Progressive restriction of otic fate: the role of FGF and Wnt in resolving inner ear potential

The inner ear of all vertebrates develops adjacent to the hindbrain from a thickened disk of epithelium known as the otic placode. The otic placode forms in stereotypical locations in almost all vertebrates, forming in the non-neural ectoderm just rostral to the level of the first somite and on either side of the hindbrain. Otic induction is mediated by the localised action of signalling factors (Baker and Bronner-Fraser, 2001). In both chick and mouse, the inner ear is initiated by the action of localised fibroblast growth factor (FGF) signals(Kil et al., 2005; Ladher et al., 2005). In the case of chick, the combined actions of Fgf3 and Fgf19 acting from the mesoderm, initiate a cascade of events that ultimately result in the induction of the inner ear. One important step in this cascade is the induction in the overlying neural plate of the Wnt signalling protein Wnt8c(Ladher et al., 2000). FGF and Wnt signalling are then thought to induce otic fate in adjacent, non-neural ectoderm. However, these studies do not resolve the temporal requirements for these interactions nor do they assign individual roles to each signal.

The role of Wnt signalling in otic induction has been subject to considerable speculation. Experiments from zebrafish seem to discount a direct role for Wnt signalling in the early specification of the inner ear(Phillips et al., 2004),although data from the mouse contradict this view(Ohyama et al., 2006). Modulation of canonical Wnt signalling in the mouse suggests that this signalling actually directs a fate choice between otic and epidermis tissue within a progenitor field. This progenitor region, which was termed the`pre-otic field', is marked by the expression of Pax2. However, data from both genetic labelling in the mouse and vital dye labelling in the chick show that the Pax2 expression domain should, more properly, be considered to encompass the inner ear and the epibranchial placodes, a group of neurogenic placodes that will give rise to the geniculate, petrosal and nodose ganglia (Ohyama and Groves,2004b; Streit,2002). To reflect these derivatives, and to prevent any ambiguity with a definition for `pre-otic' that could be understood as the region of the embryo rostral to the otic placode, we have termed the Pax2-expressing progenitor domain the otic-epibranchial progenitor domain (OEPD). Although the mechanisms of OEPD induction are not fully understood, in zebrafish an FGF signal has been proposed as being sufficient and necessary for the induction of both otic and epibranchial placodes(Nechiporuk et al., 2007; Nikaido et al., 2007; Sun et al., 2007). Such a common precursor domain may also reflect a common evolutionary relationship between the inner ear and epibranchial derivatives(Baker et al., 2008).

Inner ear cells must be specified from the precursor domain, and the view that otic fate restriction is progressive is suggested by careful examination of the timing of particular genes within the chick otic placode. Pax2is expressed at around the 4 somite stage (ss). However, the otic placode is not actually morphologically visible until 8/9ss. At this stage, inner ear markers, such as Nkx5.1, Soho1 and Bmp7, begin to be expressed (Baker and Bronner-Fraser,2001; Groves and Bronner-Fraser, 2000). Experiments that isolate the otic placode from potential extrinsic signals suggest a further dimension; the progressive restriction of inner ear fate is due to changes in either the nature or the duration of signalling interactions. Otic ectoderm isolated at 5ss and cultured for 24 hours will express Pax2; however, it will not express later markers, such as Bmp7. Only otic explants isolated after 7-8ss express both Pax2 and Bmp7(Groves and Bronner-Fraser,2000). The relevance of such a restriction is not clear, and indeed these experiments do not identify the nature of the signals or the mechanisms by which they act.

In this report, we provide definitive evidence for the progressive restriction of developmental fate in inner ear induction. Overexpression and inhibition studies allow us to determine the clear functional significance of FGF and Wnt signalling during early inner ear induction. We find that an initial pulse of FGF establishes a mitotically active progenitor domain to both the inner ear and epibranchial placodes, the OEPD. Subsequently, the attenuation of FGF together with the action of canonical Wnt signalling allow the medial region of the OEPD to commit to an inner ear fate. Conversely,epibranchial placode differentiation is enhanced by continued FGF signalling but inhibited by canonical Wnt signalling. Thus, the progressive restriction of inner ear and epibranchial potential results from the interplay of the FGF and Wnt signals that form the basis of signalling checkpoints that determine the correct and orderly differentiation of the inner ear and the epibranchial placodes.

