Notch signaling augments the canonical Wnt pathway to specify the size of the otic placode

Inner ear development is an excellent example of how a Darwinian `organ of extreme perfection and complication' can arise from simple origins. The inner ear derives from a patch of thickened ectoderm, the otic placode, lying next to the posterior hindbrain. Signals that induce the otic placode are present in the hindbrain and cranial paraxial mesoderm, although the relative contribution of these tissues to the induction process varies between species(Barald and Kelley, 2004; Groves, 2005; Riley and Phillips, 2003; Torres and Giraldez, 1998). Members of the fibroblast growth factor (FGF) family play a crucial role in inducing the otic placode in all vertebrates examined(Friesel and Brown, 1992; Ladher et al., 2005; Leger and Brand, 2002; Mackereth et al., 2005; Maroon et al., 2002; Phillips et al., 2001; Vendrell et al., 2000; Wright et al., 2004; Wright and Mansour, 2003). FGF signaling induces the expression of genes, such as Pax2 and Pax8, in a broad region of cranial ectoderm stretching from rhombomeres 3 to 6 (Maroon et al.,2002; Martin and Groves,2006; Wright and Mansour,2003). Evidence from lineage tracing in chick(Streit, 2002) and mouse(Ohyama and Groves, 2004b; Ohyama et al., 2006) suggest that this broad Pax2<sup>+</sup> domain, which we have referred to as the `pre-otic field', contains cells fated to become otic and epibranchial placodes, as well as cranial epidermis.

We recently showed that Wnt signaling plays an important role in defining the size of the otic placode within this Pax2⁺ pre-otic field. Wnt signaling is activated in a medial subset of the Pax2⁺ domain closest to the hindbrain(Ohyama et al., 2006). Inactivation of Wnt signaling in Pax2⁺ cells by conditional deletion of the β-catenin gene (Ctnnb1; also known as Catnb) leads to a large reduction in the size of the otic placode and a corresponding expansion of cranial epidermis. Conversely, activation of Ctnnb1 in Pax2⁺ cells expands the otic placode at the expense of cranial epidermis (Ohyama et al., 2006). To date, however, it is not clear how Wnt signals direct cranial ectoderm towards an otic fate. It is possible that Lef/Tcf/β-catenin transcriptional complexes activated by Wnt signaling directly regulate otic genes. Alternatively, Wnt signals might act indirectly by upregulating short-range signals that partition cranial ectoderm into otic placode and epidermis.

There is growing evidence that Wnt and Notch signaling pathways co-operate during cell fate determination in many tissues(Crosnier et al., 2006; Estrach et al., 2006; Fre et al., 2005). Notch signaling plays various roles in patterning the inner ear, ranging from specification of neurons and prosensory patches to the generation of the stereotypical pattern of mechanosensory hair cells and supporting cells(Adam et al., 1998; Brooker et al., 2006; Daudet et al., 2007; Daudet and Lewis, 2005; Haddon et al., 1998; Kiernan et al., 2005; Lanford et al., 1999; Shi et al., 2005). Both the Notch1 receptor and several of its ligands, such as jagged 1(Jag1) and delta-like 1 (Dll1), are expressed in the otic placode from very early stages (Abello et al., 2007; Adam et al.,1998; Daudet et al.,2007; Groves and Bronner-Fraser, 2000; Haddon et al., 1998). Notch signaling might therefore also have an early function during otic placode development. We now provide evidence that elements of the Notch pathway are positively regulated by Wnt signaling, and that Notch1 signaling can in turn modulate the canonical Wnt signaling pathway. We also show that while some aspects of otic placode identity are regulated only by Wnt signals, other features of placodal differentiation can be regulated independently by Wnt or Notch pathways.

The following lines of mice were used in this study: Pax2-Cre(Ohyama and Groves, 2004b)(available from the Mutant Mouse Regional Resource Center; www.mmrrc.org/strains/10569/010569.html);conditionally activated Notch1 intracellular domain [cN1ICD(Murtaugh et al., 2003)]; Notch1-null mutants (Conlon et al., 1995); conditionally activated β-catenin Catnb^lox(ex3) [cAct(Harada et al., 1999)];conditional β-catenin^floxed mutants[β-cat-CKO (Brault et al.,2001)]; Tcf/Lef Wnt reporter(Mohamed et al., 2004);conditional Rbpj/Rbsuh mutants(Tanigaki et al., 2002); and a GFP-expressing Cre reporter [Z/EG(Novak et al., 2000)]. To generate cN1ICD animals, N1ICD^floxed homozygotes were crossed with Pax2-Cre animals. Age-matched heterozygotes and wild types were used as controls for Notch1 mutant embryos. Detailed mating strategies for cAct and β-cat-CKO mice have been described previously (Ohyama et al.,2006). To generate Notch1; cAct mutants, a line that was heterozygous for Notch1; Pax2-Cre was crossed to animals that were heterozygous for Notch1; cAct. To generate cN1ICD;β -cat-CKO mutants, a line that was heterozygous forβ -cat-null; Pax2-Cre was crossed to animals that were heterozygous for N1ICD and homozygous for a floxed allele ofβ-catenin. For each mutant genotype, at least three embryos were analyzed, except for Notch1; cAct mutants (n=2). All animal experiments were done in accordance with the guidelines of the institution's Animal Care and Use Committee.

