Atoh1 expression defines activated progenitors and differentiating hair cells during avian hair cell regeneration


  • Jon Cafaro,

    1. Virginia Merrill Bloedel Hearing Research Center, Department of Otolaryngology and Head and Neck Surgery, University of Washington, Seattle, Washington
    Search for more papers by this author
  • Gi Soo Lee,

    1. Virginia Merrill Bloedel Hearing Research Center, Department of Otolaryngology and Head and Neck Surgery, University of Washington, Seattle, Washington
    Search for more papers by this author
  • Jennifer S. Stone

    Corresponding author
    1. Virginia Merrill Bloedel Hearing Research Center, Department of Otolaryngology and Head and Neck Surgery, University of Washington, Seattle, Washington
    • CHDD CD176 Box 357923, Virginia Merrill Bloedel Hearing Research Center, Department of Otolaryngology, University of Washington, Seattle, WA 98195-7923
    Search for more papers by this author


In the avian inner ear, nonsensory supporting cells give rise to new sensory hair cells through two distinct processes: mitosis and direct transdifferentiation. Regulation of supporting cell behavior and cell fate specification during avian hair cell regeneration is poorly characterized. Expression of Atoh1, a proneural transcription factor necessary and sufficient for developmental hair cell specification, was examined using immunofluorescence in quiescent and regenerating hair cell epithelia of mature chickens. In untreated birds, Atoh1 protein was not detected in the auditory epithelium, which is quiescent. In contrast, numerous Atoh1-positive nuclei were seen in the utricular macula, which undergoes continual hair cell turnover. Atoh1-positive nuclei emerged in the auditory epithelium by 15 hr post-ototoxin administration, before overt hair cell damage and supporting cell re-entry into the cell cycle. Subsequently, Atoh1 labeling was seen in 15% of dividing supporting cells. During cell division, Atoh1 was distributed symmetrically to daughter cells, but Atoh1 levels were dramatically regulated shortly thereafter. After cellular differentiation, Atoh1 labeling was confined to hair cells regenerated through either mitosis or direct transdifferentiation. However, Atoh1 expression in dividing progenitors did not necessarily predict hair cell fate specification in daughter cells. Finally, predominant modes of hair cell regeneration varied significantly across the radial axis of the auditory epithelium, with mitosis most frequent neurally and direct transdifferentiation most frequent abneurally. These observations suggest a role for Atoh1 in re-specifying supporting cells and in biasing postmitotic cells toward the hair cell fate during hair cell regeneration in the mature chicken ear. Developmental Dynamics 236:156–170, 2007. © 2006 Wiley-Liss, Inc.


Sensory hair cells are mechanoreceptors in auditory, vestibular, and lateral line sensory systems. Exposure to high doses of aminoglycoside antibiotics, such as gentamicin, causes injury and death in hair cells. In mammals, sensory deficits resulting from hair cell loss are permanent, because new hair cells are not formed (reviewed in Stone et al., 1998). In contrast, nonmammalian vertebrates, including birds, are able to regenerate a complete set of hair cells after injury, thereby restoring sensory function (reviewed in Corwin and Oberholtzer, 1997; Stone and Rubel, 2000a).

In the avian auditory epithelium (basilar papilla), it has been established that hair cells are regenerated through two distinct processes, both of which are initiated in supporting cells, the nonsensory cells residing in the sensory epithelium. After treatment with ototoxic drugs, some supporting cells initially respond by undergoing a phenotypic conversion into hair cells (Adler and Raphael, 1996, Roberson et al., 1996, 2004). This process, which also occurs in amphibian hair cell epithelia (Baird et al., 1996; Taylor and Forge, 2005), has been described as direct transdifferentiation because it does not involve cell division (Beresford, 1990; Opas and Dziak, 1998). After some delay, additional supporting cells undergo cell division, giving rise to new supporting cells and additional new hair cells (Corwin and Cotanche, 1988; Ryals and Rubel, 1988; Lippe et al., 1991; Raphael, 1992; Hashino and Salvi, 1993; Stone and Cotanche, 1994; Bhave et al., 1995).

It is not clear if all supporting cells in hair cell epithelia are capable of both direct transdifferentiation and cell division or if there are different subsets of supporting cells specialized for each response. Furthermore, signals directing each supporting cell response have not been characterized. One candidate for regulating progenitor cell behavior in the chicken inner ear is the proneural basic helix-loop-helix (bHLH) transcription factor Atoh1. The initial description of this transcription factor was provided by discovery of its homolog, atonal, in Drosophila (Jarman et al., 1993). Atonal is required for sensorineural progenitor specification in chordotonal organs, ommatidia, and olfactory sense organs in Drosophila (Jarman et al., 1995; Gupta and Rodrigues, 1997). Subsequent studies of Atoh1 in mice showed that it is also necessary for development of granule cells in the cerebellum (Ben-Arie et al., 1997) and proprioceptive neurons in the spinal cord (Bermingham et al., 2001). Most relevant to this study, Atoh1 has been implicated in mammalian hair cell development and regeneration. Atoh1-null mice fail to differentiate hair cells in auditory and vestibular epithelia (Bermingham et al., 1999). Furthermore, misexpression of Atoh1 is sufficient to induce re-specification of nonsensory cells into hair cells in the developing auditory epithelium (Zheng and Gao, 2000). Finally, viral transduction of Atoh1 into nonsensory cells in mature auditory (Kawamoto et al., 2003; Izumikawa et al., 2005) and vestibular (Shou et al., 2003) epithelia appears to trigger hair cell regeneration.

In embryonic mice, Atoh1 is first transcribed before hair cell and supporting cell differentiation in the auditory epithelium, apparently in postmitotic precursor cells as well as in a small number of mitotically active progenitors (Lanford et al., 2000; Zine et al., 2001; Chen et al., 2002; Woods et al., 2004; Matei et al., 2005). Atoh1 protein is first detected in differentiating hair cells, suggesting it may not have a role in auditory progenitor cells (Chen et al., 2002). Atoh1 expression has also been demonstrated in early developing hair cell epithelia in the chicken inner ear (Stone et al., 2003; Pujades et al., 2006) and in the zebrafish lateral line (Itoh and Chitnis, 2001). Additionally, an increase in hair cells in lateral line neuromasts of the zebrafish mind bomb mutant correlates with an increase in Atoh1-positive cells (Itoh and Chitnis, 2001), suggesting a potential role in hair cell fate specification in fish.

We used immunological methods to examine Atoh1 expression in quiescent and regenerating hair cell epithelia in mature chickens. Regeneration was induced by exposure of the epithelium to the ototoxic drug gentamicin (Janas et al., 1995). Our results show that, in normal birds, nuclear Atoh1 expression is not seen in the quiescent auditory epithelium, but Atoh1-positive nuclei are abundant in the vestibular epithelium as it undergoes hair cell turnover. After experimentally induced hair cell loss, Atoh1 expression is reactivated in the auditory epithelium, in directly transdifferentiating supporting cells, dividing supporting cells, early noncommitted postmitotic cells, and differentiating hair cells.


Atoh1 Antibody Correlates With Expected Expression Patterns

We used a rabbit polyclonal antibody to Atoh1 (from Dr. Jane Johnson, University of Texas Southwestern Medical Center; Helms and Johnson, 1998) to examine Atoh1 protein expression in control and regenerating inner ear epithelia in posthatch chickens. The immunogen used to generate this antibody was a bacterially derived full-length protein from the mouse coding sequence. Nucleotide sequences for mouse and chicken Atoh1 are highly homologous (95% identity) in the bHLH domain and less so (46% identity) in regions outside the bHLH domain (Ben-Arie et al., 1996). Western analysis of specificity has not been possible with this antibody (Dr. Johnson, personal communication). Chen et al. (2002) used this antibody to examine Atoh1 protein distribution in the auditory epithelium of developing mice. The immunolabeling pattern shown in their study is similar to reporter gene expression in Atoh1 knock-in mice (Ben-Arie et al., 1997; Bermingham et al., 1999; Woods et al., 2004) and in mice in which GFP is driven by an Atoh1 enhancer (Lumpkin et al., 2003). In addition, the immunolabeling pattern mirrors Atoh1 mRNA expression patterns, as determined by in situ hybridization, in the developing mouse auditory epithelium (Lanford et al., 2000). We previously applied this antibody to cryosections of the developing chicken inner ear (otocyst) at Hamburger and Hamilton stage 21 (E3.5) and saw Atoh1 labeling in sensory patches but not in the cochleovestibular ganglion or extrasensory regions (Stone et al., 2003). This expression pattern resembles that seen for Atoh1 mRNA in the chicken otocyst at the same time (Pujades et al., 2006).

