SEARCH

SEARCH BY CITATION

Keywords:

  • Prox;
  • transcription factor;
  • hair cell regeneration;
  • chicken;
  • sensory progenitor cell

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

In birds, mature sensory hair cells are regenerated continually in vestibular epithelia and after damage in the auditory basilar papilla. Molecular mechanisms governing the cellular processes associated with hair cell regeneration are poorly understood. Transcription factors are critical regulators of cell proliferation and differentiation in developing tissues. We examined immunoreactivity for cProx1 during both ongoing and damage-induced hair cell regeneration in chickens. Homologues of this divergent homeobox transcription factor are required for cell cycle withdrawal and differentiation in several vertebrate and invertebrate tissues. In the mitotically quiescent basilar papilla, a population of resting progenitor cells (supporting cells) shows faint nuclear immunoreactivity for cProx1. When auditory hair cell regeneration is triggered by experimental damage, nuclear cProx1 immunolabel is highly elevated in approximately 50% of dividing progenitor cells. Shortly after cytokinesis, all sibling pairs show symmetric patterns of nuclear cProx1 labeling, but pairs with asymmetric labeling emerge shortly thereafter. Strongly immunoreactive cells acquire the hair cell fate, whereas cells with low nuclear immunoreactivity differentiate as supporting cells. By contrast, cProx1 is not detected in any dividing progenitor cells during ongoing regeneration in the utricle. However, nuclear cProx1 immunoreactivity becomes asymmetric in postmitotic sibling cells, and as in the basilar papilla, cells with elevated cProx1 label differentiate as hair cells. In conclusion, cProx1 immunolabeling varies across sensory epithelial progenitors and distinguishes early differentiating hair cells from supporting cells. cProx1 may regulate the proliferative or differentiative capacities of progenitor cells and specify hair cell fate in postmitotic cells during avian hair cell regeneration. Developmental Dynamics 230:597–614, 2004. © 2004 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

The perception of sound and movement in vertebrates depends upon mechanosensory cells located in the inner ear. These cells, called hair cells because of their apical bundles of actin-filled stereocilia, transduce auditory and vestibular stimuli into neurochemical signals that are relayed to the brainstem. They are organized in an orderly array along with nonsensory supporting cells in the sensory epithelia of the inner ear (Fig. 1A). Hair cells are highly susceptible to damage from environmental agents, including intense or prolonged noise, and some therapeutic drugs (e.g., aminoglycoside antibiotics), and they are lost progressively during aging. In mammals, hair cell production is limited to the embryonic period (Ruben, 1967). Therefore, injury or loss of hair cells after birth results in permanent sensory deficits. By contrast, in birds and many cold-blooded vertebrates, hair cell loss triggers the proliferation of supporting cells (progenitor cells; reviewed in Stone and Rubel, 2000a). Their progeny differentiate into hair cells and supporting cells that later become properly patterned within the epithelium and innervated (reviewed in Cotanche, 1999; Stone and Rubel, 2000a). This process results in the restoration of hearing and vestibular function (Carey et al., 1996; Goode et al., 1999; reviewed in Smolders, 1999; Bermingham-McDonogh and Rubel, 2003).

thumbnail image

Figure 1. Organization of inner ear sensory epithelia in chickens. A: The general histological organization of the chicken's auditory and vestibular epithelia is shown in this illustration. Hair cells (dark gray cytoplasm) reside near the apical or luminal side of the epithelium. Hair cell nuclei (HC-Nu) are distributed in an approximate monolayer at a location between 1/3 (auditory) and 2/3 (vestibular) of the distance between the lumen and the basal lamina (BL). Supporting cells (white cytoplasm) extend from the lumen to the basal lamina. Their nuclei (SC-Nu) usually reside below the hair cell nuclei, between approximately 1/2 (auditory) and 4/5 (vestibular) of the distance between the lumen and the basal lamina. B: This figure is a rendering of a surface view of a gentamicin-treated basilar papilla at 3–4 days postgentamicin, using our damage protocol. The proximal (left) end and distal (right) end and the locations of the damaged and undamaged regions of the epithelium are indicated.

Download figure to PowerPoint

In birds, there are differences in the manner in which hair cells are regenerated in mature vestibular and auditory epithelia. In vestibular epithelia, hair cells die spontaneously at a slow, steady rate and in small foci (Jørgensen, 1991; Kil et al., 1997). This ongoing hair cell loss is mirrored by a comparable degree of progenitor cell proliferation (Jørgensen and Mathiesen, 1988; Roberson et al., 1992; Kil et al., 1997; Wilkins et al., 1999). In the utricle, the best-studied vestibular organ in the chicken to date, most progenitor cell divisions lead to the production of a pair of sibling cells that differentiate asymmetrically, as a hair cell and a supporting cell (Roberson et al., 1992; Stone and Rubel, 1999; Stone et al., 1999). By contrast, cells in the bird's auditory epithelium (the basilar papilla) are quiescent with respect to cell death and proliferation until exposed to a damaging stimulus (Corwin and Cotanche, 1988; Ryals and Rubel, 1988; Oesterle and Rubel, 1993). Treatment with a single injection of the ototoxic antibiotic, gentamicin, results in the total destruction of a large field of hair cells in the proximal, high frequency end of the basilar papilla (Bhave et al., 1995; Janas et al., 1995; Roberson et al., 1996, 2000; Stone et al., 1996; Stone and Rubel, 2000b; Fig. 1B). Accordingly, progenitor cells in this area enter the cell cycle and divide, forming new cells. Unlike the normal utricle, the drug-damaged basilar papilla shows little rigidity with respect to how new cells that are born at the peak of cell production differentiate. Three days after a single gentamicin injection, when most progenitor cells are dividing (Bhave et al., 1995; Stone et al., 1999), there is equal likelihood that any given mitotic event will form a hair cell and a supporting cell, two supporting cells, or two hair cells (Stone and Rubel, 2000b).

The molecules that regulate cell fate acquisition during avian hair cell regeneration have not been identified. However, considerable insight has been gained into the molecules required for hair cell development (e.g., see Eddison et al., 2000; Fekete and Wu, 2002). Cell fate decisions in developing inner ear sensory epithelia are controlled in part by transcription factors (for review see Fekete, 1999; Cantos et al., 2000; Fritzsch et al., 2000; Fekete and Wu, 2002). For example, in mice, two basic helix-loop-helix (bHLH) transcription factors, mouse atonal homolog 1 (Math1) and Hairy/Enhancer of Split 1 (Hes1), act antagonistically to regulate acquisition of the hair cell fate. Math1 is the murine homolog of atonal, which acts in fruit flies to promote neural specification (Jarman et al., 1993; Ben-Arie et al., 1996, 2000). It is expressed in young hair cells in the mouse inner ear and zebrafish lateral line organs (Bermingham et al., 1999; Shailam et al., 1999; Lanford et al., 2000; Itoh and Chitnis, 2001; Chen et al., 2002). Mice with targeted deletion of Math1 fail to form hair cells in vestibular and auditory organs (Bermingham et al., 1999). Furthermore, ectopic hair cell production is stimulated in the neonatal cochlea when it is transfected with Math1 (Zheng and Gao, 2000). By contrast, Hes1, a bHLH transcription factor homologous to Enhancer of Split in fruit flies (de la Pompa et al., 1997), is expressed in murine supporting cells (Lanford et al., 2000; Shailam et al., 1999), and hair cells are overproduced when the Hes1 gene is knocked out in mice (Zheng et al., 2000; Zine et al., 2001).

In the current study, we examined expression of the homeobox-like transcription factor chicken Prox1 (cProx1; Tomarev et al., 1996) during hair cell regeneration in the chicken inner ear. cProx1 is homologous to Prospero (Pros) in fruit flies (Chu-Lagraff et al., 1991; Doe et al., 1991; Vaessin et al., 1991; Matsuzaki et al., 1992; Tomarev et al., 1996) and to Prox1 in mouse (Oliver et al., 1993; Tomarev et al., 1998), zebrafish (Glasgow and Tomarev, 1998), and humans (Zinovieva et al., 1996). The homeodomain sequences of cProx1 and murine Prox1 are identical, and the homeodomain regions of cProx1 and Pros share 66% sequence homology (Oliver et al., 1993; Tomarev et al., 1996, 1998). Pros activity in flies is important for several cellular processes associated with sensory and neural development: cell cycle withdrawal, cellular specification, and cellular differentiation (Myster and Duronio, 2000). In the fruit fly central nervous system (CNS), Pros drives cells to withdrawal from the cell cycle, presumably by inhibiting transcription of positive regulators of cell cycle progression (Li and Vaessin, 2000). Pros is also necessary for the differentiation of several cells types in the developing CNS, including ganglion mother cells, or GMCs (Doe et al., 1991), neurons (Doe et al., 1991), glia (Akiyama-Oda et al., 2000; Freeman and Doe, 2001), and cells in the developing peripheral nervous system (PNS; Vaessin et al., 1991; Salzberg et al., 1994). Pros activity is also required for proper differentiation of cells in the CNS and PNS of adult fruit flies (Manning and Doe, 1999; Reddy and Rodrigues, 1999; Ceron et al., 2001).

