2001). Complete absence of RA signaling leads to a complete loss of r5-r7 accompanied by an expansion of r3-r4, as observed in mouse embryos mutant for Aldh1a2 (also known as Raldh2 - Mouse Genome Informatics). Aldh1a2 is an aldehyde dehydrogenase that is responsible for the majority of RA production in the early embryo(Niederreither et al., 1999; Niederreither et al., 2000). Mutations in aldh1a2 in zebrafish are less profound and show only a loss of r7 accompanied by a slight expansion of r5 and r6 similar to weak vitamin A deficiency syndrome (VAD) in amniote embryos(Begemann et al., 2001; Grandel et al., 2002; Maves and Kimmel, 2005). However, a loss of r5-r7 accompanied by an expansion of r3 and r4 is observed after knockdown of RA signaling by pharmacological treatment, suggesting that RA is produced by something in addition to Aldh1a2(Grandel et al., 2002; Maves and Kimmel, 2005). Treatment of vertebrate embryos with excess RA posteriorizes the anterior neural plate with the transformation of r2-r3 to r4-r5 identity and expansion of posterior hindbrain at the expense of presumptive fore- and mid-brain structures (Marshall et al.,1992; Kudoh et al.,2002).

Otic defects generated by changes in RA signaling are considered mostly secondary consequences due to changes in Fgf signaling, because hindbrain patterning is disturbed in RA gain- and loss-of-function studies. In embryos with no RA signaling, such as mouse Aldh1a2 mutants, expression of Fgf3, which is specifically expressed in the presumptive r5-r6 in wild-type embryos, is weak and not properly restricted, and otocysts are hypoplastic and abnormally distant from the hindbrain(Niederreither et al., 1999). The same otic phenotype can be observed in zebrafish after knockdown of RA by pharmacological treatment and it has been suggested, but not yet shown, that RA is required for normal fgf3 expression in the presumptive r4 and for the induction of the otic placode(Perz-Edwards et al., 2001; Whitfield et al., 2002). A moderate loss of RA signaling, such as in weak quail or rat VAD embryos,however, leads to the formation of supernumerary otic vesicles caudal to the main otocyst, which correlates with the expansion of posterior hindbrain and,in particular, with the caudal expansion of Fgf3 expression(White et al., 2000; Kil et al., 2005). In zebrafish, the opposite result has been reported; excess, rather than reduced,RA results in the formation of supernumerary and ectopic otic vesicles in an Fgf-dependent manner (Phillips et al.,2001). In zebrafish embryos treated with teratogenic doses (1μM) of RA, hindbrain expression domains of fgf3 and fgf8are greatly expanded into the anterior neural plate, inducing preotic pax8 expression surrounding the anterior neural plate. Inactivation of Fgf3 and/or Fgf8 by morpholino injection blocks the effects of exogenous RA(Phillips et al., 2001).

Here, we show that the opposite otic phenotypes generated by more or less RA signaling are produced by different mechanisms. We used a 50-fold lower dose of RA than that of previous studies, and show that RA affects otic induction without expanding fgf3 and fgf8 expression in the hindbrain. Instead, this low dose of excess RA leads to an increase in otic competence due to expanded expression of foxi1 throughout the preplacodal domain. By contrast, reduced RA signaling results in lower induction and impaired maintenance of otic fates, due to compromised fgf3 and wnt8b expression in the hindbrain. We further show that fgf8 is the primary inducing factor in RA-deprived embryos and that the induction but not the maintenance phenotype can be rescued by an increase in Fgf-signaling.

Embryos were obtained from the University of Oregon zebrafish facility,produced using standard procedures(Westerfield, 2000) and staged according to standard criteria (Kimmel et al., 1995) or by hours post fertilization at 28°C (hpf). The wild-type line used was AB. The lines acerebellar^ti282a (a strong hypomorphic allele of fgf8); lia^t24149 (a hypomorphic allele of fgf3); tcf2^hi2169 (a null or strong hypomorphic allele of tcf2) and the transgenic line Tg(hsp:fgf8) (to misexpress fgf8) have been described previously (Brand et al., 1996; Herzog et al., 2004; Sun and Hopkins, 2001; Hans et al., 2007), and we refer to the homozygous mutants as fgf8, fgf3 and tcf2mutants, respectively. Homozygous mutants and double mutants were scored either by their morphological phenotype and/or by PCR or loss of pou1f1 expression (Herzog et al.,2004; Sun and Hopkins,2001). Heat-shock-treatment embryos were transferred, still in their chorions, into fresh 37-39°C embryo medium in a 1.5 ml tube (20 embryos per tube) and maintained for 30 minutes in a heating block.