Here, we examined tyramide signal amplification (TSA) -enhanced immunofluorescent labeling for Atoh1 protein in cryostat sections through the head and the trunk in chicks at stage 19 (E3–E3.5) for further confirmation of specificity of the antibody. Distribution of Atoh1 mRNA has been described at this stage in the central nervous system (Ben-Arie et al., 1996). We saw identical patterns of immunolabeling in the spinal cord and hindbrain (Supplementary Figure S1, which can be viewed at In addition, we saw no Atoh1 immunolabeling in the otocyst, retina, or lens at this time. These results were anticipated; Atoh1 expression is not detected in the chicken otocyst at stage 19 (Pujades et al., 2006) or in the early developing mouse retina (Ben-Arie et al., 2000). Negative controls (run with primary antibody omitted) generated no nuclear labeling in the sections (data not shown). These controls provide additional support for the antibody's specificity for Atoh1 in chicken tissue.

Atoh1 Is Up-regulated During Early Hair Cell Regeneration in the Basilar Papilla

The auditory epithelium of birds (basilar papilla) is composed of two primary cell types, sensory hair cells and nonsensory supporting cells (Takasaka and Smith, 1971; Tanaka and Smith, 1978). Hair cells are tonotopically organized; those responsive to higher frequencies are located in the basal (proximal) region, while those responsive to lower frequencies are positioned in the apical (distal) region (Fig. 1A). Hair cells are morphologically distinct from supporting cells. They make contact with the luminal surface but not the basal lamina, and they have an apical bundle of stereocilia (Fig. 1G). In contrast, supporting cells contact both the luminal surface and the basal lamina, and they lack stereocilia. In addition, hair cell nuclei are large, round, and located near the luminal surface, while supporting cell nuclei are smaller, oval, and positioned under the hair cell nuclei and in some cases close to the basal lamina.

Figure 1.

A–F,I: Hair cell damage and regeneration in the chicken basilar papilla. This figures illustrates features of the normal (A–C), gentamicin-damaged (D–F), and regenerating (I) basilar papilla. A: The basic anatomical relationships in a control basilar papilla (BP) are shown. The box in A shows the approximate region magnified in B,C,E,F, and I. B: Nuclei in the control basilar papilla exclude incorporation of bromodeoxyuridine (BrdU), as illustrated after a 2-hr pulse/fix labeling paradigm. C: Hair cells are visible on the luminal surface of the epithelium using antibodies to Calmodulin (CaM). D: By 3 days postgentamicin, three distinct areas of hair cell damage are visible: complete loss of hair cells in the basal third (D), fully intact hair cells in the apical third (N), and a transitional zone in the middle region containing both surviving and dying hair cells (T). The hair cell lesion consistently shows an inferior leading edge. E: Supporting cells in the area of damage incorporate BrdU around 3 days postgentamicin, as shown after a 2-hr pulse/fix labeling paradigm. F–H: CaM immunolabeling reveals the hair cell lesion. New hair cells arise in the damaged region through two distinct pathways: mitotic regeneration (G) or direct transdifferentiation (H). I: Regenerated CaM-positive hair cells are evident in the damaged region 9 days after gentamicin. SC, supporting cell; HC, hair cell; Stc, hair cell stereocilia; M, mitotic supporting cell. Scale bar = 78 μm applies to B,C,E,F,I.

In chickens, all auditory hair cells are normally generated between embryonic day 4.5 and 9 (Katayama and Corwin, 1989). After hatching, the hair cell population is stable, and cells in the basilar papilla are growth-arrested (Fig. 1B,C; Oesterle and Rubel, 1993; Bhave et al., 1995). Short treatments with the ototoxin gentamicin cause hair cell injury and extrusion from the basilar papilla's basal end (Bhave et al., 1995; Hyde and Rubel, 1995; Janas et al., 1995; Stone et al., 1996; Roberson et al., 2000a). Hair cell loss is first evident around 2 days postgentamicin, in the basal tip of the epithelium. Subsequently, the area of hair cell loss rapidly spreads apically, reaching its maximal size—approximately half of the epithelium with the present treatment conditions—by 5 days postgentamicin (Fig. 1D,F). Supporting cells in the area of hair cell loss initially respond by directly transdifferentiating into new hair cells (Fig. 1H; Adler and Raphael, 1996; Roberson et al., 1996, 2004). While the exact timing of this cellular conversion is not known, it is initiated before 3 days postgentamicin, at which time hair cells formed by this nonmitotic process are initially detected (Roberson et al., 2004). Additionally, supporting cells in the damaged area re-enter the cell cycle and regenerate new hair cells, in some cases renewing themselves (Fig. 1G; Corwin and Cotanche, 1988; Ryals and Rubel, 1988; Girod et al., 1989; Raphael, 1992; Hashino and Salvi, 1993; Stone and Cotanche, 1994; Stone and Rubel, 2000b). A significant increase in the number of dividing supporting cells over normal levels occurs around 3 days postgentamicin (Fig. 1E; Bhave et al., 1998; Stone et al., 1999; Roberson et al., 2004), and by 15 days postgentamicin, rates of cell division have reached near-control levels (Bhave et al., 1998). By 9 days postgentamicin, many newly formed hair cells fill the lesion (Fig. 1I).

In posthatch chickens, we examined Atoh1 immunoreactivity in the normal (quiescent) basilar papilla and in basilar papillae at different times after gentamicin treatment (15 hr, 1–5, 8, 10, 15, 16, 19, 27, and 47 days). In normal organs, nuclear Atoh1 protein was not detected in any region of the basilar papilla (Fig. 2A,B), in either the supporting cell layer or the hair cell layer, using either NiCl2 or TSA enhancement. However, some presumed background staining was seen in the hair cell cytoplasm in all regions of the epithelium (e.g., see Fig. 2A). As early as 15 hr postgentamicin, a few nuclei with low-intensity Atoh1 labeling were detected in the supporting cell layer in the basal half of the basilar papilla using TSA-enhanced immunofluorescence (Fig. 2E). Labeled cells were not evident at this time using NiCl2 enhancement (Fig. 2D), which may indicate that TSA provided a slightly lower threshold of detection than did NiCl2. By 24 hr postgentamicin, numerous Atoh1-positive supporting cell nuclei were evident in the basal half of the epithelium, using either TSA or NiCl2 enhancement (Fig. 2G–H). For most subsequent experiments, we used TSA to enhance Atoh1 immunolabeling.

Figure 2.

Atoh1 up-regulation in the chicken basilar papilla after gentamicin damage. A,B,D,E,G,H,J,K,M: Atoh1 immunolabeling using either NiCl2 enhancement of diaminobenzidene (DAB; A,D,G,J) or tyramide signal amplification (TSA) of Alexa 488 (B,E,H,K,M) is shown in a normal basilar papilla and in basilar papillae at different times postgentamicin in the basal tip of the epithelium. C,F,I,L: Images show MyosinVI immunofluorescence in the hair cell layer overlying the regions shown in B,E,H, and K, respectively. A,B: No Atoh1-positive nuclei were seen in control tissue. D: At 15 hr postgentamicin, Atoh1-positive nuclei were not seen with NiCl2/DAB enhancement but a false-positive blood cell is evident (arrowhead). E,F: However, when TSA was used, Atoh1-positive nuclei (arrow) were seen in the supporting cell layer (E) underlying normal-appearing hair cells (F, compare with C). G–I: At 24 hr postgentamicin, Atoh1-positive nuclei (arrows) were evident with both NiCl2/DAB (G) and TSA (H) labeling methods, again under intact hair cells (I). Insets in B and H show double labeling for Atoh1 (green) and Sox2 (magenta) in the supporting cell nuclear layer. J–L: By 3 days postgentamicin, numerous Atoh1-postive nuclei (J,K) were seen in the basal end, now devoid of hair cells (L). M: A low-magnification view of the basilar papilla at 5 days postgentamicin is shown to illustrate the distribution of Atoh1-positive nuclei (green) in the basilar papilla. Normal hair cells are shown through immunolabeling for MyosinVI (magenta). In all panels, neural (superior) is toward the top, and apical is toward the right. Arrows mark the edge of the lesion. pG, postgentamicin. Scale bar = 26 μm in A–L, 193 μm in M.