Whole-mount in situ hybridization was performed as previously described(Ohyama et al., 2006). The following probes were used: Notch1 (Jeffrey Nye), Dll1(Achim Gossler), Jag1 (Tim Mitsiadu), Hes1 and Hes5(Ryoichiro Kageyama), lunatic fringe (Lfng; Thomas Vogt) and Wnt6 (Andrew McMahon). Probes for Pax2, Pax8, Foxi2, Dlx5,Krox20, Hoxb1, Fgf3, and Epha4 have been previously described(Ohyama et al., 2006). Embryos were embedded in 15% sucrose and 7% gelatin in phosphate-buffered saline(PBS), as previously described (Groves and Bronner-Fraser, 2000), and 15- to 30-μm thick sections cut using a Leica CM 1850 cryostat. Immunostaining and detection ofβ-galactosidase on cryostat sections and embryos was performed as previously described (Ohyama et al.,2006). The following primary antibodies were used: β-catenin(Zymed) at 1:200; activated Caspase-3 (R&D Systems) at 1:1000; green fluorescent protein (GFP) conjugated to fluorescein (Abcam) at 1:250 to 1:500;β-galactosidase (ICN/MP Biochemicals) at 1:100; jagged 1 (Jag1; Santa Cruz) at 1:50 to 1:100; Pax2 (Zymed) at 1:500; and phospho-histone-H3 (PH3;Upstate/Millipore) at 1:1000. Secondary goat anti-rabbit antibody conjugated to Alexa 594 (Molecular Probes) was used at 1:200. Sections were counterstained with the nuclear marker DAPI (Molecular Probes). All images were captured using a Zeiss Axiocam digital camera and Axiophot2 or M² Bio microscopes, and were processed using Adobe Photoshop CS software.

The thickened otic placode was defined as the two- to three-cell layer of ectoderm located adjacent to rhombomere 5/6 (as identified morphologically with DAPI staining and/or by lack of for Foxi2 expression). Quantifications of placode size were made from 15-μm serial sections from Notch1 mutants and age-matched control embryos. Length measurements were made using Image J software. To allow for direct comparisons along the anteroposterior (AP) axis of control and mutant mice, measurements were binned into five categories: 0-20% (being the most anterior sections), 21-40%,41-60%, 61-80% and 81-100% (being the most posterior sections). For a given genotype, each bin consisted of multiple sections from several embryos. The mean and standard error of the mean (s.e.m.) were calculated for each bin. Non-parametric Mann-Whitney U-tests were performed to test for significance between genotypes. The cranial region of Notch1 mutants was comparable in size to controls. To confirm this, we measured the dorsoventral (DV) length of the neural tube adjacent to the otic placode. The measurements were processed as described for the otic placode. We found no differences in neural tube length between Notch1 mutants and controls(data not shown). For average density measurements, serial 15-μm thick sections stained with DAPI and/or hybridized with Dlx5 or Foxi2 probes were used. The cell density for each section was calculated as follows: number of cells/μm²×500 and was pooled for each genotype.

Cell proliferation counts were performed as described previously(Ohyama et al., 2006). To account for variations in the staging of embryos, the mediolateral length of the otic cup or Wnt domain was standardized against the DV neural tube length adjacent to the otic cup and expressed as a percentage. Only mid-sections from otic cups were used for quantification and Student's t-tests were performed to test for significance between genotypes.

The pre-otic field destined to give rise to the otic placode and surrounding epidermis is marked by Pax2 expression from the 0 somite stage [0ss; E8 in the mouse (Ohyama and Groves, 2004b; Ohyama et al.,2006)]. Pax2 expression later becomes restricted to the otic placode, which is morphologically visible as a thickening patch of ectoderm next to rhombomeres 5 and 6 from the 8ss (E8.5) onwards(Ohyama and Groves, 2004b). To see whether elements of the Notch pathway were expressed at an appropriate place and time to participate in otic placode induction, we compared the expression patterns of Notch1, Jag1, Dll1, Hes1, Lfng and Hes5 to Pax2 (Fig. 1). At 0-1ss, no Notch pathway transcripts were detected in the pre-otic field (Fig. 1B,C). Onset of Notch1 expression was observed as early as the 4ss, becoming stronger by 5-7ss (Fig. 1C). Scattered cells in the anterior Pax2 domain adjacent to the neural tube expressed the Notch ligand Jag1 from around 5ss, although posterior cells did not express Jag1 until 8-9ss (arrowhead, Fig. 1C,D). Dll1 was also expressed adjacent to the neural tube at 4-5ss(Fig. 1B,C), and was restricted to the otic placode from 9ss (Fig. 1C). Between 12 and 14ss, Dll1 expression was gradually restricted to differentiating neuroblasts in the anteroventral placode (data not shown) (Adam et al., 1998). We were unable to detect Lfng at the pre-otic field and placode stages in whole mounts (data not shown), although, at later otic vesicle stages, Lfng is expressed in the anteroventral portion of the otic cup destined to produce the vestibuloacoustic ganglion and the utricular and saccular maculae (Morsli et al.,1998; Raft et al.,2004).

