Hes1 is a negative regulator of inner ear hair cell differentiation

The bony labyrinth of the inner ear develops from the otic vesicle (Van de Water, 1983; Fekete, 1996) and consists of the cochlea and vestibular end organs including the utricle, saccule and three semicircular canals. Each of these structures contains a sensory epithelium in which hair cells, mechanosensory cells, which convert sound or motion signals into electrochemical energy, and supporting cells are located. Currently, little is known about the signaling events underlying inner ear development, specifically, differentiation of hair cells. Classical tritiated thymidine incorporation studies have shown that hair cells become postmitotic between E11.5 and E17.5, with a peak at E13.5 in rodents (Ruben, 1967; Sans and Chat, 1982). Hair cells in the mammalian vestibular end organs appear to be derived from the progenitor cells or supporting cells located within the sensory epithelium (Forge et al., 1993; Warchol et al., 1993; Li and Forge, 1997; Zheng and Gao, 1997; Kuntz and Oesterle, 1998) in a similar manner as shown in birds and lower vertebrates (Corwin and Cotanche, 1988; Ryals and Rubel, 1988; Balak et al., 1990; Fekete et al., 1998). However, the exact origin of mammalian cochlear hair cells is still unclear because no cell lineage studies using lineage tracers have been performed. Previous histological studies suggest that the inner and outer hair cells in the mammalian cochlea are probably derived from the greater epithelial ridge (GER) and the lesser epithelial ridge (LER) cells, respectively, during embryogenesis (see Lim and Rueda, 1992).

Hair cell differentiation in the inner ear is likely to be controlled by specific genes (for recent reviews see Fekete, 1996, 1999). So far one of the most crucial genes for the control of inner ear hair cell differentiation appears to be the mouse basic helix-loop-helix (bHLH) transcription factor Math1 (Atoh1 – Mouse Genome Informatics), a mammalian homolog of Drosophila atonal. Math1 has been shown to be a positive regulator for the differentiation of cerebellar granule neurons (Ben-Arie et al., 1997), dorsal commissural interneurons (Helms and Johnson, 1998) and inner ear hair cells (Bermingham et al., 1999; Zheng and Gao, 2000). Targeted deletion of the Math1 gene leads to a failure of hair cell differentiation (Bermingham et al., 1999). Overexpression of Math1 in postnatal rat cochlear explant cultures induces production of extra hair cells in the GER (Zheng and Gao, 2000).

A few other bHLH transcription factors can influence cell fate determination by acting as negative regulators (Kageyama and Nakanishi, 1997). To identify a bHLH transcription factor that may interact with Math1 to inhibit hair cell differentiation, we turned our interest to Hes1, a mammalian hairy and enhancer of split homolog, which is a negative regulator of neurogenesis (Ishibashi et al., 1995; Nakamura et al., 2000). Targeted disruption of the Hes1 gene results in a precocious neuronal differentiation in the brain (Ishibashi et al., 1995) and in the retina (Tomita et al., 1996). The Hes1 null mutant mice display severe neural tube defects and die before or immediately after birth (Ishibashi et al., 1995). More recently, data collected from Hes1 and Hes5 double null mutant mice has shown that Hes1 and Hes5 act together as Notch effectors for the control of mammalian neuronal differentiation (Ohtsuka et al., 1999).

To determine whether Hes1 influences inner ear hair cell differentiation, we examined the inner ear phenotypes of Hes1-deficient mice. We found supernumerary hair cells in the cochlea and in the utricle, a vestibular end organ, of the Hes1-deficient mice. In addition, to find out the spatiotemporal expression patterns of Hes1 in the inner ear, we performed RNA in situ hybridization and RT-PCR analysis. We discovered that Hes1 was expressed in the inner ear during hair cell differentiation period and its expression became elevated around birth and persisted into adulthood. Within the vestibular sensory epithelium, Hes1 was expressed in supporting cells, but not hair cells. In the cochlea, Hes1 was selectively expressed in GER and LER regions, which are adjacent to inner and outer hair cells, respectively. Moreover, we performed co-transfection experiments with plasmids expressing Hes1 and Math1 and obtained evidence that Hes1 can block the induction of ectopic hair cells in the presence of Math1. This study suggests that Hes1 is involved in hair cell differentiation as a negative regulator possibly by antagonizing Math1.

Hes1 gene disruption and genotyping of the mice were performed as previously reported (Ishibashi et al., 1995). The inner ear phenotypes were compared between Hes1−/− embryos and their Hes1^+/− and Hes1^+/+ littermates. Cochlear explants were prepared from E17.5 mice (a close time point to birth, as Hes1−/− mice die at birth) and fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) before they were processed with FITC-conjugated phalloidin (Zheng and Gao, 1999) or anti-myosin VIIa antibody labeling (Hasson et al., 1995; Zheng and Gao, 2000). Some of the cochlear tissue was fixed in 10% neutral buffered formalin. The preparations were then washed, dehydrated in ascending graded alcohols, cleared in xylene and embedded in paraffin (Zheng et al., 1999b). Three micrometer paraffin sections of the preparations were cut with a microtome. The sections were processed for Hematoxylin and Eosin staining.

Utricular whole-mount tissue was also dissected out from E17.5 mice and fixed in 4% paraformaldehyde before they were cryoprotected in 30% sucrose solution and embedded in OCT (Miles). Serial sections (20 μm) were cut from the utricular preparations and collected on microscopic slides for myosin VIIa immunocytochemistry (Zheng et al., 1999a). Labeled preparations were washed in PBS, mounted in Fluoromount-G (Southern Biotechnology, AL) and viewed using a Zeiss Axiophot epifluorescent. Images were captured with Compix imaging systems using a cold RGB CCD camera

After fixation in 10% neutral buffered formalin, some of the cochlear preparations were post-fixed in 1% aqueous osmium tetroxide for 2 hours at room temperature, washed in distilled H₂O, dehydrated in graded ethanol, and dried using Hexamethyldisilazane (HMDS). The samples were mounted on scanning electron microscopy stubs using carbon double sticky tabs (Ted Pella, Redding, CA), sputter coated with 10 nm gold-palladium (Hummer XP, Anatech, Alexandria, VA) and viewed in a Philips SEM 525M at an accelerating voltage of 10 kV. Digital images were captured using Semicaps Genie version 1.22 software.

The pRK5-Math1-EGFP plasmid was constructed as previously described (Zheng and Gao, 2000), by inserting the EcoRI Math1 fragment spliced from the pCMV-Math1 plasmid (Akazawa et al., 1995) into the multiple cloning site of pRK5-EGFP plasmid (Murone et al., 1999). The pSV2CMV-Hes1 plasmid (Akazawa et al., 1995) contains the Ssp1(171)-EcoRI (the 3′ end) fragment of rat Hes1 cDNA. Middle turn cochlear explants (with the stria vascularis removed) were dissected from P0-P1 rats as previously described (Zheng and Gao, 1996, 2000). The cultures were transfected with the pRK5-Math1-EGFP plasmid, a mixture of pSV2CMV-Hes1 and pRK5-EGFP plasmids at a ratio of 5:1 or a mixture of equal amount of pSV2CMV-Hes1 and pRK5-Math1-EGFP plasmids using an electroporator (Model CUY-21, BEX, Tokyo) with a train of eight pulses: 25 V, 50 mseconds duration and 100 mseconds interval. The tissue was placed in a groove freshly made of 1% agarose gel, containing 3 mg/ml plasmid DNA. The top surface of the explants was positioned to face to the cathode. After electroporation, the explants were plated on a collagen-coated (80 μg/ml rat tail collagen I in 0.02 N HCl) eight-well LabTek slide in serum-free medium as described (Zheng and Gao, 1999). The cultures were then fixed at 7 days after transfection and processed for double immunocytochemistry with anti-myosin VIIa and anti-EGFP (Chemicon) antibodies, mediated by Texas Red- and FITC-conjugated secondary antibodies as described (Zheng and Gao, 1997, 2000).

To count total myosin VIIa-positive cells from the serial cryostat utricular sections, we used an ocular grid in a Zeiss Axiophot microscope with 20× and 40× lenses as previously described (Zheng et al., 1999b). Phalloidin-positive and myosin VIIa-positive cells were counted from cochlear surface preparations dissected from E17.5 Hes1−/−, Hes1^+/− and Hes1^+/+ mice, essentially in the same way as previously described (Zheng and Gao, 1996, 1999). At E17.5, the cochlea has formed only two turns, the basal and apical turns, and hair cells in the apical turn are not yet well-developed. We therefore performed hair cell counts only in the basal turn. For the inner hair cells numbers, cell counts were performed along 1 mm length starting from the basal end of the cochlea. For outer hair cell numbers, cell counts were obtained from a randomly selected 100 μm length in the middle region of the basal turn of the cochlea. Data collected from each experimental group are expressed as mean±s.e.m. Student’s t-test was used for statistical analysis. For the co-transfection experiments, total numbers of EGFP-positive and myosin VIIa/EGFP-double-positive GER cells were counted as previously described (Zheng and Gao, 2000).

Quantitative RT-PCR analysis was performed with the 5′-exonuclease assay using fluorescent non-extendible oligonucleotide probes (TaqMan PCR detector, Perkin Elmer Applied Biosystems) as described (Gibson et al., 1996). We used specific probes and primers for Hes1 (probe, CGGCTTCCAAGTGGTGCCGG; forward primer, GGAGAGGCTGCCAAGGTTTT; reverse primer, GCAAATTGG-CCGTCAGGA) and Math1 (probe, CTGAACCACGCCTTCGAC-CAGCTG; forward primer, AACGGCGCAGGATGCA; reverse primer TTGAAGGACGGGATAACGTTG). The specific probe and primers for the control housekeeping gene, Gapdh, was same as reported (Zheng et al., 1999a). Data were collected from three to five pairs of cartilageous capsules containing the entire inner ear labyrinth tissue dissected at various developmental stages and are expressed as mean±s.e.m.

E17.5 Wistar rat ears were dissected and immediately fixed in 4% paraformaldehyde in 0.1M phosphate buffer, pH 7.4, for 3.5 hours, cryoprotected in 20% sucrose in PBS and embedded in OCT (Miles, Elkhart, IN) for cryostat sectioning. Sections (20 μm) were dried (50°C for 20 minutes) and processed for in situ hybridization with digoxigenin-labeled (Boehringer Mannheim, Indianapolis, IN) probes as described previously (Hynes et al., 1995). Rat Hes1 cDNA templates were generated by PCR using T3 or T7 RNA promoter sequence coupled primers. The primer sequences are as follows: forward, CTTCGATAACAGCGGAATCCCC; reverse, TGTGCTCA-GAGGCTGTCTTTGG; T3-forward, TTATTAACCCTCACTAAAG-GGAAGCTTCGATAACAGCGGAATCCCC; T7-reverse, TTGTAA-TACGACTCACTATAGGGCGATGTGCTCAGAGGCTGTCTTTGG. For Hes5 in situ hybridization, a 992 bp rat Hes5 cDNA, including the whole coding region and 3′ UTR, was subcloned into the EcoRI and Hind III sites of pBluescript-SK and used as the template for generating digoxigenin-labeled probes. Sense and antisense digoxigenin-labeled RNA probes were made by in vitro transcription using T3, T7 RNA polymerase accordingly. Some of the sections were double labeled with anti-myosin VIIa antibody, as described above, following in situ hybridization procedures.

We performed phalloidin labeling to determine the number of hair cells in the cochlea as previously reported (Zheng and Gao, 1999). As shown in Fig. 1A, there were four rows of hair cells including one row of inner hair cells and three rows of outer hair cells in the cochlear surface preparations dissected from E17.5 wild-type mice. However, examination of cochlear surface preparations obtained from E17.5 Hes1^−/− mice revealed the presence of supernumerary inner hair cells (arrows in Fig. 1C). There was a mild increase in the number of inner hair cells in the Hes1^+/− mice, suggesting a gene-dose effect (arrows in Fig. 1B). Such a phenotypic change in the Hes1-deficient mice was also observed when a cytoplasmic hair cell-specific marker, anti-myosin VIIa antibody (Hasson et al., 1995; Xiang et al., 1998), was used (Fig. 1D-F). Cell counts performed from cochlear surface preparations of different genotypes are summarized in Fig. 2. There was a significant increase in the number of doublet inner hair cells along 1 mm length starting from the basal end of the cochlea in the Hes1^−/− mice (Fig. 2A; 17.92 ±3.69 s.e.m., n=13) as compared to the Hes1^+/+ (0.20±0.16, n=8, P<0.01) or Hes1^+/− litter mates (5.00±0.55, n=15, P<0.01 between Hes1^−/− and Hes1^+/+ mice). There was also a statistical difference between Hes1^+/− and Hes1^+/+ mice (P<0.01). In contrast, no statistical difference was found in outer hair cell numbers among the three types of animals (Fig. 2B, Hes1^−/−, 63.69±1.82, n=13, Hes1^+/+, 58.87±2.18, n=15, Hes1^+/+, 58.25±2.18, n=8, P=0.058 between Hes1^−/− and Hes1^+/+ mice). In general, the Hes1^−/− mice still showed three evenly patterned rows of outer hair cells (Fig. 1C,F), although sometimes a few extra hair cells were seen in the third outer hair cell row (see Fig. 3D). The entire length of the cochleae ranged from 3.6 to 4.1 mm for all genotypes. No apparent shortening of the cochlea was seen in Hes1^−/− mice.

Examination of paraffin sections of the cochlear tissue prepared from Hes1^−/− and Hes1^+/+ mice showed that while there were three outer hair cells and only one inner hair cell in Hes1^+/+ preparations (Fig. 3A), doublet inner hair cells were seen in Hes1^−/− preparations (arrows in Fig. 3B). Similarly, scanning electron microscopy revealed that instead of regular four rows of hair cells observed in Hes1^+/+ mice (Fig. 3C), there were extra inner hair cells produced in the tissue prepared from Hes1^−/− mice (arrows in Fig. 3D). These results therefore confirmed our observations with phalloidin and anti-myosin VII antibody labeling (Fig. 1) in cochlear surface preparations.

To determine whether hair cell production in the vestibular organs is also regulated by Hes1, we made serial cryostat sections of the utricles prepared from the Hes1^−/−, Hes1^+/+, and Hes1^+/− mice at E17.5, as previously described (Zheng and Gao, 1997). We then immunostained these sections with anti-myosin VIIa antibody (Fig. 4A,B) and performed total cell counts (Fig. 4C). We found a significant increase in the total number of hair cells in the utricles of Hes1^−/− mice (3193.50±143.94, s.e.m., n=4, P<0.05), as compared with those in the Hes1^+/+ mice (2347.17±80.99, n=6) (Fig. 4C). The number of hair cells in Hes1^+/− was in-between (2594.5±163.19, n=4), but the difference was not statistically significant relative to that of the Hes1^+/+ mice (P=0.17, Fig. 4C). Because there are no reliable supporting cell-specific markers available for the mammalian inner ear, the numbers of supporting cells could not be determined with sufficient accuracy in these cryostat sections and we did not perform cell counts for supporting cells.

To find out the cellular expression patterns of the Hes1 gene in the inner ear, nonradioactive RNA in situ hybridization was performed with sections prepared from E17.5 rat inner ear tissue. As shown in Fig. 5A1 and 5A2, specific labeling was seen in the supporting cell layer but not hair cell layer of the vestibular sensory epithelium. Double labeling the sections with anti-myosin VIIa antibody, a hair cell-specific marker (Hasson et al., 1995; Xiang et al., 1997), confirmed that hair cells were devoid of labeling (Fig. 5A2). Nonsensory epithelial cells in the transitional zone and in the roof showed no signals (Fig. 5A1). In the cochlea, Hes1 signal was seen in the GER and LER areas, which are adjacent to inner and outer hair cells, respectively (Fig. 5A3, 5A4). The sensory epithelium in which hair cells and supporting cells such as Deiter’s cells and pillar cells are located showed minimal labeling (5A4). Double labeling the sections with anti-myosin VIIa antibody confirmed that hair cells were essentially devoid of Hes1 signal (Fig. 5A4).

Given that Hes1 is expressed in both LER and GER, but extra inner hair cells are only formed in the GER, we wondered that expression of other genes such as Hes5 may also play a role in the control of hair cell differentiation. Hes5 has been shown to be another negative regulator of neurogenesis (Ohtsuka et al., 1999). We carried out Hes5 in situ hybridization on E17.5 rat inner ear tissue and found an overlapping, but distinct expression pattern, as compared with that of Hes1. In the cochlea, Hes5 signal was observed in the LER and supporting cells, including Deiter’s cells, and in pillar cells in the sensory epithelium, but not in the GER (Fig. 5B3, 5B4). In the utricle, strong Hes5 signal was seen in the supporting cells (Fig. 5B1). However, unlike Hes1 which is expressed in supporting cells throughout the sensory epithelium, Hes5 was expressed at high levels in the supporting cells in the striola region. Minimal signals or much lower levels were seen in the non-striola region (Fig.5B1, 5B2).

Owing to RNA degradation resulting from bone decalcification necessary for processing the mature ear for sectioning, we were unable to determine by in situ hybridization analysis the expression profile of Hes1 in the mature ear. To address whether Hes1 was developmentally regulated at late postnatal stages and in adult, we performed real-time quantitative RT-PCR analysis with RNA extracted from the entire inner ear labyrinth tissue prepared from E13.5, E15.5, E17.5, P0, P5, P15 and adult mice. As shown in Fig. 6A, Hes1 was expressed as early as E13.5. Its expression became elevated around birth and was maintained in the adult. To explore possible relationship between Hes1 and Math1 during hair cell differentiation, we also performed TaqMan analysis of Math1 expression in the inner ear at various developmental stages (Fig. 6B). In contrast to Hes1, Math1 was expressed mainly at embryonic and early postnatal stages with its expression being greatly downregulated and became minimal in the adult (Fig. 6B).

Our recent study showed that the GER cells that normally give rise to inner sulcus epithelial cells in postnatal rat cochlear explant cultures can be induced to become hair cells when forced to express Math1 (Zheng and Gao, 2000). These results are consistent with gene targeting experiments (Bermingham et al., 1999), indicating that Math1 is crucially involved in the control of hair cell differentiation. To attempt to understand the mechanisms by which Hes1 influences hair cell differentiation and whether there is any functional interaction between Hes1 and Math1, we co-transfected postnatal rat cochlear explant cultures with an equal amount of Hes1-expressing (pSV2CMV-Hes1) and Math1-expressing plasmids (pRK5-Math1-EGFP) and compared them to cultures transfected with pRK5-Math1-EGFP plasmid only or co-transfected with a mixture of pSV2CMV-Hes1 and pRK5-EGFP plasmids. Because the pRK5-Math1-EGFP and pRK5-EGFP plasmids contain the EGFP reporter gene, cells transfected with the pRK5-Math1-EGFP plasmid, a mixture of pSV2CMV-Hes1 and pRK5-Math1-EGFP plasmids, or a mixture of pSV2CMV-Hes1 and pRK5-EGFP plasmids can be easily identified in the cultures. We found that Math1 overexpression led to robust production of ectopic hair cells in the GER region (Fig. 7A-C) as shown previously (Zheng and Gao, 2000). However, the induction of hair cell differentiation by Math1 expression was dramatically prevented by co-expression of Hes1 (Fig. 7D-F). Overexpression of Hes1 alone, however, did not show any apparent effects and none of the Hes1 transfected cells became myosin VIIa positive (Fig. 7G-I). Previously, we have shown that there is a morphological change of the GER cells in the presence of Math1: conversion from an elongated, process-bearing morphology to a pear-shaped morphology (Fig. 7C) (Zheng and Gao, 2000). Consistently, the majority of the Hes1/Math1 co-transfected GER cells showed a process-bearing morphology (in Fig. 7D-F) similar to those of normal GER cells (Fig. 7G-I), suggesting that the presence of Hes1 also prevents the GER cells from undergoing a morphological change. Quantitative analysis of the cultures transfected with various plasmids revealed that while 98.6%±0.6 of the Math1-transfected GER cells became hair cells, only about 8.9%±1.1 Hes1/Math1 co-transfected GER cells converted into hair cells (Table 1). All of the Hes1 transfected GER cells remained myosin VIIa negative and did not differentiate into hair cells (Table 1).

The present study shows direct experimental evidence that Hes1 is involved in hair cell differentiation. When the Hes1 gene is deleted, there is formation of supernumerary hair cells in both the cochlea and the utricle of the inner ear. When Hes1 is overexpressed, it can block hair cell differentiation induced by a positive regulator, Math1. Therefore, Hes1 is a negative regulator of hair cell differentiation, which is similar to its role during neurogenesis (Kageyama and Nakanishi, 1997; Nakamura et al., 2000). A balance between negative regulators such as Hes1 and positive regulators such as Math1 appears to be crucial for production of an appropriate number of hair cells in the inner ear.

Our RT-PCR and in situ hybridization analyses show distinct expression patterns of Hes1, Hes5 and Math1 during normal hair cell development. Previous experiments have shown that Math1 is expressed in the presumptive sensory epithelium at early embryonic stage (E12.5), but its expression becomes restricted to hair cells, downregulated in supporting cells and absent in other non-sensory epithelial cells, including GER and LER cells at late embryonic stages (Bermingham et al., 1999). Expression of Math1 becomes minimal in the adult inner ear (Fig. 6B). Hes1, however, is expressed at low levels at early embryonic stages, and its expression becomes elevated at late embryonic and early postnatal stages, and is maintained in the adult. Within the utricular sensory epithelium, Hes1 is expressed selectively in all supporting cells, but not hair cells. Hes5, however, appears to be expressed in striola supporting cells with much lower levels in non-striola supporting cells. The presence of high levels of Hes1 and downregulation of Math1 in all supporting cells at late embryonic stages may be one of the mechanisms used to prevent formation of supernumerary vestibular hair cells; this may also explain why extra hair cells are produced in Hes1^−/− utricles.

It is interesting that Hes1 was expressed in the GER and LER cells, which are adjacent to inner and outer hair cells, respectively, yet no or minimal signals were seen in supporting cells within the cochlear sensory epithelium of the cochlea. Hes5, however, is expressed in the LER and supporting cells, including Deiter’s cells and pillar cells, but not in the GER. Classical histological studies suggest that inner hair cells are derived from the most distal GER cells and outer hair cells arise from the most proximal LER cells (Lim and Rueda, 1992). Our recent finding that the GER cells can directly convert into hair cells under experimental conditions (Zheng and Gao, 2000) provides further supporting evidence for this model. The expression of Hes1 in the GER and LER cells may serve as a mechanism to fine-tune the appropriate number of inner and outer hair cells. While Hes1 clearly contributes to a mechanism that fine tunes the number of inner hair cells, its role in regulating the number of outer hair cells is less clear and may be masked by the cooperative influence of Hes5, as Hes5 is expressed in the cochlear sensory epithelium and the LER (Fig. 5B4).

It is important to note that overexpression of Math1 can convert either GER cells of the cochlea or supporting cells of the utricle into hair cells (Zheng and Gao, 2000), even though the two populations of cells express Hes1. One possible interpretation for these results is that transfection-induced Math1 expression is much higher (see Zheng and Gao, 2000) than the physiological levels of Hes1. These results suggest that Math1 plays a dominant role in driving the cells to become hair cells. Hes1 only acts as a negative regulator to balance the activity of Math1. It remains to be determined whether overexpression of Hes1 in mature hair cells in which Math1 is downregulated or absent would influence the identity of hair cells: converting hair cells back into supporting cells.

Our findings provide further supporting evidence that Notch signaling is involved in hair cell fate determination, because Hes1 is believed to be a downstream target gene of the Notch signaling pathway based on previous findings that constitutively active form of Notch can activate Hes1 promoter and induce Hes1 expression (Jarriault et al., 1995; Hsieh et al., 1997; Nishimura et al., 1998). Activation of Notch by its exogenous ligand, Delta, is reported to induce either Hes1 or Hes5 expression in neighboring cells (Jarriault et al., 1998; Wang et al., 1998). Recent studies have provided supporting evidence for the involvement of Notch signaling mediated lateral inhibition in the control of hair cell fate determination (Corwin and Warchol, 1991; Lewis, 1991). During embryonic development, Notch1 and Notch ligands are expressed in the presumptive sensory epithelium at the onset of hair cell differentiation (Lewis et al., 1998; Lanford et al., 1999, Morrison et al., 1999; Zine et al., 2000). When the jagged2 gene which encodes one of the Notch ligands, is deleted, supernumerary inner hair cells are produced in the cochlea (Lanford et al., 1999). Similarly, interference in Notch signaling by either Notch1 or jagged1 antisense oligonucleotides results in an increase in the number of hair cells in cultures of developing cochlear tissue (Zine et al., 2000). In addition, during hair cell regeneration in chicks, Delta1, another Notch ligand, is upregulated in regenerating hair cells but downregulated in cells that do not acquire a hair cell fate (Stone and Rubel, 1999). Moreover, in the mind bomb zebrafish mutants, in which Notch signaling is disrupted, the ear sensory patches consist solely of hair cells, which are produced precociously in great excess, while supporting cells are absent (Haddon et al., 1998). These data considered together demonstrate that Notch signaling may influence the choice of a precursor cell to adopt a hair cell fate after it exits the cell cycle. Our RT-PCR analysis indicates that Hes1 is expressed during hair cell differentiation and maintained in the adult, which appears to follow the expression of Notch1. Hes1-deficient mice show a similar inner ear phenotype as compared to the jagged 2 null mutant mice, but less severely in terms of the production of total cochlear hair cells (Lanford et al., 1999). This could be due to the compensation from Hes5 in the Hes1 mutant mice, because Hes5 is expressed in supporting cells of the cochlear sensory epithelium and in the LER (Fig. 5B3, 5B4). It is also possible that expression of other genes that modulate Notch signaling (Zhang et al., 2000) in the sensory epithelium and the boundary areas can play important roles in the control of hair cell differentiation. Interactions from multiple genes may contribute together to the complex morphogenesis of the mammalian cochlea.

Based on the present experiments and previous findings mentioned above, we favor the following model for hair cell differentiation: Notch → Hes1/Hes5 –| Math1 → hair cell differentiation (Kegeyama and Nakanishi, 1997, Shailam et al., 1999). Activation of Notch would lead to the expression of Hes1/Hes5, which antagonizes Math1. Math1 acts as a dominant, positive regulator for hair cell differentiation. In the absence of or at low levels of negative regulators such as Hes1 and Hes5, Math1 would be sufficient to drive the epithelial cells within and nearby the sensory epithelium to differentiate into hair cells. When Hes1 is expressed at high levels, Math1 activity can be inhibited and hair cell differentiation is prevented. Support for the interaction between Hes1 and Math1 also comes from a previous biochemical study (Akazawa et al., 1995), in which Hes1 is shown to inhibit the Math1/E47-induced transcriptional activation. Although the exact mechanism by which Hes1 interacts with Math1 is still unknown, there could be several possibilities (Akazawa et al., 1995). First, it is believed that Math1 functions through a heterodimer with another HLH factor, E2a, which then binds to the E-box to activate transcription. Hes1 may prevent the binding of Math1 to E2a (Sasai et al., 1992) by competition, to suppress gene transcription that is necessary for cell differentiation to occur. Second, Hes1 might bind directly to promoter DNA and restrain transcription (Ohsako et al., 1994; Van Doren et al., 1994). In addition, Hes1 may bind to Groucho (McLarren et al., 2000), a transcription co-repressor that can prevent cells from differentiation (Paroush et al., 1994).

Considered together with our previous work (Zheng and Gao, 2000), we believe that the postnatal cochlear explant cultures could serve as a convenient model system for hair cell differentiation studies. When the GER cells in the cultures are forced to express Math1, they differentiate into hair cells. Transfection of the same type of cells with other genes such as Hes1 (Fig. 7G-I) or Brn3c (Pou4fc – Mouse Genome Informatics; Zheng and Gao, 2000) does not lead to production of extra hair cells. One can use this model system to determine whether another specific gene is sufficient to induce hair cell differentiation by using Math1, Hes1 and Brn3c as positive and negative controls. Moreover, when the cultures are co-transfected with Hes1- and Math1-expressing plasmids, the induction of hair cell differentiation by Math1 is blocked. Theoretically, one can co-transfect the cultures with two or more plasmids to dissect genetic pathways between genes that might regulate hair cell differentiation.

Finally, understanding the mechanisms of hair cell differentiation would be helpful for us to stimulate hair cell regeneration following injury and could eventually lead to a therapeutic treatment of hearing and balance impairments. Over the past several years, a lot of efforts have been mainly made to identify mitogenic growth factors that can stimulate proliferation of supporting cells (Lambert, 1994; Yamashita and Oesterle, 1995; Corwin et al., 1996; Gu et al., 1996; Oesterle et al., 1997; Zheng et al., 1997, 1999a; Kuntz and Oesterle, 1998; Staecker and Van De Water, 1998) and to understand the mechanisms controlling cell proliferation (Navaratnam et al., 1996; Chen and Segil, 1999; Lowenheim et al., 1999) in the inner ear sensory epithelium. On the one hand, cell proliferation appears to be needed to compensate the number of lost hair cells and the supporting cells that are capable of converting into new hair cells. On the other hand, additional signals or strategies might be required for the proliferative and non-proliferative supporting cells to successfully convert into hair cells. Our recent findings that overexpression of specific genes such as Math1 can induce hair cell differentiation in the cochlea and facilitate the conversion of supporting cells into hair cells in the utricle (Zheng and Gao, 2000) have suggested an additional approach for the production of new hair cells. In this regard, it is conceivable that downregulation of Hes1 in addition to upregulation of Math1 in the inner ear could be another strategy by which to stimulate hair cell regeneration in adults, given that Hes1 expression persists in the adult inner ear. A combination of mitogenic supporting cell growth factors and upregulation/ downregulation of specific genes involved in the control of hair cell differentiation could be a more effective way to stimulate hair cell regeneration in the inner ear.

We thank Arnon Rosenthal for his helpful discussions, Tama Hesson for her gift of anti-myosin VIIa antibody, Linda Rangell and Gilbert Keller for scanning electron microscopy, Meg Kelly for animal breeding, Patti Tobin for paraffin sectioning, and Allison Bruce for preparation of the figures.