Studies of Prox1 in mice demonstrate it is also important for cell cycle exit and/or cellular differentiation in several tissues in vertebrates as well. In the developing lens, loss of Prox1 function leads to decreased expression of negative regulators of the cell cycle, prolonged cell division, and arrested differentiation of lens fibers (Wigle et al., 1999). Prox1 is also necessary for cell cycle exit and differentiation of horizontal cells in the developing retina (Dyer et al., 2003). In lymphatic vessels, Prox1 acts to induce cell proliferation and promote differentiation of lymphatic endothelial cells (Wigle and Oliver, 1999, 2002; Petrova et al., 2002). In the liver, Prox1 is required for normal developmental migration of hepatocytes (Sosa-Pineda et al., 2000). The spatial and temporal pattern of Prox1 expression in the murine brain suggests the transcription factor may regulate the differentiation of neural progenitors and terminally mitotic neurons (Torii et al., 1999; Pleasure et al., 2000).

We recently reported that nuclear cProx1 immunoreactivity is restricted to the nuclei of neuronal and sensory progenitors and precursors in the developing inner ear of chickens (Stone et al., 2003). Prox1 is also expressed in the zebrafish otocyst (Glasgow and Tomarev, 1998) and in hair cells of the cavefish lateral line (Jeffery et al., 2000). In studies described here, we show that nuclear cProx1 immunoreactivity distinguishes subsets of hair-cell progenitors across the mature inner ear as well and that nuclear cProx1 immunoreactivity is elevated in postmitotic hair cells but not supporting cells. Our data also demonstrate that molecular mechanisms underlying hair cell regeneration in mature chickens share important features with sensory organ formation in adult fruit flies.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

cProx1 Immunoreactivity in the Mitotically Quiescent Basilar Papilla

The specificity of the anti-Prox1 antibody used to detect chicken cProx1 in this study has been established in previous studies (Belecky-Adams et al., 1997; Stone et al., 2003). We used confocal laser scanning microscopy in whole-mount preparations to examine cProx1 immunolabeling in the untreated basilar papilla of 7-day-old birds, which is mitotically quiescent (Corwin and Cotanche, 1988; Ryals and Rubel, 1988; Oesterle and Rubel, 1993). In supporting cells, there was little or no cytoplasmic labeling, but some nuclei were cProx1-immunoreactive (Fig. 2A,B). Levels of cProx1 labeling in supporting cell nuclei ranged from undetectable to low, with supporting cells along the superior, neural edge of the basilar papilla showing the highest degree. No proximal-to-distal gradients in labeling were evident. We found a similar variation in cProx1 immunoreactivity among supporting cell nuclei in 1-month-old birds (data not shown).

thumbnail image

Figure 2. cProx1 immunolabeling in the normal and the damaged basilar papilla. A,B: The supporting cell nuclear layer of proximal (A) and distal (B) papilla is shown. The neural edge of the epithelium is toward the top of each panel. Thin arrows point to faintly cProx1-positive (green) nuclei. Nuclei are counterstained with propidium iodide (red). Arrowheads indicate cProx1-negative nuclei. Insets in A,B show cProx1 immunolabel at higher magnification. C,D: cProx1 immunolabeling at 3 days postgentamicin in proximal damaged (C) and distal undamaged (D) area. The neural edge is toward the top. Arrows point to strongly cProx1-positive nuclei, and arrowheads point to cProx1-negative nuclei. A cProx1-labeled Vibratome section of undamaged basilar papilla (middle region) is shown in E. Hair-cell cytoplasm and nuclei are indicated by thick and thin arrows, respectively. F–H: cProx1-labeled Vibratome sections through damaged region at 3 days (F), 8 days (G), and 14 days postgentamicin (H). Arrows in G and H indicate cProx1-positive cells (hair cell nucleus in G, HC cytoplasm in H). Double-arrowed line in E–H indicates the depth of the basilar papilla. Asterisks in F–H indicate connective tissue beneath the basilar papilla. In E–H, the neural edge is toward the left. HC-Nu, hair cell nuclear layer; SC-Nu, supporting cell nuclear layer. Scale bar in H = 15 μm in A–H, 8 μm in insets.

Download figure to PowerPoint

Hair cells throughout the basilar papilla were immunolabeled for cProx1 (Fig. 2E), but, in contrast to supporting cells, their labeling was cytoplasmic in nature, not nuclear, and it was highly uniform across the hair cell population. There was no immunoreactivity in the neural processes in the basilar papilla (data not shown). Negative control tissue (primary antibody omitted) showed no immunoreaction above background levels (Fig. 2, green dots; data not shown).

cProx1 Immunoreactivity in the Regenerating Basilar Papilla

To stimulate hair cell regeneration in the basilar papilla, chicks were administered a single injection of the aminoglycoside antibiotic gentamicin. This treatment causes complete loss of hair cells in the proximal one third of the basilar papilla by 3–4 days after the injection and no hair cell loss in the distal two thirds (Bhave et al., 1995; Janas et al., 1995; Roberson et al., 1996, 2000; Stone et al., 1996; Fig. 1B). Proliferation of supporting cells is initiated in the damaged region by 1 day after gentamicin injection, peaks by 3 days, and falls to near-normal levels by 7–9 days (Bhave et al., 1995; Stone et al., 1999). Most supporting cells appear to go through a single round of cell division during the course of auditory hair cell regeneration after gentamicin exposure (Stone et al., 1999). Very few, if any, supporting cells are stimulated to divide in the undamaged region of the basilar papilla (Bhave et al., 1995).

Relative to untreated birds, cProx1 immunolabeling in cell nuclei increased dramatically in the damaged (proximal) region of the basilar papilla at 3 days after the gentamicin injection (Fig. 2C). This increase was evident in nuclei in both the hair cell and supporting cell nuclear layers. No significant change in labeling was detected in the undamaged (distal) portion of the basilar papilla (Fig. 2D) or in the connective tissue underlying the damaged basilar papilla (Fig. 2F–H). We examined the time course of the up-regulation in nuclear cProx1 immunoreactivity in the damaged region after gentamicin treatment. At 2 days postgentamicin, we detected only a few strongly labeled nuclei (data not shown). At 3 days postgentamicin, there were many cProx1-positive nuclei in both the hair cell nuclear layer and supporting cell nuclear layer (Fig. 2F). At 8 days postgentamicin, there was a marked decline in the number of labeled nuclei (Fig. 2G). At 14 days postgentamicin, only a few labeled nuclei were detected, and heavy cytoplasmic labeling was seen among some cells in the hair cell nuclear layer (Fig. 2H). These cells were probably regenerated hair cells, which are detectable using hair cell-selective markers in the damaged region by 3–4 days after gentamicin injection and which attain mature morphologies by 10–15 days postgentamicin (Stone et al., 1996; Stone and Rubel, 2000b).