Approved gene and protein names that follow the zebrafish nomenclature conventions(http://zfin.org/zf_info/nomen.html)are used.

cDNA probes that detect the following genes were used: cldna(Kollmar et al., 2001); dlx3b (Ekker et al.,1992); fgf3 (Phillips et al., 2001); fgf8(Reifers et al., 1998); foxi1 (Solomon et al.,2003); otx2 (Li et al., 1994); pax8(Pfeffer et al., 1998); pou1f1 (Nica et al.,2004); stm(Söllner et al., 2003);and wnt8b (Kelly et al.,1995). Probe synthesis and single- or double-color in situ hybridization was performed essentially as previously described(Thisse et al., 1993; Jowett and Yan, 1996; Whitlock and Westerfield,2000).

The dlx3b-MO, foxi1-MO and wnt8b-MO have been previously described (Liu et al.,2003; Solomon and Fritz,2002; Kim et al.,2002). For pharmacological treatments, the following stock solutions were made and stored at -80°C: 100 mM 4-(Diethylamino)-benzaldehyde (DEAB; Sigma) in DMSO and 1 mM all-trans retinoic acid (RA; Sigma) in DMSO. For embryo treatments, dilutions of these chemicals were made in embryo medium as follows: DEAB, 10 μM; and RA, 10,20 and 1000 nM. Prior to gastrulation (30% epiboly or 4.7 hpf) embryos were removed from their chorions and transferred into Petri dishes containing the treatment solution. Treatments with a teratogenic dose of RA were carried out as described previously (Kudoh et al.,2002); dechorionated embryos at 40% epiboly stage were treated with 1 μM RA for 80 minutes followed by thorough rinses with embryo medium. For control treatments, sibling embryos were incubated in corresponding dilutions of DMSO. All incubations were conducted in the dark.

It has been shown previously that excess or deficient RA signaling leads to opposite otic phenotypes, but the underlying mechanisms have remained obscure(Perz-Edwards et al., 2001; Phillips et al., 2001). To study the role of RA signaling in otic development, we analyzed the expression of several otic markers at the vesicle, placodal and preplacodal stages. To interfere with RA signaling, we used the pharmacological inhibitor 4-(Diethylamino)-benzaldehyde (DEAB), a potent retinaldehyde dehydrogenase inhibitor (Russo et al.,1988). Compared with the morphology of controls at 50 hpf, loss of RA signaling led to strongly reduced otic vesicles with only one small otolith and impaired semicircular canal protrusions, whereas excess RA at a concentration of 10 nM led to the formation of enlarged otic vesicles containing three otoliths and multiple semicircular canal protrusions(Fig. 1A-C). Assessed at 22 hpf by expression of the otic marker starmaker (stm), which was expressed throughout the entire epithelium of control otic vesicles, otic vesicle size was reduced by up to 50% in RA-signaling-depleted embryos, but increased up to 40% in the presence of 10 nM RA(Fig. 1D-F). At the 12-somite stage, when the placode was outlined by cldna expression in control embryos, the otic placode was reduced in size in RA-signaling-depleted embryos, whereas excess RA led to a size increase(Fig. 1G-I). Use of pax8 expression, which is initiated in the otic anlagen prior to the formation of the placode, showed that the opposite phenotypes generated by excess or deficient RA signaling are already evident at preotic stages. In RA-signaling-depleted embryos, the preotic expression of pax8 is reduced, whereas, in RA-treated embryos, the preotic pax8 expression domain is enlarged and weak expression of pax8 surrounding the entire anterior neural plate can be detected (Fig. 1J-L). An increase of RA to 20 nM amplifies the pax8expression surrounding the anterior neural plate(Fig. 2A), consistent with results published previously (Phillips et al., 2001). However, in contrast to Phillips et al., who reported that 20-30% of embryos treated with teratogenic doses (1 μM) of RA produce ectopic otic vesicles at the anterior limit of the head, we observed only randomly distributed small patches of ectopic otic cells in embryos treated with 20 nM RA (Fig. 2B-D). To demonstrate that our RA treatment did not posteriorize the anterior neural plate in the same manner as typical teratogenic doses of RA(Phillips et al., 2001; Kudoh et al., 2002), we treated embryos with either 20 nM RA or a teratogenic dose of RA and compared them to controls (see Fig. S1 in the supplementary material). Subsequent marker analysis of the anteroposterior axis showed that teratogenic doses, but not 20 nM, of RA significantly altered the expression of early anteroposterior-axis-specific genes. In comparison to controls, expressions of six3a and otx2 was unaffected by 20 nM RA but are completely abolished in the presence of a teratogenic dose (see Fig. S1A-F in the supplementary material). Expression of the RA-controlled gene hoxb1b(Alexandre et al., 1996) in the caudal hindbrain of control embryos was mildly expanded in embryos treated with 20 nM RA, but it was strongly misexpressed throughout the entire neural plate of embryos treated with a teratogenic dose of RA (see Fig. S1G-I in the supplementary material). In embryos treated with 20 nM RA, expression of gbx1 in the rostral hindbrain was indistinguishable from control embryos, whereas it was misexpressed in the anterior neural plate of embryos treated with a teratogenic dose of RA (see Fig. S1J-L in the supplementary material). Together, our results show that the opposite phenotypes generated by gain or loss of RA signaling can be traced back to otic induction and that ectopic otic induction occurs with low doses of RA that do not apparently affect other aspects of embryonic development.

