The Notch ligands DLL1 and JAG2 act synergistically to regulate hair cell development in the mammalian inner ear

The Notch signaling pathway plays myriad roles in both developing and adult multicellular organisms (for reviews, see Baron, 2003; Gridley, 2003; Lai, 2004; Schweisguth, 2004). One of its best known roles, originally studied in Drosophila melanogaster and Caenorhabditis elegans, is in mediating lateral inhibition, a patterning mechanism in which a cell differentiating as one type prevents its neighboring cells from doing the same, thus creating a mosaic of cell types from an initially equivalent epithelium(Bray, 1998; Lai, 2004). The cellular arrangement of the sensory epithelium in the inner ear, in which each hair cell is surrounded by supporting cells, is suggestive of this type of patterning mechanism (Corwin et al.,1991; Lewis, 1991; Kiernan et al., 2002). Moreover, hair cells and supporting cells have been shown to arise from a common progenitor in lower vertebrates(Fekete et al., 1998),consistent with a lateral inhibitory mechanism in which different cell types arise from an initially equivalent epithelium. Expression studies have shown that the genes encoding the Notch ligands JAG2 and DLL1 are both expressed in nascent hair cells as they begin to differentiate(Adam et al., 1998; Lanford et al., 1999; Morrison et al., 1999). Several different Notch receptors are also expressed in the ear, supporting a role for Notch-mediated lateral inhibition in the inner ear(Lindsell et al., 1996; Lanford et al., 1999). Direct functional evidence for lateral inhibition in the ear is supported by studies of Notch pathway mutants in zebrafish, mouse and chicken(Haddon et al., 1998; Lanford et al., 1999; Riley et al., 1999; Zhang et al., 2000; Zine et al., 2001; Daudet and Lewis, 2005). In the mouse, a deletion of the Jag2 gene resulted in extra rows of both inner and outer hair cells in the cochlea(Lanford et al., 1999). Similarly, mice deleted for Hes1 and Hes5, downstream targets of Notch signaling, as well as mice lacking one copy of the Notch1 gene, demonstrated increased hair cell numbers(Zhang et al., 2000; Zine et al., 2001). Although consistent with the model, it was unclear why only relatively modest increases in hair cell numbers in the cochlea were observed. Furthermore, it was not known whether supporting cell numbers were concomitantly reduced, indicating a cell fate switch. Thus, it was not clear whether Notch signaling played only a minor role in cell patterning in the organ of Corti, or whether the milder phenotype was observed due to genetic redundancies in the pathway.

Because both the Dll1 and Jag2 genes are expressed in nascent hair cells (Lanford et al.,1999; Morrison et al.,1999), the prospect of redundancy was a real possibility, as both ligands may be necessary to fully deliver a lateral inhibitory signal. To test whether the Dll1 gene also plays a role in lateral inhibition and to investigate potential genetic interactions with Jag2 mutations, we generated embryos that carried various combinations of Jag2 and/or Dll1 mutant alleles. Our results show that Dll1 functions synergistically with Jag2, demonstrating that both ligands are required to regulate the numbers of hair cells that form in the mammalian cochlea. Using conditional gene inactivation, we also show that both the JAG2 and DLL1 ligands are likely to signal through the NOTCH1 receptor. Consistent with the proposed lateral inhibition model, most supernumerary hair cells in the Dll1/Jag2 double mutant cochleae did not arise through excess proliferation, suggesting instead a switch in cell fate. However, supporting cells did exhibit abnormal proliferation, implicating a novel role for the Notch pathway in regulating cellular proliferation in the ear.