Changes in the number of cProx1-positive nuclei in the damaged region were quantified in whole-mounts at different times after gentamicin injection (Fig. 3). We counted heavily labeled cProx1-positive nuclei in three areas of the damaged region: the proximal edge, the middle zone, and the distal edge (along the zone of transition from damaged to undamaged tissue). Labeled cells in both the hair cell and supporting cell layers were included in the analysis. In all three areas, we found a similar trend (data not shown), so these numbers were averaged to obtain a representative value for the damaged region, which is shown in Figure 3. This value peaked at 3 days postgentamicin and returned to low levels by 8 days. The time course we observed is very similar to that of 5′-bromo-2-deoxyuridine (BrdU) incorporation after gentamicin treatment (Bhave et al., 1995; Stone et al., 1999). Interestingly, a few heavily labeled cProx1-positive nuclei were detected as late as 30 days after gentamicin injection.

thumbnail image

Figure 3. Time course of cProx1 up-regulation during regeneration in the basilar papilla. This graph compares the time course of changes in cProx1 immunoreactivity in the lesioned region of the basilar papilla after gentamicin treatment with the time course of bromodeoxyuridine (BrdU) incorporation in a similar experimental paradigm. cProx1 data were derived at each time point by determining the number of cProx1-positive nuclei in three different areas (41,209 μm2 each) within the lesioned region and averaging the numbers across the areas. BrdU data were published previously (Stone et al., 1999) and were obtained for each time point by counting the total number of BrdU-positive nuclei present in the lesioned region after a short pulse/fix paradigm. It is important to note that different sampling methods were used to obtain the cProx1 and BrdU data sets, so they are not directly comparable from a quantitative perspective.

Download figure to PowerPoint

cProx1 Immunoreactivity Is Elevated in Some Dividing Progenitor Cells

Because most progenitor cells divide at 3 days postgentamicin, when the number of cProx1-positive nuclei peaks, we examined basilar papillas at this time to determine whether cells that up-regulate nuclear cProx1 immunoreactivity are mitotically active. Chickens received a single injection of gentamicin and recovered for 3 days. Then they received a single injection of BrdU, which is incorporated into DNA during S phase. Birds were killed at different times after BrdU injection to permit cells that become labeled in S phase to progress to later stages of the cell cycle. Whole-mount basilar papillas were fixed and double-labeled to detect BrdU and cProx1. Killing of chickens at 30 min post-BrdU injection allowed us to label progenitor cells in S phase (Fig. 4A). Of 165 BrdU-positive cells counted in two basilar papillas at 30 min post-BrdU, 79 (48%) were strongly cProx1-positive and 86 (52%) were weakly labeled or unlabeled. We noted a similar variation in cProx1 labeling in BrdU-positive progenitors when birds were killed 4 hr after a single BrdU injection (data not shown), suggesting that differences were not dependent upon a cell's position in the cell cycle.

thumbnail image

Figure 4. cProx1 is labeled in some dividing progenitor (supporting) cells in the damaged basilar papilla. All panels show cProx1 labeling in cells at different stages of the cell cycle in whole-mount basilar papillas at 3 days postgentamicin. cProx1 is shown in green in all panels. The label shown in red is indicated in each panel. A: A bromodeoxyuridine (BrdU)-positive nucleus that is cProx1-positive at 30 min post-BrdU injection is indicated by the thin arrow. A BrdU-positive and cProx1-negative cell is indicated by the thick arrow. The arrowhead shows a BrdU-negative, cProx1-positive cell. B: pH3 immunolabeling reveals two sets of chromosomes in this dividing cell (arrowhead); the cytoplasm is cProx1-negative. C: Two M-phase cells that are BrdU-positive and cProx1-positive are shown (arrowheads) at 10 hr post-BrdU. D: The two sets of chromosomes in this cProx1-positive cell in late telophase are well separated and condensed, as shown by BrdU labeling at 10 hr post-BrdU. E: MPM2 immunolabeling identifies cells in M phase. In this panel, two MPM2-positive cells (orange, thick arrows) are cProx1-positive, and one MPM2-positive cell (red, thin arrow) is cProx1-negative. cProx1 distribution appears uniform within the cytoplasm of these labeled cells. Scale bar = 11.5 μm in E (applies to A–E).

Download figure to PowerPoint

We also examined cProx1 immunolabeling in progenitor cells in M phase in whole-mount preparations of the basilar papilla. To do this, we used two methodological approaches. We examined BrdU-positive cells in birds killed at 10 hr after BrdU injection administered on day 3 postgentamicin (Fig. 4C,D). Many BrdU-labeled cells at this time had progressed to M phase and were located near the lumen, where mitosis occurs (Katayama and Corwin, 1989; Raphael, 1992; Stone and Cotanche, 1994), although the cell's position relative to the lumen is not apparent in the figure. We also labeled tissue with antibodies to M phase-specific antigens, pH3 (Fig. 4B) and MPM2 (Fig. 4E; Westendorf et al., 1994; Wei et al., 1998) and then counterlabeled it with the anti-cProx1 antibody.

We detected no systematic pattern of variability in cProx1 labeling among M-phase cells; even cells at similar stages of M phase showed different levels of cProx1 protein. Furthermore, none of the cProx1-labeled cells in M phase that we examined displayed asymmetric distribution of the cProx1 protein within their cytoplasm. Most of the cells in this figure were fixed and imaged while they were in anaphase or telophase; the specific submitotic stage of cells in Figure 4E is not known because the chromosomes are not rendered visible with the MPM2 label. A dividing cell is shown in Figure 4B, and the cytoplasm around both of its chromosomal sets appears to have little or no cProx1 immunoreactivity. Two dividing cells are shown in Figure 4C, and the cytoplasm within these cells is uniformly cProx1-positive. A dividing cell is shown in Figure 4D, and the cytoplasm around both of its chromosomal sets appears uniformly cProx1-positive. In Figure 4E, three MPM2-positive cells in M phase are shown. Two of these are cProx1-positive, and one is cProx1-negative.

cProx1 Immunoreactivity Becomes Asymmetric in Some Postmitotic Sibling Pairs

We examined cProx1 immunoreactivity in postmitotic cells at different stages of maturation in the regenerating auditory epithelium. To do this, chicks received an injection of BrdU at 3 days postgentamicin and were killed between 7 hr and 3 days later. Basilar papillas were dissected and immunoreacted to detect cProx1 and BrdU. We observed newly regenerated, BrdU-positive cells over time, as they differentiated, focusing our attention on cProx1 immunolabeling in pairs of BrdU-labeled cells assumed to be postmitotic siblings based on their proximity and similar BrdU labeling patterns (Stone and Cotanche, 1994; Stone and Rubel, 1999, Stone and Rubel, 2000b; Stone et al., 1999).

At 3 days post-BrdU, which corresponds to 5–6 days postgentamicin, numerous regenerated cells throughout the damaged area are in the process of differentiating as hair cells or supporting cells (Stone et al., 1996; Stone and Rubel, 2000b). At this time, we determined that sibling pairs within the damaged region display three distinct patterns of nuclear cProx1 labeling (Fig. 5A): (1) both cells with similar, high levels of nuclear cProx1 (symmetrically high); (2) both cells with low or no apparent nuclear cProx1 (symmetrically low); and (3) each cell in the pair with a very different level of nuclear cProx1 (asymmetric; also see Fig. 5B–D″). Analysis of basilar papillas (n = 5) revealed that these combinations appeared with similar frequency (Table 1; Fig. 6A): 35% of sibling pairs had symmetrically high levels of nuclear cProx1 immunoreactivity, 30% of sibling pairs had symmetrically low levels, and 35% of sibling pairs had highly asymmetric levels. Means are also provided in Table 1 and Figure 6A. The relatively small number of pairs in this analysis (Table 1) reflects our caution in ensuring that pairs of BrdU-labeled cells were sufficiently isolated from other cells, showed similar patterns of BrdU labeling and were, therefore, likely to have been generated from a common cell division.

thumbnail image

Figure 5. cProx1 immunoreactivity in early postmitotic and differentiating cells in the regenerating basilar papilla. A–D: Whole-mount basilar papillas immunolabeled for cProx1 (green) and bromodeoxyuridine (BrdU, red) are shown. A: BrdU labeling defines pairs of sister cells labeled with cProx1 at 2 days post-BrdU injection (equivalent to 5 days postgentamicin). Arrowheads point to pairs of cells with asymmetric levels of nuclear cProx1; short arrows point to pairs of cells with symmetric, low levels of nuclear cProx1; and the long arrow points to a pair of cells with symmetric, high levels of nuclear cProx1. B–D: cProx1 labeling patterns in pairs of BrdU-positive sibling cells shortly after mitosis. B: Sibling cells with symmetric high levels of cProx1 (+/+) at 10 hr post-BrdU. C: Sibling cells with symmetric low levels of cProx1 (−/−) at 8 hr post-BrdU. D: Sibling cells with asymmetric levels of cProx1 (+/−) at 17 hr post-BrdU. B′–D″: The same cells as shown in B–D, respectively, with the red channel (BrdU labeling) shown separately on top (B′–D′) and the green channel (cProx1 labeling) shown separately on the bottom (B″–D″). Arrows in C″ and D″ indicate the positions of the BrdU-labeled nuclei. Scale bar in A = 17.5 μm in A, 10 μm in B–D″.