Our results indicate that loss of RA signaling impairs the formation of the otic vesicle, whereas increased RA results in enhanced otic development. This could be due to changes in the mechanisms underlying the induction of otic fates or in the competence of cells to respond to the inductive signals, or both. To distinguish among these possibilities, we examined the expression of the factors that regulate otic induction and competence. In zebrafish, Fgf3 and Fgf8 have been implicated to have overlapping functions in otic placode induction (Phillips et al.,2001; Maroon et al.,2002; Léger and Brand,2002). At 95% epiboly, fgf3 was expressed strongly in the r4 primordium of control embryos, but its expression was severely reduced in embryos depleted of RA signaling (Fig. 3A,B). Because the complete loss of RA signaling led to a loss of r5-r7 accompanied by an expansion of r3-r4, we consistently observed enlarged fgf3 and fgf8 expression domains in DEAB-treated embryos at the three-somite stage, although the level of fgf3 expression appeared to still be reduced in comparison with control embryos(Fig. 3D,E,G,H). By contrast,expression of both fgf3 and fgf8 was completely unaffected in embryos treated with 10 or 20 nM RA(Fig. 3C,F,I and data not shown). We and others recently proposed that Foxi1 and Dlx3b provide competence for cells to form the inner ear(Hans et al., 2004; Solomon et al., 2004). In comparison to control embryos, foxi1 expression was unchanged in the absence of RA signaling, whereas in embryos treated with 20 nM RA, foxi1 was misexpressed in a stripe surrounding the anterior neural plate (Fig. 3J-L). By contrast,we observed no change in dlx3b expression after the complete loss or ectopic activation of RA signaling (Fig. 3M-O). These results show that the complete loss of RA signaling impairs early fgf3 expression in the developing hindbrain, which cannot be compensated for by expanded hindbrain-specific domains of fgf3 and fgf8 expression at later stages. Excess RA signaling, on the other hand, causes an expansion of foxi1 expression throughout the entire preplacodal ectoderm.