Targeted disruptions of the Jag2 and Dll1 loci were described previously (Hrabe de Angelis et al., 1997; Jiang et al.,1998). The construction and characterization of the Dll1^hyp allele will be described in detail elsewhere (R.C. and A.G., unpublished). Briefly, the full-length Dll1 cDNA along with a neomycin phosphotransferase selection cassette was inserted into the Dll1 locus by gene targeting. Mouse embryos doubly heterozygous for this allele and the Dll1 null allele (i.e. Dll1^hyp/- embryos) survived past the period of vascular lethality at E10.5 that causes the death of Dll1^-/-embryos. The Dll1^hyp/- embryos exhibit segmentation defects consistent with the previously described role for the Dll1gene during somite formation (Hrabe de Angelis et al., 1997). All data collected to date support the model that the Dll1^hyp allele is a hypomorphic Dll1 mutant allele. Jag2 and Dll1 mice were maintained on a mixed 129S1/SvImJ; C57BL/6J and C57BL/6J backgrounds,respectively. To produce Foxg1-Cre Notch1^flox/- mice, Notch1^flox/flox (Yang et al., 2004) mice were mated to mice heterozygous for both Foxg1-Cre (Hebert and McConnell,2000) and Notch1(Swiatek et al., 1994). The Foxg1-Cre mice were maintained on an outbred Swiss Webster background, whereas the Notch1^+/- mice and the Notch1^flox/flox mice were maintained on a C57BL/6J background.

For whole-mount preparations, inner ears were dissected and fixed overnight in 4% paraformaldehyde. The bony shell and the stria vascularis were removed,and the ears were incubated with a biotinylated lectin (Griffonia simplifonia I, Vector Laboratories) diluted 1:100. A FITC-labeled avidin secondary reagent was used to visualize the hair cells. All other immunocytochemistry was performed on standard 7-μm paraffin-embedded sections. Antibodies used included anti-myosin VIIa (1:1000, a gift from Drs A. EL-Amraoui and C. Petit,Institut Pasteur, Paris, France), anti-p27^kip1 (1:100, Neomarkers)and anti-BrdU (1:500, Roche). For anti-p27^kip1 labeling, an antigen-retrieval step was performed by boiling the sections for 10 minutes in 10 mM citric acid. For BrdU labeling, two different protocols were used: in the first, antigen retrieval was performed as described previously for p27^kip1, followed by pepsin digestion (100 μg/ml for 20 minutes at 37°C) and acid treatment (2N HCl for 30 minutes at 37°C). For doubly labeled sections, heat-activated antigen retrieval was performed followed by DNAse I digestion (5 U/ml for 30 minutes at 37°C). Cell death was examined by TUNEL staining of paraffin-embedded sections, using the TMR in situ cell detection kit (Roche).

Hair cell counts from lectin-stained wholemounts Hair cell counts were performed on mid-basal regions of the lectin-stained cochleae, extending between 800 and 1400 μm. Each genotype contained counts from three or four different embryos. After capture of high-resolution images of the cochleae,counts and measurements were performed using Zeiss Axiovision software.

For each ear, images from 32 sections through the mid-modiolar region of the cochlea that were triply labeled for 4′-6-Diamidino-2-phenylindole(DAPI; to stain nuclei), myosin VIIa (to label hair cells) and p27^kip1 (to label supporting cells) were captured using Zeiss Axiovision software. Nuclei from cells that were doubly labeled with DAPI and either myosin VIIa (hair cells) or p27^kip1 (supporting cells) in the basal and middle turns of the organ of Corti were counted.

To examine cell proliferation, pregnant female mice were injected with a BrdU solution (10 mg/ml in PBS; final dose, 50 μg per gram of body weight),three times daily (at 4 hour intervals), between E14.5 and E17.5. Animals were euthanized and half heads were fixed and embedded for paraffin sectioning.

Inner ears were prepared for scanning electron microscopy, as described previously, using a version of the osmium tetroxide-thiocarbohydrazide (OTOTO)method (Kiernan et al., 1999). Specimens were examined with a Hitachi 3000N scanning electron microscope.

For sample preparation, inner ears were dissected from the head and fixed overnight in 4% paraformaldehyde. After washing in PBS, the bony shell and stria were removed from the cochleae and the samples were dehydrated in methanol. In situ hybridization was performed as described(Stern, 1998), with the exception of the post-hybridization washes, which were done according to Rau et al. (Rau et al., 1999). After the reactions were judged to be complete, cochleae were flat mounted on glass slides in 70% glycerol. Probes for α-tectorin and β-tectorin(gifts from Drs K. Legan and G. Richardson, University of Sussex, UK) were as described (Rau et al.,1999).