Fertilised hens' eggs (Shiroyama Farms, Kanagawa, Japan) were incubated in a humidified chamber at 38°C. Embryos were staged by counting somites, or,if prior to somitogenesis, by using the stage series of Hamburger and Hamilton(Hamburger and Hamilton,1992).

Embryos at 1ss/HH7-16ss/HH12 were removed and washed in Ringer's solution. Two types of explant were performed. In both cases, the inner ear region was dissected by making two transverse cuts, just rostral to the first somite and just caudal to rhombomere 3. Presumptive otic explants were dissected away from neural ectoderm and underlying paraxial mesendoderm. Presumptive otic regions are similar but were just bisected at the midline. The explanted tissue was transferred immediately to a chilled 10-μl drop of collagen. Once set, collagen drops were flooded with pre-warmed medium [DMEM+10%knockout serum replacement (KSR; Invitrogen, CA]. Otic regions were additionally incubated in media supplemented with 100 ng/ml recombinant human DKK1 (R&D Systems). Explants were grown for 1 or 7 days in a humidified CO₂ incubator, at which time they were rinsed in PBS and fixed in 4% PFA for one hour.

Collagen drops were processed for in situ hybridisation and immunohistochemistry as described previously(Wright et al., 2004).

Knockdown of Fgf3 and Fgf19 was performed using shRNA-encoding constructs. Briefly, putative shRNA sequences designed against 20 bases starting from +251 and +324 of Fgf3 and Fgf19,respectively were inserted into pSilencer 1.0 (Ambion). In addition, a scrambled oligonucleotide based on the Fgf19 +324 sequence was inserted in pSilencer (shScrambled) to act as a control for non-specific effects. To overexpress Fgf3 and Fgf19, coding regions were cloned downstream of the Ef1α promoter. A stabilised, and thus, constitutively active β-catenin and a construct expressing mouse Dickkopf1 (Dkk1) were the kind gift of Drs Fumi Kubo and Shinichi Nakagawa(RIKEN ASI, Wako, Japan).

DNA was unilaterally co-electroporated with a tracer (mCherry) into either the anterior streak of HH4 embryos or the ectoderm of HH5 embryos that were cultured ex ovo (Uchikawa et al.,2003). The embryos were then incubated in a humidified CO₂ incubator at 37°C for 10-49 hours.

Embryos were fixed in 4% paraformaldehyde and rinsed in PBS. The following probes were used for whole-mount in situ hybridisation: Fgf3, Fgf19,Nkx5.1, Pax2, Phox2b and Soho1. These probes were as described previously (Begbie et al.,2002; Ladher et al.,2000). A probe recognising Foxi2 was obtained through the BBSRC chick EST database (ChEST 884m4)(Boardman et al., 2002). In situ hybridisation was performed as described previously(Ladher et al., 2005). Some stained embryos were cryosectioned.

Double-fluorescent in situ hybridisation was based on published protocols(Denkers et al., 2004). Digoxigenin-labelled probes were detected using an alkaline phosphatase-conjugated antibody and revealed using VectorRed (Vector Laboratories). The second probe was labelled using fluorescein. This was detected using a peroxidase-conjugated antibody, and revealed using a fluorescein-tyramide kit (Perkin Elmer).

The following antibodies were used: anti-hair-cell-antigen (a kind gift of Prof. Guy Richardson, University of Sussex, UK), anti-phospho-histone H3(Upstate) and anti-Ds-Red (Living Colors), which was used to detect mCherry protein.

Collagen cultures were analysed after in situ hybridisation for Soho1 and immunohistochemistry for hair-cell antigen (HCA). Explants were grouped based on stage and the proportion of positive staining. We used statistical tests (Smith's statistical package) for two samples with two possible probabilities (positive or negative) to determine the earliest stage at which the expression of Soho1 and the development of inner ear hair cells were autonomous.

