ABSTRACT
The purpose of the present study was to examine the spatio-temporal pattern of cell proliferation in the chick cochlea in response to the sensory hair cell loss induced by a 1.5 kHz pure tone at 120 dB SPL (1 dB=20 Pa) for 48 h. DNA replication was evaluated with the bro-modeoxyuridine (BrdU) pulse-fix technique. One group of birds was given multiple injections of BrdU (50 mg/kg) over a period of 8 h at various starting times during or after the exposure. Afterwards, their cochleas were removed and processed as whole mounts for BrdU immunohistochemistry. The cochleas of a second group of acoustically traumatized chicks were evaluated by scanning electron microscopy in order to determine the spatio-temporal pattern of hair cell loss. Hair cell loss was first observed 12 h after the start of the exposure and DNA replication started near the inferior edge of the hair cell lesion 24–32 h after the start of the expo-sure, i.e. 12–20 h after the first sign of hair cell loss. The site of hair cell loss and DNA replication shifted toward the superior edge of the basilar papilla as the exposure continued. The rate of DNA replication accelerated and reached its peak near the end of the 48 h exposure. The estimated latency of cell proliferation after hair cell loss was faster and the duration of DNA replication shorter than that observed in other sensory systems. The spatio-temporal pattern of DNA replication follows the spatio-temporal gradient of hair cell loss, suggesting that cell proliferation is triggered by hair cell loss itself rather than by intrinsic positional cues or gradients.
INTRODUCTION
The avian cochlea has become the focus of intense study in recent years after it was discovered that the sensory hair cells could regenerate after being destroyed by intense sound exposure or ototoxic drugs. The time course and pat-tern of hair cell replacement in the basilar papilla after acoustic trauma has been extensively described using scan-ning electron microscopy (Cotanche, 1987; Corwin and Cotanche, 1988; Girod et al., 1989; Marsh et al., 1990; Cotanche and Corwin, 1991). Physiological measures, such as the auditory evoked response and the compound action potentials, recover during the process of hair cell regener-ation, suggesting that regenerated hair cells become func-tional (Corwin and Cotanche, 1988; McFadden and Saun-ders, 1989; Salvi et al., 1991). Two possible sources of regenerated hair cells have been identified from studies in which tritiated thymidine was used as a marker of DNA replication. Possible precursors are the supporting cells (Corwin and Cotanche, 1988) or the cuboidal/hyaline cells, which lie outside the auditory epithelium (Girod et al., 1989).
In spite of the progress made in understanding the process of hair cell regeneration, there are several fundamental questions that remain to be addressed. First, at what point in time following acoustic trauma does the first phase of cell reproduction actually begin? Girod et al. (1989) demonstrated that labeling with tritiated thymidine was completely absent in the chick cochlear epithelium 6 h after an 18 h traumatizing exposure, but was present 15 h after the exposure. Similarly, Raphael (1992) first observed S-phase supporting cells 24 h after the end of a 4 h noise exposure. Since cell proliferation was not observed during the exposure, one might conclude that cell reproduction is initiated after the end of acoustic over-stimulation (Stone and Cotanche, 1992). However, Stone and Cotanche (1992) suggested that the initiation of hair cell regeneration is inde-pendent of the duration of acoustic exposure. If cell pro-liferation is triggered by loss of hair cells and/or support-ing cells, then precursor cells could conceivably enter the cell cycle even during the period when the traumatizing stimulus is on. In order to assist researchers in identifying the molecules that regulate hair cell proliferation and differ-entiation, it is important to determine the exact time of onset and the duration of cell proliferation during or after acoustic overstimulation.
Another important issue related to hair cell regeneration is whether there are any intrinsic spatial gradients or positional cues underlying the process of cell proliferation. After aminoglycoside ototoxicity, we observed that hair cell regeneration proceeds along a base-to-apex gradient in neonatal chicks (Hashino et al., 1991) and adult budgeri-gars (Hashino et al., 1992). The underlying mechanism responsible for this pattern of regeneration is unknown. It is conceivable that it is related to the spatio-temporal pat-tern of hair cell degeneration or it could arise from intrin-sic positional gradients or hot zones of cell proliferation. During normal development, several hot zones of cell pro-liferation run transversely across the epithelium (Katayama and Corwin, 1989). Clearly, it is important to determine the pattern of cell proliferation and relate it to the progression of hair cell loss.
In order to examine the temporal and spatial patterns of DNA replication in the avian inner ear, we applied the 5-bromo-2′-deoxyuridine (BrdU) pulse fix paradigm to the chick cochlea during and after a traumatizing sound expo-sure. BrdU is a thymidine analog that is incorporated into single-stranded DNA during the S-phase of the cell cycle (Gratzner, 1982) and is visualized using a monoclonal anti-body against BrdU.