The Dll1 gene is expressed in nascent hair cells in the inner ear(Morrison et al., 1999), but a role for the Dll1 gene in regulating cochlear hair cell differentiation has not been examined because of the early embryonic lethality of embryos homozygous for a targeted Dll1(Dll1^-/-) null mutation (Hrabé de Angelis et al.,1997). However, mouse embryos that carried both the Dll1null allele and a newly constructed Dll1 hypomorphic allele(Dll1^hyp/- mice) survive until birth (R.C. and A.G.,unpublished; see Materials and methods). We therefore examined cochlear hair cell differentiation and patterning in Dll1^+/-,Dll1^+/hyp and Dll1^hyp/- mutant embryos. Embryos were harvested at embryonic day 18.5 (E18.5), and hair cell patterning in their cochleae was examined by staining with a lectin that binds to hair cell stereocilia. Similarly to Jag2^+/- cochleae, Dll1^+/- and Dll1^+/hyp cochleae did not show a significant increase in hair cells, although occasional second row inner hair cells and fourth row outer hair cells were observed(Fig. 1B-D). Surprisingly,cochleae from Dll1^hyp/- mutant embryos did not show a statistically significant increase in hair cell numbers(Table 1), although supernumerary inner and outer hair cells were observed more frequently than in wild-type embryos.

We then tested whether the Dll1 and Jag2 genes functioned synergistically during cochlear hair cell differentiation. As neither Jag2^-/- homozygous mice nor Dll1^hyp/-compound heterozygous mice survive postnatally, we set up crosses between Dll1^+/- Jag2^+/- and Dll1^+/hypJag2^+/- mice to generate embryos with varying Dll1and Jag2 gene dosages. Cochleae that were mutated for only a single allele of a Notch ligand (either Dll1^+/-,Dll1^+/hyp or Jag2^+/-) demonstrated no significant increases in hair cell numbers when compared with wild-type embryos, although occasional extra inner and outer hair cells were observed(Fig. 1A-D). However, cochleae that were doubly heterozygous for both Dll1 and Jag2(Dll1^+/- Jag2^+/-) showed many extra inner hair cells and a nearly complete fourth row of outer hair cells, indicating a genetic interaction between mutations in these two genes(Fig. 1E). Hair cell counts from the middle turn of the cochleae demonstrated that Dll1^+/-Jag2^+/- double heterozygous cochleae have significantly more hair cells than wild type (Table 1). Dll1^+/- Jag2^+/- cochleae had nearly as many supernumerary hair cells as Jag2^-/-cochleae, although there were not as many extra inner hair cells. This indicates that the Dll1 gene may not play as large a role in the inhibition of inner hair cell development as the Jag2 gene does. Counts of Dll1^+/hyp Jag2^+/- cochleae revealed a more modest increase in hair cell numbers, reflecting the hypomorphic nature of the Dll1^hyp allele and demonstrating the graded response of the cochlea to the dosage of genes encoding Notch ligands.

Cochleae that lacked both copies of Jag2, and that carried either a single null allele of Dll1 (Dll1^+/-Jag2^-/-) or a null allele combined with the Dll1hypomorphic allele (Dll1^hyp/- Jag2^-/-), showed even larger increases in hair cell numbers than Jag2^-/-cochleae (Fig. 1G,H). In Dll1^hyp/- Jag2^-/- cochleae, two to four rows of inner hair cells were present and four to six rows of outer hair cells could be identified. However, unlike the Dll1^+/-Jag2^+/- or Jag2^-/- cochleae, the rows of hair cells were more disorganized and the hair cells were very densely packed,making accurate determination of hair cell numbers difficult in whole-mount preparations.

We used scanning electron microscopy to examine in more detail the morphological defects in the double mutant cochleae(Fig. 2). Dll1^hyp/- Jag2^-/- mutant cochleae were extremely disorganized, both in the patterning of the sensory cell rows and,at a single hair cell level, in the loss of polarity and disorganization of many hair cell stereocilia bundles (Fig. 2F). When compared with Jag2^-/- and Dll1^+/- cochleae (Fig. 2A-D), it is clear the amount of disorganization correlates with the amount of Notch ligand that is present, as some disorganization is present even in Jag2^-/- cochleae. However, it is not clear from these data whether the Notch pathway plays a direct role in planar polarity signaling (Barald and Kelley,2004) and/or patterning of the hair cells, or whether the observed disorganization may be a secondary effect of the increased numbers of hair cells present in the mutant cochleae.