To gain further support for the interpretation that Atoh1 labeling emerges in supporting cell nuclei, we co-labeled basilar papillae at 24 hr postgentamicin with antibodies to Atoh1 and Sox2, a transcription factor that is expressed in all supporting cell nuclei, but not in hair cells, in control posthatch basilar papilla (Fig. 2B, inset; our unpublished observations). Sox2 is also expressed in embryonic progenitor cells in the developing mammalian organ of Corti (e.g., Kiernan et al., 2005). Atoh1 labeling was indeed seen in Sox2-positive nuclei at 24 hr postgentamicin (Fig. 2H, inset), which supports the interpretation that, after gentamicin treatment, Atoh1 expression first emerges in supporting cells.

At 15 and 24 hr postgentamicin, there was no overt evidence of hair cell damage or disruption of the hair cell pattern in the area where Atoh1-positive nuclei had emerged (Fig. 2F,I). Previous studies showed that cells in S phase were only rarely detected in the basal region of the basilar papilla at 1 and 2 days postgentamicin (Bhave et al., 1998; Stone et al., 1999; Roberson et al., 2004). Therefore, our findings indicate that Atoh1 protein levels are elevated in supporting cells before hair cell extrusion and passage of a substantial number of supporting cells into the cell cycle.

By 3 days postgentamicin, nearly all native hair cells in the basal third of the basilar papilla have been extruded (Stone et al., 1996; Stone and Rubel, 2000b; Fig. 2L). At this point, numerous Atoh1-positive nuclei were distributed throughout the damaged region (Fig. 2J,K). At 3 days postgentamicin and at later time points, Atoh1 immunoreactivity was confined to the lesion and was not observed in the apical undamaged region (Fig. 2M). No Atoh1 immunoreactivity was ever seen in dying hair cells (data not shown). While the number of Atoh1-positive nuclei in the damaged region appeared to decline after 5 days postgentamicin, rare labeled nuclei were seen as late as 47 days postgentamicin (data not shown).

To assess the time course of Atoh1 up-regulation, we measured the density of Atoh1-positive nuclei in the hair cell lesion, or in the corresponding region before hair cell loss, at 1, 3, 4, 5, 8, and 16 days postgentamicin. This measuring was accomplished by sampling the basal 2,000 μm of the basilar papilla at each time point. In addition, we analyzed controls at 7–10 days posthatch. To permit simultaneous assessment of whether Atoh1 becomes up-regulated in a basal-to-apical manner, mirroring hair cell loss, we subdivided the lesion into three regions: the basal tip (1–675 μm), the middle of the lesion (675–1,330 μm), and the apical end of the lesion (1,330–2,005 μm). Our analyses demonstrated that total numbers of Atoh1-positive nuclei in the lesion were elevated over controls by 1 day postgentamicin, peaked around 4–5 days postgentamicin, and then steadily declined (Fig. 3). Analysis of variance (ANOVA) of total numbers showed significant differences in Atoh1 numbers between all sequential time points, except between 4 and 5 days postgentamicin. Rather surprisingly, we detected no significant regional gradients in the average density of Atoh1-positive nuclei, except at 1 day postgentamicin, when the apical region lagged the basal and middle regions. Thus it appears that initial Atoh1 up-regulation occurs rapidly and globally throughout the lesion, except for the apical portion at 1 day postgentamicin, which may reflect a slight delay in hair cell injury in that region.

Figure 3.

Quantification of Atoh1 up-regulation. The graph plots the average densities of Atoh1-positive nuclei in three positions in the lesion against different times postgentamicin. Basal lesion = 1–675 μm from basal tip; middle lesion = 675–1,330 μm from basal tip; apical lesion = 1,330–2,005 μm from basal tip. Error bars show standard deviations of the mean.

Hair Cell Markers Emerge in Atoh1-Positive Cells

To assess the fate of Atoh1-postive cells that emerge in the lesion, we counter-labeled basilar papillae with antibodies to the unconventional myosin, MyosinVI. This antigen is abundant in the cytoplasm of immature and mature hair cells but is not detected in dividing or postmitotic supporting cells (Hasson et al., 1997; J.S. Stone, unpublished observations). Atoh1-positive cells in the damaged basal region were first seen to co-label for MyosinVI at 3 days postgentamicin (Fig. 4A,B). Only a small number of double-labeled cells could be seen at this time. Most Atoh1-positive cells were not MyosinVI-positive, most likely reflecting a lag between increased Atoh1 translation and increased MyosinVI translation in transdifferentiating supporting cells. None of the MyosinVI-positive cells seen in the damaged area at 3 days postgentamicin were Atoh1-negative, suggesting that all hair cells express Atoh1 during an early phase of differentiation.

Figure 4.

Hair cell markers emerge in Atoh1-positive cells. A–D: Atoh1/NiCl2-positive nuclei (A,C) were localized to MyosinVI-positive cells (B,D) at 3 days (A,B) and at 5 days (C,D) postgentamicin. Arrows point to double-labeled cells, and arrowheads point to a MyosinVI-positive cell lacking Atoh1 labeling. pG, postgentamicin. Scale bar = 11 μm.

By 5 days postgentamicin, numerous Atoh1-positive/MyosinVI-positive cells were detected in the damaged region (Fig. 4C,D). We calculated that 93.7% (± 2.0 SD) of all Atoh1-positive cells were MyosinVI-positive, and 99.5% (± 0.6%) of MyosinVI-positive cells were Atoh1-positive. The strong colocalization of Atoh1 and Myosin VI labeling provides good evidence that Atoh1 is expressed primarily in regenerated hair cells at this time. A small percentage of MyosinVI-positive cells (<1%) did not show nuclear Atoh1 labeling (Fig. 4C,D). Because of their more mature morphology compared with other Atoh1-positive/MyosinVI-positive cells in the region, Atoh1-negative/MyosinVI-positive cells most likely represent hair cells at a more advanced stage of differentiation that had down-regulated Atoh1 translation previous to 5 days postgentamicin.

Atoh1 Expression in Dividing Supporting Cells and in Hair Cells Regenerated Through Mitosis

To assess Atoh1 expression during mitotic regeneration, chickens at 4 days postgentamicin were administered a single injection of bromodeoxyuridine (BrdU) to label a subset of dividing supporting cells. This time point was chosen because large numbers of supporting cells are in S phase at that time (Bhave et al., 1995, 1998; Stone et al., 1999). Chickens were killed at 2 hr post-BrdU to catch supporting cells in S phase or at later time points post-BrdU to allow mitosis to occur and postmitotic cells to differentiate. Because BrdU appears to be cleared from the chicken inner ear around 6 hr postinjection (Stone and Cotanche, 1994), a discrete cohort of progenitor cells can be labeled with a single BrdU injection. Basilar papillae were collected and immunoreacted to detect BrdU, Atoh1, and in some cases, the hair cell-selective marker Calmodulin, which is expressed in mature hair cells and is first detected in regenerated hair cells around 24 hr post-BrdU (Stone et al., 1996; Stone and Rubel, 2000b).

The following data points are provided to demonstrate that similar numbers of BrdU were sampled across organs. As anticipated, there was a doubling in the number of BrdU-positive nuclei per basilar papilla between 2 hr post-BrdU (premitosis; 140 ± 23.6 SD) and 24 hr post-BrdU (postmitosis; 290 ± 86.5). Total numbers of BrdU-positive nuclei per basilar papilla at 4 days post-BrdU (232 ± 33.7) and 12 days post-BrdU (229 ± 23.9) did not appear to vary substantially from those at 24 hr post-BrdU.