Hes1 and Hes5 are effectors of the Notch pathway that function in many processes, including in the regulation of cell fate decisions(Bray, 1998; Kageyama et al., 2007; Lai, 2004). Hes1expression was scattered throughout the pre-otic field and by 10-11ss was restricted to the otic placode (Fig. 1C). We found no evidence for Hes5 expression in the pre-otic field (data not shown). These data suggest that at least some transcriptional targets of the Notch pathway are expressed during early phases of otic placode development.

We previously used a transgenic Wnt reporter mouse line(Mohamed et al., 2004) to show that the canonical Wnt signaling pathway is activated in the pre-otic field between 3 and 5ss (Ohyama et al.,2006). Several Wnt family members are expressed in an appropriate location to trigger the observed Wnt reporter activity - for example, Wnt8 is expressed in rhombomere 4(Ohyama et al., 2006). We also observed Wnt6 expression in the Pax2⁺ pre-otic field at 0ss. It continues to be expressed in the neural folds at 5-7ss and in the dorsal-most region of the otic placode at 11ss(Fig. 1C)(Lillevali et al., 2006). As the onset of Notch pathway gene expression closely corresponded to Wnt reporter activity in the pre-otic field(Fig. 1B,C), we hypothesized that the Notch1 pathway might interact with the canonical Wnt signaling pathway in mediating the fate decision between otic placode and epidermis.

Previous studies suggest that Notch pathway components can be regulated byβ-catenin (e.g. Estrach et al.,2006; Katoh and Katoh,2006). We therefore examined expression of Notch pathway genes in embryos carrying gain- or loss-of-function mutations of the canonical Wnt pathway in the Pax2⁺ pre-otic field. We crossed Pax2-Cre transgenic mice (Ohyama and Groves, 2004b) with mice in which β-catenin is constitutively activated in Cre-expressing cells [cAct(Harada et al., 1999)], and examined expression of Jag1, Notch1 and Hes1. In cAct mutants, Jag1 expression was expanded ventrally to the level of the pharynx at the 9-10ss (bracket, Fig. 2A) and this ectopic expression continued until at least E9.0. Jag1 is thought to be a direct target of β-catenin, as its promoter region contains five, three and six consensus Tcf/Lef-binding sites in mouse, human and rat, respectively(Estrach et al., 2006; Katoh and Katoh, 2006). The domain of Notch1 and Hes1 expression was also expanded,although only after a delay (from the 14-15ss; Fig. 2A, brackets). Such a delayed induction of Notch1 and Hes1 relative to Jag1 has also been observed in epidermis in which β-catenin is activated (Ambler and Watt,2007). Other Notch pathway genes, such as Dll1, Hes5 and Lfng, were not expressed in cAct mutants (data not shown).

To determine whether Wnt signaling is necessary for the expression of Notch pathway components, we analyzed the expression of Jag1, Notch1 and Hes1 in the Pax2⁺ pre-otic field of mice lackingβ-catenin [β-cat-CKO(Brault et al., 2001)]. The Jag1 domain was significantly reduced at 10-11ss, as we have previously reported for Pax2 and Pax8(Ohyama et al., 2006). Many cells within the vestigial β-cat-CKO otic vesicle wereβ-catenin^-;Jag1^-, suggesting that Wnt signaling is directly responsible for Jag1 induction in the placode (bracket, Fig. 2B). Close examination of the vestigial mutant vesicles at E9-E9.5 revealed that cells expressing Jag1 were β-catenin⁺ and had therefore failed to undergo Cre recombination (Fig. 2B). The domains of Notch1 and Hes1 were also significantly reduced, although the resolution of the whole-mount in situ technique made it difficult to determine whether all Notch1- and Hes1-expressing cells also expressed β-catenin protein(Fig. 2B).