One prediction of the lateral inhibition model is that additional sensory hair cells would be generated at the expense of nonsensory supporting cells. In order to examine whether supporting cells were decreased in Dll1/Jag2 double mutant cochleae, hair and supporting cell counts were performed on mid-modiolar sections of Dll1^hyp/-Jag2^-/- mutant and control cochleae that had been labeled with both a hair cell marker (Myosin VIIa) and a supporting cell marker (p27^kip1)(Fig. 3A-D). Hair cell counts revealed a 1.7-fold increase in Dll1^hyp/-Jag2^-/- cochleae when compared with controls. However, supporting cell counts revealed a milder 1.2-fold decrease in supporting cells (Fig. 3I).

p27^kip1 immunostaining of the Dll1^hyp/-Jag2^-/- mutant and control cochleae indicated that many of the missing cells resided beneath the outer hair cells and were presumably Deiter's cells (Fig. 3C,D). In order to assess which supporting cell populations were decreased, Dll1/Jag2 double mutant and control cochleae were processed for whole-mount in situ hybridization using probes that mark specific populations of supporting cells, α-tectorin andβ -tectorin (Rau et al.,1999). Expression of α-tectorin in wild-type cochleae marks the supporting cell populations surrounding the organ of Corti proper (including Koelliker's organ), the supporting cells surrounding the inner hair cells (inner phalangeal cells), and Hensen's cells. In Dll1^+/- Jag2^-/- cochleae,α -tectorin expression appeared to be similar to in controls,suggesting that none of these surrounding supporting cell populations were converted to hair cells (Fig. 3E,F). β-Tectorin marks several supporting cell regions that lie in the organ of Corti proper, including the inner and outer pillar cells,the last row of Deiter's cells, and a smaller domain of cells in Koelliker's organ. In Dll1^+/- Jag2^-/- cochleae, most of these expression domains appeared to be similar to those in controls, with the exception of the Deiter's cell expression domain(Fig. 3G,H). Here, theβ-tectorin expression was punctate rather than being maintained in a continuous stripe, and appeared to be reduced, suggesting that many of the missing supporting cells were derived from the Deiter's cell population.

Because we could only detect mild supporting cell losses, we next examined whether any of the hair and/or supporting cells were produced via excess cell division. The majority of hair cells and supporting cells in the organ of Corti undergo their final division between E12 and E14 in a spatio-temporal gradient from apex to base (Ruben,1967). The Dll1 and Jag2 genes are not expressed in the cochlea until approximately E14.5, when cells in the organ of Corti have exited the cell cycle and are beginning to differentiate(Lanford et al., 1999; Morrison et al., 1999). One possibility is that Notch signaling is important for the suppression of continued cell division in the cochlea, rather than for mediating lateral inhibition. Therefore, we examined whether there was continued proliferation occurring between E14.5 and E17.5 in Dll1/Jag2 double mutants. Pregnant Dll1^+/hyp Jag2^+/- females that had been mated to Dll1^+/- Jag2^+/- males were injected with bromodeoxyuridine (BrdU) between E14.5 and E17.5. Embryos were taken at E18.5 and their cochleae were processed to detect BrdU incorporation. Between E14.5 and E17.5, there was little proliferation in the control mouse ventral cochlear epithelium with the exception of cells in Koelliker's organ and scattered cells in the lesser epithelial ridge(Ruben, 1967), although occasionally labeled cells were observed in the organ of Corti(Fig. 4 and Table 2). However, mutant cochleae (either Dll1^+/- Jag2^-/- or Dll1^hyp/- Jag2^-/-) frequently displayed labeled cells in the organ of Corti. The most commonly labeled cell types were supporting cell types; amongst these, pillar cells were most often labeled,followed by Deiter's cells and Hensen's cells(Fig. 4B-D and Table 2). Hair cells were only rarely labeled (Table 2). These data show that the majority of the supernumerary hair cells are not arising through continued proliferation, consistent with the model of Notch-mediated lateral inhibition. However, supporting cells were frequently labeled, which may explain why only modest decreases in supporting cell numbers were observed in Dll1^hyp/- Jag2^-/- mutant cochleae.