Download figure to PowerPoint

Table 1. cProx1 Expression in Sibling Cell Pairs Born in the Drug-Damaged Basilar Papillas 3 Days Postgentamicina
 Sample#+/−+/+−/−
  • a

    cProx1 immunolabeling in pairs of BrdU-labeled cells was examined at 8 hours, 17 hours, 24 hours, and 3 days after a single BrdU injection administered at 3 days postgentamicin. There are 3 possible combinations of labeling patterns in pairs: one cell is cProx1+, one is cProx− (+/−), both cells are cProx1+ (+/+), or both cells are cProx1− (−/−). Numbers reflect pairs of cells with each combination. SD = standard deviation from the mean. hpBr, hours postbromodeoxyuridine; dBr, days postbromodeoxyuridine; BrdU, bromodeoxyuridine.

8h pBr1157
 2076
 3011
 4012
 5069
 Mean (SD)0.2 (0.4)4 (2.8)5 (3.4)
 % of total24454
17h pBr1536
 261218
 33910
 40611
 5256
 6145
 Mean (SD)2.8 (2.3)6.5 (3.4)9.3 (4.9)
 % of total153550
24h pBr1765
 2420
 37310
 41378
 58313
 6626
 7132
 8321
 Mean (SD)6.1 (3.6)3.5 (1.9)5.6 (4.6)
 % of total402337
3d pBr1653
 29108
 3441
 4879
 5123
 Mean (SD)5.6 (3.2)5.6 (3.1)4.8 (3.5)
 % of total353530
thumbnail image

Figure 6. Quantification of cProx1 labeling in pairs of bromodeoxyuridine (BrdU)-labeled cells at different times after BrdU injection. A,B: The numbers of sibling cell pairs that are symmetrically cProx1-positive (+/+), asymmetric with respect to cProx1 labeling (+/−), or symmetrically cProx1-negative (−/−) are shown in mean values (plus SD; left) or as a percentage of total pairs (right) for the drug-damaged basilar papilla at 3 days postgentamicin (A) and for the undamaged utricle (B). Samples were analyzed at 8 hr, 17 hr, 24 hr, or 3 days after a single BrdU injection given to chickens at 3 days postgentamicin (for basilar papillas) or to untreated chickens (for utricles). Numbers for each time point are provided in the Experimental Procedures section.

Download figure to PowerPoint

We looked at differentiating sibling cells at earlier time points post-BrdU to determine when the different patterns of cProx1 immunoreactivity emerge after mitosis. Between 7 and 10 hr post-BrdU, pairs of postmitotic sibling cells typically displayed either symmetrically high or symmetrically low levels of nuclear cProx1 labeling (Fig. 5B–C″). We rarely identified pairs that had asymmetric cProx1 labeling until 17 hr post-BrdU (Fig. 5D–D″), or 7–10 hr after cytokinesis. At 8 hr post-BrdU, 2% of sibling pairs were asymmetrically labeled, 44% were symmetrically high, and 54% were symmetrically low (Fig. 6A; Table 1). At 17 hr post-BrdU, there was a substantial increase in the percentage of pairs that were asymmetrically labeled (15%) and a decrease in the percentage of pairs that were symmetrically high (35%) or symmetrically low (50%). At 24 hr post-BrdU, the percentage of asymmetrically labeled pairs reached 40%, which was comparable to 3 days post-BrdU (35%). Concomitantly, the percentage of pairs that had symmetrically high or low levels of cProx1 immunoreactivity fell to 23% and 37%, respectively. These findings demonstrate that the nuclei of both cells in the majority sibling pairs are either cProx1-positive or cProx1-negative shortly after mitosis, but over time, a single cell in some sibling pairs alters its relative level of nuclear cProx1 protein and establishes an asymmetric pattern.

cProx1 Immunoreactivity Is Elevated in Regenerated Hair Cells

Data presented in the previous section led us to examine whether the distribution of cProx1 among postmitotic cells correlates with a particular cell fate. We triple-labeled basilar papillas at 4 days postgentamicin with antibodies to cProx1, calmodulin, and class III beta tubulin, using the TuJ1 antibody (Lee et al., 1990). The latter two proteins are detected in abundance in the cytoplasm of mature and differentiating hair cells in chickens but are not detected in supporting cells (Stone at al., 1996; Stone and Rubel, 2000b). At 4 days postgentamicin, numerous new hair cells and supporting cells at early stages of differentiation are scattered throughout the damaged region. Because our damage protocol reliably results in complete hair cell loss in the proximal one third of the epithelium, we felt confident that cells with cytoplasm immunoreactive for calmodulin are newly formed hair cells rather than hair cells that survived the damage protocol. Additional assurance was provided by the relative disorganization and morphological immaturity of hair cells in the damaged area when compared with an undamaged region of the same organ (data not shown). By 4 days postgentamicin, many calmodulin- and/or TuJ1-positive cells were present in the hair cell nuclear layer of the damaged epithelium. These cells had morphologic features of immature regenerated hair cells (Stone and Rubel, 2000b). Several of these young hair cells had strongly cProx1-positive nuclei (Fig. 7A). Cells in the supporting cell nuclear layer did not exhibit significant labeling for calmodulin, TuJ1, or cProx1 (Fig. 7B). By injecting chicks with BrdU at 3 days postgentamicin and examining basilar papillas at different times post-BrdU, we determined that new hair cells show immunoreactivity for both cProx1 and calmodulin by 2 days after their birth dates (Fig. 7C).

thumbnail image

Figure 7. Predominant localization of cProx1 to regenerated hair cells in the basilar papilla. Confocal images of whole-mount basilar papillas, with cProx1 labeling green in all panels. Colors of other labels are indicated. A,B: Basilar papilla, 4 days postgentamicin. A: A scan through hair cell nuclear (HC-Nu) layer. Arrow indicates cProx1-positive nucleus in calmodulin (CaM)/TuJ1-positive (both red) hair cell. B: Supporting cell nuclear (SC-Nu) layer from the same field is shown. TuJ1-positive neural processes (arrowhead) from eighth cranial nerve are evident. C: Basilar papilla, 2 days postbromodeoxyuridine (BrdU; 5 days postgentamicin). Thin arrow indicates cell labeled for BrdU and cProx1 in nucleus and calmodulin in cytoplasm. Thick arrow shows BrdU-positive, cProx1-negative nucleus surrounded by calmodulin-negative cytoplasm. D–G: Basilar papillas from chicks given BrdU at 3 days postgentamicin and killed 9 (D–F) or 16 (G) days later. Labels are the same as in C (same color scheme). D is a scan through the HC-Nu layer, and E is a scan through the SC-Nu layer in the same field. One pair of BrdU-positive sibling cells is shown, with its nuclei split between the two layers (arrows indicate each nucleus). D: The cell in the HC-Nu layer is labeled for cProx1 and calmodulin. E: The cell in SC-Nu layer is cProx1-negative and calmodulin-negative. F: Pair of BrdU-labeled sibling cells (arrows) in HC-Nu layer. Both cells are positive for cProx1 and calmodulin. G: Pair of BrdU-positive sibling cells (arrows) in SC-Nu layer. Both cells are negative for cProx1 and calmodulin. Scale bar in C = 18 μm in A–C, 15 μm in D–G.

Download figure to PowerPoint

These observations strongly suggested that cells with high immunoreactivity for cProx1 shortly after mitosis acquire the hair cell fate while cells with low levels become supporting cells. We explored this possibility further by studying whether the fate of each cell in a sibling pair is correlated with its intensity of nuclear cProx1 labeling. Chicks received a single injection of BrdU at 3 days postgentamicin and were killed at 3, 9, or 16 days later. Basilar papillas were fixed and triple-labeled to detect BrdU, cProx1, and calmodulin. As before, siblings were identified by their proximity and their patterns of BrdU labeling. Within a given cell, relative levels of cProx1 and calmodulin tended to be comparable; pairs with asymmetric levels of cProx1 tended to have asymmetric levels of calmodulin, and the cell in the pair with the highest degree of cProx1 labeling also had the strongest calmodulin labeling (Fig. 7D,E). In pairs of cells with symmetric, high levels of cProx1, both cells tended to have high levels of calmodulin (Fig. 7F). In pairs of cells with symmetric, low levels of cProx1, both cells usually had low or undetectable calmodulin labeling (Fig. 7G). For samples at 3 days post-BrdU, the relative intensity of calmodulin and cProx1 immunolabeling in each cell in a pair was quantitatively analyzed. Cells with medium-to-high immunolabeling were scored as positive, and cells with immunolabeling ranging from undetectable to low in intensity were scored as negative. This analysis revealed a systematic relationship with respect to calmodulin and cProx1 immunoreactivity among sibling pairs at 3 days post-BrdU (Table 2; P < 0.0001; Chi squared = 82.2; contingency coefficient = 0.798; df = 4). These data were further supported by our observations that the cProx1-positive cell(s) in the pair was usually located in the hair cell layer and the cProx1-negative cell(s) in the pair was usually located in the supporting cell layer. Collectively, these data demonstrate that cProx1 immunoreactivity is elevated in the nuclei of nascent hair cells but is present in small amounts, or is absent from, the nuclei of nascent supporting cells.