Our observation that foxi1 is misexpressed in RA-treated embryos suggested the possibility that the preplacodal domain is posteriorized in these embryos, similar to fate changes in the developing neural plate in embryos treated with a teratogenic dose of RA(Marshall et al., 1992; Kudoh et al., 2002). Subsequently, other placode-derived structures might be affected. To examine this possibility, we tested molecular markers for the formation of anterior pituitary, olfactory, lens, trigeminal, lateral line and epibranchial placodes. Except for the lens, all of these structures formed in embryos treated with 20 nM RA (data not shown). Compared with controls at 50 hpf,embryos treated with 10 nM RA developed a much smaller lens and embryos treated with 20 nM RA completely failed to show any morphological sign of a lens (Fig. 4A-C). Consistent with this, expression of connexin 44.1, which is expressed only in the ocular lens (Cason et al.,2001), was changed. In comparison to controls at 24 hpf, embryos treated with 10 or 20 nM RA showed severe downregulation or no expression of connexin 44.1, respectively (Fig. 4D-F). Labeling for pitx3, which is expressed in the ventral and posterior diencephalon, in the lens and in the pituitary placode at 24 hpf in wild-type embryos (Dutta et al., 2005), corroborated our results, showing expression changes,after RA treatment, specifically in the lens placode without affecting other expression domains (data not shown). At early segmentation stages, pitx3 expression defines an equivalence domain for the lens and anterior pituitary placodes (Dutta et al.,2005), and we found that its expression was unaffected in the presence of 10 or 20 nM RA (data not shown). By contrast, expression of pax6b, a specific lens placode marker(Thisse et al., 2001), was downregulated or almost absent in embryos treated with 10 or 20 nM RA,respectively, in comparison to control embryos(Fig. 4G-I). Together, our results show that increasing amounts of RA impair lens placode formation,which subsequently leads to a loss of the lens.

In zebrafish, Fgf3 and Fgf8 are necessary for otic induction and, because loss of RA signaling led to a significant reduction in fgf3expression in the developing hindbrain, we hypothesized that the remaining otic induction is dependent primarily on fgf8. In comparison to controls, otic vesicles in the absence of RA signaling or in fgf8mutants are reduced in size and typically form only one otolith(Fig. 5A-C). By contrast, loss of RA signaling in fgf8 mutants produced a different and highly penetrant phenotype (n=25 out of 25 fgf8 mutants); embryos never formed an otic vesicle (Fig. 5D). Nevertheless, the otic marker stm revealed the presence of some residual otic cells in fgf8 mutants (n=10 out of 10 fgf8 mutants) depleted of RA signaling compared to untreated fgf8 mutants, RA-depleted or control embryos(Fig. 5E-H). Analysis of pax8 expression showed that this phenotype was already evident at preotic stages. Loss of fgf8 or RA signaling alone led to reduced pax8 expression in the preotic region, but combined loss of fgf8 and RA signaling caused a severe loss of pax8(Fig. 5I-L). Our observation of some indications of residual otic specification in fgf8 mutants depleted of RA signaling is consistent with the delayed and reduced expression of fgf3 in the hindbrain. Furthermore, loss of RA signaling in fgf3 mutants had no effect on otic vesicle size (data not shown),supporting our hypothesis that, in the absence of RA signaling, residual otic induction depends primarily on Fgf8.

Our results indicate that competence provided by Foxi1 and Dlx3b to initiate otic fate is not impaired in RA-depleted embryos. This suggests that reduced otic induction caused by the delayed onset of fgf3 expression is responsible for the observed otic phenotype, which we hypothesize could be rescued by ectopic Fgf signaling. To activate ectopic Fgf signaling, we used a stable transgenic line that allows us to express fgf8 uniformly under the control of the zebrafish temperature-inducible hsp70 promoter(Hans et al., 2007). Misexpression of fgf8 at late gastrulation stages led to ectopic otic induction and caused expanded marker expression, such as that of pax8,pax2a and sox9a, within the preotic region in comparison to non-transgenic siblings (Fig. 6A,B) (Hans et al.,2007). Subsequently, many more cells embarked on the otic development pathway and much larger otic vesicles formed(Fig. 6E,F)(Hans et al., 2007). By contrast, we also observed an expanded pax8 expression domain after misexpression of fgf8 at late gastrulation stages in embryos depleted of RA signaling, but the subsequent formation of enlarged otic vesicles was impaired (Fig. 6C,D,G,H). In embryos depleted of RA signaling, otic vesicles were only slightly bigger after ectopic activation of Fgf signaling at late gastrulation stages than in non-transgenic siblings depleted of RA signaling, and they were still much smaller than in untreated controls. We also frequently observed small ectopic otic vesicles located anteriorly and posteriorly to the endogenous otic vesicle after misexpression of fgf8 at late gastrulation stages in RA-depleted embryos (data not shown), but, in general, otic specification was always reduced in comparison to control embryos. Taken together, our results demonstrate that, in the absence of RA signaling, cells in the preotic region can undergo ectopic otic induction, similar to control embryos, after the ectopic activation of Fgf signaling at late gastrulation stages. However, otic fate cannot be properly maintained in the absence of RA signaling.