Cell counts for phospho-histone H3 (PHH3)-positive cells were performed after imaging whole embryos, using Adobe Photoshop software. Briefly, the OEPD was first defined as being an area just rostral to the first somite adjacent to the hindbrain, and that extended rostrally for three somite lengths. All subsequent counts were made of positive cells within this area. PHH3-positive cells were counted and the normal difference between the left and right sides of 10 control embryos between 10ss and 13ss calculated. Such values provide a strong control for differences among embryos. The difference in PHH3-positive cells between left, control sides and right, electroporated sides of six pEF-Fgf electroporated embryos were similarly calculated. P-values were calculated using Student's t-test.

We have previously shown that FGF and Wnt signalling induce the inner ear;however, we were aware of a lack of temporal resolution in these studies(Ladher et al., 2000). To understand individual roles that FGF and Wnt play during inner ear development, the particular steps in the developmental program of the inner ear required clarification. Studies had hinted at multiple states during inner ear induction. Explants of presumptive otic tissue are competent to express Pax2, but not Bmp7 when isolated at 5ss; explants can express both when isolated at 7-8ss(Groves and Bronner-Fraser,2000). We hypothesised that these studies reflected a progressive commitment to otic differentiation. Thus, to assess whether presumptive otic ectoderm, isolated at 5ss, is committed to a definitive inner ear fate, we tested the ability of this tissue to express the otic marker Soho1and to develop inner ear hair cells, the mechanosensory receptor of the inner ear (Fig. 1A)(Bartolami et al., 1991; Deitcher et al., 1994). Few(18%) presumptive otic explants isolated at 5ss and cultured for 24 hours expressed Soho1 (Fig. 1B,C). Similarly, few presumptive otic explants were able to undergo differentiation, as judged by immunoreactivity to HCA after seven days of culture (Fig. 1B,D). These numbers were not statistically significant. Thus, we concluded that, although 5ss explants are able to express Pax2, they lack the ability to develop definitive otic character. By contrast, a significant number of presumptive otic explants isolated at 7-8ss (HH9) showed expression of Soho1 and could develop hair cells(Fig. 1B,E,F).

The fibroblast growth factor family members Fgf3 and Fgf19 act from caudal cephalic paraxial mesoderm in the chick to induce otic ectoderm. We have previously shown that Fgf19 is downregulated from the mesoderm and neuroectoderm at 8ss (Ladher et al.,2000). Soon after, at 9ss, we find that mesodermal Fgf3expression is also downregulated. The coincidence of the ability of presumptive otic ectoderm to differentiate with this downregulation led us to suspect that the lowering of FGF expression was a pre-requisite for otic differentiation. To test the necessity of FGF attenuation for otic commitment,we introduced constructs of Fgf3 and Fgf19 driven by the Ef1α promotor (pEF-Fgf3 and pEF-Fgf19 respectively,collectively called pEF-Fgf). This promotor acts constitutively to drive sustained FGF expression in the early chick embryo.

We first verified the function of the pEF-Fgf constructs by assessing their effects on Pax2 expression. Several groups have identified a role for FGF signalling in the induction of Pax2expression in the OEPD of zebrafish(Nechiporuk et al., 2005; Nikaido et al., 2007; Sun et al., 2007). We thus unilaterally introduced pEF-Fgf3, pEF-Fgf19, or both, into the anterior streak region of HH4 embryos by electroporation(Fig. 2A-C). Such electroporation targets the mesoderm as well as the neural ectoderm adjacent to the presumptive otic region (Iimura et al., 2007; Psychoyos and Stern, 1996). Overexpression of pEF-Fgf constructs caused expansion of the normal OEPD Pax2-expression domain(Fig. 2D-I). This expansion was seen as early as 8-14 hours after electroporation, when embryos were at 9-16ss. The effect of electroporating either pEF-Fgf3 or pEF-Fgf19 alone was indistinguishable; however, introducing both constructs caused a larger expansion and stronger expression; we therefore presumed that the two ligands are additive during OEPD induction. These results are consistent with those reported in zebrafish.

We next investigated Soho1 expression in response to sustained FGF action. As shown in the preceding section, Soho1 expression correlates with the ability of the otic placode to differentiate, and is thus a marker for definitive otic specification(Fig. 1B). In contrast to Pax2 expansion, the otic expression of Soho1 was reduced or absent in embryos electroporated with pEF-Fgf3, pEF-Fgf19,or both (Fig. 2J-O). To verify this, we tested a second marker of the inner ear, Nkx5.1. This gene,like Soho1, is expressed once the inner ear has become morphologically apparent (Adamska et al.,2001). Like Soho1, expression of Nkx5.1 was similarly reduced in response to sustained FGF signalling(Fig. 2P-S). These results strongly suggest that definitive otic specification was inhibited in embryos when FGF signalling was sustained, and that, during normal development, FGF attenuation is necessary for otic differentiation.