Table 2. Patterns of Calmodulin and cProx1 Labeling in Sibling Cell Pairs Born in Drug-Damaged Basilar Papillas at 3 Days Postgentamicina
Sample C+/C−C+/C+C−/C− 
  • a

    cProx1 and calmodulin immunolabeling was examined in pairs of BrdU-labeled cells days after a single BrdU injection administered at 3 days postgentamicin. Combinations of labeling of cells within each pair for each marker are shown. P, cProx1; C, calmodulin; +, positively labeled; −, weakly labeled or unlabeled. Numbers reflect pairs of cells with each combination. Chi square = 82.2, df = 4; chi square P value < 0.0001. BrdU, bromodeoxyuridine.

1P+/P−1001 
 P+/P+120 
 P−/P−008 
2P+/P−500 
 P+/P+020 
 P−/P−004 
3P+/P−600 
 P+/P+040 
 P−/P−004 
MeanP+/P−700 
 P+/P+02.60 
 P−/P−005.3 
Chi squared analysis of contingency
     Totals
 P+/P−210122
 P+/P+1809
 P−/P−001616
 Totals22817 

cProx1 Immunoreactivity in the Undamaged Utricle

For comparison, we examined cProx1 immunolabeling in the untreated posthatch chicken utricle, which exhibits ongoing spontaneous hair cell death and regeneration (Jørgensen and Mathiesen, 1989; Jørgensen, 1991; Roberson et al., 1992; Kil et al., 1997). Unlike the drug-damaged basilar papilla, the majority of mitotic events in the normal utricle result in the production of a hair cell and a supporting cell (Stone and Rubel, 1999; Stone et al., 1999). Thus, this tissue serves as a good model in which to further test the hypothesis that high levels of nuclear cProx1 shortly after mitosis predict hair cell fate.

In the utricle, cProx1 immunoreactivity was detected in the cytoplasm of hair cells (Fig. 8A). Strongly labeled cell nuclei were concentrated in the hair cell nuclear layer (Fig. 8B), and some were seen in or at the top of the supporting cell nuclear layer (Fig. 8C). To examine if progenitor cells in the process of dividing are immunoreactive for cProx1, normal, untreated chickens at posthatch day 7 were given a single injection of BrdU and killed 2 hr later. Unlike the regenerating basilar papilla after drug damage, none of the BrdU-labeled cells (progenitor cells in S phase) showed significant cProx1 immunoreactivity (Fig. 8D,E). This was also the case in utricles that were administered a single injection of gentamicin and examined 3 days later, when cell division is highly up-regulated (Stone and Rubel, 1999; data not shown). Analysis of cProx1 expression in M-phase progenitors in the utricle was not feasible due to the large number and high density of mature hair cells, which exhibit strong cProx1 immunoreactivity at the luminal surface, where mitosis occurs.

thumbnail image

Figure 8. cProx1 immunoreactivity in the undamaged utricle. Confocal images of cProx1 immunolabeled whole-mount utricles, with cProx1 labeling in gray scale in A–C and as indicated, and in green in F–I. Other labels are indicated for each panel. A–C: cProx1 labeling is evident in hair cell tops (HC-top, A), hair cell nuclei (HC-Nu, B) and supporting cell nuclei (SC-Nu, C). D,E: Bromodeoxyuridine (BrdU) -positive (D) and cProx1-positive (E) nuclei in the SC-Nu layer from utricle fixed 2 hr post-BrdU. Arrows in D,E point to positions of BrdU-labeled cells in D. F,G: Labeling for BrdU and cProx1 in two sibling cell pairs at 17 hr post-BrdU. One pair has symmetrically low cProx1 immunoreactivity (F), and one pair has asymmetric cProx1 immunoreactivity (G). In F′–G″, labeling for each antigen is shown separately. H–I′″: Labeling for BrdU, cProx1, and myosinVI in a sibling pair (arrows) at 5 days post-BrdU. The BrdU-positive cell in HC-Nu layer (H) has elevated nuclear cProx1 and cytoplasmic myosinVI immunolabeling relative to the cell in SC-Nu layer (I). Arrowheads in H indicate additional myosinVI-positive hair cells. In H′–I′″, labeling for each antigen is shown separately. Note that myosinVI labeling appears erroneously in the hair cell nucleus. This occurs because both the anti-cProx1 and anti-myosinVI antibodies are raised in rabbit; some secondary antibody directed against the anti-myosinVI antibody bound to the anti-cProx1 antibody. Scale bar in A = 14 μm in A–E,H,I, 7 μm in F–G″, H′–I′″.

Download figure to PowerPoint

We examined immunolabeling in postmitotic sibling pairs in control utricles at different times post-BrdU to determine how cProx1 expression changes over the course of hair cell differentiation. Normal posthatch day 7 chicks received a BrdU injection, and utricles were fixed 8 hr, 17 hr, 24 hr, or 3 days later and immunoreacted to detect BrdU and cProx1. Pairs were identified based on their proximity and similarity of BrdU labeling, and cProx1 labeling was analyzed at each time point. At 8 hr post-BrdU, the vast majority of siblings (90%) were symmetrically cProx1-negative (Table 3; Fig. 6B; means are also shown), 4% were symmetrically positive, and 6% were asymmetric. Figure 8F–G″ shows examples of symmetrically negative and asymmetric pairs at 17 hr post-BrdU. Six percent of pairs had symmetrically high levels of cProx1, 69% had symmetrically low or undetectable levels, and 25% had asymmetric levels. At 24 hr post-BrdU, 6% of pairs were symmetrically cProx1-positive, 41% were symmetrically cProx1-negative, and 53% were asymmetric. At 3 days post-BrdU, 7% of pairs had symmetrically high levels of cProx1, 17% had symmetrically low or undetectable levels, and 76% were asymmetric. At this time, the cProx1-positive nucleus in the pair tended to be rounder and bigger and located closer to the lumen than the cProx1-negative nucleus (data not shown). These features of cProx1-positive nuclei are characteristic of hair cells, but not supporting cells, in the chicken utricle (Roberson et al., 1992; Stone and Rubel, 1999; Stone et al., 1999). To further address this, we injected chicks with BrdU, fixed utricles at 5 days post-BrdU, and triple-labeled utricles with antibodies to BrdU, cProx1, and myosinVI. MyosinVI is an unconventional myosin that has been detected immunologically in hair cells in several nonavian species (e.g., Hasson et al., 1997). Regenerated cells with elevated nuclear cProx1 immunoreactivity at 5 days post-BrdU were located in the hair cell layer and had myosinVI-positive cytoplasm (Fig. 8H–I′″), confirming their identity as hair cells. Attempts to examine cProx1 immunoreactivity in regenerated hair cells at earlier time points relative to cell birth were unsuccessful due to the lack of a hair cell-specific cytoplasmic antigen that is detectable in conjunction with the cProx1 immunolabeling.

Table 3. cProx1 Expression in Sibling Cell Pairs Born in Normal Utriclesa
 Sample#+/−+/+−/−
  • a

    cProx1 immunolabeling in pairs of BrdU-labeled cells was examined 3 days after a single BrdU injection administered to untreated birds. There are three possible combinations of labeling patterns in pairs: (1) one cell is cProx1-positive, one is cProx-negative (+/−); (2) both cells are cProx1-positive (+/+); or (3) both cells are cProx1-negative (−/−). Numbers reflect pairs of cells with each combination. SD, standard deviation from mean. hpBr, hours post- bromodeoxyuridine; dBr, days postbromodeoxyuridine; BrdU, bromodeoxyuridine.