At 2 hr post-BrdU, Atoh1 labeling was detected in a small percentage (15%) of BrdU-positive nuclei (Fig. 5A,D). Most double-labeled nuclei at this time had low-intensity Atoh1 labeling compared with some BrdU-negative/Atoh1-positive cells (Fig. 5A, inset on right). At 15 hr post-BrdU, corresponding to 5–9 hr postmitosis (Stone and Cotanche, 1994; Stone et al., 2004), BrdU-positive progenitors had divided, and pairs of BrdU-positive sister nuclei were evident (Fig. 5B). The nuclei in these pairs were in close association and exhibited different patterns of Atoh1 immunolabeling. In some cases, sister pairs showed asymmetric Atoh1 labeling, with one nucleus strongly Atoh1-positive and one nucleus Atoh1-negative (Fig. 5B). Additional pairs were symmetrically Atoh1-positive or symmetrically Atoh1-negative (data not shown). By 24 hr post-BrdU, 21% of BrdU-positive cells expressed Atoh1, which ANOVA confirmed is a significant increase over 2 hr post-BrdU. At 4 and 12 days post-BrdU, 22% and 16% of BrdU-positive cells expressed Atoh1, neither of which is significantly different from 24 hr post-BrdU. By 24 hr post-BrdU, postmitotic cells expressing Atoh1 were Calmodulin-positive (Fig. 5C), indicating they were differentiating as hair cells. At 4 days post-BrdU, co-labeling with Calmodulin was seen in the majority of BrdU/Atoh1-positive nuclei (data not shown). However, we also noted BrdU/Calmodulin-positive nuclei in cells that were Atoh1-negative (data not shown), which suggests that some regenerated hair cells had down-regulated Atoh1 by this time.

Figure 5.

Atoh1 immunoreactivity in dividing progenitors and postmitotic cells. Birds received a single bromodeoxyuridine (BrdU) injection at 4 days postgentamicin, and basilar papillae were dissected and were fixed at various times post-BrdU. BrdU-positive cells (magenta) were assessed for Atoh1/tyramide signal amplification (TSA) immunoreactivity (green), and in some cases, for immunoreactivity for the hair cell-specific antigen Calmodulin (blue). A: At 2 hr post-BrdU, a BrdU-positive nucleus with low-intensity Atoh1 labeling is indicated by the arrow, and a BrdU-negative/Atoh1-positive–positive nucleus is indicated by the arrowhead. Each color channel is shown separately in insets to better illustrate labeling patterns. B: At 15 hr post-BrdU, a sister pair of BrdU-positive nuclei (arrowheads), with an asymmetric pattern of Atoh1 labeling is shown. Each color channel is shown separately in insets to better illustrate labeling patterns. C: By 24 hr post-BrdU, some BrdU-positive/Atoh1-positive cells co-labeled for the hair cell marker Calmodulin (arrow). D: The graph shows the mean percentage of BrdU-positive nuclei with low- or high-intensity Atoh1 labeling at different times post-BrdU. Error bars represent standard deviations of the mean. E: Spatial distributions of Atoh1 labeling and BrdU incorporation within the mid-lesion area are very different, as shown in this low-magnification image of the basilar papilla at 2 hr post-BrdU on day 3 postgentamicin. The superior side (toward the top of the image) has a higher density of BrdU-positive nuclei than the inferior side (toward the bottom). A reciprocal pattern is seen with respect to Atoh1-positive nuclei. The apical region is toward the right. The white line indicates the approximate position of the boundary between the superior and inferior hlves of the epithelium. Scale bar = 10 μm in A–C, 57 μm in E.

During this analysis, we noted considerable variability in the intensity of Atoh1 labeling among BrdU-positive nuclei. At each time point and throughout the BP, we scored BrdU-positive nuclei as having either low- or high-intensity Atoh1 labeling relative to other nuclei in each respective field. This measure was purely subjective; nuclei with intense, saturated labeling were scored as high-intensity, while those with pale labeling were scored as low-intensity. At 2 hr post-BrdU, most BrdU-positive nuclei showed low-intensity Atoh1 labeling. In contrast, at 24 hr post-BrdU and at later time points, more BrdU-positive nuclei showed high-intensity Atoh1 labeling than low-intensity labeling. ANOVA demonstrated that there was a significantly higher proportion of BrdU-positive nuclei with low-intensity Atoh1 labeling at 2 hr post-BrdU than at 24 hr, 4 days, or 12 days post-BrdU. However, proportions of low-intensity labeled nuclei did not change significantly after 24 hr post-BrdU. In contrast, there was a significant increase in the proportion of BrdU-positive nuclei showing high-intensity Atoh1 labeling between 2 hr post-BrdU and 24 hr, 4 days, and 12 days post-BrdU. A significant decrease in the proportion of BrdU-positive nuclei with high-intensity Atoh1 labeling was seen between 4 and 12 days post-BrdU. These findings reveal that, after cell division, there is a rapid decrease in low-intensity Atoh1 labeling and a steady increase in high-intensity Atoh1 labeling.

Pulse-fix delivery of BrdU at 3 days postgentamicin demonstrated a significantly higher density of BrdU-labeled nuclei in the superior (neural) half of the epithelium than in the inferior (abneural) half (Fig. 5E). This difference was most evident in the middle portion of the lesion. In contrast, Atoh1 immunoreactivity showed a reciprocal pattern from BrdU's pattern, appearing more intense in the inferior half of the epithelium than in the superior half (Fig. 5E). This observation raised the question of whether the distinct spatial segregation of Atoh1-positive cells from BrdU-positive cells corresponds to a spatial segregation of regenerated hair cells made by means of direct transdifferentiation from those regenerated through cell division. To address this question, we implanted birds with mini-osmotic pumps to deliver BrdU continuously to the perilymphatic fluid of the inner ear for 6 days after gentamicin treatment. This method, which has been characterized in previous studies for the chicken inner ear (Roberson et al., 1996, 2000a, b, 2004), effectively identifies all hair cells formed by mitotic division because of the incorporation of BrdU into their nuclei. On the other hand, all new hair cells formed by direct transdifferentiation show no BrdU labeling. After BrdU infusion, basilar papillae were immunoreacted to detect BrdU, Atoh1, and Calmodulin, or BrdU and MyosinVI. Continuously infused basilar papillae showed a heavy concentration of BrdU labeling in the superior half of the epithelium (Fig. 6A), similar to basilar papillae after a short pulse of BrdU at 3 days postgentamicin (Fig. 5E). As anticipated based on previous studies using continuous BrdU infusion (Roberson et al., 1996, 2004), we observed regenerated hair cells (Calmodulin- or MyosinVI-positive) that were BrdU-positive (mitotically regenerated) and regenerated hair cells that were BrdU-negative (regenerated through direct transdifferentiation; Fig. 6B). Atoh1 labeling was strong in Calmodulin-positive cells with or without BrdU, indicating that Atoh1 is expressed in hair cells regenerated through either mitotic or nonmitotic mechanisms.

Figure 6.

Continuous infusion of bromodeoxyuridine (BrdU) to the regenerating basilar papilla. For this figure, inner ears were infused continuously with BrdU for 8 days in vivo to a time equivalent to 6 days postgentamicin. A: Strong cumulative incorporation of BrdU in the superior half of the basilar papilla (toward the top) relative to the inferior half (toward the bottom) is shown. The apical region is toward the right. The white line indicates the approximate position of the boundary between the superior and inferior halves of the epithelium. B: Calmodulin (CaM)-positive regenerated hair cells (blue) have nuclei that are either BrdU-positive and Atoh1-positive (magenta and green; arrows) or BrdU-negative, Atoh1-positive (green only, arrowheads) nuclei. C: The graph shows the mean percentage of MyosinVI-positive cells that was BrdU-positive (postmitotic) or BrdU-negative (transdifferentiated) in the mid-lesion area in superior and inferior regions of basilar papillae that received continuous BrdU. D: The graph shows the mean percentage of BrdU-positive cells that was MyosinVI-positive (differentiated as hair cells, HCs) or MyosinVI-negative (differentiated as supporting cells, SCs) in superior and inferior regions of the mid-lesion area of basilar papillae that received continuous BrdU. For both graphs, error bars represent standard deviations of the mean. Scale bar = 87 μm in A; 17 μm in B.