The expression of Notch pathway genes in the pre-otic field and otic placode, together with the regulation of these genes by Wnt signaling suggested that Notch signaling might participate in the fate decision between otic placode and epidermis. To test this, we conditionally activated Notch1 in the pre-otic field using mice in which the active, intracellular domain of Notch1 receptor (N1ICD) was knocked into the ROSA26 locus with a transcriptional STOP cassette flanked by LoxP sites(Murtaugh et al., 2003). We drove expression of N1ICD in the Pax2⁺ pre-otic field using Pax2-Cre mice (Ohyama and Groves, 2004b). The Pax2-Cre mouse line expresses Cre recombinase in the midbrain and rhombomere 1 (R1) of the hindbrain(Ohyama and Groves, 2004b). Conditionally activated N1ICD (cN1ICD) mutants displayed an open neural tube phenotype at the level of the midbrain-R1 region, which is likely to result from overproliferation of precursor cells induced by Notch activation. However, the patterning of the posterior hindbrain next to the ear was normal at E8.5-E9.5, based on the expression of Hoxb1 (rhombomere 4), Fgf3 (rhombomeres 5 and 6), Epha4 and Krox20(rhombomeres 3 and 5; see Fig. S1B in the supplementary material), suggesting that any otic placode phenotype in cN1ICD mutants is not due to changes in the adjacent hindbrain.

We examined embryos inheriting both the Cre-inducible N1ICD and Pax2-Cre transgenes for otic placode and epidermal markers. The N1ICD transgene also harbors an IRES-nGFP sequence, allowing the visualization of N1ICD-expressing cells by GFP fluorescence. GFP-expressing cells were observed throughout the pre-otic field from 5-6ss(data not shown). E9.5 cN1ICD mutant embryos displayed GFP expression throughout a thickened placode-like structure that expanded to the level of the ventral pharynx (see Fig. S1A in the supplementary material). The analysis of cN1ICD embryos at 10-11ss revealed that the Pax8 domain was expanded ventrally (arrowheads and brackets, Fig. 3A,A′). We previously showed that Foxi2 is an epidermal marker expressed in a complementary manner to Pax2 and Pax8 during otic placode development (Ohyama and Groves,2004a). By E8.75-E9, Foxi2 expression was reduced dramatically in cN1ICD mutants when compared with controls (dotted outline and brackets, Fig. 3C,C′), complementing the expansion of the thickened epidermis. To determine whether cell proliferation was responsible for the expanded placode, we examined expression of the M-phase marker phosphohistone-3 (PH3) in cN1ICD embryos (n=10 placodes) and control embryos (n=8 placodes) produced by crossing the Pax2-Cre line with the Cre-inducible Z/EG GFP-expressing line (Novak et al., 2000). We saw no significant differences in total PH3⁺ or PH3⁺;GFP⁺ cell counts per section (see Fig. S1D in the supplementary material).

The expansion of Pax8 at the expense of Foxi2 in cN1ICD embryos is strikingly similar to that seen in embryos in which the canonical Wnt pathway is activated [cAct embryos(Ohyama et al., 2006)]. However, in contrast to cAct embryos, we saw only a modest expansion of the Pax2 domain (bracket and arrowhead, Fig. 3B,B′; see also Fig. S1C in the supplementary material), and no expansion of the otic markers Gbx2 or Sox9 (Fig. 3D). Finally, a marker of the dorsolateral otocyst, Hmx3,which does not require either Wnt or Hedgehog signaling for its expression(Ohyama et al., 2006; Riccomagno et al., 2002) was also not expanded in cN1ICD mutants(Fig. 3D).

These results suggest that different aspects of otic placode development are differentially regulated by Wnt and Notch signaling. Placode markers such as Pax2, Gbx2 and Sox9 appear to be regulated by Wnt signaling (Ohayama et al., 2006; Saint-Germain et al., 2004),but not Notch signaling, whereas markers such as Pax8, the morphological thickening of epithelium and the repression of the epidermal marker Foxi2 can be regulated by both Notch and Wnt signals. To determine whether Notch signaling can regulate these markers independently of Wnt signaling, we analyzed β-cat-CKO;cN1ICD mutant embryos in which β-catenin was inactivated and Notch1ICD was activated throughout the pre-otic field. Mutant embryos displayed greatly expanded regions of thickened placode-like epithelium that expressed both Pax8 and Jag1 (Fig. 3E). This expanded region of thickened epithelium was largely devoid of Foxi2expression (Fig. 3E), although occasional Foxi2⁺ patches of cells could sometimes be detected. These results show that Notch and Wnt signals can independently regulate some aspects of otic placode development.