Previously, we demonstrated that mice heterozygous for a Notch1null allele exhibited increased numbers of outer hair cells, suggesting that the NOTCH1 protein was required for proper differentiation in the organ of Corti (Zhang et al., 2000). However, we were unable to assess the ear phenotype of Notch1^-/- embryos as they die at midgestation(Swiatek et al., 1994). Therefore, we conditionally inactivated Notch1 function in the otic epithelium using the Foxg1-Cre mouse line(Hebert and McConnell, 2000; Pirvola et al., 2002) and a Notch1^flox allele(Yang et al., 2004). Embryos with the genotype Foxg1-Cre Notch1^flox/- and littermate controls were isolated at E18.5, and whole-mount preparations of the cochleae were stained with lectin. These experiments showed a large increase in the hair cell population in the Foxg1-Cre Notch1^flox/-cochleae (Fig. 5B,D,E) that resembled the hair cell increases observed in Dll1/Jag2 double mutant cochleae. Measurements of the lengths of the entire cochleae in Foxg1-Cre Notch1^flox/- mutants (mean=5359 μm±260; n=3) versus controls (mean=5311 μm±492; n=3) were not significantly different (P=0.92), indicating that the supernumerary hair cells are a result of real increases in hair cell numbers and not simply a redistribution of the cells due to a shortened cochlea. Interestingly, the increase in hair cells in Foxg1-Cre Notch1^flox/- cochleae was far more dramatic than that in the Dll1^hyp/- Jag2^-/- cochleae. In fact, Foxg1-Cre Notch1^flox/- cochleae displayed about a 3-fold increase in the number of hair cells (compared with the 1.6-fold increase in the Dll1^hyp/- Jag2^-/- double mutants), whereas the decrease in supporting cell numbers was only slightly more severe than in the double mutants (1.6-fold in the Foxg1-CreNotch1^flox/- cochleae versus 1.2-fold in the Dll1/Jag2 double mutant cochleae; Fig. 5E). The difference in the severity of the phenotypes could be explained by the fact that the Dll1^hyp/- Jag2^-/- cochleae still retained some residual DLL1 function due to the hypomorphic nature of one of the alleles,or, alternatively, there may be yet another Notch ligand, such as JAG1, that plays a role in the lateral inhibitory process. However, the overall similarity between the Dll1^hyp/- Jag2^-/- double mutant and Foxg1-Cre Notch1^flox/- phenotypes strongly suggests that both the DLL1 and the JAG2 ligand signal through the NOTCH1 receptor during sensory cell differentiation in the cochlea. Our genetic data is supported by experiments demonstrating that both the DLL1 and the JAG2 ligand can bind cell lines expressing the NOTCH1 receptor and activate a Notch reporter construct (Hicks et al.,2000; Shimizu et al.,2000).

We demonstrate here a role for the Notch ligand DLL1 in patterning of the inner ear, where, along with the JAG2 ligand, it regulates the numbers of sensory hair cells that form in the organ of Corti. We also show that both ligands are likely to signal through the NOTCH1 receptor. Cochleae that lack both copies of Jag2 and have a reduction in Dll1 show an even greater increase in hair cell numbers than Jag2^-/-mutant cochleae do, indicating that their roles are synergistic. Interestingly, although there were increased numbers of hair cells in Dll1^hyp/- mutant cochleae, these increases were not statistically significant (Table 1). This could be due to the fact that, because a Dll1hypomorphic allele was used rather than a complete null, DLL1 protein levels may not have been reduced sufficiently to affect cochlear patterning. Thus, it was only when the mutant Dll1 alleles were combined with either one or two copies of the Jag2 null allele that significant effects on hair cell formation were achieved. These data suggest there is a threshold of Notch ligand expression (either Dll1 or Jag2) that must be achieved for proper patterning to take place.