Analysis of superior and inferior halves of the mid-lesion basilar papilla demonstrated substantial spatial segregation of hair cells regenerated by each mechanism (Fig. 6C), when analyzed with MyosinVI as the hair cell marker. In the superior half, 19% of regenerated hair cells resulted from direct transdifferentiation and 81% resulted from cell division. In contrast, in the inferior half, 66% of regenerated hair cells resulted from direct transdifferentiation and 34% resulted from cell division. These findings suggest that direct transdifferentiation is the predominant mechanism of hair cell production in the inferior half of the epithelium and raises the question of whether a higher rate of supporting cell renewal in that region occurs to repopulate the subpopulation of supporting cells that has converted into hair cells. To address this question, we examined the fates of BrdU-positive cells in the superior and inferior halves of the epithelium from the same tissue as analyzed for Figure 6C. We scored BrdU-positive nuclei as MyosinVI-positive (“hair cell”) or MyosinVI–negative (“supporting cell”). Data shown in Figure 6D show that similar proportions of postmitotic cells differentiate as hair cells and supporting cells in each region, indicating that supporting cell renewal is not more prevalent in the inferior half.

Atoh1 Immunoreactivity in Normal Utricles During Ongoing Hair Cell Regeneration

The utricle contains a macular epithelium specialized for vestibular function. The cellular organization of the utricle is similar to the basilar papilla in that nonsensory supporting cells surround sensory hair cells, and the two cell types' nuclei are arranged in laminae, with hair cell nuclei aligned near the lumen and supporting cell nuclei aligned near the basal lamina. Unlike the basilar papilla, hair cell production in the avian utricle normally continues after embryogenesis. After hatching, a relatively small number of supporting cells distributed evenly throughout the epithelium is in S-phase at any given time, and this proliferative activity is mirrored by a similar degree and distribution of apoptotic hair cells (Jørgensen and Mathiesen, 1988; Roberson et al., 1992; Weisleder and Rubel, 1993; Kil et al., 1997; Matsui et al., 2003). Exposure of chickens to gentamicin causes a significant increase in hair cell death, and up-regulation in hair cell production, particularly in the striolar region (Oesterle et al., 1993; Weisleder and Rubel, 1993). The spatial and temporal progression of hair cell loss and regeneration in the chicken utricle after gentamicin treatment has not been well defined (Stone and Rubel, 1999). However, after short streptomycin treatments, significantly increased supporting cell division is evident in the striolar region by 1 day poststreptomycin, and levels of division peak between 2 and 4 days and return to near-control levels by 15 days (Bhave et al., 1998). It is anticipated that the location and the time of onset of the hair cell lesion and the nature of the cellular response in the utricle would be similar after gentamicin or streptomycin treatment.

In the undamaged utricle, numerous Atoh1-positive nuclei were distributed throughout the epithelium in a seemingly uniform pattern throughout striolar and extrastriolar regions (Fig. 7A). At 4 days postgentamicin, the density of Atoh1-positive nuclei appeared to increase in the striola but remained unchanged in extrastriolar areas (data not shown). Similar to the basilar papilla, Atoh1-positive nuclei were detected in presumed hair cells, which are immunolabeled for MyosinVI (data not shown).

Figure 7.

Atoh1 expression in normal utricles. A: Atoh1/tyramide signal amplification (TSA) -positive nuclei are evenly distributed throughout the normal utricle. The white C-shaped outline defines the approximate location of the striolar region. Normal birds were given injections of bromodeoxyuridine (BrdU) and killed at various times post-BrdU to assess Atoh1 labeling (green) in BrdU-positive (magenta) nuclei. B: Low-intensity Atoh1 labeling was evident in some BrdU-positive nuclei at 2 hr post-BrdU (arrowhead). C: A mitotic figure in metaphase, labeled with propidium iodide (PI, magenta), is Atoh1-positive (green). D: Sister chromosomes in telophase showed symmetric levels of Atoh1 labeling. E: At 15 hr post-BrdU, asymmetric Atoh1 labeling in a pair of BrdU-positive sister nuclei is shown (arrowhead) and symmetrically absent Atoh1 labeling in a pair is also shown (arrow). The inset shows symmetrically positive Atoh1 labeling in another pair for reference. F: The graph shows the mean percentage of BrdU-positive nuclei (pre-mitotic/mitotic) or pairs (postmitotic) with a given Atoh1 labeling pattern, which for single nuclei is either Atoh1+ or Atoh1− and for pairs is Atoh−/− (two Atoh1-negative nuclei), Atoh1+/+ (two Atoh-positive nuclei), or Atoh1+/− (one Atoh1-positive nucleus and one Atoh1-negative nucleus). Error bars show standard deviations of the mean. Scale bar = 360 μm in A, 13 μm in B–E.

We used pulse/fix BrdU labeling in normal utricles to examine whether dividing supporting cells expressed detectable levels of Atoh1 protein and when cell progeny become Atoh1-positive relative to cell division. For quantitative analysis, the following numbers of BrdU-labeled cells were assessed in each group of utricles (2 hr: 53 nuclei ± 18.7; 15 hr: 52 pairs ± 11; 24 hr: 37 pairs ± 12.2; and 4 days: 27 pairs ± 11.3). Examination of utricles at 2 hr post-BrdU showed that 30% of BrdU-positive supporting cells, which are in or near S phase, have Atoh1-positive nuclei (Fig. 7B,F). Most Atoh1 labeling in BrdU-positive nuclei was low-intensity compared with that seen in some neighboring BrdU-negative nuclei (data not shown). Co-labeling of utricles with the DNA dye propidium iodide (PI) allowed us to identify supporting cells in mitosis. PI-labeled mitotic figures were seen exclusively at the luminal surface and were typically the only labeled structures in this layer. Quantitative analysis showed that, similar to S-phase cells, 27% of mitotic cells were Atoh1-immunoreactive (Fig. 7C,D,F), while the remainder showed no Atoh1 immunoreactivity (Fig. 7F). Analysis of telophase cells (Fig. 7D) demonstrated that Atoh1 protein, when detected, always appeared to be symmetrically distributed to daughter cells during mitosis. This observation was confirmed in postcytokinetic cells identified by PI labeling (data not shown).

If the fate of cell progeny were to depend solely on Atoh1 levels in progenitor cells and Atoh1 protein is evenly distributed during mitosis, then we would predict that all divisions would lead to symmetric differentiation. Furthermore, because nearly one third of progenitors are Atoh1-positive and two thirds are Atoh1-negative, we would anticipate that one third of divisions would result in the production of two hair cells and two thirds would yield two supporting cells. However, in the chicken utricle during ongoing regeneration, the majority (∼78%) of divisions generate a hair cell and a supporting cell (Stone et al., 1999, 2004; Stone and Rubel, 1999). Therefore, Atoh1 expression in progenitor cells cannot be the sole determinant of daughter cell fate. Analysis of Atoh1 expression in sister cells over time after mitosis showed a dramatic change in Atoh1 labeling (Fig. 7F). At 15 hr post-BrdU (approximately 5–9 hr postmitosis), we detected numerous sister pairs with asymmetric labeling for Atoh1 (one nucleus with high Atoh1 labeling and the other with no detectable Atoh1 labeling; Fig. 7E). ANOVA demonstrated that the percentage of asymmetrically labeled pairs increased significantly between 15 and 24 hr post-BrdU as well as between 24 hr and 4 days post-BrdU. During this time, there was a significant decrease in the proportion of pairs that were either symmetrically Atoh1-positive or symmetrically Atoh1-negative. By 4 days post-BrdU, the majority of sister pairs (77%) had established asymmetric Atoh1 labeling, which resembles the percentage of pairs that ultimately differentiates asymmetrically, as supporting cell/hair cell pairs (see above). Furthermore, at 4 days post-BrdU, the nuclei of most pairs showing asymmetric Atoh1 labeling had segregated into distinct layers, with the Atoh1-positive nucleus in the hair cell layer and the Atoh1-negative nucleus in the supporting cell layer. These observations reveal that postmitotic Atoh1 levels, which appear to be established through inheritance, are significantly altered after mitosis, achieving predicted patterns by 4 days post-BrdU.