Our results suggested that an Fgf-independent mechanism is required to maintain otic fate and, because loss of RA signaling led to the expansion of r3-r4 accompanied by loss of r5-r7, we hypothesized that the source might be located within r5-r7. Analysis of fgf3;tcf2 double mutants confirmed this hypothesis (Fig. 7A-E). The homeobox gene tcf2 (previously vhnf1) is expressed in the posterior hindbrain and anterior spinal cord in an RA-dependent manner at early segmentation stages. The anterior borders of the tcf2expression domain and of r5 coincide (Sun and Hopkins, 2001; Hernandez et al., 2004). In strong tcf2 mutant alleles, the r5-r6 region of the developing hindbrain is lost and partially transformed into an r4 identity (Hernandez et al.,2004). Although otic induction is unaffected at early stages in these mutants (Sun and Hopkins,2001), otic vesicles are reduced in size at later stages indicating that maintenance of otic fate is compromised(Fig. 7A,D)(Sun and Hopkins, 2001). In comparison to either wild type or tcf2 or fgf3 single mutants, combined loss of tcf2 and fgf3 resulted in otic vesicles that formed only one otolith (n=10 out of 10 fgf3;tcf2 double mutants), similar to RA-depleted embryos(Fig. 7A-E). Semicircular canal formation, however, which was severely impaired in RA-signaling-depleted embryos, was less severely affected in fgf3;tcf2 double mutants,indicating that, similar to mouse (Romand,2003), RA signaling is involved in pattern formation of the zebrafish otic vesicle. Recent studies in mouse have shown that Wnt signaling is required early for the maintenance of otic fate and later in the otic vesicle for expression of dlx5 and for the acquisition of a dorsal fate (Ohyama et al., 2006; Riccomagno et al., 2005). We found that expression of dlx3b in the dorsal portion of the otic vesicle was completely lost in RA-depleted embryos at 22 hpf(Fig. 7F,G), suggesting that Wnt signaling may be affected. Wnt8b is a likely candidate because it is expressed strongly in r1, r3 and r5 from mid-somitogenesis onwards(Kelly et al., 1995). Compared to controls, both tcf2 mutants and RA-depleted embryos showed a loss of wnt8b expression specifically in r5(Fig. 7H-J). To test whether wnt8b acts to maintain otic fate, we compromised Wnt8b function using morpholino injection into wild type. Furthermore, we compromised Wnt8b function in fgf3 mutants to test whether loss of fgf3 and wnt8b together mimics the otic vesicle size reduction of RA-signaling-depleted embryos, which have defects in fgf3 and wnt8b expression (Fig. 7K-O). Loss of Fgf3 or Wnt8b function alone led to the formation of smaller otic vesicles, but the size reduction was not as profound as in embryos depleted of RA signaling (Fig. 7K-N). However, compromised Wnt8b function in fgf3mutants showed an additive effect, producing smaller otic vesicles similar to vesicles in RA-depleted embryos (n=14 out of 18)(Fig. 7L,O). Depletion of RA signaling and Wnt8b function together resulted in a slight further reduction in otic vesicle size (data not shown). However, because of the expansion of r3-r4, otic vesicles were located closer to r3 in RA-depleted embryos (data not shown), and expression of wnt8b in r3 might partly compensate for the loss of wnt8b from r5 in these embryos. Together, our results demonstrate that the hindbrain-derived factor Wnt8b from r5 maintains otic fate, and that compromised Fgf3 expression and loss of Wnt8b function together mimics the size reduction of the otic vesicle observed in RA-depleted embryos.