8h pBr12012
 21117
 30114
 41123
 Mean (SD)1 (0.8)0.75 (0.5)16.5 (4.8)
 % of total6490
17h pBr17122
 26119
 37215
 410325
 Mean (SD)7.5 (1.7)1.75 (1.0)20.25 (4.3)
 % of total25669
24h pBr11538
 2609
 316112
 Mean (SD)12.3 (5.5)1.3 (1.5)9.7 (2.1)
 % of total53641
3d pBr12211
 22536
 32427
 41525
 Mean (SD)21.5 (4.5)2 (0.8)4.75 (2.6)
 % of total76717

These findings demonstrate that, in the normal utricle, S-phase supporting cells are cProx1-negative and the majority of early postmitotic sibling pairs have very low or undetectable levels of cProx1 immunoreactivity in their nuclei. By 17 hr post-BrdU, there is a substantial increase in the percentage of sibling pairs that display asymmetric nuclear cProx1 immunoreactivity. By 3 days post-BrdU, the majority of sibling pairs exhibit dramatically asymmetric levels of nuclear cProx1 immunolabel. Furthermore, our data demonstrate that nascent hair cells acquire relatively high levels of nuclear cProx1 immunoreactivity after they are born, while nascent supporting cells retain low or undetectable levels.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Chicken Prox1 is a divergent homeobox transcription factor whose homologues in fruit flies and mice regulate cellular proliferation, fate determination, and differentiation in sensory tissues. We examined cProx1 immunolabeling in mitotically quiescent and active progenitor cells and in postmitotic, differentiating cells during regeneration of auditory and vestibular hair cells in mature chickens. We found that immunoreactivity for the transcription factor in the auditory epithelium is (1) absent from some mitotically quiescent supporting cells but detected at low levels in the nuclei of others, in undamaged tissue; (2) absent from half of dividing supporting cells at 3 days after gentamicin treatment but elevated in the nuclei of the other half; (3) symmetric in sibling cells shortly after cell division; and (4) asymmetric in some postmitotic sibling cells as they differentiate distinct phenotypes (hair cell versus supporting cell), being restricted to the new hair cell. In the vestibular epithelium undergoing spontaneous hair cell regeneration, cProx1 immunoreactivity is (1) absent from all dividing supporting cells; (2) absent from both sibling cells shortly after cell division; and (3) asymmetric in the majority of sibling pairs as they differentiate and, as in the basilar papilla, restricted to the new hair cell. These observations demonstrate a here-to-fore unknown molecular diversity among progenitor cells and between differentiating hair cells and supporting cells in the inner ear sensory epithelia. The results are discussed in detail below.

cProx1 and Supporting Cells

The differential immunoreactivity of cProx1 detected across mitotically quiescent supporting cells in the basilar papilla is intriguing because there are few morphological or molecular features that define subsets of supporting cells in the avian auditory epithelium. Auditory supporting cells have many functions: they help to maintain the ionic integrity of the inner ear, they provide mechanical support within the sensory epithelium, and they generate the tectorial membrane in developing and mature ears (Cotanche, 1987; Shiel and Cotanche, 1990; Epstein and Cotanche, 1995). Furthermore, supporting cells serve as progenitors to new hair cells in mature birds (reviewed in Stone and Rubel, 2000a). It is unclear, however, if all supporting cells, or only a subset, serve these different functions. In fact, the term “supporting cell” is used widely to describe any cell in avian sensory epithelia that lacks hair cell features. Several markers have been reported to label all supporting cells within the basilar papilla (for review see Stone and Rubel, 2000a). By contrast, very few proteins are expressed in subsets of supporting cells (e.g., proliferating cell nuclear antigen [Bhave et al., 1995], fibroblast growth factor receptor 3 [Bermingham-McDonogh et al., 2001]). The distribution of these proteins does not appear to correlate with the spatial pattern of cProx1-positive supporting cells that we have observed.

In response to gentamicin treatment, hair cells in the proximal basilar papilla die and supporting cells in that region undergo mitosis. The greatest number of supporting cells enters the cell cycle at 3 days after gentamicin treatment (Bhave et al., 1995; Stone et al., 1999). At this time, there is a large increase in nuclear cProx1 immunoreactivity in the damaged area relative to the corresponding proximal region in untreated birds and relative to the apical, undamaged area in treated birds. Strong cProx1 labeling is seen in only half of the dividing cells at this time; the other half is unlabeled or only weakly labeled. As discussed in the previous section, levels of nuclear cProx1 immunoreactivity are also variable in resting supporting cells. It is not clear if these relative levels are preserved in supporting cells as they transition from a resting state to a mitotic state. In contrast to the drug-damaged basilar papilla, all S-phase supporting cells in the undamaged or damaged utricle show little or no cProx1 immunoreactivity in their nuclei. Therefore, levels of nuclear cProx1 immunolabel are highly variable in dividing supporting cells across and within sensory epithelia of the mature avian inner ear.

The roles of cProx1 in mitotically quiescent and dividing supporting cells during postembryonic hair cell production are not clear. Studies in other tissues and species demonstrate cProx1's homologues promote terminal mitosis. In the fruit fly CNS, neuroblasts divide to give rise to a ganglion mother cell (GMC), which is a progenitor for neurons and glia, and another neuroblast. After this division, Pros protein is elevated in the nucleus of the GMC, which becomes terminally mitotic, but is absent from the nucleus of the neuroblast, which continues to divide. In the GMC, Pros suppresses transcription of positive regulators of the cell cycle and thereby diminishes the cell's capacity for further mitotic activity (Li and Vaessin, 2000; Myster and Duronio, 2000). Examination of Prox1 function in the murine lens suggests that it regulates mitotic activity there as well. Lens progenitors in Prox1 knockout mice show delayed expression of negative regulators of the cell cycle, such as p27kip1, and undergo terminal mitosis at later-than-normal times (Wigle et al., 1999).

This knowledge leads to the hypothesis that nuclear cProx1 immunoreactivity in the mature auditory and vestibular epithelia delineates progenitor cells with different capacities for cell proliferation. Supporting cells with a low level of cProx1 may act like stem cells, having a relatively unlimited capacity for cell division, while supporting cells with a high level of cProx1, like GMCs in fruit flies, may be on the verge of undergoing terminal mitosis. Accordingly, proliferative supporting cells in the normal utricle, which express very low levels of or no cProx1 in S phase, would have the capacity to divide continuously. This interpretation is compatible with the finding that hair cells are regenerated on a continual basis in vestibular epithelia and that each supporting cell division leads to its self-renewal (Jørgensen and Mathiesen, 1988; Roberson et al., 1992; Kil et al., 1997; Stone and Rubel, 1999; Stone et al., 1999).

It is notable that cProx1 is also expressed in sensory and neural progenitors and precursors during development of the avian inner ear (Stone et al., 2003). Nuclear cProx1 immunoreactivity is elevated in the sensory primordial patches around the time when these regions are first detectable in the otocyst. Levels remain elevated in all progenitor cells during periods of intense proliferation. Some time after hair cells and supporting cells have differentiated many of their mature features, nuclear cProx1 immunoreactivity is largely down-regulated. Thus, hair cell regeneration clearly involves re-expression of a developmentally relevant gene. However, the temporal pattern of cProx1 expression in the sensory primordial patches does not support a function for the protein in directing terminal mitosis during development.

cProx1 and Regenerated Cells

As described above, dividing progenitor cells in the damaged basilar papilla and in the control utricle display differing levels of nuclear cProx1 immunoreactivity, and cProx1 protein appears to be inherited symmetrically. Consistent with these findings, pairs of sibling cells show a range of levels of cProx1 immunolabeling shortly after mitosis. In the regenerating basilar papilla, where dividing cells are either strongly cProx1-positive or cProx1-negative, very young sibling pairs (at 8 hr post-BrdU) display either symmetrically high or symmetrically low levels of nuclear cProx1 immunoreactivity. By contrast, in the regenerating utricle, where all dividing cells are cProx1-negative, all young sibling pairs at 8 hr post-BrdU are symmetrically cProx1-negative. Sibling pairs with asymmetric levels of cProx1 protein emerge in both tissues by 17 hr post-BrdU.