The standard model of lateral inhibition predicts that, if supernumerary sensory cells are produced via a cell fate switch, a concomitant loss of nonsensory supporting cells should be observed along with the increase in sensory hair cells. However, an examination of supporting cell populations revealed only a modest loss of p27^kip1-positive supporting cells that was significantly different from the increase in hair cell numbers (only about a 1.2-fold loss in supporting cells compared with a 1.7-fold increase in hair cells; Fig. 3I). This affect was even stronger in Foxg1-Cre Notch1^flox/- mutant cochleae, where there was a dramatic 3-fold increase in hair cells accompanied by only about a 1.6-fold drop in p27^kip1-positive supporting cells(Fig. 5E). Deiter's cells appeared to be the most dramatically reduced supporting cell population,suggesting a cell fate switch from Deiter's cells to outer hair cells. Cell death does not appear to account for the missing supporting cells, as the number of apoptotic profiles was not increased in either Dll1/Jag2double mutant cochleae at E15.5 or Foxg1-Cre Notch1^flox/-mutant cochleae at E18.5, as determined using the TUNEL assay (data not shown). BrdU incorporation studies in Dll1/Jag2 double mutant cochleae revealed ectopic proliferation of supporting cells with very few labeled hair cells in the double mutant cochleae. Interestingly, pillar cell numbers were neither decreased nor increased in any of the Notch mutants described here, despite the fact that they were frequently observed to be dividing. These data suggest that pillar cells may have unique stem cell-like properties in the developing organ of Corti. Stem cell-like cells have been identified in the adult mammalian utricle(Li et al., 2003), although it is not known whether similar cells are present in the cochlea. Taken together,these data indicate that the majority of the supernumerary hair cells arise via a cell fate switch, and not through continued cell proliferation. However,continued proliferation of the remaining nascent supporting cells compensates for the loss of supporting cell precursors, resulting in only modest decreases in the supporting cell population. These data support a role for the Notch pathway in mediating lateral inhibition in the inner ear, and also reveal a role for Notch signaling in the control of cell proliferation within the developing organ of Corti (Fig. 6).

A previously described role for the Notch pathway in regulating cell proliferation has been to maintain cells in an undifferentiated state, thereby promoting cell proliferation in many contexts. However, it has become clear in recent years that the effects of Notch signaling on cell proliferation are complex and context dependent (Weng and Aster, 2004). For example, an emerging role for the Notch pathway in promoting differentiation and cell cycle withdrawal has been revealed by studies in the skin and nervous system. Conditional Notch1 deletion in mouse skin causes hyperplasia and deregulation of differentiation, leading to the development of basal cell carcinoma-like tumors(Rangarajan et al., 2001; Nicolas et al., 2003). Similarly, specific Notch1 downregulation is found in aggressive cervical cancers (Talora et al., 2002). These results have led to the suggestion that, in some contexts, Notch signaling may act as a tumor suppressor by promoting differentiation and cell cycle withdrawal. In the nervous system, the Notch pathway has been implicated in promoting the glial cell fate (Furukawa et al., 2000; Hojo et al.,2000; Morrison et al.,2000). Given that inner ear supporting cells share some characteristics with glia, these results raise the possibility that the Notch pathway plays an instructive role in supporting cell differentiation in the cochlea. When Notch signaling is downregulated, as in this study, some progenitor cells may fail to differentiate properly and continue dividing. Thus, Notch signaling may play a dual role in sensory differentiation in the inner ear, preventing adoption of the hair cell fate through a lateral inhibitory mechanism while promoting cell cycle withdrawal and supporting cell differentiation. Alternatively, the continued progenitor/supporting cell proliferation may be an indirect effect caused by the loss of contact-mediated inhibition due to the supporting cell fate conversion or other cellular changes in the epithelium. Regeneration studies using ototoxic drugs or acoustic trauma in the avian inner ear and in the mammalian vestibular regions have shown that hair cell death triggers proliferation of the supporting cells(Corwin and Cotanche, 1988; Ryals and Rubel, 1988; Warchol et al., 1993; Matsui et al., 2002). This suggests that, under normal circumstances, the hair cell exerts an anti-proliferative effect on the surrounding support cells, although the molecular identity of this signal is not known. It is interesting to note that analysis of the zebrafish Mind bomb mutant, a mutation in a gene encoding a ubiquitin ligase involved in Notch signaling, demonstrated a 10-fold excess hair cells in the inner ear, far more than could be explained by a simple cell fate conversion (Haddon et al.,1998). These results suggest that excess cell division may also occur in these mutant ears, although proliferation was not specifically examined. Taken together, these data suggest there may be a direct role for Notch signaling in the control of cell proliferation in the organ of Corti.

Previous studies have shown that control of cell proliferation in the organ of Corti involves the cyclin-dependent kinase inhibitors, p27^kip1and p19^Ink4d (Chen and Segil,1999; Lowenheim et al.,1999; Chen et al.,2003), and the retinoblastoma protein (Rb)(Mantela et al., 2005; Sage et al., 2005). Studies of mice that lack these proteins in the mouse inner ear have shown that both p19^Ink4d and Rb primarily appear to play a role in maintaining the postmitotic state of the hair cells, whereas p27^kip1 appears to be more involved in preventing the continued proliferation of the progenitor cells and supporting cells. Because we observed mainly dividing supporting/progenitor cells, we were interested in determining whether p27^kip1 expression was affected in the Notch mutants. We examined p27^kip1 expression in both Dll1/Jag2 double mutant and Foxg1-Cre Notch1^flox/- mutant cochleae at E18.5(Fig. 3B-D, Fig. 5D) but found no differences in expression between mutants and controls.

Both Dll1/Jag2 double mutant cochleae and Foxg1-Cre Notch1^flox/- cochleae displayed severely disorganized hair cell rows, and loss of organization and polarity of the hair cell stereociliary bundles. It is not clear whether this disorganization is a direct consequence of reduced Notch signaling, or whether it is a secondary event resulting from the abnormal cellular composition of the cochlea. Recent work has shown that an evolutionarily conserved mechanism for generating cell polarity within epithelial cell layers, termed planar cell polarity (PCP), is involved in regulating the polarity of inner ear hair cells and the orientation of their stereocilia bundles(Barald and Kelley, 2004). A role for Notch signaling in PCP has not been reported in any vertebrate system. However, a role for the Notch pathway in planar polarity has been shown during eye development in Drosophila, where Notch signaling specifies the R4 photoreceptor cell fate(McNeill, 2002). Proper specification of the R3 and R4 cell fates is essential for the loss of symmetry within the ommatidial clusters and thus the proper genesis of PCP within the eye. Similarly, correct specification of the hair cell and supporting cell fates may be required to establish PCP within the inner ear. Further insights into a possible role for Notch signaling in planar polarity in vertebrates may come from gene expression or genetic interaction studies with recently identified planar polarity genes in the inner ear(Curtin et al., 2003; Montcouquiol et al., 2003; Lu et al., 2004). However, it should be noted that, unlike the previously identified PCP mouse mutants where hair cell bundles are intact but disoriented, many of the hair bundles in the Notch mutants appear to have lost their organization completely. This suggests that reduced Notch signaling may affect a more basic level of bundle organization than PCP. Moreover, similarly disorganized bundles have been observed in mutant cochleae that lack the Rb gene, which exhibit a large overproliferation of hair cells(Mantela et al., 2005; Sage et al., 2005). Taken together, these data suggest that the disorganization may be an indirect effect of having too many hair cells in the epithelium, leading to a disruption in the signals that polarize and organize the stereocilia bundles.

These results highlight the complexity of Notch signaling in the inner ear,and demonstrate that the Notch pathway plays a dual role in regulating cellular differentiation and patterning in the cochlea, acting both through lateral inhibition and the control of cellular proliferation(Fig. 6). Unlike birds and amphibians (Corwin and Cotanche,1988; Ryals and Rubel,1988; Corwin and Oberholtzer,1997), mammals demonstrate little regenerative potential in the inner ear (Johnsson et al.,1981; Forge,1985), which may be related, at least in part, to a failure of supporting cell proliferation after injury(Roberson and Rubel, 1994). Our results suggest that modulation of the Notch pathway may represent an important avenue for regeneration studies in the mammalian inner ear.

We thank Drs A. EL-Amraoui, C. Petit, K. Legan, G. Richardson and R. Burgess for reagents, and Weidong Zhang of the Computational Biology Resource for help with the statistical analysis. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to A.G. (Go 449/9-1), and by grants from the NIH to A.E.K (DC05865), R.K. (HD044056), T.G. (NS036437 and DK066387)and the Jackson Laboratory (CA034196).