Treatment of vertebrate embryos with RA causes a posteriorization of the anterior neural plate in a concentration-dependent manner, which leads to transformation of r2-r3 into r4-r5 and to the expansion of the posterior hindbrain at the expense of the presumptive fore- and mid-brain structures(Marshall et al., 1992; Kudoh et al., 2002). Furthermore, teratogenic doses of RA completely transform forebrain and midbrain into hindbrain and enlarge the preotic expression domain of pax8 (Phillips et al.,2001). Our results demonstrate that much lower concentrations of RA that do not interfere with the patterning of the anterior neural plate are sufficient to produce ectopic expression of foxi1, which expands the domain of preotic pax8 expression. Consistent with recent reports showing that preotic pax8 expression depends on Foxi1(Solomon et al., 2003; Nissen et al., 2003), we found that the enlarged pax8 expression observed in 20 nM RA-treated wild-type embryos was severely reduced to a small, residual domain of expression at the anterior end in RA-treated foxi1 mutants(Hans et al., 2007). Fate changes after the application of 20 nM RA were observed only in the preplacodal domain, but not within the anterior neural plate, whereas a teratogenic dose of RA affected both tissues. This difference is presumably due to the presence of RA-degrading enzymes of the Cyp26 class that are expressed within the anterior neural plate but not in the preplacodal domain(Dobbs-McAuliffe et al., 2004; Kudoh et al., 2002; Hernandez et al., 2006). Consistent with this interpretation, a recent study has shown that 5 nM RA is sufficient to posteriorize cyp26a1 mutants, whereas 200 nM RA is required to obtain the same effect on wild-type embryos(Hernandez et al., 2006).

Our observation that excess RA leads to ectopic expression of foxi1 throughout the entire preplacodal domain provides a consistent explanation for the findings of Woo and Fraser (Woo and Fraser, 1997). Transplantation of ventral and lateral germring tissue into the prospective forebrain region can induce ectopic otic vesicles,whereas the transplantation of dorsal germring tissue (the embryonic shield)does not. Although germring grafts also lead to the induction of ectopic hindbrain tissue, the direct transplantation of hindbrain into the prospective forebrain region never induces ectopic otic vesicles(Woo and Fraser, 1997). RA is generated by the enzymatic activity of aldehyde dehydrogenases, which are encoded by Aldh genes, and, during zebrafish gastrulation, aldh1a2 is expressed in the ventral and lateral germring but not dorsally in the embryonic shield (Begemann et al.,2001; Grandel et al.,2002). Thus, we propose that RA produced by the grafted ventral and lateral germring increases the competence of anterior preplacodal cells to respond to otic-inducing signals because of ectopic activation of foxi1. The observed ectopic otic induction might also benefit from ectopic Fgf signaling provided by the graft, because fgf3, fgf8,fgf17b and fgf24 are all expressed in a dorsoventral gradient in the zebrafish germ ring, with highest levels on the dorsal side(Fürthauer et al., 1997; Fürthauer et al., 2004; Reifers et al., 1998; Cao et al., 2004).

Although RA is sufficient to induce foxi1 expression, our data also show that loss of RA signaling during zebrafish gastrulation has no effect on the expression of foxi1. Thus, impaired otic induction in RA-depleted embryos is not due to an effect on Foxi1 level and, hence, we conclude that RA signaling is not required for the early events of otic induction. The observed effect of excess RA on foxi1 expression might rather reveal a regulatory mechanism employed during other aspects of development. In zebrafish at 48 hpf, foxi1 is expressed in the retina, in the pharyngeal pouches and in the dorsal septum of the otic vesicle(Thisse et al., 2005), and RA signaling has been implicated in the development of all of these tissues(Drager et al., 2001; Mark et al., 2004; Romand, 2003). It is currently unknown, however, whether foxi1 expression depends on RA signaling in these tissues.

Our results further indicate that early fgf3 expression is crucial during the relatively short time-window for otic induction in zebrafish. Impaired fgf3 expression in the developing hindbrain of RA-signaling-depleted embryos delayed otic induction and could not be compensated for by an expanded hindbrain domain of fgf3 and fgf8 expression at later stages. Consistently, depletion of fgf3 by morpholino injection severely reduces pax8expression at the tailbud stage (Phillips et al., 2001; Léger and Brand, 2002). Based on experiments with a stable transgenic line that expresses fgf8 uniformly under the control of the zebrafish temperature-inducible hsp70 promoter, we recently proposed that otic induction in zebrafish occurs between the end of gastrulation and the early segmentation stages (Hans et al.,2007). Using the same approach here, we show that even in the absence of RA signaling, cells in the preotic region can undergo ectopic otic induction after ectopic activation of Fgf signaling at late gastrulation stages, similar to control embryos. Thus, cells are still competent to adopt otic fate after exposure to a high uniform dose of Fgf8 but are unable to respond to the endogenous Fgf3 and Fgf8 emanating from the developing hindbrain after the loss of RA signaling. Incorrect spatial positioning of inducing and induced tissue might also contribute to impaired otic induction,because, in RA-signaling-depleted embryos, the r3 primordium is greatly expanded, displacing the inducing r4 from the preotic region.