MATERIALS AND METHODS
Acoustic exposure
Ten-to fifteen-day-old White Leghorn chicks were used as sub-jects. All the experimental animals, except the age-matched con-trols, were exposed to a 1.5 kHz pure tone at 120 dB SPL (sound pressure level; 1 dB=20 mPa). The traumatizing stimulus was pro-duced by a signal generator (Wavetek, model 132), a power ampli-fier (Technics, SE-9060) and a loudspeaker with exponential horn (Altec 808–8A). The loudspeaker was suspended from the ceiling of a sound booth approximately 45 cm above the floor of the cage housing the animals. The sound-pressure level was monitored at the beginning, middle and the end of the 48 h exposure using a Bruel & Kjael type 2606 measuring amplifier and 1.27 cm con-denser microphone. The sound level varied by less than ±2 dB at various locations within the cage. Most experimental animals were exposed to the traumatizing stimulus for 48 h.
BrdU immunohistochemistry
Birds received 5 intramuscular injections of BrdU (Sigma, 50 mg/kg) at intervals of 1.5 h. BrdU injections began at 24, or 40 h after the onset of the 48 h exposure, or 8, 24 or 40 h after the completion of the exposure (a minimum of 6 animals per group). Two hours after the final injection of BrdU (injection period: 8 h), the birds were given an overdose of sodium pentobarbital, decapitated and their temporal bones removed and fixed in 3.5% paraformaldehyde for 4 h at 4°C. The group of animals that received the first injection of BrdU 24 h after the start of the expo-sure was killed 8 h following the initial injection. Therefore, the total duration of the acoustic exposure for this group was 32 h. Age-matched control animals were injected with the same amount of BrdU as experimental animals.
The cochlear duct of experimental and control animals was dis-sected out of the temporal bone as a whole mount in phosphate buffered saline (PBS). To inactivate endogenous peroxidase, the specimen was immersed in 90% methanol containing 0.3% H2O2 for 20 min, denatured by immersion in 2 N HCl for 30 min, and rinsed with 0.5% bovine serum albumin (BSA) in PBS. Non-specific binding was inhibited by immersing the specimens for 20 min in a blocking solution containing 2% normal horse serum, 2% BSA and 5% nonfat dry milk. Subsequently, the specimens were incubated overnight in a 1:50 dilution of mouse anti-BrdU monoclonal antibody (Becton Dickinson, Mt. View, CA), 2% normal horse serum and 1% Triton X-100 in PBS. After washing in BSA solution, specimens were treated with biotinylated horse anti-mouse IgG (Vector Labs., Burlingame, CA) for 1 h, followed by a complex of avidin-biotinylated horseradish peroxidase (HRP; Vector Labs, Elite ABC Kit) for 45 min. Specimens were immersed for 4 min in a 0.2% solution of 3,3′-diaminobenzidine in sodium citrate buffer (pH 5.1) with 0.01% H2O2. After stain-ing, the tegmentum vasculosum, tectorial membrane and cochlear ganglion were dissected away. Whole mounts of the cochlear epithelia were mounted in glycerol and viewed on a Zeiss Axiovert light microscope equipped with DIC attachment. To preserve the integrity of the cochlear epithelium, the specimens were viewed without a coverglass except for observations with the ×40 objec-tive. Specimens were photographed (Kodak Technical Pan, ASA 100), and a photographic montage was constructed from several photographs taken at different focal planes. The total number of BrdU-labeled cells in the epithelium was counted from the pho-tographic montage. Counts were obtained from 5 cochleas at each survival time. Some of the cochleas were embedded in JB-4 Plus (Polysciences, Warrington, PA) and transversely sectioned at a thickness of 2 μm. Sections were lightly counter-stained with tolu-idine blue.
Scanning electron microscopy
The spatio-temporal pattern of hair cell loss was evaluated by scanning electron microscopy according to the procedure described previously (Hashino et al., 1991, 1992). Briefly, 12 ani-mals were exposed to the traumatizing tone for 12, 24 or 48 h. Immediately after the exposure, the animals were given an over-dose of sodium pentobarbital, decapitated and their cochleas removed and fixed with 1% osmium tetroxide for 45 min. The fixed tissues were dehydrated through a graded series of ethanol up to 70% and then dissected out to expose the luminal surface of the basilar papilla. The specimens were dehydrated up to 100% ethanol, which was replaced with tert-butyl alcohol, and subse-quently freeze-dried. Afterwards, the tissue was mounted on an aluminum stub, sputter coated with gold/palladium and viewed on HITACHI S-800 scanning electron microscope at an accelerating voltage of 15 kV.