Our results show that Notch1 activation throughout the Pax2⁺ pre-otic field expands some otic placode markers at the expense of epidermis. In complementary experiments, we examined Notch1 mutants, in which a substantial portion of the Notch1gene is deleted [amino acids 1056-2049(Conlon et al., 1995)]. This deletion encompasses RAM and Ankyrin repeats required for RBPJκsignaling (Conlon et al.,1995; Fortini and Artavanis-Tsakonas, 1994; Kurooka et al., 1998a; Kurooka et al., 1998b; Lamar et al., 2001; Nam et al., 2003; Tani et al., 2001). We confirmed that posterior hindbrain patterning was normal in Notch1mutants by assaying for Hoxb1, Fgf3 and Krox20 expression(see Fig. S2A in the supplementary material). All three genes were expressed normally, suggesting that any defects observed in otic placode development are due to deficiency in Notch1 signaling in the placode, rather than in the hindbrain.

To determine whether Notch signaling was necessary for the expression of otic markers, we examined Pax2 and Pax8 expression in Notch1 mutants. By 9-11ss there was a dramatic downregulation of Pax2 expression in mutants in both the otic region and the epibranchial placodes (Fig. 4A). Although the anteroposterior limits of Pax8expression in the otic region was reduced, expression in the hyoid arch was relatively unchanged (Fig. 4B). In Notch1 mutant whole mounts, the limits of the Pax2 and Pax8 domains in the anteroposterior axis were reduced (brackets, Fig. 4). Sections through Notch1 mutants also revealed a reduction in the mediolateral extent of Pax2 and Pax8 expression (brackets, Fig. 4).

The reduction in the size of the otic placode in Notch1 mutants may result from increased apoptosis, increased cell density or a change in cell fate. We measured the size of the placode by examining Foxi2expression, which is precisely excluded from the thickened placode region. The Notch1 mutant otic placode was indeed smaller at 9-13ss, on the basis of Foxi2 expression (dotted outline, Fig. 5A,B). We compared the mediolateral extent of the thickened otic placode in Notch1 mutants and controls at 9-11ss and 12-13ss (see Materials and methods; Fig. 5C,D) Notch1mutants (9-11ss, n=25 placodes; 12-13ss, n=10 placodes) had significantly smaller placodes than did controls (9-11ss, n=13 placodes; 12-13ss, n=6 placodes), regardless of the axial level of the section (P<0.05-0.005; Fig. 5C,D). There were no significant changes in placode cell density at 9-11ss (n=10 mutant placodes; n=6 control placodes) and 12-13ss (n=5 mutant placodes; n=4 control placodes; P>0.05; Fig. 5F),or in apoptosis when analyzed for activated caspase 3 expression(Fig. 5G)(Conlon et al., 1995; Del Monte et al., 2007). We also confirmed that the smaller placode was not caused by the precocious generation of neurons by analyzing Ngn1 expression (data not shown).

Collectively, our data show that many otic placode precursors undergo a fate change to epidermis in Notch1-deficient embryos. It is possible that the other Notch receptors are active during otic placode development in addition to Notch1. We confirmed our results by examining conditional mutants of Rbpj/Rbsuh, a transcriptional co-factor of NICD. The otic placode still forms in these mice (see Fig. S2C in the supplementary material) (see also Oka et al.,1995; de la Pompa et al.,1997), confirming that Notch signaling can modulate the size of the otic placode but is not necessary for its induction.

Daudet and colleagues recently suggested that initiation, but not maintenance, of Jag1 expression in the chick otic placode is regulated independently of Notch1 signaling(Daudet et al., 2007). We confirmed this result in mice: Jag1 continued to be expressed in the placode of Notch1 mutants, but the intensity of expression was reduced when compared with controls (see Fig. S2B in the supplementary material). It has been previously reported that Jag1 continues to be expressed in a morphologically distinct otic placode in mice carrying mutations of Pofut1, an O-fucosyltransferase essential for Notch signaling (Shi and Stanley,2003). We confirmed that Jag1 and Hes1expression can be initiated in the absence of canonical Notch signaling by examining conditional mutants of Rbpj/Rbsuh. Both genes continue to be expressed in a morphologically visible otic cup, although Hes1 was expressed at significantly reduced levels compared with controls (see Fig. S2C in the supplementary material). This is consistent with Hes1 expression being initiated by Notch signaling, but Jag1expression being initiated independently of Notch signaling.

Canonical Wnt signaling plays an important role in defining the size of the otic placode by driving medial Pax2⁺ pre-otic cells towards an otic rather than a cranial epidermis fate(Ohyama et al., 2006). Similarly, conditional activation of Notch1 in the Pax2⁺pre-otic field expands some, but not all, otic markers at the expense of epidermis (Fig. 3). Additionally, Notch pathway gene expression can be activated by canonical Wnt signaling (Fig. 2). These results suggested the possibility of reciprocal interactions between the Notch and Wnt pathways. To test whether Wnt signaling is modulated by the Notch pathway in the developing otic placode, we crossed Wnt reporter mice expressing a β-galactosidase reporter gene under the control of six Tcf/Lef DNA-binding sites (Mohamed et al.,2004) to either cN1ICD or Notch1 mutant lines.