Atoh1 Is Up-regulated During Hair Cell Fate Specification in Regenerating Sensory Epithelia

We used immunolabeling to examine Atoh1 protein expression in quiescent and regenerating hair cell epithelia of mature chickens. Atoh1-positive nuclei were not detected in the normal basilar papilla, in which no new hair cells are formed. However, they emerged in areas of hair cell loss after gentamicin treatment, in transdifferentiating supporting cells, dividing hair cell progenitors, early postmitotic cells, and differentiated hair cells. These observations demonstrate that Atoh1, which is critical for hair cell development, is reactivated in mature chickens after hair cell damage, and they suggest that Atoh1 may play a role in specifying postmitotic cells as hair cells, and in re-specifying supporting cells as hair cells, during avian hair cell regeneration.

Atoh1 Distinguishes Transdifferentiating Supporting Cells

Our observations of the drug-damaged basilar papilla indicate that Atoh1 has a role in both mitotic and nonmitotic hair cell regeneration. Several observations suggest that the earliest Atoh1-positive cells to emerge in the lesion (at 15–24 hr postgentamicin) are supporting cells undergoing direct transdifferentiation. Early Atoh1-positive nuclei reside in the supporting cell nuclear layer, near the basal lamina and below normal-appearing hair cells. These Atoh1-positive cells are clearly neither injured hair cells nor mitotic supporting cells (significant numbers of mitotic supporting cells do not emerge until the third day postgentamicin, Bhave et al., 1995, 1998; Stone et al., 1999; Roberson et al., 2004). Continuous delivery of BrdU demonstrates that hair cells regenerated by means of direct transdifferentation have Atoh1-positive nuclei, providing further support that they arise from Atoh1-positive supporting cells. Although direct transdifferentiation has been implicated since the mid-1990s as a mechanism for hair cell regeneration (Adler and Raphael, 1996; Roberson et al., 1996), Atoh1 appears to be the first identified marker to distinguish transdifferentiating supporting cells from other supporting cells in birds.

Two important questions are raised by these observations: Which signals lead to increased Atoh1 expression in supporting cells, and why is Atoh1 up-regulated in only a subpopulation of supporting cells? Clearly, signals triggering Atoh1 up-regulation arise shortly after gentamicin treatment, but before hair cell extrusion. The confinement of Atoh1-positive nuclei to the lesion area indicates that signaling leading to Atoh1 up-regulation is dependent upon hair cell injury. Previous studies show that significant ultrastructural and molecular changes occur in damaged hair cells by 15 hr postgentamicin (e.g., Hirose et al., 2004; Mangiardi et al., 2004). Such changes may negatively impact direct cell–cell contact between hair cells and supporting cells. Signaling through the protein Notch negatively regulates transcription of Atoh1 and other proneural genes (reviewed in Bertrand et al., 2002) and, therefore, represents a potential mechanism for inhibiting supporting cell transdifferentiation in quiescent epithelia. Notch is an extracellular receptor that functions as a transcriptional regulator (reviewed in Selkoe and Kopan, 2003; Schweisguth, 2004). When Notch is bound by ligand located on neighboring cells, its cytoplasmic portion is cleaved and enters the nucleus, where it triggers a cascade of events leading to increased transcription of the bHLH transcription factor, Hairy/Enhancer of Split (HES) and related proteins. HES1 and 5 repress Atoh1 transcription in developing murine hair cells (Zine and Ribaupierre, 2002). In the mature chicken auditory epithelium, Notch1 transcripts are expressed in supporting cells but not in hair cells (Stone and Rubel, 1999). Transcripts for the Notch ligand Serrate1 are expressed in hair cells and supporting cells (Stone and Rubel, 1999). Serrate2 is expressed in hair cells during late development (Eddison et al., 2000; Eddison, 2001) and may be retained in hair cells after development. Transcripts for a third ligand, Delta1, are not expressed in normal tissue but are highly up-regulated after hair cell damage (Stone and Rubel, 1999). In normal epithelia, Serrate1/2 in hair cells could activate Notch in neighboring supporting cells, thereby maintaining them in a quiescent state, with low Atoh1 expression. Gentamicin could cause early degradation of Serrate1/2 ligand and decreased Atoh1 suppression in supporting cells. As a result, supporting cells in the damaged area would begin to up-regulate Atoh1, and cells with high Atoh1 expression levels would phenotypically convert into hair cells. In Drosophila, proneural transcription factors similar to Atoh1 activate transcription of the Notch ligand, Delta (Hinz et al., 1994). Therefore, once Atoh1 expression is up-regulated in transdifferentiating supporting cells, the resulting increase in Notch ligand production in the same cells could restore the signaling required to inhibit transdifferentiation in neighboring supporting cells. As a result, supporting cells that surround transdifferentiating hair cells with high Delta1 expression would be inhibited from generating high levels of Atoh1. This model could account for why only some supporting cells complete transdifferentiation and why regenerated hair cells emerge in a salt-and-pepper pattern throughout the lesion.

An additional clue to mechanisms regulating supporting cell behavior is that there appears to be a greater propensity for hair cells to be regenerated through direct transdifferentiation than through mitosis in the inferior (abneural) half of the epithelium, and the converse is seen in the superior (neural) half. This finding has several implications, the first of which is that some signals guiding transdifferentiation or cell division in the damaged basilar papilla may be unevenly distributed and thereby exert regional effects. The cytoarchitecture of the organ is asymmetric along the superior–inferior axis. In the sensory epithelium, there are two morphologies of hair cells, tall and short, which are confined to the superior half and inferior half of the epithelium, respectively, and which receive distinct patterns of innervation (Takasaka and Smith, 1971; Hirokawa, 1978; Fischer, 1992). After short gentamicin treatments, hair cell loss is seen in the inferior half before the superior half (e.g., Hyde and Rubel, 1995), so it is possible that slightly different timing or patterns of hair cell loss in each region leads to regionally distinct supporting cell responses. In addition, signals derived from outside the epithelium may have differential effects in either half of the epithelium after damage, due to asymmetric delivery or asymmetric distribution of required receptors in supporting cells. Two additional implications are made by the finding that mitotic and nonmitotic regeneration are regionally segregated. First, the process of re-entry into the cell cycle may antagonize Atoh1 up-regulation/transdifferentiation, or visa versa. Second, supporting cells in different regions of the basilar papilla may be specialized or intrinsically biased to undergo a specific cellular response to hair cell damage.

Despite a significantly higher rate of direct transdifferentiation in the inferior half of the epithelium during the first 5 days of regeneration, there is no apparent difference in the rate of supporting cell renewal in that region during that period. Thus it is not clear how numbers of supporting cells are restored there. In our analysis, we only examined the fates of cells born during the first 5–6 days postgentamicin. While previous studies show that most substantial supporting cell division appears to be limited to the period between 3 and 7 days postgentamicin (Bhave et al., 1998; Stone et al., 1999), it is possible that undocumented cell divisions that occur after this time act to restore supporting cells in the inferior half.

Atoh1 Is Expressed in Dividing Supporting Cells

In the regenerating utricle and basilar papilla, between15 and 30% of dividing progenitors have Atoh1-immunoreactive nuclei. This finding is not surprising when one considers that several mitotically active neural progenitors in the central nervous system express Atoh1 (e.g., Ben-Arie et al., 1997; Helms and Johnson, 1998) or similar proneural genes (reviewed in Bertrand et al., 2002). Furthermore, during sensory epithelial development in the chicken inner ear, Atoh1 mRNA is seen in BrdU-labeled progenitor cells 6 hr after a BrdU pulse (Pujades et al., 2006), indicating that Atoh1 is expressed in dividing or newly postmitotic cells. In contrast, a study of the developing auditory epithelium in mice reported no Atoh1 immunoreactivity in dividing progenitor cells but rather its first emergence in early postmitotic hair cell precursors (Chen et al., 2002). This finding suggests that Atoh1 expression in hair cell progenitors is quite different between chickens and mice. However, Chen et al. did not use signal amplification to detect Atoh1, as we did, so it is also possible that we were able to detect significantly lower levels of Atoh1 protein than they were.