The expansion of the OEPD and the block in otic commitment caused by sustained FGF expression could be the result of the maintenance of proliferation within the OEPD; the failure to exit cell cycle in this case may result in a block in commitment. To test this, we used an antibody to the serine-10 phospho-form of histone H3 (PHH3), which labels mitotic cells between late G2 to anaphase (Hans and Dimitrov, 2001; Hendzel et al., 1997). In normal, non-electroporated embryos, there is a slight difference in the number of PHH3-positive cells between the left and right sides of the embryo. On average there are 3% fewer PHH3-positive cells on the right side of non-electroporated embryos. By contrast, embryos unilaterally electroporated with pEF-Fgf constructs, have 23% more mitotic cells on the right, electroporated, side than on the left,non-electroporated side (Fig. 3). These data strongly suggest that one of the functions of FGF during early OEPD induction is to maintain the proliferation of at least a portion of these progenitors.

Many studies have described a role for FGF signalling in otic induction(reviewed by Schimmang, 2007),thus the finding that sustained FGF signalling actually inhibited otic differentiation was surprising. In many of these studies, suppression of FGF expression was used to show necessity for inner ear development. We hypothesised that, in these studies, FGF suppression actually inhibited inner ear development by blocking OEPD formation. To confirm this, we investigated the effect of removing Fgf3 and Fgf19 on OEPD induction and otic commitment. Constructs encoding short-hairpin interfering RNA were designed to knockdown Fgf3 and Fgf19 (known as shFgf3 and shFgf19, respectively, and, collectively, as shFgf). Electroporation of either shFgf3 or shFgf19caused a reduction of Fgf3 and Fgf19 expression,respectively (Fig. 4A,B). Such electroporated embryos were assessed after 12 hours for Pax2expression (Fig. 4C-F) and after 24 hours for Soho1 expression(Fig. 4G-J). Knockdown of both Fgf3 and Fgf19 substantially reduced the expression of both Pax2 and Soho1 (Fig. 4F,J). Reducing the levels of either Fgf3 or Fgf19 individually had some effect on OEPD induction(Fig. 4D,E) and otic development (Fig. 4H,I),indicating that during normal development the two act redundantly. These data show that FGF signalling is necessary for OEPD induction and, consequently,otic fate.

The early expression of Pax2 defines the otic-epibranchial precursor domain (OEPD), encompassing the otic and epibranchial placodes(Ohyama and Groves, 2004b; Streit, 2002). Even though late otic markers are inhibited by sustained FGF signalling, Pax2 is not. Thus, it is possible that sustained FGF signalling, as well as maintaining proliferation, also altered specification within the OEPD by converting the presumptive otic portion into non-otic cell types arising from the OEPD. Thus, we assessed Foxi2 expression. In chick, as has been described in mouse (Ohyama and Groves,2004a), Foxi2 expression is excluded from the otic part of the OEPD (Fig. 5A). At 22ss, Foxi2 is detected in epibranchial placodes. Thus, we hypothesised that early Foxi2 expression overlapped with Pax2 to label non-otic OEPD derivatives. This was confirmed using double fluorescence in situ hybridisation (Fig. 5D). Such analysis identified regions of the ectoderm that are either Pax2or Foxi2 positive, as well as a domain of Pax2 and Foxi2 co-expression at the periphery of the normal Pax2expression domain.

Downregulation of Fgf3 and Fgf19 using shFgfconstructs, resulted in a diminution of Foxi2 expression(Fig. 5B). By contrast,electroporation of pEF-Fgf constructs resulted in slightly stronger Foxi2 expression (Fig. 5C), and the region of Pax2/Foxi2 overlap was slightly broader (Fig. 5F). However, we noted that Foxi2 expression was still excluded from the presumptive otic region. These data suggest that FGF signalling is required for the development of the non-otic portion of the OEPD. In addition,sustained expression does not force the presumptive otic ectoderm to adopt a non-otic fate.