Surprisingly, although the thickened Pax8⁺ placode was dramatically expanded to the level of the pharynx in cN1ICD embryos(Fig. 3A), Wnt reporter activity showed a much more modest expansion, extending a little beyond the lateral edge of the otic cup (Fig. 6A). We observed similar results with Dlx5, a known Wnt-responsive marker of the otic placode (bracket, Fig. 6A). To verify these results, we made use of the fact that cN1ICD mutants also express nuclear GFP after Cre recombination(Murtaugh et al., 2003). We co-immunostained cN1ICD;Wnt reporter embryos with anti-β-galactosidase and anti-GFP antibodies to mark the extent of the Wnt reporter and the expanded otic placode, respectively(Fig. 6B). By E9-E9.25, Wnt activity was elevated in the lateral regions of the mutant otic cup, which normally demonstrate moderate or low Wnt activity (red arrowhead; Fig. 6A). Furthermore, the otic cup region was larger in cN1ICD mutants than in controls (see Fig. S3 in the supplementary material, n=13 mutant placodes, n=14 control placodes; P<0.005). However, the ectopic placode region lateral to the otic cup, which expressed N1ICD and GFP, did not expressβ-galactosidase (bracket, Fig. 6B). These results suggest that Notch signaling can augment Wnt signaling, but that the active Notch1 ICD does not directly regulate Wnt-responsive genes containing Tcf/Lef DNA-binding sites.

To test whether Wnt signaling can also be modulated by loss of Notch1 activity, we examined Notch1 mutant mice crossed to a Wnt reporter mouse background. As expected, Wnt reporter activity was detected in Notch1 mutant placodes (Fig. 6C). However, the intensity of Wnt activity, as measured by time-matched β-galactosidase reactions was weaker than in controls. Additionally, the mediolateral extent of the Wnt reporter and expression of the Wnt-responsive gene Dlx5 was slightly reduced(Fig. 6C), reflecting the observed reduction in the placode size caused by Notch1 deficiency(Fig. 4). Taken together with our data showing that Wnt signaling can upregulate Notch pathway components,our results are consistent with a model in which the Wnt pathway can positively regulate components of the Notch pathway, and can, in turn, be augmented by Notch signaling. One prediction of this model is that maximal activation of Wnt signaling by a constitutively activated β-catenin mutation will be unaffected by a Notch1 mutation. To test this, we analyzed Pax8 and Foxi2 expression in Notch1 mutant embryos that also carried the activated β-catenin (cAct)mutation. As expected, the size of the expanded Pax8 domain seen in cAct embryos was not significantly different from that in Notch1;cAct mutants (Fig. 6D). Similarly, the reduced domain of epidermal Foxi2 expression seen in cAct mutants was not significantly different from that in Notch1;cAct mutants (Fig. 6D).

Notch signaling plays multiple roles in inner ear patterning, from the specification of neurons and prosensory patches to the generation of the stereotypical pattern of hair cells (Adam et al., 1998; Brooker et al.,2006; Daudet et al.,2007; Daudet and Lewis,2005; Haddon et al.,1998; Kiernan et al.,2005; Lanford et al.,1999; Shi et al.,2005). Here, we have uncovered new roles for Notch and Wnt signaling in the early development of the ear. Conditional activation of Notch1 in the Pax2⁺ pre-otic field causes the expansion of some, but not all, otic markers at the expense of epidermis. Conversely, in the absence of Notch1 signaling, the otic placode is significantly smaller. We have also shown that Wnt signaling regulates components of the Notch pathway, such as Jag1, and that Notch signaling positively regulates Wnt signaling. Our results suggest that Notch augments the Wnt signaling pathway to help define the size of the otic placode.