By 3 days post-BrdU in the basilar papilla, there is near-equal likelihood that differentiating siblings display symmetrically high, symmetrically low, or asymmetric levels of cProx1 immunoreactivity. This distribution is reflective of the near-equal probability in this tissue that any given supporting cell division will lead to the production of two hair cells, two supporting cells, or one of each cell type (Stone and Rubel, 2000). By contrast, in the undamaged utricle, most (76%) sibling pairs show asymmetric levels of cProx1 immunoreactivity at 3 days post-BrdU, reflecting that the majority (∼70%) of supporting cell divisions lead to the formation of a hair cell and a supporting cell in this tissue (Stone and Rubel, 1999; Stone et al., 1999). Thus, in auditory and vestibular epithelia, the distribution of cProx1 labeling across sibling pairs at 3 days post-BrdU mirrors the manner in which they differentiate (i.e., symmetrically vs. asymmetrically), but earlier patterns do not. The relative level of nuclear cProx1 in a postmitotic cell correlates with its fate; cells with a high degree of nuclear cProx1 immunoreactivity differentiate as hair cells, while cells with low or undetectable levels become supporting cells. Based on our data, we suspect that cProx1 protein moves from the nucleus to the cytoplasm in most regenerated hair cells between 8 and 14 days postgentamicin, which would correspond to between 5 and 11 days after they are generated.

In fruit flies, Pros is also expressed asymmetrically in sibling cells that acquire distinct fates (Hirata et al., 1995; Knoblich et al., 1995; Spana and Doe, 1995). This asymmetry is necessary to produce distinct cell types and/or to direct their differentiation (Chu-Lagraff et al., 1991; Doe et al., 1991; Vaessin et al., 1991; Matsuzaki et al., 1992; Manning and Doe, 1999; Reddy and Rodrigues, 1999). The mechanisms used to establish asymmetric Pros expression between sibling cells vary developmentally. For example, external sensory (ES) organs are composed of four different cells whose common progenitor is the sensory organ progenitor cell (SOP). ES cells are generated in adult flies as well as during embryogenesis. In both cases, the SOP divides and forms two daughter cells, the SOPIIa and the SOPIIb. The IIa cell divides and generates a hair cell and a socket cell, and the IIb cell divides to form a neuron and a sheath cell (but see Gho et al., 1999). In embryonic flies, each SOPII cell inherits different levels of Pros protein and/or mRNA during mitosis from the SOP cell (Knoblich et al., 1995; Spana and Doe, 1995). By contrast, in adult flies, Pros protein is not detected in the adult SOP. Instead, asymmetric levels of the protein are established in the two SOPII cells shortly after mitosis (Manning and Doe, 1999; Reddy and Rodrigues, 1999; but see Gho et al., 1999). Our results here demonstrate a similarity in Pros/Prox expression patterns between cell formation in the mature fly external sense organ and in the mature chicken hearing and balance organs. Previous studies have acknowledged the similarities in structure, function, and developmental regulation of sense-organ development across fruit flies and vertebrates (Adam et al., 1998; Eddison et al., 2000; Fritzsch et al., 2000; Hassan and Bellen, 2000). While postmitotically established asymmetry in Pros activity is critical for proper cell fate acquisition in the external sense organs of mature flies (Manning and Doe, 1999; Reddy and Rodrigues, 1999), it remains to be tested if this is also the case for cProx1 in the inner ear of mature chickens or other vertebrates.

What might be the role of cProx1 during hair cell differentiation? Because Prox1 knockout mice die before hair cells are born (Wigle et al., 1999), it has not been possible to characterize the gene's role in sensory development in the inner ear. Our finding that cProx1 immunoreactivity is elevated in hair cells but absent from supporting cells suggests the transcription factor promotes acquisition of the hair cell fate and/or directs the differentiation of committed hair cells. This proposed function is supported by studies that show Prospero and Prox1 control cellular differentiation in other species (see Introduction section). Variation in cProx1 expression among mitotic supporting cells may endow daughter cells with varying differentiative potentials (Fig. 9). Because cProx1 protein is not segregated asymmetrically in mitotic supporting cells, both sibling cells generated by a supporting cell with a relatively high level of cProx1 will inherit substantial protein. By contrast, mitotic supporting cells with low levels of cProx1 will generate two sibling cells with even lower cProx1 levels. Assuming cProx1 functions to drive specification or differentiation of hair cells, different levels of cProx1 in mitotic supporting cells would endow them with an initial bias toward generating two cell types of similar fate, with the fate dependent upon the relative level. The very low expression of cProx1 in utricular progenitors would favor the formation of sibling pairs in which both cells differentiate as supporting cells. This mechanism could confer some assurance that hair cells are not easily overproduced in the normal utricle, which has a nearly full contingent of hair cells at any given time. By contrast, in the drug-damaged basilar papilla in which there is complete hair cell loss in the area of regeneration and progenitors are either cProx1-positive or -negative, this mechanism would confer an equal likelihood of forming daughter cell pairs composed of two hair cells or two supporting cells.

thumbnail image

Figure 9. Summary of cProx1 immunoreactivity during hair cell regeneration. This model depicts four scenarios that interpret our observations on cProx1 immunoreactivity during avian hair cell regeneration. A,B: The model contends that mitotically quiescent supporting cells (SCs, or progenitor cells) with weakly cProx1-positive nuclei (light gray) become highly cProx1-positive (dark gray) while dividing. A: Some of these cells generate two siblings, both of which have cProx1-positive nuclei and differentiate as hair cells (HCs, stippled). B: Additional cProx1-positive SCs form siblings that differentiate asymmetrically, as an HC and an SC. In these pairs, nuclear cProx1 immunoreactivity is down-regulated in the cell that acquires the SC fate. C,D: Quiescent SCs that are cProx1-negative remain so while dividing. D: Some of these cells generate two siblings, both of which have cProx1-negative nuclei and differentiate into SCs. C: Additional cProx1-negative SCs form siblings that differentiate asymmetrically. In these pairs, immunoreactivity for nuclear cProx1 is up-regulated in the cell that becomes an HC. Dividing SCs that are cProx1-positive are more likely to generate HCs than are cProx1-negative SCs. A–D: In the drug-damaged basilar papilla, all mechanisms of cell production shown here may be used. C: In the undamaged utricle; however, the mechanism used in the majority of cell divisions is shown.

Download figure to PowerPoint

If cell fate outcomes during avian hair cell regeneration were to depend solely upon cProx1 levels in progenitor cells, there would be no evidence of asymmetric differentiation between sibling cells. In fact, 33% of sibling pairs generated in the basilar papilla at 3 days postgentamicin differentiate as a hair cell and a supporting cell (Stone and Rubel, 2000b), and approximately 70% of sibling pairs generated in the undamaged utricle do so as well (Stone and Rubel, 1999; Stone et al., 1999). In our model, therefore, asymmetric differentiation must result from additional signals that modify gene expression in postmitotic sibling cells, despite the early bias conferred upon them by maternal cProx1 expression. These signals may be derived from the cellular environment, which as discussed above is quite different between the drug-damaged basilar papilla and the undamaged utricle.

In fruit flies, two extracellular signaling pathways are believed to influence Pros transcription: one by means of the tyrosine kinase receptors sev and DER, and one by means of the Notch receptor (Manning and Doe, 1999; Reddy and Rodrigues, 1999; Xu et al., 2000). In mice, signaling through FGF receptors appears to be required for Prox1 expression in the CNS (Govindarajan and Overbeek, 2001). Multiple studies have implicated tyrosine kinase receptors or Notch receptors in the regulation of hair cell production in vertebrates, either during embryogenesis or regeneration (reviewed in Oesterle and Hume, 1999; Eddison et al., 2000; Stone and Rubel, 2000a; Fekete and Wu, 2002; Bermingham-McDonogh and Rubel, 2003), but it is not known if activation of these receptors modulates Prox1 transcription in vertebrates.

It is also possible that sensory epithelial progenitors throughout the mature avian inner ear are inherently different from one another, in that some are biased toward generating unlike cell types. Some supporting cells, such as those in the utricle and some in the basilar papilla, may contain cell fate determinants that are asymmetrically distributed in their cytoplasm before mitosis and passed on differentially to daughter cells, thereby ensuring that cells differentiate asymmetrically (reviewed in Jan and Jan, 1998). One such protein, Numb, has been shown recently to be asymmetrically distributed in progenitor cells in the developing otocyst (Eddison et al., 2000) and regenerating basilar papilla (J.S. Stone, unpublished observations), but its function in hair cell specification remains to be proven.