We next asked whether, like the otic portion of the OEPD, sustained FGF signalling also inhibited the differentiation of the non-otic portion of the OEPD. To investigate this, we assessed the development of the epibranchial-derived neurons. At 13ss, Pax2 expression begins to partition into separate otic and epibranchial regions(Fig. 5G). After pEF-Fgf electroporation, epibranchial Pax2 expression was expanded (Fig. 5I). We note that segregation of Pax2 into an `otic' domain also occurred normally after pEF-Fgf electroporation, despite the inhibition of otic commitment. To further analyse the effect of sustained FGF signalling upon epibranchial differentiation, we investigated the expression of Phox2b. This homeodomain protein is expressed in the sensory neurons of the epibranchial placodes, with expression apparent in all epibranchial placodes by 23ss/HH14+ of development(Begbie et al., 2002)(Fig. 5J). In response to sustained FGF signalling, epibranchial neurogenesis, as marked by Phox2b expression, is increased(Fig. 5L).

To test the necessity of FGF signalling for epibranchial development,knockdown constructs were electroporated to reduce Fgf3 and Fgf19 expression. These resulted in a marked reduction in epibranchial placodes, as revealed by Pax2 and Phox2bexpression (Fig. 5H,K). Together these data suggest that, like the otic placode, FGF signalling is necessary for epibranchial development, most likely through the induction of the common progenitor domain. However, in contrast to otic development,sustained FGF signalling stimulates epibranchial neurogenesis.

FGF signalling is sufficient and necessary for the early expression of Pax2 in the OEPD, but it has profound effects on the differentiation of its derivatives, repressing Soho1 and Nkx5.1, markers of otic commitment, and stimulating Phox2b, a marker for epibranchial neurogenesis. However, as evidenced by the continued exclusion of Foxi2 from the presumptive otic region(Fig. 5C), sustained FGF expression does not alter the patterning of the otic-epibranchial progenitor domain. This suggests that additional signals are necessary to direct definitive otic specification.

Wnt signalling is strongly implicated in otic development(Jayasena et al., 2008; Ohyama et al., 2006; Riccomagno et al., 2005). To test whether Wnt signalling partitions the OEPD into otic and epibranchial fates, we used a stabilised mutant of β-catenin. Canonical Wnt signalling is mediated through the repression of cytoplasmic β-catenin degradation. Thus, the stabilised β-catenin acts as a constitutively active effector of canonical Wnt signalling. We introduced this construct into the ectoderm of HH5/HH6 chick embryos and then allowed these embryos to develop until 10-15ss.

Ectodermal overexpression of constitutively active (CA) β-catenin did not alter the expression of Soho1(Fig. 6A,B). Similarly, the dorsal expression domain of Nkx5.1, an additonal inner ear marker,was unchanged (Fig. 6C,D). We investigated the effect of Wnt activation on OEPD development using Pax2 expression. At 7ss, Pax2 expression is unaffected by the overexpression of CA-β-catenin(Fig. 6E-G); however, by 13ss,the Pax2 expression domain was more rounded, without the typical lateral flares (Fig. 6H-J). We postulated that this lateral domain represented the non-otic part of the OEPD. This was confirmed after observing the reduction of Foxi2 expression in these embryos (Fig. 6K-M). The reduction of lateral Pax2 OEPD expression and the loss of Foxi2 expression suggests that ectopic Wnt signalling might also inhibit the differentiation of non-otic regions of the OEPD. Expression of the epibranchial-derived neuronal marker Phox2b was reduced 36 hours after the electroporation of constitutively active β-catenin(Fig. 6N,O).

Stimulation of Wnt signalling by the introduction of activatedβ-catenin inhibits epibranchial placode development but has little or no effect on otic development. Although these data do not support the idea that Wnt signalling acts to apportion otic fate within OEPD, they do not rule out the possibility that Wnt acts permissively for inner ear commitment. Thus, we investigated whether Wnt signalling was necessary for inner ear commitment. Unilateral electroporation of the Wnt inhibitor Dkk1 into the ectoderm resulted in a dramatic reduction of Soho1 and Nkx5.1expression (Fig. 7A-D).

The above experiments indicated that Wnt signalling was necessary for otic commitment, but did not indicate whether Wnt signalling was necessary for OEPD induction. To verify this, we analysed Pax2 expression. We investigated the genesis of the Pax2 domain at several time points after Dkk1 overexpression. Nine hours after electroporation (7ss), Pax2 expression was unchanged in response to Wnt signalling inhibition (Fig. 7F-H). However, by 12 hours after electroporation (13ss), the expression domain was altered, missing the medial-most expression(Fig. 7I-K). Consistent with the downregulation of Soho1 and dorsal Nkx5.1, this medial portion is most likely to be the region fated to give rise to the inner ear.

To understand the effect of Wnt inhibition on patterning within the OEPD,we next analysed Foxi2 expression. Foxi2 is normally excluded from the presumptive otic region. After Dkk1 overexpression, the region of Foxi2 exclusion was reduced, but Foxi2 never invaded the whole of this presumptive otic portion of the OEPD(Fig. 7L-N). To determine whether the expansion of Foxi2 expression also correlated with an expansion of epibranchial neurogenesis, we investigated the expression of Phox2b. We could detect no clear difference in Dkk1 overexpressing embryos, although in a few cases the domain of epibranchial neurogenesis was slightly larger (n=4/11) (Fig. 7P). When considered together, these data suggest that canonical Wnt signalling is not required for the initial induction of the OEPD; however,Wnt signalling is necessary for otic commitment.

The induction and specification of otic and epibranchial fates are directed by the actions of FGF and Wnt signals from adjacent mesoderm and neural ectoderm. Here, we ascribe exact roles for these factors during otic and epibranchial induction. We show that FGF provides an initial signal that induces a progenitor domain to both. The subsequent commitment to inner ear or epibranchial fates depends, in part, on the action of Wnt signals and the continued function of FGF. More specifically, we make the unexpected observation that sustained FGF signalling is inhibitory to otic differentiation. Thus, whereas otic differentiation is instructed by the combination of canonical Wnt signalling and a downregulation of FGF signalling, epibranchial differentiation is stimulated by sustained FGF signalling and inhibited by Wnt signalling. The use of these multiple signals act as checkpoints to ensure the proper co-ordination of otic and epibranchial induction, guiding the transition between progenitor and commitment states and ensuring their correct localisation within the cranial ectoderm(Fig. 8).

A role for FGF signalling in early inner ear development has been firmly established (Adamska et al.,2001; Alvarez et al.,2003; Hans et al.,2007; Ladher et al.,2005; Ladher et al.,2000; Leger and Brand,2002; Liu et al.,2003; Maroon et al.,2002; Phillips et al.,2001; Vendrell et al.,2000; Wright and Mansour,2003; Zelarayan et al.,2007). More recent data from zebrafish suggest that FGF signalling additionally induces the epibranchial placodes(Nechiporuk et al., 2007; Nikaido et al., 2007; Sun et al., 2007). This is consistent with the data reported here, showing that FGF overexpression expands the precursor domain of both. Conversely, knockdown using a shRNA strategy results in repression of the OEPD and, consequently, a loss of both committed otic and epibranchial precursors. The idea that only early FGF signalling is sufficient for inner ear development is supported by studies using pharmacological inhibition. Here, Pax2 expression is inhibited if the inner ear is treated with the FGF inhibitor SU5402 prior to 5ss. If the treatment is performed between 5-8ss, Pax2 expression is unaffected(Martin and Groves, 2006).

An unexpected finding is the suppression of otic commitment when FGF expression is sustained. However, FGF signalling is under tight regulation;the mesodermal expression of Fgf3 and Fgf19 is rapidly downregulated at 7-8ss. Furthermore, sprouty2, an inhibitor of FGF signalling is expressed in presumptive inner ear tissue from 8ss(Chambers and Mason, 2000). Thus, the observation that otic commitment is suppressed when FGF expression is experimentally sustained is not so surprising. When considered with the behaviour of isolates of otic ectoderm, a clearer understanding of the role of FGF signalling during inner ear development emerges. Isolates express Pax2, but can neither express Soho1 nor show inner ear hair cell differentiation when removed from their embryonic environment at 5ss. During this time, FGF signalling induces the OEPD. Otic commitment only takes place at 7-8ss, concomitant with the downregulation of Fgf3 and Fgf19 expression, and the suppression of FGF signals by sprouty2.

The observation that the OEPD of FGF-expressing embryos contains a larger number of cells labelled with phospho-histone H3 suggests that mitotically active otic progenitors are maintained by FGF signalling. Thus, the attenuation of such signalling is necessary for cell cycle exit and,consequently, for the transit to a committed state that is able to differentiate definitive inner ear character. Interestingly, the downregulation of FGF expression can also be detected in the mouse. Fgf8 and Fgf10, which with Fgf3 are the putative murine inner ear inducers, are also downregulated in the peri-otic mesoderm at 8ss (Ladher et al., 2005). This raises the intriguing possibility that FGF attenuation might also be required in mice for otic commitment to take place.

A role for FGF in blocking otic commitment has not been previously described. Indeed, previous data appear to contradict this model. However, one caveat is that in many cases only early OEPD markers have been investigated(Leger and Brand, 2002; Maroon et al., 2002; Sun et al., 2007). Some experiments have investigated the effect of FGF overexpression on otic commitment and differentiation. In Xenopus, the implantation of Fgf2-soaked beads resulted in ectopic otocysts(Lombardo and Slack, 1998). In chick, the implantation of Fgf2-soaked beads resulted in larger otocysts(Adamska et al., 2001). Similarly, treatment of chick ectodermal explants with Fgf19-soaked beads resulted in the induction of Soho1 expression(Ladher et al., 2000). In all cases, it is likely that the position of the beads has shifted during development, and, when considered with protein half-life, it is possible that the tissue can adopt a committed otic fate when it escapes the influence of exogenous protein. A more stable method of gene transfer was performed in chick embryos using an Fgf3-expressing retrovirus(Vendrell et al., 2000). In these experiments, even though viral infection was widespread, ectopic otocysts formed adjacent to only some infected cells. A similar study overexpressed Fgf3 in the hindbrain adjacent to the neural tube. Again, ectopic otocysts appeared adjacent to the region of overexpression(Zelarayan et al., 2007). In both studies, only a single FGF was overexpressed, and it is possible, as we have shown in this study, that the sustained action of two or more Fgfs is necessary for efficient repression of otic commitment. Finally, a zebrafish transgenic line that expresses a heat-inducible Fgf8 construct showed larger,well-patterned, otocysts only when Fgf8 was overexpressed at the late gastrula stage (Hans et al., 2007). Such experiments might indicate a difference in the action of particular FGF molecules.

Similar to the effect of sustained FGF action, inhibition of Wnt signalling also results in the suppression of otic commitment. However, there are differences in the effect of each on Pax2 expression. As described,FGF overexpression expands the early OEPD domain. At later stages, and despite the inhibition of otic commitment, Pax2 expression is detected in the otic placode. By contrast, Wnt inhibition causes only a reduction of the medial portion of the Pax2 expression domain in the OEPD after 7-8ss,at the onset of otic specification. This requirement for Wnt signalling may be transient: treatment of otic regions with recombinant DKK1 protein does not inhibit otic formation after 9/10ss (S.F. and R.K.L., unpublished). The ectopic activation of canonical Wnt signalling using a stabilised,constitutively active β-catenin does not overtly affect otic development and differentiation; however, epibranchial development is affected, again only after 7-8ss (see below). These data strongly suggest that Wnt signalling is not necessary for OEPD induction, but is required for otic specification,acting after FGF signalling has established the OEPD.

Using conditionally active or mutant lines of β-catenin, Ohyama et al. proposed a model for mouse inner ear induction that strongly suggested an involvement of Wnt signalling in OEPD patterning(Ohyama et al., 2006). Similar to our data in the chick, these authors found that the removal of canonical Wnt signalling from the OEPD resulted in the loss of the inner ear; however,the absence of a detailed time course of the changes to the OEPD meant that the possibility that the expression of Pax2 and Pax8 (the markers that define the mouse OEPD) was initially normal was not ruled out.

As described above, sustained expression of Fgf3 and Fgf19 blocks otic differentiation. We suggest that otic differentiation is blocked because these cells remain in a proliferative state; however, in the absence of otic differentiation, does this region also develop a non-otic-OEPD epibranchial fate? Our analysis of the expression of Foxi2 and Phox2b suggests that it does not. Foxi2is normally excluded from the region of the OEPD fated to the otic lineage,and its co-expression with Pax2 marks the non-otic OEPD region. Even when FGF signalling is sustained, and otic differentiation inhibited, this region of Foxi2 exclusion is maintained. However, the region of overlap between Foxi2 and Pax2 is slightly broader, and epibranchial-derived neurogenesis (as marked by Phox2b) is slightly enhanced. These data suggest that early mesodermal FGF signalling is not involved in patterning the OEPD, instead a second phase of FGF signalling,probably from the endoderm, might permit epibranchial fate. This is suggested by recent data showing that endodermal Fgf3 is involved in zebrafish epibranchial development (Nechiporuk et al., 2005). It should be noted that additional signals also play a role in the development of the epibranchial placodes. Endodermal Bmp7 has been shown to stimulate neurogenesis of these precursors(Begbie et al., 1999).

Our data also show a negative role for Wnt signalling in the development of the epibranchial placodes. Constitutively active β-catenin overexpression results in the repression of epibranchial derivatives. Ohyama and colleagues have touched on the role of Wnt signalling in epibranchial development in the mouse (Ohyama et al., 2006). Similar to our data in chick, the stimulation of canonical Wnt signalling in mouse results in a reduction of epibranchial-derived precursors. However, the result of removing β-catenin in mouse is somewhat confusing: here it results in a reduction of epibranchial development. It is possible that the mouse and chick use different mechanisms for epibranchial development, and that a low basal level of Wnt signalling may be required for mouse epibranchial development. However, a TOP-gal reporter that detects canonical Wnt signalling was not active within epibranchial precursors(Ohyama et al., 2006; Riccomagno et al., 2005). This argues against canonical Wnt signalling acting within epibranchial placodes. Alternatively, the known participation of β-catenin in tight junctions could suggest that mutants in this gene have impaired epithelial integrity. In such a situation, the epibranchial placodes might be induced normally in mouse embryos lacking β-catenin; however, continued development and morphogenesis might be aberrant.

Our data suggest a sequential role for first FGF signalling and then Wnt signalling in the development of the inner ear. In this revised model, we propose that FGF signalling, from subjacent mesoderm and adjacent neural ectoderm, establishes a mitotically active progenitor domain, the OEPD. The OEPD is then influenced by Wnt signals from the neural tube. By acting on the OEPD, Wnt enables otic differentiation of a subset of OEPD cells, while repressing epibranchial development in others. These data point to a crucial role for canonical Wnt signalling in the lineage choice between otic and non-otic portions of the OEPD. However, an important question is whether other signals are also involved. The failure of constitutively active β-catenin to expand the otic domain could suggest a lateral signal that antagonises the pro-otic Wnt signal. Similarly, the failure of Wnt inhibition to cause the complete conversion of putative otic precursors to non-otic Foxi2-expressing cells, despite the downregulation of otic Pax2,Nkx5.1 and Soho1, may suggest that the lateral `anti-otic'signal regulates the expression of Foxi2, but that its limited range renders it insufficient to cause complete conversion of the OEPD to Foxi2-positive precursors. Thus, the progenitor region is the site of competing signals: a medial otic-promoting/epibranchial-repressing Wnt signal and a lateral, as yet unidentified, signal that promotes epibranchial development and is likely to repress otic development(Fig. 8). An important feature of this model is that the inner ear and epibranchial placodes can be positioned only at sites of intersecting permissive interactions. Furthermore,by controlling the timing of these interactions, the transition between cell states can be controlled, ensuring the correct balance between progenitor proliferation and cell-type differentiation.

We thank Marie Paschaki for patient advice during this study. We are also grateful to Tom Becker, Paul O'Neill and members of the Laboratory for Sensory Development for critical reading of this manuscript, and to Yoshiko Kondoh for technical support. We thank Fumi Kubo and Dr Shinichi Nakagawa for Dkk1 and constitutively active β-catenin constructs. This work was supported by a RIKEN CDB intramural grant, the CDB Director's fund and a JSPS pre-doctoral fellowship (S.F.).