Our expression data show that several components of the Notch signaling pathway - Notch1, Jag1, Dll1 and Hes1 - are expressed in a medial subset of the mouse Pax2⁺ pre-otic field from the 5ss onwards, and that Wnt signaling initiates expression of at least some Notch pathway components. Notch1, Jag1, Dll1 and Hes1 are all expressed in the pre-otic field after Wnt6, Wnt8 and the first signs of Wnt reporter activity (Fig. 1C) (Ohyama et al.,2006). Expression of these Notch pathway genes occurs only within the region of the medial Pax2⁺ pre-otic field that responds to Wnt signaling (Fig. 1C). Consistent with previous reports(Duncan et al., 2005; Espinosa et al., 2003; Estrach et al., 2006), we found that ectopic activation of the canonical Wnt pathway induced the expression of Jag1, Notch1 and Hes1(Fig. 2A), whereas conditional deletion of β-catenin greatly reduced their expression(Fig. 2B). Wnt signaling can control the transcription of Notch pathway genes by directly acting on elements located in their promoters(Duncan et al., 2005; Espinosa et al., 2003; Estrach et al., 2006; Katoh and Katoh, 2006). In the case of the Jag1 promoter, there are multiple Tcf/Lef-binding sites that are conserved between mouse and human(Estrach et al., 2006; Katoh and Katoh, 2006). Putative Tcf/Lef-binding sites have also been identified in the Notch1 promoter (Galceran et al.,2004); however, although the Dll1 promoter also has Tcf/Lef-binding sites (Galceran et al.,2004), its expression was not expanded in embryos expressing activated β-catenin. Recent evidence suggests that factors distinct from Notch signaling are required to initiate Jag1 expression in the chick otocyst (Daudet et al., 2007),although maintenance of Jag1 is Notch dependent. Our results suggest that Jag1 initiation in the developing ear might be regulated directly by Wnt signaling (Fig. 2B, Fig. 7),whereas Notch1 and Hes1 expression might be initiated by Wnt signaling, and possibly also by FGFs(Norgaard et al., 2003; Zhou and Armstrong, 2007), in addition to by Notch signaling itself.

We recently showed that the mouse pre-otic field defined by the expression of Pax2 undergoes a fate decision to give rise to the cranial epidermis and the otic placode (Ohyama et al., 2006). The placode-epidermis fate decision is mediated by the canonical Wnt pathway, such that conditional deletion of β-catenin in Pax2⁺ cells drastically reduces the otic placode and expands the epidermis, whereas conditional activation of β-catenin in Pax2⁺ cells expands the otic placode at the expense of epidermis (Ohyama et al.,2006). In the light of the expression of many components of the Notch signaling pathway in the developing otic placode(Fig. 1), we hypothesized that Notch signaling might act with the canonical Wnt pathway to specify otic placode identity.

The activation of Notch1 signaling in the pre-otic field leads to a massive expansion of thickened, placode-like epithelium expressing Pax8 at the expense of Foxi2⁺ epidermis, in a manner very similar to the activation of β-catenin (Fig. 3). By contrast, although Pax2 expression can be expanded by the activation of Wnt signaling (Ohyama et al., 2006), it showed only a modest expansion in cN1ICD mutants compared with Pax8 (compare Fig. 3A with 3B). Pax2 and Pax8 are known to be differentially regulated by FGF signaling and the foxi1 transcription factor during induction of the zebrafish ear(Hans et al., 2004; Nissen et al., 2003; Solomon et al., 2003; Solomon et al., 2004), and our results suggest that these genes might also be differentially regulated by Notch signaling. In particular, Pax8 can be regulated either by canonical Wnt signaling or by Notch signaling. However, it is not clear whether the two pathways regulate Pax8 in entirely different ways or whether they converge on a nodal point, such as the binding of Lef/Tcf complexes to the Pax8 promoter(Schmidt-Ott et al., 2007). Pax8 expression correlates with epithelial thickening in all experiments in our study. However, further experiments are required to determine whether Pax8 is directly responsible for regulating this morphological change in the otic placode.

The examination of Notch1 mutants consistently showed a significant reduction in the size of the otic placode(Fig. 4B,C; Fig. 5). This small reduction is unlikely to be due to redundancy with other Notch genes, as there is no detectable expression of Notch2-Notch4 in the otic placode(Lewis et al., 1998; Williams et al., 1995). A similar persistence of the otic placode is seen after treating chick otic ectoderm with DAPT, a γ-secretase inhibitor that abolishes Notch signaling (Abello et al., 2007; Daudet et al., 2007), in mice carrying mutations in Pofut1, an O-fucosyltransferase that is an essential component of the Notch pathway(Shi and Stanley, 2003)(C.S.J., unpublished), and in mice lacking Rbpj/Rbsuh/CSL(Oka et al., 1995; de la Pompa et al., 1997) (see Fig. S2C in the supplementary material). In all of these experiments, any reduction in placode size in the absence of Notch signaling is much more modest than that seen in mice in which Wnt signaling is blocked by the conditional deletion of β-catenin(Ohyama et al., 2006)(Fig. 2B).