It is important to acknowledge that, in this study, we have only examined immunoreactivity for cProx1 protein and not protein levels. Therefore, it is not clear at this point whether changes in immunoreactivity reflect different levels of the protein or different degrees of its antigenicity. Changes in cProx1 immunoreactivity associated with cellular proliferation or differentiation have been described in the developing vertebrate lens (Duncan et al., 2002), hippocampus (Pleasure et al., 2000), and retina (Dyer et al., 2003), but how these changes are controlled has not been explored. Changes in protein levels can be achieved by altering protein transcription and translation, degradation, or subcellular localization, all of which have been postulated to regulate Pros protein levels in fruit flies (Hirata et al., 1995; Ikeshima-Kataoka et al., 1997; Li et al., 1997; Shen et al., 1997; Schuldt et al., 1998; Srinivasan et al., 1998; Manning and Doe, 1999; Reddy and Rodrigues, 1999; Ceron et al., 2001; Demidenko et al., 2001). Changes in antigenicity can reflect posttranslational changes in the protein structure (e.g., phosphorylation) or masking of the antigen site by binding of other molecules to the transcription factor. In fact, there are several phosphoisoforms of Pros (Srinivasan et al., 1998), and domains of Pros and Prox1 have been shown to bind to other proteins, such as Miranda and Pax6, respectively (e.g., Shen et al., 1998, Mikkola et al., 2001). The cellular mechanisms leading to the postmitotic changes in nuclear cProx1 immunoreactivity in the chicken's inner ear sensory epithelia require further exploration.

EXPERIMENTAL PROCEDURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

Animal Care

White Leghorn chicken eggs were purchased from Hyline International (Graham, WA) and maintained in a heated and humidified incubator until hatching. Hatchlings were housed in heated brooders with freely available food and water. All procedures for animal use were approved by the University of Washington's Animal Care Committee and complied with guidelines of the National Institutes of Health.

Gentamicin Injections

Chicks between posthatch day 7 and 10 received a single subcutaneous injection of gentamicin (400 mg/kg [Fugisawa] or 600 mg/kg [Elkins Sinn]) to induce hair cell death in the inner ear epithelia (Stone et al., 1996). These high doses of gentamicin cause illness in some birds for several hours after the injection, and they are lethal in 0 to 25% of the birds. After the injection, chicks were returned to the brooder, where they survived for different periods before killing.

BrdU Injections

The nucleotide analog BrdU was administered to chicks to identify proliferative and postmitotic cells in the sensory epithelia. BrdU (12 mg/ml dissolved in sterile phosphate-buffered saline [PBS]) was injected intraperitoneally (final dose, 100 mg/kg) into chicks at 3 days after gentamicin injection and into chicks that received no gentamicin treatment. Chicks were killed between 30 min and 30 days after the BrdU injection.

Tissue Preparation

Posthatch chicks were killed by overdose of sodium pentobarbital (intraperitoneal injection, 100 mg/kg). Cochlear ducts (containing the basilar papilla) and utricles were removed and fixed with 4% buffered paraformaldehyde for 20 min to 1 hr. Most organs were processed further as whole-mounts, but some organs were embedded in agar and sectioned on a Vibratome (60–100 μm).

Immunofluorescence

Free-floating whole-mounts or Vibratome sections were processed by using standard indirect immunofluorescence methods. The following primary antibodies were used: rabbit anti-Prox1 (S.T., Belecky-Adams et al., 1997), mouse anti-calmodulin (Sigma), mouse anti-class III beta tubulin (TuJ1, Anthony Frankfurter, University of Virginia), rabbit anti-myosinVI (Tama Hasson, University of California, San Diego, CA), rabbit anti-pH3 (Upstate Biotechnology), mouse anti-MPM2 (Upstate Biotechnology), and mouse anti-BrdU (Becton Dickinson). Primary antibodies were detected with secondary antibodies conjugated to Bodipy/FL (Molecular Probes), Lissamine Rhodamine, Cy3, or Cy5 (Jackson Immunoresearch Laboratories). Blocking and diluting solutions consisted of normal goat serum diluted 1/10 in 0.05% Triton X-100 in PBS. Sections were rinsed with PBS between steps. BrdU detection was conducted as per Stone and Rubel (2000b). For double-labeling and triple-labeling experiments, primary/secondary antibody pairs were added sequentially. Some tissues were counterstained with propidium iodide (Sigma), a fluorescent label for nucleic acids, to provide information about the location of nuclei. For negative controls, primary antibody solution was omitted from the reaction series. Finally, tissue (whole-mounts and sections) was mounted onto glass microscope slides in Vectashield anti-fade medium (Vector Laboratories).

Imaging and Analysis

Immunoreacted tissue was imaged by using a Bio-Rad 1024 ultraviolet confocal scanning laser microscope and Bio-Rad Lasersharp acquisition software. Images were processed by using NIH Image and Adobe Photoshop and printed on an inkjet (Epson or Hewlett Packard) printer.

Quantitative analyses were performed. We counted cProx1-labeled nuclei in the damaged region of basilar papillas at 2 days (n = 5 basilar papillas), 3 days (n = 4), 5 days (n = 5), 8 days (n = 3), 15 days (n = 7), and 30 (n = 2) days postgentamicin. For samples at 3 to 30 days postgentamicin, three areas (203 × 203 μm) within the damaged region of whole-mount organs were chosen for counts; one at the proximal pole of the papilla, one in the middle of the lesion, and one at the distal edge of the lesion. For samples at 2 days postgentamicin, we only counted nuclei in the proximal and distal edges of the lesion, because it was very small in size. For each area, we generated a Z-series scan by using a ×60 objective through the depth of the papilla at 2-μm intervals. The first slice of each Z-series was taken at the luminal surface, and the last slice was taken at the start of the basilar membrane, which was morphologically distinguishable due to background fluorescence. Using Object Image, we stepped through the slices, counting strongly labeled nuclei through the entire depth of the scan and marking off each nucleus as it was counted to avoid including the same nucleus more than once.

We also examined BrdU-labeled nuclei in DNA synthesis (S phase) in the drug-damaged basilar papilla to determine whether they were cProx1-positive or -negative. This analysis was performed on tissue from birds that were killed 30 min or 4 hr after a single BrdU injection at 3 days postgentamicin and double-immunolabeled to detect BrdU and cProx1. Z-series were generated in similar regions and in a similar manner as described above. We selected three slices (one from the top, one from the middle, and one from the bottom) from the series in which to examine the cellular labeling patterns. By using Object Image, we stepped through the slices, examining each BrdU-positive nucleus to determine whether it was cProx1-negative or -positive and marking it off to avoid analysis of the same nucleus in the subsequent section. A total of 165 BrdU-positive cells were examined in two basilar papillas.

Finally, by using a ×60 objective, we examined cProx1 labeling in BrdU-labeled sibling pairs in drug-damaged basilar papillas (whole-mounts) from birds that received a single BrdU injection at 3 days postgentamicin and were killed after 8 hr (n = 5 basilar papillas), 17 hr (n = 6), 24 hr (n = 8), or 3 days (n = 5). We performed a similar analysis on utricles from untreated birds that received a single BrdU injection at postnatal day (P) 7 and were killed at similar time-points (for 8 hr, n = 4 utricles; for 17 hr, n = 4; for 24 hr, n = 3; and for 3 days, n = 4). Additionally, we examined cProx1 and calmodulin immunoreactivity in BrdU-labeled sibling pairs in drug-damaged basilar papillas (whole-mounts; n = 3) from birds that received a single BrdU injection at 3 days postgentamicin and were killed 3 days later. For these analyses, the drug-damaged area of the basilar papilla and the entire area of the undamaged utricle were scanned, and sibling pairs were identified and scored based on their levels of immunoreactivity for the antigen(s). Two BrdU-labeled nuclei were defined as “paired” if they were located within 1 nuclear width of each other and more than three nuclear widths from other BrdU-labeled nuclei. In order for cells to be considered “BrdU-positive,” “cProx1-positive,” or “calmodulin-positive,” they needed to have substantial, above-background signal for each antigen. Examples are shown in the figures.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES

We thank Dr. Edwin Rubel for sharing facilities in which these experiments were conducted and for thoughtful discussions of the data. We thank Glen MacDonald for technical assistance with confocal microscopy and Kevin Whitham for assistance with computing. We also thank Dr. Clifford Hume for comments on the manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES