Atoh1 and Neurog1 belong to an ancient family of basic Helix-Loop-Helix (bHLH) genes that are involved in cell fate determination across phyla and systems, typically being expressed in proliferating precursors. For example, in coelenterates, an ATONAL-like protein determines fate between muscle and neurons (Seipel et al.,2004). In mammals, various bHLH genes determine the relative frequency of distinct retinal cell types (Akagi et al.,2004; Kim et al.,2005), regulate the size and progression of precursor fate in the olfactory epithelium (Wu et al.,2003), are involved in cell fate determination of neurons and switch between neuron and glial cell fates in the brain (Anderson et al.,1997; Bertrand et al.,2002), and determine the fate of progenitors of the gut epithelium (Yang et al.,2001). Except for continuously proliferating systems, these bHLH genes seem to be downregulated in neonatal mammals within days after birth (Akazawa et al.,1995). As in these other neuronal and non-neuronal systems, at least three bHLH transcription factors are involved in the development of hair cells and sensory neurons of the ear: Neurog1 is essential for sensory neuron development (Ma et al.,1998), Atoh1 is essential for hair cell differentiation (Bermingham et al.,1999), and Neurod1 is required for sensory neuron survival (Liu et al.,2000; Kim et al.,2001). In apparent agreement with other developing systems in which cell fate alterations have been reported after one or more of these genes have been mutated (Akagi et al.,2004), Neurog1 null mutants show not only loss of all sensory neuron formation but in addition a severe reduction of hair cells in several sensory epithelia (Ma et al.,2000). As reported for seemingly comparable bHLH gene interactions in the eye, spinal cord, and the olfactory system (Gowan et al.,2001), it was hypothesized that this hair cell reduction indicates a molecular interaction of NEUROG1 with undisclosed proteins (Ma et al.,2000) in some common precursors of both cell types (Fritzsch et al.,2000; Fritzsch and Beisel,2003).
In apparent contradiction to this idea are recent data from an elegant study, which suggest that Atoh1 is a differentiation factor of postmitotic hair cells that plays no role in hair cell precursor selection in the cochlea (Chen et al.,2002). These conclusions were obtained using an Atoh1 promoter fragment to drive eGFP (Gowan et al.,2001) as well as immunocytochemistry for ATOH1; both sets of data did not show Atoh1 signal prior to cessation of proliferation of hair cell precursors. However, the transgenic construct used in this study requires the presence of Atoh1 protein to be upregulated and may not show the earliest expression of Atoh1 transcript. Low levels of protein may not be detected using immunocytochemistry. Indeed, previous and more recent work has shown somewhat earlier Atoh1 expression using other techniques (Bermingham et al.,1999; Lanford et al.,2000; Zine et al.,2001; Woods et al.,2004). Other data showed an even earlier upregulation of Atoh1-eGFP in some vestibular hair cells at a time when both sensory neurons and hair cells become postmitotic in the vestibular system (Radde-Gallwitz et al.,2004). However, the onset of expression of Atoh1 in the ear has not been investigated using more sensitive techniques such as RT-PCR or Q-PCR. It seems, therefore, fair to say that at the moment we do not know when Atoh1 is first upregulated and what cell type is initially expressing Atoh1.
While Atoh1 is clearly involved in hair cell differentiation (Izumikawa et al.,2005), it appears that sensory epithelia composed of undifferentiated cells that express Atoh1-LacZ and Bdnf-LacZ form even in Atoh1 null mice and receive afferent and efferent innervation (Fritzsch et al.,2005a). In addition, recent data (Radde-Gallwitz et al.,2004) support previous results on expression of the transcription factor Islet1 in differentiating vestibular sensory neurons and hair cells (Adam et al.,1998). These Islet1 data agree with information gathered with the neurotrophin BDNF; Bdnf mRNA is also expressed in differentiating hair cells and sensory neuron precursors that appear to delaminate from the future sensory epithelia (Farinas et al.,2001; Fritzsch et al.,2002). Together these data are compatible with the idea that some sensory neurons and hair cells may be derived from common neurosensory precursors (Fritzsch and Beisel,2004). Indeed, recent lineage tracing work in chicken has shown some lineage relationships of sensory neurons and hair cells (Satoh and Fekete,2005).
These data leave several questions unanswered: It remains unclear whether the detection of the earliest appearance of Atoh1 expression is compromised by the sensitivity of the employed technique, leading to false-negative results. For example, it is possible that the delay in generating eGFP in the Atoh1 transgenic mice may not show the earliest expression of ATOH1 protein (needed to drive the promoter) and certainly will not show the earliest upregulation of Atoh1 message. Likewise, limited expression of Atoh1 mRNA may be below the detection threshold of in situ based techniques and thus may lead to false-negative results suggesting, for example, the absence of Atoh1 expression in neurosensory precursors.
For this study we had six aims: First, we wanted to compare Bdnf expression in Atoh1 and Neurog1 simple and double null mice. Second, we wanted to determine terminal mitosis as revealed by BrdU in wildtype and Neurog1 null mice. Third, we wanted to identify the earliest expression of Atoh1 in control and Neurog1 null mice using various techniques. Fourth, we wanted to study the expression of Atoh1 as revealed by Atoh-Cre expression in Rosa 26 mice to provide evidence supporting the idea that some hair cells and sensory neurons are progeny of common neurosensory cells. Fifth, we wanted to show the expression of Atoh1 in vestibular and spiral sensory neurons as revealed by Atoh1-eGFP, Atoh1-LacZ staining, and Q-PCR, suggesting that this labeling is not an integration artifact (Lumpkin et al.,2003). Sixth, we wanted to determine the significance of this expression in sensory neurons using a biological readout system and show whether absence of Atoh1 may affect path-finding properties of sensory neurons. Lastly, we wanted to develop a model that can explain how Neurog1 absence mediates hair cell reduction. For this model, we assume a possible NEUROG1-ATOH1 protein interaction in common precursors of hair cells and sensory neurons (neurosensory precursors) and that Neurog1 absence skews cell fate decisions of common neurosensory precursors toward hair cells.
MATERIALS AND METHODS
Mouse Lines Used to Generate Single and Compound Mutant Mice
Two Atoh1 null alleles were used in this study. One allele carried a Hprt cassette (Atoh1-Hprt) and the other one carried the LacZ reporter gene (Atoh1-LacZ) instead of the coding sequence (Ben-Arie et al.,1997,2000). To analyze the ATOH1/NEUROG1 interaction further, we generated doubly heterozygotic mice that carried the Neurog1 null allele (Ma et al.,1998) in combination with the Atoh1-Hprt allele and crossed this line with a line carrying the Bdnf null allele combined with LacZ replacement (Bdnf-LacZ) previously described (Jones et al.,1994; Fritzsch et al.,2005a).
Mice with a LacZ or Hprt marker were bred from heterozygotes as previously described (Bermingham et al.,1999; Fritzsch et al.,2005a). Embryos were collected at embryonic day 11.5 (E11.5), E12.5, E16.5, and E18.5 and several heterozygotic animals were fixed at 2, 5, and 9 months of age. Older embryos and adults were perfusion or immersion fixed, after appropriate anesthesia, with 4% PFA for 30 min. Ears were rapidly dissected in PBS, and subsequently reacted for Lac-Z histochemistry overnight at room temperature (Farinas et al.,2001). Some of these ears of heterozygotic and homozygotic Atoh1 mutants were also immunostained for nerve fiber stain using tubulin antibodies as previously described (Fritzsch et al.,1997) and anti β-galactosidase. All animal breeding and treatment was approved by an IACUC protocol no. 0630.
In older embryos, the neurotrophin Bdnf is expressed only in differentiated hair cells (Farinas et al.,2001) but may play a role in bHLH gene upregulation in earlier stages (Ito et al.,2003). We studied the expression of Bdnf-LacZ distribution in control, Atoh1, and Neurog1 null littermates. To achieve this, we crossed Atoh1-Hprt/Neurog1 doubly heterozygots with Bdnf heterozygote animals. In the offspring, we selected an Atoh1-Hprt/Neurog1/Bdnf-LacZ compound heterozygote male and bred this animal with Atoh1-hprt/Neurog1 heterozygote females. The resulting offspring produced Atoh1-Hprt null mutants with Bdnf-LacZ heterozygosity at the expected frequency of 1 in 8 embryos. Ears of E11.5, E12.5, and E18.5 perfusion fixed animals were reacted for β-galactosidase histochemistry as described above. Embryonic mice were from timed pregnancies and were, in addition, staged (Theiler,1989).
In these animals, we also investigated the distribution of cells in which the terminal mitosis was visualized with BrdU injections (100 μg/g). Injections were administered for each litter only once, at noon, to a given pregnant mouse of days E9.5–17.5 (with noon after vaginal plug considered as embryonic day 0.5). Animals were perfusion fixed at E18.5, genotyped and the ears processed for BrdU imaging. Briefly, ears were treated for 1 h with 1N HCl, washed, blocked with bovine serum, and incubated for 1 h with anti-BrdU antibody conjugated to Alexa 647 (Invitrogen). Ear epithelia were mounted flat on a slide in glycerol, coverslipped, and viewed with a confocal microscope (Biorad Radiance 2000 or Zeiss LSM). Assessment of labeled cells was achieved by first focusing on the surface of the organ of Corti using the 488-nm laser band. Subsequently, stacks of images were collected through the hair cells and supporting cells to include the nucleus of hair cells, which appeared as a black hole against the autofluorescence of the cytoplasm. In a second channel, the emission of the Alexa 647 was recorded, showing the distribution of BrdU in the nuclei. Nuclei were scored as labeled when 50% or more of the surface was stained in the Z-axis collapsed stacks. Identification of nuclei was supported by the fact that in most cases at least some minor specks of BrdU labeling appeared. Following this high-power analysis using a 40× 1.3 NA lens, we took overview images to illustrate the distribution of BrdU-labeled cells for illustration purposes only. In selected cases, we verified that the cells identified as hair cells were indeed hair cells by showing co-localization of BrdU in nuclei of cells labeled with Bdnf-β-galactosidase reaction product. To enhance imaging, the β-galactosidase reaction product was photoactivated to yield a 488-nm excitable fluorescing product (Matei et al.,2005).
Some wildtype and Atoh1 null mutant embryos were also used for neuronal tract tracing as previously described (Fritzsch et al.,2005a) to study the possible effect of Atoh1 absence on the pattern of innervation. In addition, some ears of Atoh1 null and wildtype animals were reacted with acetylated tubulin to reveal the pattern of innervation. Ears were whole mounted in glycerol and viewed with the Zeiss LSM confocal system.
The bHLH gene Neurod1 (formerly NeuroD) is expressed in all sensory neurons and in some differentiated hair cells in embryos older than E14.5 (Kim et al.,2001). We studied the Neurod1-LacZ expression in control and Neurog1 null littermates. For this, we crossed Neurog1 heterozygotes with Neurod1-LacZ heterozygotic mice. In the offspring, we selected Neurog1/Neurod1-LacZ compound heterozygote males and bred these animals with Neurog1 heterozygote females. We bred two litters of E12.5– E18.5 embryos in which we had two offspring that were Neurog1 null mutants and that were also heterozygous for Neurod1-LacZ. Ears of these E18.5 embryos were perfusion fixed and reacted for β-galactosidase histochemistry as described above.
After β-galactosidase reaction, ears were dissected and mounted flat for analysis of the sensory epithelia area. Some ears were embedded in soft epoxy resin, sectioned at 20 μm thickness, and analyzed in a compound light-microscope (Nikon Eclipse 800). Images were captured using a CCD camera and processed using ImagePro or using the Zeiss LSM multiphoton confocal system and 2 photon activation (Matei et al.,2005). Images were compiled into plates using CorelDraw.
To generate this line, we used the JQ2-Atoh1 promoter fragment originally described by (Gowan et al.,2001) and used to analyze Atoh1 upregulation in developing hair cells (Chen et al.,2002) provided by Dr. J. Johnson. This fragment was ligated to the bacteriophage P1 Cre recombinase cDNA and injected into mouse embryo pronuclei to generate two transgenic mouse lines that carried the Atoh1-Cre. Integrity of the construct was checked by Southern blot. Transgenic animals were mated with Rosa 26 lines to demonstrate expression of Cre using the conditional LacZ reporter system of this line. We generated two transgenic strains of mice with this construct. Mice of one strain were fixed in 4% paraformaldehyde at E12.5 and E18.5, mice of the second strain were fixed at E11 and at 1 month of age, the latter after lethal anesthesia by perfusion. These stages were chosen to capture the earliest expression detectable with these techniques as well as the long-term expression. Ears were dissected and either reacted as whole mounts for β-galactosidase or sectioned and reacted for β-galactosidase. Ears and sections were viewed in an Olympus EZ dissection scope or a Nikon Eclipse 800 and images were taken using a Coolsnap camera and Metamorph software. We also used a newly developed technique to photoactivate the BCI reaction product of X-Gal to enhance visibility (Matei et al.,2005).
RT-PCR was done on dissected inner ear tissues at E10.5 and E11.5 and Q-PCR was done on vestibular ganglia dissected from P0 and P10 ears using gene-specific primers according to a published procedure (Beisel et al.,2000) to determine the absence or presence of Atoh1 transcripts. Care was taken that this tissue was not contaminated with adjacent brain tissue known to express Atoh1. The samples were treated with DNase to eliminate any trace genomic DNA. Afterwards, the quality of the RNA was assessed using the Agilent 2100 Bioanalyzer. Primer/probe sets were designed over introns to eliminate any possible amplification from contaminating genomic DNA. In the reactions themselves, no Template Control samples were used and none of these amplified a PCR product. First-strand synthesis was done using a T7-dT(24) primer (Beisel et al.,2000). A primary amplification was done using mAtoh1-896 FOR (5′ CATCACCTTCGCACCGCCTC CTC 3′) and a T7-specific oligomer at standard PCR conditions with 35 cycles. A secondary amplification step was done using the nested primer sets of mAtoh1-1094 FOR (5′ CTAACA GCGATGATGGCACAGAAG 3′) and mAtoh1-1777 REV (5′ AGTGATGAAGTGCGTGTATT CTGG 3′) with 30 cycles. A primary amplification was done using mAtoh1-896 FOR (5′ CATCACCTTCGCACCGCCTCCTC 3′) and a T7-specific oligomer. A secondary amplification step was done using the nested primer sets of mAtoh1-1094 FOR (5′ CTAACAGCGATGATGGCACAGAAG 3′) and mAtoh1-1777 REV (5′ AGTGATGAAGTGCGTGTATTCTGG 3′. We also dissected vestibular ganglia of newborn (PO; N = 3) and 7-day-old mice (P7; N = 2) and processed these ganglia for Q-PCR. We used the primers outlined above and compared the Atoh1 transcripts with Neurod1 (a bHLH gene present in sensory neurons), Neurog1 (a bHLH gene absent in postnatal sensory neurons), and ribosomal RNA.
Comparison of Bdnf Expression in Atoh1 and Neurog1 Simple and Double Null Mice
We previously showed that Neurog1 null mice lose not only all sensory neurons but also show loss and reorganization of hair cells in all sensory epithelia (Ma et al.,2000). We also showed that undifferentiated cells that express Atoh1-LacZ develop in Atoh1 null mice and attract afferent fibers likely because many of these cells co-express Bdnf (Fritzsch et al.,2005a), a known major attractor for growing ear sensory neuron processes (Tessarollo et al.,2004). Here we wanted to analyze how the temporal expression of Bdnf is altered in the absence of Atoh1, Neurog1, or both bHLH genes using Bdnf-LacZ as a marker for hair cells and sensory neuron precursors, possibly including common neurosensory precursors.
The following differences and similarities in Bdnf-LacZ expression were apparent in the E12.5 and E18.5 ears of these mutants. First, the canal cristae and the apex of the cochlea showed roughly identical expression of Bdnf, regardless of age or genetic background and whether differentiated hair cells existed or not (Fig. 1A–K). Second, delaminating sensory neurons showed expression of Bdnf as they exited from the utricle, saccule, and ductus reunions/basal turn of the cochlea of Atoh1 null and wildtype embryos (Fig. 1E,F). As expected, no Bdnf-positive sensory neurons were visible outside the otocyst in Neurog1 null mice at any time during development (Fig. 1C,G). Interestingly, delaminating cells were clearly positive for Bdnf-β-galactosidase in Atoh1-Hprt null mice despite the fact that no Bdnf was expressed in the areas they delaminated from (Fig. 1F). This Bdnf positivity suggests that sensory neuron precursors are capable of Bdnf upregulation as they delaminate, but does not exclude the possibility that Bdnf is already present in wildtype neurosensory precursors, giving rise to Bdnf-positive hair cells and sensory neurons. Lastly, there was a conspicuous lack of expression of Bdnf in the basal turn, utricle, and saccule of Atoh1 null mutants (Fig. 1B,F). This absence of Bdnf expression in Atoh1 null mutants contrasted sharply with the profound expression of Bdnf in distinctly labeled cells of the utricle in Neurog1 null mice (Fig. 1F,G). We show Bdnf-LacZ-positive cells in the utricle of E12.5 Neurog1 null mice (Fig. 1G) and Bdnf-LacZ-positive cells in the ductus reuniens of E18.5 Neurog1 null mice (Fig. 1C). The frequency of these cells and their distribution was variable and we are currently investigating whether such cells are hair cells by crossing Neurog1 with Atoh1-LacZ.
Since sensory neurons, but not hair cells of utricle and saccule, express Bdnf in Atoh1 null mutants (Fig. 1B,F), we generated mice that were homozygotic null for both Atoh1 and Neurog1 combined with Bdnf-LacZ heterozygosity. These mice showed the combined phenotype of Neurog1 null mice (shortened cochlea, reduction of saccular recess, absence of sensory neurons; Fig. 1D; Table 1) and Atoh1 (absence of Bdnf expression in utricle, saccule, and basal turn of cochlea; Fig. 1D).
Table 1. Mutation of Atoh1, Neurog1, and double Atoh1/Neurog1 Affects the Size and Length of Cochlea and Anterior Crista as Judged by Measuring the Area of Bdnf Expression and/or Total Length
Atoh1/WT (% of length)
Neurog1/WT (% of length)
Atoh1-Neurog1 double null
Atoh1-Neurog1 T° ler (% of length)
5,495 ± 68
3,722 ± 37
2,356 ± 29
1,329 ± 18
5,776 ± 86
5,407 ± 78
3,083 ± 48
3,003 ± 76
237 ± 21
125 ± 19
193 ± 44
79 ± 32
608 ± 35
451 ± 38
413 ± 35
225 ± 86
107,350 ± 193
36,487 ± 283
58,003 ± 175
12,755 ± 88
N = 6
N = 6
N = 4
N = 4
The data on more prominent expression of Bdnf in cells of the utricle of Neurog1 null mice at E12.5 are compatible with at least three scenarios: (1) hair cell precursors exit the cell cycle in younger animals and upregulate Bdnf; (2) neurosensory precursors convert to a hair cell fate and remain positive for Bdnf; (3) sensory neurons fail to delaminate, remain as Bdnf-positive cells, that eventually die and are replaced by hair cells forming later. In each case, more cells should be positive for Bdnf in the utricle of Bdnf null mice at an earlier stage, precisely what we demonstrated (Fig. 1G). The assumption of earlier exit from the cell cycle of hair cells can be verified by demonstrating alterations in the time of terminal mitosis of hair cells in the Neurog1 and Atoh1null mice, our next experiment. Such data can also show whether cells that exit the cycle at this early stage remain or die. If such cells would die, BrdU-positive hair cells should not be found at later stages after an early embryonic injection of BrdU. Such hair cells would have replaced the dead cells by cells without BrdU labeling, thus allowing us to distinguish between the three possible scenarios.
Terminal Mitosis as Revealed by BrdU in Wildtype and Neurog1 Null Mice
We injected BrdU once in pregnant mice ranging between E9.5 and E17.5 and analyzed the distribution of cells that showed BrdU-Alexa 647 antibody labeling of at least 50% of the area of a given nucleus at E18.5. We used the simple fact that differentiated hair cells form at this stage one row of inner and three rows of outer hair cells to identify hair cells. As is obvious from sections of the cochlea using our hair cell marker Bdnf-LacZ (Fig. 2A), hair cell nuclei are distinctly distributed from the supporting cell nuclei. Thus careful focusing of a flat mounted cochlea allows clear identification of hair cells and their nuclei using autofluorescence of the cytoplasm in combination with nuclear size and distribution to distinguish between the large hair cell nuclei and the smaller nuclei of cells of the greater epithelial ridge (GER; Fig. 2A,B) and lesser epithelial ridge (LER; Fig. 2A,B). Such a criterion provided information on cells that had their terminal mitosis shortly after the time of injection as clearance of BrdU takes around 2 h. Analysis in other systems has shown that the cell cycle length changes over the time of our analysis from about 8 h around E9.5 to about 12 h around E16.5 (Calegari and Huttner,2003). Any continued cell cycling, therefore, would have reduced the BrdU content of the nuclei to less than 50% (Fig. 2B) within 8–12 h after the injection. Such reduction of BrdU results in specks of BrdU-Alexa 647 labeling that amounted to less than 50% of a given nucleus (Fig. 2B). This technique provides a clear picture of the distribution of cells that underwent terminal mitosis shortly after BrdU injection and remained mitotically quiescent afterwards. This technique will also reveal decreasingly labeled cells derived from precursors labeled at the time of BrdU application but that continue to cycle. Such decrease in labeling will allow detailing further the topological progression of cells that undergo terminal mitosis at earlier or later stages. Of course, cell death would eliminate such cells and their BrdU labeling. The whole mount approach used here will allow identification of cells using histological criteria established over 100 years ago (Ruben,1967). In our hands, some of the commonly used antibodies to highlight hair cells showed artifacts or no staining above background, probably due to the acid treatment necessary to uncover the BrdU epitope for the antibody recognition of BrdU. In addition, we also used in selected cases Bdnf-LacZ heterozygotic animals, photoactivated the β-galactosidase reaction product that is only in hair cells in the basal turn, and showed directly that the cells we identified as hair cells based on topology and cytology criteria are indeed hair cells (Fig. 2A,E).
Using these criteria in whole-mounted ears, we detected labeled spiral sensory neurons in a base-to-apical progression and hair cells in an apex-to-base progression as previously described (Ruben,1967). Specifically, injections of BrdU at E9.5 and E10.5 and imaging at E18.5 showed label in only a few neurons in the basal turn with only punctate labeling in the apex (data not shown). Injections at E11.5 and imaging the distribution of nuclear staining at E18.5 showed that neurons were labeled in the basal and middle spiral turn (Fig. 2C). In contrast, neurons in the apex had less than 50% filled nuclei, suggesting that they had at least one or more additional mitosis after the BrdU application. Using the same criterion, we found only a few labeled hair cells in the apical-most tip of the organ of Corti. However, a trail of less labeled hair cell nuclei extended throughout the middle turn almost to the basal hook (Fig. 2C).
Injections of BrdU at E12.5 labeled only neurons in the apex of the cochlea and many hair cells in the apical half turn of the organ of Corti (Fig. 2B,D). As in earlier stages, there was a sharp boundary toward the apex of labeled hair cells whereas the boundary toward the basal turn was fuzzy with many cells showing spotty BrdU labeling. Injections at E13.5 labeled very few spiral sensory neurons and showed labeled hair cells extending throughout the middle and basal turn. These injections also showed profound but spotty labeling of other cell types throughout the cochlea, making it difficult to identify the labeled hair cells at low magnification. However, using the above-described procedure of generating a single focal stack through the hair cells alone, we were able to identify a few hair cells near the basal tip that showed profound nuclear labeling. Injections at E14.5 labeled no sensory neurons and only a few hair cells in the most basal tip of the organ of Corti where the rows of outer hair cells become reduced (data not shown).
These data suggest that in the apex, the first hair cells become postmitotic around the time the last spiral sensory neurons become postmitotic, with an apparent spatial and temporal overlap. In contrast, in the base, hair cells become postmitotic up to three days after spiral sensory neurons. These data suggest that, overall, sensory neurons and hair cells are among the first cells to exit cell cycling in opposing gradients as previously demonstrated (Ruben,1967).
Neurog1 Null Mice
We next investigated the effects of Neurog1 on time of hair cell terminal mitosis in Neurog1 null littermates of our wildtype analyses presented above. In animals injected at E9.5 to E11.5 and analyzed at E18.5, we found no labeling in sensory neurons, consistent with previous data that suggest complete abrogation of any sensory neuron formation in Neurog1 null mice (Ma et al.,1998,2000). However, in the middle to apical turn of the shortened and widened organ of Corti of these mutants, we found numerous hair cells in which the nuclei were at least 50% filled with the BrdU antibody fluorescence (Fig. 3B,C). Injections after E13.5 did not reveal any hair cells with labeled nuclei, suggesting that the entire phase of hair cell proliferation was shifted in Neurog1 null mutants to an approximately 1–2 days earlier start of hair cell terminal mitosis compared to wildtype littermates (Fig. 3A). Data in the vestibular endorgans were less obvious, presumably because of the prolonged time over which hair cells are generated in these epithelia (Ruben,1967) and the lack of a simple linear progression of terminal mitosis along one axis, as in the cochlea. In summary, the data on the cochlea suggest that abrogation of Neurog1 leads to a terminal mitosis of hair cells in the cochlea that temporally overlaps with the time of terminal mitosis of sensory neurons in wildtype littermates.
We next wanted to confirm and extend this finding with other markers that show a delayed expression in hair cells compared to sensory neurons. Neurod1 is upregulated in sensory neurons immediately after Neurog1 (Ma et al.,1998) but appears only late in development in some hair cells, progressing from base to apex (Kim et al.,2001). We, therefore, bred mice that were null for Neurog1 and heterozygotic for Neurod1. Wild type E18.5 littermates showed no Neurod1-positive hair cells in the apex, and had only few inner hair cells displaying Neurod1-mediated β-galactosidase staining (Fig. 3D,F). In contrast, Neurog1 null mice showed numerous Neurod1-β-galactosidase-positive hair cells in the apex that were scattered across the multiple rows of hair cells of these mutants (Fig. 3E,G). As previously shown (Fig. 1A,C), the absence of Neurod1-positive sensory neurons (Fig. 3D) results in the Neurog1 null mice in a much tighter coiling of the cochlea. At the same magnification, we can therefore show the entire apex of the Neurog1 null mice but not of the wildtype littermate (Fig. 3D,E). These data show upregulation of Neurod1 in many more hair cells throughout the organ of Corti of E18.5 Neurog1 null mice (Fig. 3F,G). Such data are consistent with the idea that at least some of the hair cells that undergo 1–2-day earlier terminal mitosis are derived from neurosensory precursors that cannot differentiate into neurons in the absence of Neurog1 but become hair cells instead. Other genes with similar expression profiles are needed to verify whether this is peculiar to Neurod1 or is a general effect of NEUROG1 protein abrogation.
In conclusion, these data confirm previous findings that absence of Neurog1 affects not only sensory neuron formation but also hair cell formation (Ma et al.,2000). We now show that absence of Neurog1 leads to a terminal mitosis of hair cells up to two days earlier and results in upregulation of Neurod1 in many more hair cells of E18.5 embryos throughout the cochlea. These data are consistent with the notion of a Neurog1/Atoh1 interaction in the common neurosensory (hair cell/sensory neuron) precursors to regulate the terminal mitosis and cell fate of cell types derived from these proliferating precursors. A simple way to explain these effects of Neurog1 null mutation would be that ATOH1 protein-mediated onset of differentiation interferes with the cell cycle progression thus resulting in premature terminal mitosis of hair cells and truncation of the proliferative capacity of those sensory neuron/hair cell precursors. More complex interactions would also be compatible with these data such as paracrine interactions between neuronal and sensory precursors. We, therefore, next wanted to see whether Atoh1 is expressed early enough to be in mitotically active neurosensory precursors and whether this expression changes in the Neurog1 null mice, thus supporting the idea that some sensory neurons of the mammalian ear show coexpression of Neurog1 and Atoh1, which interact either directly or through yet-to-be-determined intermediaries to allow full clonal expansion of their common and/or separate precursors.
Earliest Expression of Atoh1 in Control and Neurog1 Null Mice as Revealed by Various Techniques
We investigated whether the temporally shifted terminal mitosis of hair cells of Neurog1 null mice also lead to an upregulation of Atoh1. This would indicate not only that the onset of terminal mitosis of hair cells is accelerated but also that ATOH1-mediated hair cell differentiation in Neurog1 null mice is accelerated compared to wildtype mice. We studied the expression of Atoh1 in the ear of E11- to E13.5-old mice. In the wildtype littermates, the E12.5 in situ data showed a very faint staining near the forming vestibular epithelia. No differences other than a reduction in size of vestibular epithelia were observed in the Neurog1 null littermate and wildtype control animals. However, a difference was observed in the cochlea of E13 Neurog1 null mice compared to age-matched littermates. In the Neurog1 null mice, we obtained an Atoh1 signal in the cochlear duct that was strongest near the upper middle turn, fell off steeply near the apex, and tapered off in the lower basal turn (Fig. 4C). In contrast, the wildtype littermates showed a fainter in situ signal exclusively in the basal turn (Fig. 4B). Our β-galactosidase stain in Atoh1-LacZ heterozygotic animals showed a comparable temporal and spatial onset of Atoh1 expression to Atoh1 detected by in situ hybridization (Fig. 4E). We also obtained some Atoh1-eGFP expressing embryos (courtesy of Dr. P. Chen and N. Segil) and could not detect any eGFP signal in the cochlea of mice that were at least one day older than the mice of our lines (Fig. 4A), thus confirming their reported observations (Chen et al.,2002).
We next checked whether absence of eGFP signal in these lines is due to a different timetable of Atoh1 upregulation or due to delayed eGFP upregulation owing to the construct. Our Atoh1-LacZ-positive ears showed a weak signal in the basal turn of the cochlea (Fig. 4E) in an animal in which we could not detect any eGFP, suggesting that this technique fails to detect the Atoh1 upregulation visible with the other two techniques employed here. Most importantly, these data suggest that in our mouse lines there is a potential spatio-temporal overlap of detectable Atoh1 expression in proliferating precursors only in the base of the cochlea, but not in the middle turn and apex, supporting previous suggestions (Chen et al.,2002).
We also tested whether we could detect Atoh1 signal using PCR technique. All three samples of E11.5 embryonic ears exhibited Atoh1 expression (data not shown), while 1 out of two samples from E10.5 embryos were positive for Atoh1 (Fig. 5A). The RNAse treated controls did not have any amplification products. These data suggest that the appearance of Atoh1 transcripts in the developing inner ear is at or around E10.5. This is the earliest time Atoh1 is expressed in detectable amounts in the developing ear.
In summary, these data imply that Neurog1 expression affects spatial upregulation of Atoh1 in apical hair cells that exit the cell cycle about two days earlier. In the absence of Neurog1, the hair cells that show this earlier terminal mitosis also show spatial upregulation of Atoh1. However, the gradient does not follow the previously reported pattern of upregulation of Atoh1 in wildtype mice and progresses instead from near the apex toward the base. Interestingly, Atoh1 expression shows little temporal alteration (only about ½ day in the apex) compared to terminal mitosis (two days), suggesting that those two events are not tightly coupled. We have recently identified several other bHLH genes in the ear using PCR and their expression profiles could show a more profound upregulation in Neurog1 null mice, something we are actively investigating. Compared to PCR data, our histological techniques do reveal expression of Atoh1 at least 24 h later (E10.5 vs. E11.5), suggesting that such techniques are less than appropriate to exclude expression of low levels of Atoh1 in common neurosensory precursors. We, therefore, attempted to show such expression using yet another, more sensitive technique, Atoh1-Cre expression in Rosa 26 mice.
Expression of Atoh1 as Revealed by Atoh-Cre Expression in Rosa 26 Mice: Evidence That Some Hair Cells and Sensory Neurons May Be the Progeny of Common Neurosensory Cells
Cre-mediated expression of the LacZ reporter results in the Rosa 26 line in β-galactosidase staining wherever the CRE-recombinase was and is active (Ohyama and Groves,2004). Recently, an Atoh1 promoter fragment used to drive eGFP (Gowan et al.,2001) was engineered by us to drive Cre. If Atoh1 is expressed in neurosensory precursors, the Atoh1-Cre should show β-galactosidase staining in sensory neurons as well as hair cells. We investigated the ears of E 11, E12.5, and E18.5 and P7 of two Atoh1-Cre/Rosa 26 mice. In E12.5 embryos of one Atoh1-Cre line, we found expression of β-galactosidase only in single cells in what appeared to be several vestibular sensory epithelia patches of the utricle, saccule, anterior, horizontal, and posterior crista. In the cochlea, staining was in the medial wall of the basal and middle turn (Fig. 4D). In addition to these intraepithelial cells, β-galactosidase-positive cells were found outside the inner ear epithelia in the vestibular ganglion and the forming spiral ganglion (Fig. 4D). In the hindbrain and midbrain, Atoh1-Cre-mediated β-galactosidase stain displayed a pattern consistent with the expression of Atoh1 as revealed in the same aged Atoh1-LacZ brains (data not shown). This expression was verified in a total of four ears and two brains of E12.5 mice.
The second Atoh1-Cre line showed a more profound β-galactosidase staining and we chose this line to evaluate the earliest Atoh1-expression, enhancing the β-galactosidase reaction product with a recently developed photoactivation process (Matei et al.,2005) and imaging with the confocal microscope. These data showed that as early as E11, cells are Atoh1-Cre-positive and appear in distinct areas of the ear that appear to correspond to future sensory epithelia (Fig. 4G–I). In addition to cells in the otocyst wall, we also found a few cells distributed among the differentiating vestibular sensory neurons (Fig. 4I). Labeling of neurons with acetylated tubulin revealed that developing nerve fibers were near the Atoh1-Cre-positive cells. These data suggest that this technique allows detection of Atoh1 in tissue as early as approximately 12 h after we can first detect Atoh1 using RT-PCR.
Some positive cells were also found outside the sensory epithelia in Atoh1-eGFP ears (Fig. 4A) suggesting that the expression of Atoh1 is possibly transient and thus cannot be detected with in situ hybridization or Atoh1-LacZ. Whether the Atoh1 fragment-mediated signal is due to an expression in common neurosensory precursors or reflects an independent and transient upregulation in delaminating sensory neurons in addition to hair cell precursors remains unclear as we could not detect it in proliferating cells.
At E18.5, β-galactosidase was found in almost all hair cells of all sensory epithelia, with the exception of the most apical tip of the cochlea (Fig. 6A,B). In addition, Atoh1-Cre-mediated LacZ expression was found in spiral ganglion neurons, vestibular ganglion neurons, pillar cells, and other supporting cells (Fig. 6B). We also investigated this expression of Atoh1-LacZ in hair cells of E18.5 Neurog1 null mice combined with Atoh1-LacZ heterozygosity and obtained comparable results. Combined, these data suggest that the most apical tip of the cochlea, the area in which the first hair cells become postmitotic, has a profound delay of Atoh1 upregulation of several days, in particular in wildtype animals, but also in Neurog1 null mice. Cells other than hair cells were also Atoh1-Cre-β-galactosidase positive. These cells were particularly frequent in the forming spiral prominence (Fig. 6A,B). In addition, few supporting cells of the cochlea and the vestibular epithelia were also β-galactosidase-positive. Interestingly, some cells of the non-sensory epithelium of the ear were also β-galactosidase-positive. These cells were particularly prominent in the extension of the saccule into the endolymphatic duct (Fig. 6A,B) but also as an extension of the cruciate eminence of the canal cristae into the canals, in particular the anterior and posterior canal.
The ears of the 1-month-old second Atoh1-Cre-Rosa 26 line (P31) also showed expression other than the hair cells of the sensory epithelia (Fig. 7B,C,G). This expression showed three prominent distribution patterns, sensory neurons, supporting cells in the sensory epithelia (in particular pillar cells; Fig. 7G) and non-sensory cells of the ear outside the sensory epithelia such as a few cells in the stria vascularis (Fig. 7B). Almost all sensory neurons were positive in both the spiral ganglion and the vestibular ganglion in this line with little staining elsewhere (Fig. 7B,C) with labeling being so prominent that the nerve fibers showed a faint β-galactosidase reaction product (Fig. 7C).
These data obtained from two Atoh1-Cre transgenic lines suggest that Atoh1-Cre-expressing precursors may not only give rise to hair cells but also to sensory neurons. Since positive cells are found within 12 h of the first possible detection of Atoh1 transcript inside and outside the otocyst, we suggest that these cells are progeny of Atoh1-positive precursors. Other cells found positive in both lines are some supporting cells and non-sensory epithelia of the ear. We will next focus on the possibility that Atoh1 is indeed expressed in sensory neurons where our two Atoh1-Cre lines showed strong expression (Figs. 6A,B, 7B,C).
Expression of Atoh1 in Vestibular and Spiral Sensory Neurons as Revealed by Atoh1-eGFP, Atoh1-LacZ Staining and Q-PCR Suggested That This Labeling Is Not an Integration Artifact
Atoh1-Cre mediated β-galactosidase could provide a false-positive signal that is related to the genomic integration of the promoter fragment. To test for this possibility, we investigated the ears of E12.5 and P7 Atoh1-eGFP mice (courtesy of Drs. Chen and Segil). Initially, we could not detect any staining outside the sensory epithelia. However, using the Zeiss LSM confocal system color unmixing algorithm, we were able to extract a weak eGFP signal from the background in scattered spiral sensory neurons of P7 (Fig. 7A) mice. We also found few eGFP-positive cells delaminating in E12.5 Atoh1-eGFP mice (Fig. 4A). Detailed comparison of Atoh1-Cre and Atoh1-eGFP thus confirms staining in hair cells and some sensory neurons, but also shows some differences in these three transgenic lines using the same Atoh1 promoter fragment in additional staining of cells in the cochlea. Most importantly, the signal in delaminating cells of E12.5 Atoh1-eGFP mice was stronger than in sensory neurons at P7 (Figs. 4A, 7A), suggesting that Atoh1 might be transiently upregulated. Moreover, only one of our two Atoh1-Cre transgenic lines showed labeling in a large fraction of sensory neurons, suggesting that the possibly transient expression of Atoh1 in these cells might be more widespread. While it is virtually inconceivable that this expression of Atoh1 in sensory neurons presents an integration artifact that would need to be similar in all three transgenic lines, we nevertheless wanted to obtain independent confirmation of the validity of the expression in sensory neurons indicated by the Atoh1 promoter fragment to demonstrate that this expression in sensory neurons is truly not an aberration of Atoh1 promoter integration.
We reasoned that if Atoh1 staining is indeed in sensory neurons and not an artifact of the Atoh1 promoter fragment, than we should be able to detect Atoh1-LacZ expression in the spiral and vestibular sensory neurons of juvenile and adult mice and should also detect it with Q-PCR. We investigated this possibility by staining the ears of 9-day-, 2-month-, 5-month-, and 9-month-old Atoh1-LacZ heterozygotic mice for β-galactosidase. Our data demonstrated that indeed there is an expression of β-galactosidase not only in hair cells but also in sensory neurons. Moreover, the overall distribution was not unlike that found in E18.5 Atoh1-Cre mice and showed only a small fraction of cells to be positive for Atoh1-LacZ (Fig. 7E,I). The distribution of these cells in the cochlea showed a more profound cellular staining in the apex, the area in which Neurog1 null mutants show the most profound effect on the spatial expression of Atoh1 and on temporal changes of hair cell terminal mitoses. Vestibular ganglion neurons are among the largest cells in the ear, 40–50 μm in size, and cannot be mistaken for any other cell type in the vestibular ganglia (Fig. 7I). Interestingly, all transgenic lines as well as adult Atoh1-LacZ lines showed prominent expression in virtually all hair cells and most pillar cells (Figs. 6B, 7D,G,H) implying that these cells should also be considered as Atoh1-positive cells.
We next tested whether spiral sensory neurons express Atoh1 and if so how this expression changes quantitatively over time. Previous work had shown that Atoh1 is sharply downregulated in the brain (Akazawa et al.,1995) and it could, therefore, be possible that the Atoh1-LacZ expression is unrelated to actual Atoh1 protein and gene expression. We, therefore, investigated the expression of Atoh1 using in situ hybridization and Q-PCR at P0 and P10. With both techniques, we confirmed expression of Atoh1 in neonatal and older mice in the sensory neurons of the cochlea at virtually identical levels compared to newborns (Fig. 5B). Together these data firmly establish the presence of Atoh1 in spiral and vestibular ganglion neurons.
Absence of Atoh1 May Affect Pathfinding Properties of Sensory Neurons
We next wanted to understand whether Atoh1 expression in sensory neurons is biologically meaningful. We reasoned that pathfinding properties of sensory neurons are most likely to be disturbed and investigated this issue using tract tracing and immunocytochemistry for tubulin. The fibers projecting to the posterior crista have previously been demonstrated to provide a good readout for molecular perturbation of pathfinding properties such as neurotrophin misexpression (Tessarollo et al.,2004) or semaphorin signaling (Gu et al.,2003). Specifically, Bdnf null mutant mice lose all canal crista innervation and do not receive many fibers even when sensory neurons cannot degenerate in the absence of Bax (Hellard et al.,2004), suggesting that hair cells lacking BDNF are not able to attract fibers. Loss of hair cells late in development as in Pou4f3 null mutants (Xiang et al.,2003) or absence of differentiated hair cells as in E18.5 Atoh1 null mutants is compatible with retention of canal epithelia innervation (Fritzsch et al.,2005a). Moreover, misexpression of Bdnf under Ntf3 promoter control redirects vestibular fibers away from vestibular hair cells and even to areas devoid of any hair cells (Tessarollo et al.,2004). Thus, lack of hair cell differentiation in Atoh1 null mice combined with the presence of BDNF protein in canal epithelia should have no effect on the initial growth of canal epithelia afferent fibers.
Consistent with this expectation, we could not detect any obvious alterations in pathfinding at E11.5 (Fig. 8A,B), but found aberrations at E12.5 (Fig. 8C,D) and E13.5 (Fig. 8F,G). Specifically, fibers tended to branch profusely off the main branchlet to the posterior crista with only few fibers entering the epithelium (Fig. 8D,E) in Atoh1 null mice. The aberrations observed resembled those previously reported (Gu et al.,2003) in targeted disruption of semaphorin signaling through its neuropilin 1 receptor (Fig. 8E). In contrast, wildtype mice showed bifurcation of the posterior crista branchlet to enter the two hemicristae of the posterior cristae epithelium. Similar data were obtained for the anterior and horizontal crista and suggest that around the onset of Bdnf-mediated rerouting of sensory processes (Tessarollo et al.,2004), Atoh1 null affects targeting of afferents near the sensory epithelia in a matter reminiscent of that in Sema3a signaling–disrupted mutants (Gu et al.,2003).
In summary, we show here that Neurog1 null mice have a reduction in the number of hair cells and a reduced longitudinal growth of the cochlea. This is in part mediated by a terminal mitosis of hair cells of up to two days earlier and, to a lesser extent, by topographically altered upregulation of Atoh1 in the apical turn of the cochlea and a more prominent expression of Neurod1 in Neurog1 null mice. We demonstrate the presence of Atoh1 in sensory neurons using three Atoh1 transgenic lines, Atoh1-LacZ heterozygotes, and Q-PCR of vestibular ganglion neurons and show a possible effect of Atoh1 absence on neuronal pathfinding properties. Combined, these data support the notion that Atoh1 is expressed in some common hair cell and sensory neuron precursors, in particular in the cochlea and saccule. What remains to be shown is that common neurosensory clones exist in mice as in chicken (Satoh and Fekete,2005) and that Neurog1 is also co-expressed with Atoh1 in neurosensory precursors giving rise to hair cells and neurons, a possibility that was recently suggested (Raft, unpublished data).
Past research in the ear has demonstrated that three bHLH genes are essential for the formation and/or differentiation of the two types of neurosensory cells of the ear, the hair cells, and the sensory neurons (Ma et al.,1998; Bermingham et al.,1999; Liu et al.,2000; Kim et al.,2001). Additional information on Neurog1 and Atoh1 null mutations have shown that loss of sensory neuron formation in Neurog1 null mice affects hair cell formation (Ma et al.,2000). In Atoh1 null mutants, undifferentiated and/or dying cells still form in the topology normally occupied by hair cells (Chen et al.,2002; Fritzsch et al.,2005a). It still is unclear how the interaction between Neurog1 and Atoh1, resulting in overlapping effects upon the ear neurosensory epithelium, could be reconciled, especially in light of the complex interactions of bHLH genes known in other developing sensory systems (Bertrand et al.,2002; Wu et al.,2003; Akagi et al.,2004). The Neurog1 and Atoh1 interaction also suggests a possible clonal relationship between some sensory neurons and hair cells in mammals (Fritzsch and Beisel,2004), which was recently demonstrated in chickens (Satoh and Fekete,2005). As previously pointed out, our data and those of others (Bermingham et al.,1999; Chen et al.,2002; Fritzsch et al.,2005a) do not support the recent idea that Atoh1 is not only essential for hair cell differentiation but also for sensory epithelia formation (Woods et al.,2004). In fact, our data show that even in combined Atoh1/Neurog1 double null mutants, there is a formation of Bdnf-LacZ-positive cells in a topology that closely corresponds to some of the sensory epithelia found in wildtype littermates, suggesting that neither of these two bHLH genes is essential for the formation of undifferentiated sensory epithelia precursors and their maintenance up to E18.5 (Fig. 1D). The observed reduction in total length of the cochlea in Neurog1 null mice exceeds the length of Bdnf-LacZ-expressing cells in Atoh1 null mice. The data on length and area changes in Neurog1 null, Atoh1 null and double null mice suggest a simple additive effect, implying that Neurog1 affects both hair cell and non-sensory epithelium formation (Fig. 1; Table 1). These data are more in line with the idea that other genes are responsible for setting up the neurosensory precursor domains of the ear (Zou et al.,2004; Daudet and Lewis,2005; Kiernan et al.,2005).
Atoh1 in Sensory Neurons
With almost any technique employed to reveal Atoh1 in the ear, we are able to demonstrate Atoh1 expression not only in the hair cells but also in some spiral and vestibular sensory neurons (Figs. 4–7). In the spiral ganglion, we find labeled cells in particular in the apex, an area where past research suggests a more prominent presence of type II spiral neurons that project to the outer hair cells (Ryugo,1992; Rubel and Fritzsch,2002). The reported lack of projection deeper into the undifferentiated organ of Corti of the remaining sensory neuron afferent fibers of Atoh1 null mutants at E18.5 (Fritzsch et al.,2005a) could indicate a lack of projection of type II neurons. Alternatively, it could relate to the limited attraction provided by the undifferentiated cells that express Atoh1-LacZ and Bdnf-LacZ (Fritzsch et al.,2005a) as these cells are apparently negative for any other hair cell or supporting cell protein marker tested thus far (Chen et al.,2002; Woods et al.,2004).
It remains somewhat unclear why none of the previous publications has indicated such an expression of Atoh1 in sensory neurons. In large part, this may be related to the more prominent expression of Atoh1 in hair cells combined with the known effects of Atoh1 on hair cell maturation (Bermingham et al.,1999; Izumikawa et al.,2005) that diverted a complete inner ear expression analyses. In part this appears to be inherent to some of the techniques employed, like the Atoh1-eGFP reporter, which requires additional techniques to extract the rather faint signal from the background. A compounding factor is likely the early formation of bone around the sensory neurons, which preclude proper penetration of detection solutions, thus leading to false negatives as we can detect Atoh1 in spiral and vestibular sensory neurons in older animals in which bone encloses these neurons. Lastly, without reason to suspect expression of Atoh1, one would not dissect the vestibular ganglia and process them for PCR to detect the clearly present Atoh1 signal (Fig. 5B).
In summary, Atoh1 expression is not only in hair cells but also in some sensory neurons and may be functionally meaningful for the development of specific projection patterns (Fig. 8).
Neurog1 and Hair Cell Terminal Mitosis and Differentiation
Our past work has shown that many hair cells do not form in Atoh1 null mice (Ma et al.,2000), in particular in the saccule and the cochlea. We now show that combined mutation of Atoh1 and Neurog1 leads to a simple additive effect of the Atoh1 phenotype on the Neurog1 phenotype (Fig. 1; Table 1). Our data confirm previous reports on a base-to-apex progression of sensory neuron and apex-to-base progression of hair cell terminal mitosis (Ruben,1967). Minor differences in the timing of these events relate to differences in staging the mice and differences in gestational length of the mouse strains used. Most importantly, this past work has indicated that apical hair cells may become postmitotic before apical sensory neurons. We now show that there is a widespread overlap of timing of terminal mitosis in apical sensory neurons and hair cells around E11.5–E12.5 (Fig. 2). On closer examination of terminal mitosis, Neurod1 and Atoh1 expression in Neurog1 null embryos shows both terminal mitosis and Neurod1 and Atoh1 expressions are temporally and/or spatially altered in Neurog1 null mice. In fact, terminal mitosis of hair cells occurs up to two days earlier in Neurog1 null mice. This temporal acceleration of hair cell terminal mitosis, combined with our inability to demonstrate earlier expression of Atoh1, implies that other genes are responsible for the cell fate switch from sensory neurons to hair cells in the Neurog1 null mice: instead of being able to differentiate as sensory neurons, postmitotic cells now become hair cells. The altered expression of Neurod1 and Atoh1 in the cochlea with a spatial shift toward the apex suggests that Neurog1 is not only affecting cell fate selection but also expression of other bHLH genes.
The simplest way of explaining these data as well as the expression of Atoh1 in sensory neurons would be the assumption that neurosensory precursors co-express Atoh1 and Neurog1 and switch their cell fate to hair cell differentiation in Neurog1 null mice. (Fig. 9). However, we cannot exclude other models, namely that Neurog1 absence disrupts coordinated cell cycling of nearby cells and affects their gene expression profiles, thus causing Neurod1 upregulation in hair cells. This alternative scenario was previously raised in developing dorsal root ganglion (Ma et al.,1999). Clearly, the truncation of hair cell formation may relate to either direct effects in neurosensory precursors through interactions of these bHLH genes as previously reported (Gowan et al.,2001) or may interfere with the coordinated inhibitory processes or collaboration with other factors that set up sensory epithelia (Adam et al.,1998; Daudet and Lewis,2005).
Similar effects on reduced clonal expansion and changes in cell fate assignment have been reported for null mice of the winged helix gene Bf1 (now Foxg1). In these mutants, the clonal expansion of cortical precursors is truncated through premature terminal mitosis and cell fate is limited to only one cortical neuron type (Hanashima et al.,2002,2004). It appears that the Neurog1 null mutation has comparable effects on neurosensory precursors of the ear, limiting the fate choice to one (hair cells) and initiating an earlier terminal mitosis of these cells resulting in clonal compression. Clearly, not only are fewer hair cells formed in the smaller epithelia of Neurog1 null mice, but these mutants also miss all sensory neurons, suggesting that NEUROG1 protein is essential for the complete clonal expansion of the entire neurosensory population of the ear, possibly including some hair cell precursors.
In summary, much like in the spinal cord (Gowan et al.,2001), these data could be most easily explained by assuming an inhibitory interaction of Neurog1 and Atoh1 in a set of neurosensory precursor cells of the ear (Fig. 9). Absence of this interaction in Neurog1 null mice leads to temporal shifting of hair cell terminal mitosis through a phenotype switch of postmitotic cells. Recent work in chicken has demonstrated such a clonal relationship exists at a limited scale (Satoh and Fekete,2005) and our data are compatible with such a relationship for mice but do not prove it. Direct evidence is still needed for mammals.
Atoh1 and Hair Cell Terminal Mitosis
Obviously, our Atoh1 in situ hybridization data as well as our Atoh1-LacZ data strongly suggest that Atoh1 expression is at best in a small set of cycling hair cell precursors in the basal turn, the last hair cells to become postmitotic in the cochlea. These data are in agreement with almost all other data on Atoh1 expression in the cochlea published thus far (Bermingham et al.,1999; Zine et al.,2001; Chen et al.,2002). Only some recent Atoh1-LacZ data suggest a somewhat more profound expression of Atoh1 throughout the cochlea at E13.5 (Woods et al.,2004), which partially agrees with our Atoh1-LacZ data (Fig. 4E). Combined, our data confirm the previous conclusion based on Atoh1-eGFP expression that Atoh1 is not expressed in detectable levels in cycling precursors (Chen et al.,2002), except for a few cells at the basal tip of the organ of Corti.
A low level of Atoh1 may be present but this possibility cannot be ruled out with in situ hybridization, Atoh1-LacZ, or Atoh1-eGFP. Indeed, our Atoh1-Cre data show somewhat earlier upregulation of Atoh1 expression in the future vestibular epithelia at E11 (Fig. 4G) and the cochlea basal turn at E12.5 (Fig. 4D) as well as β-galactosidase labeling in sensory neurons at these developmental stages (Fig. 4). These data imply that some Atoh1 expression in amounts not detectable by current histological techniques exist in the growing organ of Corti several hours or days prior to E12.5. Recent data on another bHLH gene, MyoD, showed that more sensitive in situ techniques can demonstrate cellular distribution prior to detection with more established in situ techniques, providing cellular distribution information for already existing RT-PCR data (Gerhart et al.,2004). Indeed, it appears that there is a delay of approximately 12 h between the first detection of Atoh1 with PCR and with Atoh1-Cre (Figs. 4,5). This Atoh1 expression revealed with the second Cre line is particularly remarkable as it is so early despite the fact that this reporter must be delayed through the multiple steps required for expression and eventual detection. This is apparent by the earlier Atoh1-Cre activation and detection in sensory neurons in Atoh1-Cre/Rosa 26 lines. In the light of these data, combined with the clear effect of Neurog1 on changes in timing of terminal mitosis of hair cells, we suggest that the spatio-temporal overlap of Atoh1 expression with proliferating hair cell precursors might be more profound than our data and those of others suggest (Chen et al.,2002) as they may be biased toward detection only of larger amounts of Atoh1 transcript or ATOH1 protein, rather than the earliest onset of expression. Using the newly developed Atoh1-Cre line combined with photoactivation of the β-galactosidase reaction product (Figs. 2, 4) should allow us to capture such expression in neurosensory precursors, if it exists.
As outlined above, it is possible that none of the employed histological techniques may detect the earliest onset of Atoh1 expression. Such early upregulation of Atoh1-Cre/Rosa 26 signal indicates that Atoh1 transcription may exist in cochlea precursor cells as early as E10.5, a time in which the majority of hair cells are not yet postmitotic. We assume that this early expression is in common neurosensory precursors of sensory neurons and hair cells, leading to the formation of enough ATOH1 protein to activate the Atoh1 promoter fragment that will drive the Cre expression that will eventually activate the LacZ reporter system of the Rosa 26 line, a notion supported by the early expression revealed in the second Atoh1-Cre line at E11 (Fig. 4G). However, while not very likely, we cannot exclude that most hair cells and sensory neurons arise from discrete progenitors and that the early Atoh1-Cre expression reflects an early upregulation of Atoh1 in some sensory neurons.
In many other systems, quantitative relationships within clones are determined by one or more bHLH genes through regulation of both cell fate and numbers of cells of a given fate (Bertrand et al.,2002; Akagi et al.,2004). Elimination of one bHLH gene may lead to replacement of a given cell type through clonal expansion and regulation of cell fate using another bHLH gene (Ma et al.,1999). In the ear, there appears to be a dichotomy of the bHLH gene channeling towards either a hair cell or sensory neuron fate. Thus, abrogation of Neurog1 may cause premature upregulation of Neurod1 and possibly other genes that can, however, in the context of the ear only cooperate in hair cell differentiation (Fig. 9). The temporal acceleration of terminal mitosis of hair cells will move these cells out of synchrony with other developing gene expression patterns, thus leading to the formation of multiple rows of hair cells and disorganization of apical specializations (Ma et al.,2000) presumably by moving hair cell differentiation out of synchrony with polarity-determining gene expressions such as Vangl (Amonlirdviman et al.,2005). Closer examination of Vangl upregulation (Montcouquiol et al.,2003) in Neurog1 null mice and control littermates is needed to support this notion. Alternatively, the more profound upregulation of Neurod1-driven LacZ reporter in these hair cells of Neurog1 null mice may affect cell polarity organization more directly.
Interestingly, our data on Atoh1 expression, regardless of any genetic background, show that apical hair cells express Atoh1 only in early neonates, but become postmitotic between E11.5 and E12.5. In contrast to Atoh1 null mice, where Atoh1 absence apparently leads to rapid degeneration of many of the undifferentiated Atoh1-lacZ-positive precursors found in these mutants (Chen et al.,2002; Fritzsch et al.,2005a), Atoh1-negative postmitotic hair cells of the apex in wildtype mice seem to do just fine for several days, even in Atoh1 null mice (Fritzsch et al.,2005a). Several explanations for this conflicting data are apparent. It is possible that cells require ATOH1 protein once the upstream regulators have initiated Atoh1 expression and failure of ATOH1 protein formation leads to cell death within several hours. Available data on base-to-apex progression of apoptosis would be compatible with such an idea (Chen et al.,2002). Since the lack of ATOH1 protein will invariably lead to apoptosis, the reported base-to-apex progression of apoptosis in Atoh1 null mutants might also relate to inherited specific maximal times postmitotic hair cells can survive without a specified amount of ATOH1 protein.
Atoh1 Expression in Sensory Neurons May Be Biologically Meaningful
The expression of Atoh1 in the vestibular ganglion neurons, sensory neurons known to project to the saccule and the posterior crista (Maklad and Fritzsch,1999), appears to relate to a possible effect of Atoh1 null mutation on targeted afferent fiber growth (Fig. 8). Examination of initial vestibular fiber growth has shown unusual deviations of fibers near the target epithelia that could not be explained by lack of neurotrophins in the target epithelia (Fritzsch et al.,2005a). Clearly, Bdnf, the most important neurotrophin to guide and maintain afferent fibers to the crista organs (Fritzsch et al.,2004; Tessarollo et al.,2004), is expressed in these epithelia even in mutants lacking both Neurog1 and Atoh1 (Fig. 1H–J). Moreover, Atoh1 null mice show a prominent innervation of the posterior crista as late as E18.5 (Fritzsch et al.,2005a) thus ruling out that Bdnf expression might be the problem as it would have resulted in loss of all crista innervation by E14.5 (Fritzsch et al.,2004; Tessarollo et al.,2004). We now have extended this initial finding and show aberrations that resemble those previously reported for Npn1sema3a-mutant mice (Gu et al.,2003). We suggest that the expression of Atoh1 in vestibular sensory neurons mediates certain path-finding properties of sensory neurons either by regulating neuropilin expression (Gu et al.,2003) or by affecting the BDNF/TRPC pathway for guiding growth cones of vestibular ganglion neurons (Li et al.,2005; Wang and Poo,2005). Such path-finding properties are essential for the connection of insect mechanosensory neurons with the CNS and may even be mediated by Atoh1 knocked into atonal in flies (Ben-Arie et al.,2000; Wang et al.,2002). Governing path-finding may also be a property of Atoh1 in CNS neurons in which it is expressed (Ben-Arie et al.,1997; Bermingham et al.,1999,2001). Such involvement of Atoh1 in regulating the molecular basis of path-finding may directly relate to the ancestry of the atonal gene family and their association with the sensory ciliated neuron (Fritzsch and Beisel,2003) and the sensory system development and evolution (Fritzsch and Beisel,2003; Fritzsch et al.,2005b). Testing this suggestion requires conditional null mutation of Atoh1 only in sensory neurons using a Neurog1-mediated Cre expression. Alternatively, analyzing the expression of Neuropilin1, TrkB, and TrpC in wildtype and Atoh1 null ears could reveal differential regulation of these genes known to be involved in path-finding.
In summary, our data support the notion that some sensory neurons express low levels of Atoh1, particularly those of the apex of the cochlea and the inferior vestibular ganglion known to innervate the saccule and posterior crista. Neurog1 abrogation leads to earlier terminal mitosis of hair cells, somewhat displaced upregulation of Atoh1 and Neurod1, and down-regulation of clonal expansion of hair cell precursors (Fig. 9). Consistent with recent data in chicken (Satoh and Fekete,2005), our data are compatible with a clonal relationship of some sensory neurons and hair cells in mice, being derived from common neurosensory precursors. However, this does by no means imply that all hair cells and all sensory neurons share a common neurosensory ancestor and more data are needed to demonstrate the extent of the postulated common neurosensory precursor. For translational research, it would be extremely helpful if the molecular switches from sensory neuron precursor to hair cell precursors are better understood and can be molecularly manipulated. Ultimately, this may lead to implantation of a single neurosensory precursor population that can give rise to both sensory neurons and hair cells. Indeed, sensory neuron precursors that have the capacity to proliferate in vitro have recently been isolated from human cochlea (Rask-Andersen et al.,2005) and thus offer a possible therapeutic approach in combination with the above outlined insights.
This work was supported by grants from NIH (RO1 DC005590 to B.F.; R01 DC04279 to K.W.B.). We acknowledge the help of Dr. P. Chen and Q. Ma to improve the manuscript and Drs. P. Chen, H. Zoghbi, N. Segil, D. Ginty, and Q. Ma for providing their mutant mouse lines. This investigation was conducted in a facility constructed with support from Research Facilities Improvement Program Grant Number 1 C06 RR17417-01 from the National Center for Research Resources, National Institutes of Health. We acknowledge the use of the confocal microscope facility of the NCCB, supported by EPSCoR EPS-0346476 (CFD 47.076).