Our results suggest a model (Fig. 7B,C) in which both Notch and Wnt signaling can specify the size of the epithelium destined to form the otic placode by virtue of their regulation of Pax8, Foxi2 and Jag1, and by the induction of a thickened epithelial morphology. Our data from mice in which Wnt signaling is activated in the absence of Notch1(Fig. 6D), or in which Notch1 is activated in the absence of β-catenin(Fig. 3C), show that the two pathways can regulate these genes independently of each other. However, unlike the Wnt pathway, Notch signaling does not regulate the expression of otic placode-specific genes such as Gbx2, Sox9 and Hmx3, as these are unchanged in cN1ICD mutants(Fig. 3D). In addition, our results, taken together with previously published studies, suggest that Notch signaling also acts to augment Wnt signaling during otic placode induction,rather than being absolutely necessary for placode induction

To integrate our gain- and loss-of-function experiments with the Notch and Wnt pathways, we propose a model in which some Notch pathway components, such as Jag1, are induced by Wnt signaling. Subsequently, activation of Notch1 by Jag1 feeds back to augment the Wnt response(Fig. 7B). This feedback activity has no effect on the most medial regions of the pre-otic field -which receive the highest levels of Wnt signaling - but acts to increase Wnt signaling in mediolateral regions of ectoderm that receive modest to low levels of Wnt signaling. Thus, Notch-mediated feedback serves to sharpen and refine the initial mediolateral gradient of Wnt activity during the pre-otic field stage (Ohyama et al.,2006) into a more binary pattern at the otic placode stage, where Wnt signaling is either active (giving rise to the otic placode) or silenced(giving rise to epidermis; Fig. 7C).

Our data support this model in four ways. First, Notch1 deficiency causes a reduction in the area and intensity of β-galactosidase activity in Wnt reporter mice, and a reduction of the domain of the Wnt-responsive gene Dlx5 (Fig. 6C). However, loss of Notch1 does not abolish the expression of either marker, consistent with the notion that Notch1 signaling augments the Wnt response but does not initiate it. Second, the reduction in Wnt signaling resulting from the loss of Notch1 (Figs 4, 5) causes a consistent reduction in the size of the otic placode, but does not eliminate it entirely. The otic placode also forms in mice lacking other crucial components of the Notch pathway, such as Pofut1 or Rbpj/Rbsuh/CSL(Oka et al., 1995; de la Pompa et al., 1997; Shi and Stanley, 2003). Third,mutation of Notch1 has no effect on the size of the otic placode in embryos also expressing constitutively active β-catenin in the entire pre-otic field (Fig. 6D),presumably because cells expressing artificially high levels of activatedβ-catenin are not dependent on Notch1 function for the stabilization of otic fate. Finally, artificial N1ICD activation throughout the pre-otic field greatly expands Pax8 to the ventral pharynx, but this is not the case for Dlx5 or Wnt activity (Fig. 6A). This suggests that ectopic activation of N1ICD in regions of the pre-otic field that receive no Wnt signals is insufficient to augment or initiate the Wnt response (Fig. 7C). Furthermore, Wnt reporter expression is enhanced in regions receiving moderate levels of Wnt activity in cN1ICD mutants(Fig. 6A). Although, the mechanism of how Notch signaling augments Wnt activity is not clear, this result suggests that it is unlikely that N1ICD can directly activate transcription of Wnt-responsive genes by itself. A growing body of evidence suggests that Wnt and Notch pathways interact during cell fate determination(Aoyama et al., 2007; Arias and Hayward, 2006; Crosnier et al., 2006; Estrach et al., 2006; Fre et al., 2005). Notch signaling can act upstream of the Wnt pathway(Balint et al., 2005; Johnston and Edgar, 1998; Neumann and Cohen, 1996), or downstream (Estrach et al.,2006). Stimulation of the Wnt pathway can either antagonize or activate the Notch pathway in different contexts - for example, dishevelled can antagonize Notch signaling(Axelrod et al., 1996), whereas the downregulation of Gsk3 activity by Wnt signaling stimulates the Notch pathway (Espinosa et al.,2003). The Notch receptor is also able to antagonizeβ-catenin activity (Nicolas et al.,2003), sometimes in an NICD-independent manner(Hayward et al., 2006; Hayward et al., 2005).

Taken together, our current and previously published data suggest a model of otic placode induction whereby FGF signaling initially establishes a Pax2⁺ pre-otic field that is then patterned by a gradient of Wnt signaling arising from the midline. Wnt signaling upregulates components of the Notch pathway, which then act locally to augment the Wnt response and to mediate the placode-epidermis fate decision in the pre-otic field.

We thank the following colleagues for generously providing transgenic mice for this study: Charles Murtaugh (N1ICD mice), Daniel Dufort (Wnt reporter mice), Ron Conlon (Notch1 mutant mice), Tasuku Honjo(conditional Rbpj mice) and Mark Taketo (cActβ -catenin mice). We also thank Jeffrey Nye, Achim Gossler, Tim Mitsiadu, Ryoichiro Kageyama, Thomas Vogt and Andy McMahon for gifts of probes; Juan Llamas, Welly Makmura, Sheri Juntilla, Francesca Della Ripa and Juemei Wang for excellent technical support; and for Groves and Segil lab members for help and advice during this project. This work was funded by the House Ear Institute, by a March of Dimes Research Grant (A.K.G.), and by RO1DC04675 (A.K.G.) and RO1DC06185 (A.K.G. and N.S.).