What might be the role of Atoh1 in dividing progenitor cells during avian hair cell regeneration? One possibility is that elevated levels of Atoh1 in dividing progenitors bias cell progeny toward the hair cell fate (Woods et al., 2004). Alternatively, elevated Atoh1 may trigger dividing supporting cells to exit the cell cycle rather than to divide again; previous studies have correlated high proneural gene expression with decreased mitotic activity (e.g., Farah et al., 2000; Leow et al., 2004; reviewed in Bertrand et al., 2002). It is also possible that dividing cells with low levels of Atoh1 immunoreactivity reflect supporting cells that initiated transdifferentiation but were subsequently stimulated to take an alternate course and divide. Finally, Atoh1 may be expressed stochastically at different levels in dividing supporting cells, and this property may have no functional significance with respect to proliferative behavior or cell fate outcome.

Early Atoh1 Expression Does Not Necessarily Predict the Hair Cell Fate

Examination of regenerated cells after several days of differentiation established that strong nuclear Atoh1 labeling is positively correlated with the hair cell fate and not associated with regenerated supporting cells. This observation, in conjunction with Atoh1′s known role in embryonic development of hair cells (Bermingham et al., 1999; Zheng and Gao, 2000), suggests that Atoh1 specifies the hair cell fate during avian hair cell regeneration. However, not all cells expressing Atoh1 in the regenerating inner ear epithelium necessarily complete maturation as hair cells. In control utricles, approximately 30% of S-phase dividing cells are Atoh1-positive. These cells appear to divide symmetrically, generating two Atoh1-positive sister cells. Approximately 5–9 hr after mitosis, the percentage of postmitotic sister pairs showing symmetric Atoh1 labeling is significantly below 30%, and numerous asymmetrically Atoh1-labeled sister pairs have emerged. The proportion of asymmetrically labeled sister pairs increases significantly over time, such that, by 4 days after mitosis, the majority of sister pairs are asymmetrically labeled. Thus some cells transiently expressing Atoh1 ultimately acquire the supporting cell fate. These observations demonstrate that Atoh1 labeling in progenitors and early postmitotic sister pairs does not reliably predict the final fate of cell progeny. These results are consistent with previous studies. For example, neural crest progenitors transiently expressing the bHLH proneural transcription factor Ngn2 give rise to glial cells as well as neural cells (Zirlinger et al., 2002). Furthermore, some cells that transiently misexpress Atoh1 in the developing mouse auditory epithelium develop into supporting cells (Woods et al., 2004).

We documented considerable variability in Atoh1 immunoreactivity among dividing and postmitotic cells and showed that, after mitosis, there is a rapid decrease in low-intensity Atoh1 labeling and a steady increase in high-intensity Atoh1 labeling. The functional significance of this trend is not clear. Furthermore, it is not known how the rapid and dramatic change in Atoh1 protein levels is achieved shortly after mitosis. Possible mechanisms include degradation of Atoh1 protein or export of Atoh1 protein from the nucleus into the cytoplasm. Furthermore, while Atoh1 protein appears to be distributed evenly to daughter cells during division, Atoh1 mRNA may be distributed asymmetrically. In this case, cells inheriting higher levels of Atoh1 mRNA than their sisters could translate and accumulate higher levels of Atoh1 protein after mitosis. Because the normal utricle in chickens has a high density of mature and maturing hair cells, it is likely that negative feedback, perhaps from the hair cells themselves, prevents nascent cells from acquiring the hair cell fate. As discussed earlier, emergence of atonal/Atoh1 expression is regulated by several signaling pathways, including through the Notch receptor (Zine and Ribaupierre, 2002), bone morphogenetic protein receptors (Wine-Lee et al., 2004), and epidermal growth factor receptor (zur Lage et al., 2004). However, mechanisms regulating Atoh1 degradation, translocation, and translation are poorly understood.



Fertilized eggs of White Leghorn chickens were purchased from Hyline International (Graham, WA) and were placed in a humidified incubator at 37°C until embryo harvesting or hatching. Hatchlings were housed in heated brooders with ample food and water in the Animal Care Facility at the University of Washington. All described procedures were approved by the University of Washington's Animal Care Committee and conform to NIH guidelines for vertebrate animals.

Drug Injections

In posthatch chickens (5–10 days of age), hair cell death was induced by a single subcutaneous administration of the ototoxic aminoglycoside antibiotic gentamicin (Sigma, 300 mg/kg), on 2 consecutive days. Gentamicin solution was made according to Husmann et al. (1998). For survival time points after gentamicin treatment, time zero was considered to coincide with the first injection. At 4 days after gentamicin injection, chickens received a single intraperitoneal injection of the nucleotide analog, 5-bromo-2′-deoxyuridine (BrdU; 12 mg/ml; Sigma Chemical Co., St. Louis, MO) dissolved in sterile phosphate buffered saline (PBS) at a dose of 100 mg/kg.

Osmotic Mini-pump Implantation

An osmotic mini-pump (Alzet, Model #2002, 14-day duration, 0.5 μl/hr; Durect Corporation, Cupertino, CA) was used to deliver BrdU (10 mg/ml) continuously to the inner ear for 8 days. Pumps were implanted 2 days before gentamicin treatment and maintained until the end of the experiment (6 days postgentamicin). Previous studies show that this type of pump delivers BrdU continuously to dividing supporting cells (Roberson et al., 1996).

Surgical methods for pump implantation followed closely those developed by Roberson et al. (1996, 2000a, b). All surgical procedures were performed under sterile conditions. Briefly, birds were anesthetized with inhaled isoflurane (2–3% isoflurane in 100% oxygen, delivered at 0.3 L/min). A small injection of lidocaine (10 μl) was delivered to the skin behind the external ear, and a small incision was made in this location using electrocautery. The postauricular musculature was spread, revealing the underlying bone. A window was created in the bone using a small bone elevator, revealing the vestibule medial to the lateral semicircular canal. A 30-gauge needle was used to generate a hole in the vestibule, and a 1.5-mm stainless steel cannula attached to a catheter (both from the brain microinfusion kit, Durect Corporation) was implanted and secured with dental acrylic. An incision was made in the skin overlying the scapula, and the pump was inserted into the subdermal space. The catheter was snaked under the skin of the neck and back and attached to the pump. Incisions near the ear and the scapula were sutured. Animals awoke from anesthesia within 5 min of ceasing isoflurane delivery. No morbidity (including vertigo) or mortality due to surgical procedures was experienced. Birds received the standard doses of gentamicin on days 2 and 3 after surgery and were killed on day 6 postgentamicin.

Tissue Dissection

Chick embryos at stage 19 (Hamburger and Hamilton, 1992, equivalent to embryonic day 3 to 3.5) were immersed in 4% buffered paraformaldehyde for 1 hr at room temperature. Embryos were immersed in sucrose (15% for 4 hr then 30% overnight), placed in OCT (Tissue-Tek), and frozen in a −80°C freezer. OCT blocks were sectioned at 14 μm and collected on chrome alum-subbed microscope slides.

Posthatch chickens were killed by intraperitoneal injection (100 mg/kg) of Beuthanasia D followed by decapitation. Control chickens were killed between posthatch day 5 and 10. Birds were killed at the following days postgentamicin (the first injection is considered day 0): 0, 1, 2, 3, 4, 5, 6, 8, 10, 15, 16, 19, 27, and 47. The external ear and middle ear were opened so as to allow fixative to reach the inner ear. Chicken heads were immersed in 4% buffered paraformaldehyde for 2 hr at room temperature. Fixed cochlear ducts (containing the basilar papilla) and utricles were dissected from the temporal bone and placed in PBS (pH 7.4). Using fine microforceps, the tegmentum vasculosum was dissected off the cochlear ducts, and the tectorial membrane was stripped from the epithelial surface by grasping it with microforceps at the apical end of the basilar papilla and pulling toward the basal end. Cochlear ducts were rinsed in PBS and stored at 4°C in PBS until labeling. Utricles were dissected previous to fixation and immersed in cold Hanks' buffered saline solution. Otoconia were gently blown off with a flow of buffer from a syringe, and the utricle was immersed in 4% paraformaldehyde for 30 min at 4°C.