Our analysis further demonstrates that defects in otic specification caused by loss of RA signaling are a secondary consequence of changes in hindbrain patterning. We found that embryos depleted of RA signaling formed small otic vesicles because of compromised otic induction due to decreased fgf3expression and because of compromised maintenance of otic fate due to loss of wnt8b expression. Reduced otic induction due to weaker expression of Fgf3 has been proposed from work in both mammals and zebrafish(Niederreither et al., 1999; Whitfield et al., 2002; Phillips et al., 2001; Léger and Brand, 2002),and a recent study in mouse shows that Wnt signaling is required after Fgf-dependent induction to maintain otic fate(Ohyama et al., 2006). In the latter study, Wnt signaling was detected using a Tcf/Lef reporter construct within the preotic region in a subset of Pax2-positive cells after Fgf-dependent induction but prior to the formation of the placode. Furthermore, disruption of the canonical Wnt signaling pathway in preotic Pax2-positive cells leads to an expansion of epidermal fate at the expense of otic fate, and otic vesicles are significantly reduced in size(Ohyama et al., 2006). Conversely, constitutive activation of canonical Wnt signaling in preotic Pax2-positive cells leads to the expansion of otic fate at the expense of epidermal fate, suggesting that Wnt signaling mediates a placode-epidermis fate decision by directing preotic cells to an otic fate(Ohyama et al., 2006). So far,it is unknown whether Wnt signaling plays the same role in zebrafish. It has been reported that, in zebrafish, the Wnt reporter gene TOPdGFP, a transgene consisting of a GFP-coding sequence downstream of a minimal promoter and four Lef-binding sites, is not expressed during early stages of otic induction (Phillips et al.,2004), but nothing is known about its expression later in development, prior to the formation of the otic placode. Consistent with a similar role for Wnt signaling, we found that, in RA-depleted embryos, otic induction can be rescued but not maintained. Furthermore, we have previously found that expression of the endogenous Wnt reporter gene axin2 is absent during the initial stages of otic induction but that axin2 is expressed within the preotic region prior to the formation of the otic region(our unpublished results). Cell-autonomous disruption or constitutive activation of the canonical Wnt signaling pathway in mouse by conditional knockout of β-catenin or by the expression of a conditionally activated form of β-catenin within preotic Pax2-positive cells led Ohyama et al. to propose that Wnt8a from r4 is the source of Wnt signaling(Ohyama et al., 2006). Our data, on the other hand, indicate that wnt8b expression in r5 is the source, consistent with findings that Wnt8b signaling is strongly deficient in the posterior hindbrain of mafb mutants and that inactivation of Wnt8b by morpholino injection results in smaller otic vesicles that mimic the size and patterning defects of mafb mutants (M. Brand, personal communication). In zebrafish, wnt8a encodes a bicistronic message encoding two complete open-reading frames, ORF1 and ORF2, and, at 75% epiboly(8 hpf), ORF2 can be detected in r5-r6 adjacent to the otic anlagen(Kelly et al., 1995; Lekven et al., 2001). Loss of Wnt8a function by morpholino injection results in greatly reduced otocysts(Phillips et al., 2004),supporting the model proposed by Ohyama et al.(Ohyama et al., 2006). However, proper expression of fgf3 and fgf8 is disturbed in the hindbrain of these embryos, thus compromising otic induction(Phillips et al., 2004) and making it difficult to address later Wnt8a function during otic development. It is very likely that several Wnt molecules are involved in otic fate maintenance after Fgf-dependent induction. Expression of wnt1, wnt3aand wnt10b has been reported in the zebrafish hindbrain in addition to wnt8a and wnt8b, and, interestingly, expression of wnt1 and wnt3a is severely reduced in the caudal hindbrain of tcf2 mutants (Lekven et al.,2003; Buckles et al.,2004; Lecaudey et al.,2006), as we have shown for wnt8b. Conditional and combinatorial knockout will be necessary to address the role of particular Wnt molecules in otic fate maintenance.