Immunohistochemistry and Immunofluorescence

Cryosections of embryos and whole-mount preparations of cochlear ducts and utricles were processed for indirect immunohistochemistry and/or for standard indirect immunofluorescence. The following primary antibodies were used: rabbit polyclonal anti-Atoh1 (from Dr. Jane Johnson, University of Texas Southwestern Medical Center; Helms and Johnson, 1998), mouse monoclonal anti-Calmodulin (clone 6D4, Sigma-Aldrich), rabbit polyclonal anti-MyosinVI (from Dr. Tama Hasson, University of California, San Diego, now available from Proteus Biosciences, Ramona, CA), goat anti-Sox2 (sc-17320, Santa Cruz Biotechnology, Santa Cruz, CA), and rat anti-BrdU (Harlan Seralab., Loughborough, UK). Secondary antibodies were purchased from Jackson Immunolaboratories (goat anti-rabbit Cy3, donkey anti-mouse Cy5), Invitrogen (goat anti-rat Alexa 594, goat anti-rat Alexa 488), and Vector Laboratories (biotinylated goat anti-rabbit IgG). BrdU immunolabeling (Gratzner, 1982) was performed according to Stone and Rubel (2000b).

Two forms of signal amplification were implemented: nickel chloride (NiCl2) enhancement of diaminobenzidene precipitation and TSA of Alexa 488 deposition (Invitrogen, Carlsbad, CA). Before all immunoreactions, endogenous peroxidases were blocked using a 1-hr incubation in 1.5% H2O2. Tissue was permeabilized and nonspecific antibody binding was inhibited using blocking solution (10% normal horse serum in 0.05% or 0.17% Triton X-100 dissolved in PBS, pH 7.4) or Invitrogen TSA kit blocking reagent for at least 30 min. Antibodies were diluted in block solution. For detection of all antigens except Atoh1, primary and secondary antibody incubations were performed overnight at 4°C and for 1.5 hr at room temperature, respectively, and all other incubations were performed at room temperature. For Atoh1 detection, all reactions were run at 4°C, with incubations in primary antibody run overnight and incubations in secondary antibody run for 2 hr. Some samples were counterlabeled with PI (Sigma-Aldrich; 1 μg/ml) to label nucleic acids. All samples were cover-slipped with Vectashield mounting medium (Vector Laboratories).

Microscopic Imaging

Immunolabeled whole-mount cochlear ducts and utricles were imaged as follows. A Zeiss Axioplan 2ie upright compound microscope connected to a Photometrics Cool Snap HQ digital camera was used for brightfield and epifluorescence microscopy. Slidebook (Intelligent Imaging Innovations, Denver, CO) acquisition and processing software was used to acquire digital images and convert them to TIFF files. TIFF files were managed in ImageJ and/or Adobe Photoshop. In addition, tissue was analyzed using a confocal laser scanning microscope (Bio-Rad 1024 MRC) with Lasersharp (Version 3.2) acquisition software. Dyes were excited with a Krypton–Argon Laser at 488, 568, and/or 647 nm and collected with 522df35, 605df32, and 680df32 filters respectively. Sequential image acquisition was performed when bleed-through between channels was an issue. Files were imported into ImageJ and/or Adobe Photoshop for processing and analysis.

Data Analysis

Quantitative analyses were conducted in either whole-mount basilar papillae or utricles using images obtained by confocal microscopy with a ×60 oil objective. Atoh1-positive nuclei were quantified in digital images of basilar papillae in control birds (P5–P10) and in basilar papillae at several times postgentamicin (1, 3, 4, 5, 8, and 16 days). Z-series images (205 × 205μm2) were taken at nine standardized positions along the organ's length, beginning from the basal tip and extending to the mid-basal region. Each z-series was taken through the entire depth of the epithelium (lumen to basal lamina), and fields were nonoverlapping. As a result, approximately 40% of each basilar papilla in the basal 2 mm was sampled. In each image, the number of Atoh1-positive nuclei was determined by stepping through the z-series using ImageJ. Four basilar papillae were analyzed for each of the following groups: 3, 4, 5, and 8 days postgentamicin. Three basilar papillae were analyzed for the 1-day and 16-day postgentamicin groups. Epithelial area at each of the nine imaged positions was assessed in four age-matched basilar papillae. These measurements were used to calculate the average epithelial area at each position along the basilar papilla, and average areas were used to calculate the density of labeled nuclei at each position. For each region measured, there was little variability in epithelial area across samples (standard deviations of the mean were less than 0.005 mm2).

Atoh1 labeling was assessed in BrdU-positive nuclei at different times after BrdU injection (2 hr, 24 hr, 4 days, and 12 days) in basilar papillae using ImageJ. The relative intensity (low or high) of Atoh1 immunolabeling was also assessed. Relative intensities were scored in eight basilar papillae by two independent “blinded” investigators. Similar proportions of BrdU-positive nuclei were scored as low or high by each investigator, lending credence that the qualitative differences were reliably distinguishable. For this analysis, four basilar papillae were examined at each time point, with the exception of samples at 12 days post-BrdU, in which three basilar papillae were analyzed.

Assessment of Atoh1/MyosinVI co-labeling in basilar papillae at 5 days postgentamicin (n = 4) was performed using epifluorescence microscopy (Axioplan 2ie) using a ×20 objective. Z-series images were generated in three locations in each sample: one at the basal tip, one located mid-lesion, and one in the apical end of the lesion, before the transitional zone. Because all samples were very similar qualitatively with respect to the number of Atoh1-positive cells, the number of Atoh1-positive cells was counted using Slidebook in only one sample and that number (1,000 cells) was applied for all samples. Then, numbers of Atoh1-positive/MyosinVI-negative cells and Atoh1-negative/MyosinVI-positive cells were assessed in each sample, yielding the estimated percentages given. Approximately half of the damaged region below the transitional region was sampled for this analysis.

Two analyses were performed in continuously BrdU-infused basilar papillae. First, to compare proportions of hair cells that regenerate through direct transdifferentiation vs. cell division in the superior and inferior halves of the lesion, basilar papillae (n = 3) labeled for BrdU and Calmodulin were examined with confocal microscopy. Z-series images were taken at ×60 in three nonoverlapping contiguous regions aligned with either the superior edge or the inferior edge of the basilar papilla. Using Image J, MyosinVI-positive cells (presumed hair cells) were scored as either BrdU-positive or BrdU-negative. Second, in the same images, each BrdU-positive nucleus was scored as “hair cell” or “supporting cell” based on whether or not it was surrounded by MyosinVI immunofluorescence.

Atoh1 labeling was assessed in BrdU-labeled nuclei in control utricles (no gentamicin treatment) at different times after BrdU injection. This analysis was conducted on the confocal microscope (not in images) by systematically moving the field across the entire utricle with care not to double-sample. In samples at 2 hr post-BrdU, single BrdU-positive nuclei were scored as Atoh1-positive or negative. In samples at 15 hr, 24 hr, and 4 days post-BrdU, BrdU-labeled sister pairs were scored. BrdU-labeled nuclei were considered to be sister pairs if they were in close proximity (<1 nuclear length apart) and separated from other BrdU-positive nuclei by >5 nuclear lengths. For each time point, five utricles were examined. Atoh1 labeling was also assessed in mitotic cells (24 mitotic cells across 4 utricles), which were identified by scanning the luminal surface of the epithelium for condensed PI-labeled chromosomes.

For all quantitative measures, ANOVA with Fisher's PLSD was performed using Statview software (SAS Institute). For all analyses, differences with confidence intervals of 5% or less were considered significant.


We thank Dr. Jialin Shang for assistance with histological preparations, Glen MacDonald for assistance with histological analyses and imaging, and Dr. Edwin Rubel for sharing facilities. We acknowledge the assistance of Dr. David Roberson with surgical minipump implantation and the gracious donation by Dr. Jane Johnson of her anti-Atoh1 antibody. We also thank Drs. David Raible and Edwin Rubel for critical comments on the manuscript.