Absent or inappropriate amounts of Wnt signaling might also be responsible for the reduced number of otic cells in RA-treated embryos at later stages. Our treatment of embryos with 20 nM RA led to the massive ectopic expression of pax8 around the anterior neural plate border, similar to the results published by Phillips et al.(Phillips et al., 2001). However, in contrast to Phillips et al., who reported that 20-30% of RA-treated embryos produce ectopic morphologically visible otic vesicles at the anterior limit of the head, we observed only randomly distributed small patches of ectopic otic cells. Phillips et al. showed that the anterior neural plate is strongly posteriorized after treatment with a teratogenic dose of RA,which leads to the anterior expansion of hindbrain fgf3 and fgf8 expression domains. Wnt signaling is presumably also shifted anteriorly in these RA-treated embryos, maintaining the fate of the ectopic otic cells. In wild-type embryos, however, Wnt signaling is absent in the early neural plate anterior to the future midbrain-hindbrain boundary(Wilson and Houart, 2004). Because our conditions to generate excess RA signaling (20 nM) did not apparently change anterior neural plate fate, pax8-positive cells anterior to the future midbrain-hindbrain boundary presumably do not receive Wnt signaling, which is essential for the maintenance of otic fates after induction. Furthermore, pathways to generate other placode-derived sensory organs were presumably intact under our experiments, which might also contribute to fewer ectopic otic cells in RA-treated embryos at later stages.

We found that the application of 20 nM RA led to the expression of foxi1 within the entire preplacodal domain. However, we did not observe any expansion of the otic tissue at the expense of other sensory organs, except for the lens. Interestingly, it was recently shown that the entire preplacodal domain is initially specified as lens tissue, implying that lens fate is a default state of the preplacodal territory and that subsequent development requires repression of lens fate in prospective non-lens domains(Bailey et al., 2006). By promoting otic fate in response to Fgf signaling, Foxi1 in the preotic region of wild-type embryos and expanded Foxi1 expression in RA-treated embryos might be involved in lens fate repression. Consistent with this interpretation,microarray analysis covering approximately 20,000 zebrafish genes showed that only seven genes, including pax6b, are significantly repressed after overexpression of Foxi1 (Yan et al.,2006). However, it is also likely that the reduction or loss of lens tissue is not due to misexpression of foxi1 or to a posteriorization of the preplacodal domain, but rather due to the ectopic expression of pax8. Reciprocal inhibition between the paired-domain proteins Pax2 and Pax6 has been implicated in the specification of mammalian eye territories and in formation of the di-mesencephalic boundary(Schwarz et al., 2000; Nakamura and Watanabe, 2005). Other studies have shown that both Pax2 and Pax6 can be converted from transcriptional activators to repressors via interaction with co-repressors of the Groucho protein family (Eberhard et al., 2000). Pax2 is highly related to Pax5 and Pax8, and gene replacement in the mouse has shown that Pax5 can functionally substitute for Pax2, indicating that Pax2, Pax5 and Pax8 proteins are biochemically interchangeable (Pfeffer et al., 1998; Bouchard et al.,2000). Combined and redundant gene function has also been shown for Pax2 and Pax8 during development of the mouse urogenital system and the zebrafish inner ear (Bouchard et al.,2002; Hans et al.,2004; Mackereth et al.,2005). Further investigation of the roles of Foxi1 and Pax2 or Pax8 in otic induction and in the repression of other fates will help clarify the functional similarities and differences among these factors.

We wish to thank Sandra Brown and Lauren Clancey for technical assistance;Michael Brand for sharing unpublished results; Lisa Maves for critical reading of the manuscript; and Gunnar Valdimarsson for materials. This work was supported by NIH grants DC04186 and HD22486. S.H. is a recipient of a Feodor Lynen fellowship of the Alexander von Humboldt foundation.