RESULTS
Progression of hair cell damage by acoustic over-stimulation
Fig. 1a-d shows a series of scanning electron micrographs of the chick basilar papilla from animals that had been exposed for 12, 24 or 48 h along with a micrograph from an unexposed control animal. In the control animal (Fig. 1a), hair cells are packed in a mosaic pattern over the entire basilar papilla as described previously (Hashino et al., 1991). Supporting cells, which are covered by microvilli, completely surround each of the hair cells, resulting in a distinct boundary between each hair cell on the basilar papilla (Fig. 1a, inset). After 12 h of exposure to the 1.5 kHz tone, hair cells with damaged stereocilia as well as a few missing hair cells were observed in a crescent-shaped patch along the inferior (abneural) edge of the papilla (Fig. 1b). After 24 h of exposure, hair cell loss was evident in all of the cochleas examined. The region of the papilla located 20 to 40% of the distance from the basal end of the cochlea was characterized by a crescent-shaped depression along the inferior edge of the papilla that was devoid of hair cells (Fig. 1c). The location of the hair cell lesion is consistent with previous results using the same exposure (Cotanche and Dopyera, 1990; Stone and Cotanche, 1992). After 48 h of exposure, the region of hair cell loss had expanded toward the superior (neural) edge of the papilla as well as toward the apical and basal end of the papilla compared to the lesions seen at shorter exposures (Fig. 1d). In the damaged region, which was located 15 to 50% of the distance from the base, only a small number of hair cells were observed. The other cells within the lesion, pre-sumably supporting cells that had expanded into the spaces vacated by the hair cells, were covered by numerous microvilli.
DNA replication during acoustic over-stimulation
The cochleas of the control animals that were injected with BrdU showed no BrdU labeling (Fig. 2a). This observation confirms the results of previous studies showing an absence of DNA labeling in control animals using either [3H]thymi-dine (Corwin and Cotanche, 1988; Girod et al., 1989) or BrdU immunohistochemistry (Raphael, 1992) labeling tech-niques. We conclude that natural turnover of hair cells and supporting cells is absent or very rare in the post-hatched chick cochlea.
A small number of BrdU-labeled cells was first observed in the cochleas of chicks injected with BrdU 24–32 h after onset of the exposure (Fig. 2b). Out of 15 cochleas exam-ined during this period, labeling was present in 11 cochleas. The average number of labeled cells was approximately 50 (Fig. 3). In preliminary experiments, we immunostained the cochleas of chicks that had received BrdU 18–24 h after the start of exposure and found no labeling in any of these cochleas (data not shown). However, a crescent-shaped region of hair cell damage was clearly distinguishable by light microscopic observation of the whole mounts. Taken together, these results suggest that DNA replication begins 24–32 h after the start of the exposure.
The distribution of BrdU-stained cells was compared with the location of hair cell loss, which, in scanning elec-tron micrograph (Fig. 1c) and also in the light micrograph (Fig. 2b), consisted of a depressed region of the basilar papilla. BrdU-labeled cells were primarily located near the inferior edge of the damaged region (Fig. 2b). No labeling was observed outside the damaged region of the sensory epithelium. However, labeled cells were present in the non-epithelial region beyond the inferior edge of the sensory epithelium. Transverse sections of the basilar papilla revealed that these labeled cells, presumably fibroblasts, were located below the basilar membrane (data not shown). No labeling was seen in the hyaline cell region.
In the cochleas of chicks injected with BrdU 40–48 h after the onset of exposure, the damaged region was filled with labeled cells (Fig. 2c). The DNA replication rate was upregulated by ninefold compared to that in the preceding period. As shown in Fig. 3, the average number of labeled cells appeared to reach its peak 40–48 h after the onset of exposure. Because BrdU labeling was extremely heavy for the 40–48 h and 56–64 h time periods, the counts for these two periods are probably underestimated, since a labeled cell near the surface of the epithelium could mask other cells beneath it. Fig. 4a,b shows a higher magnification view of a whole mount of the basilar papilla at different focal planes. Note that the normal mosaic formed by the apical surfaces of the hair cells is absent in a crescent-shaped depression in the papilla. The pair of photomicrographs demonstrates the relationship between the boundary of the hair cell lesion and the location of BrdU-labeled cells. Most of the labeled cells were located inside the edge of the cres-cent-shaped depression in the papilla. Cross-sections taken through a corresponding region of the papilla showed that the labeled cells were aligned as a monolayer on the basi-lar membrane (Fig. 4c). Throughout the period of acoustic exposure, labeled fibroblasts were much more frequently observed in the damaged region (Fig. 4c, arrow) than in the intact region.
DNA replication after acoustic over-stimulation
In the chicks injected with BrdU 56–64 h after the onset of exposure, we observed a band of labeled cells running along the superior edge of the damaged region (Fig. 2d). The depression in the papilla, which was present earlier, had disappeared, presumably due to the influx or expansion of cells into this region during the recovery process. There-fore, it was difficult to judge the precise position of the boundary between damaged and intact regions from the sur-Hours after the Onset of Exposure face preparation. Transverse sections taken from this region were used to determine the precise location of labeled cells within the epithelium. Fig. 5a shows a cluster of BrdU-labeled cells located just above the auditory nerve fibers. A few hair cells that survived the exposure were present slightly inferior to the BrdU-labeled cells (between broken lines). The inferior half of the papilla, which was severely damaged by the exposure, was composed of a single layer of unlabeled cells, which presumably had regenerated prior to the BrdU injections.
Higher-magnification analysis provides more detailed information on the status of BrdU-labeled cells in the cell cycle. Several labeled cells in each cochlea show mitotic figures characteristic of meta-or anaphase (Fig. 5b,c). The mitotic chromosomes were generally located in a focal plane above the one containing labeled, but non-mitotic, nuclei. However, few of the labeled cells appeared in pairs as described by Raphael (1992). While BrdU-labeled cells were frequently observed along the superior edge of the papilla, the total number of labeled cells was significantly less than in the preceding time period (t=3.66, P<0.01).
The cochleas of animals injected with BrdU 72-80 h after the onset of exposure (24-32 h post-exposure) contained relatively few labeled cells. The labeled cells were scattered throughout the inferior half of the basilar papilla (Fig. 2e). The number of labeled cells observed during the 72-80 h injection period was significantly less than in the 56–64 h injection period (Fig. 3; t=6.85, P<0.01). The cochleas of the animals injected with BrdU 88–96 h after the onset of exposure (40-48 h post exposure) contained only a small number of labeled cells in each cochlea (Fig. 2f). The cochleas of animals injected with BrdU at 2–4d post-expo-sure contained even fewer labeled cells than in the previ-ous injection period (data not shown).
DISCUSSION
We do not know how the concentration of BrdU in the chick cochlea changes over time. However, it is known that [3H]thymidine, a BrdU analog, is present in the blood serum of the canary as early as 30 min following an intramuscu-lar injection and that its labeling potential is maintained for approximately 90 min after the injection (Alvarez-Buylla et al., 1990). On the basis of these data, we would assume that our BrdU injections, which were given once every 1.5 h, would label the majority of the cells that enter the S-phase of the cell cycle during our 8 h injection period. Since the same injection schedule was used for all groups, and since a given percentage of S-phase cells are expected to incorporate BrdU, we believe that it is possible to compare the number of BrdU-labeled cells at different times during and after the noise exposure.
Temporal patterns of DNA replication
DNA replication started 24–32 h after the onset of the 48 h exposure and rapidly accelerated. The number of BrdU-labeled cells reached its peak near the end of the 48 h expo-sure and then gradually declined after the exposure (Fig. 6). Few proliferating cells were observed 48–96 h post-exposure. Since the peak of DNA replication occurred during the exposure, we conclude that the process of cell proliferation is not suppressed by the presence of the trau-matizing stimulus. These results are consistent with our ear-lier studies showing that immature hair cells are already present in the avian cochlea at the end of kanamycin treat-ment (Hashino et al., 1991, 1992). We do not know if regen-erating hair cells are as susceptible to acoustic trauma as mature hair cells; however, one might expect them to be less susceptible because the mechanical shearing forces applied to the short, immature stereocilia on regenerating hair cells (which are presumably unattached to the tector-ial membrane) would be expected to be less than that applied to the tall stereocilia on mature hair cells, which are attached to the tectorial membrane. This interpretation is consistent with recent reports showing that prior noise damage to the tectorial membrane and hair cells protects the ear from subsequent noise exposure (Raphael, 1991). To clarify this issue further, it would be useful to employ the BrdU pulse-chase technique in conjunction with a longer period of acoustic exposure.
Since hair cells and supporting cells were not immediately destroyed at the onset of the exposure, it is difficult to determine precisely the time between the onset of hair cell loss and the onset of cell proliferation However, the scanning electron micrographs of the basilar papilla that were obtained at various times during the exposure can be used to estimate the onset of hair cell loss, and the loca-tion and nature of the damage. Hair cell loss was evident near the inferior edge of the basilar papilla as early as 12 h after the onset of exposure (Fig. 1b). These results are consistent with those of Cotanche and Dopyera (1990), who showed that hair cells were ejected from the sensory epithe-lia 12 h after the onset of an exposure similar to the one used in this study. If hair cell loss is the trigger signal for cell proliferation, then it must take another 12–20 h before the precursor cells reach the S-phase of the cell cycle. The 12–20 h latency observed in this study is consistent with the latency of 8–24 h obtained from in vitro observations of the basilar papilla in which the onset of hair cell destruc-tion was precisely determined using a laser ablation tech-nique (Warchol and Corwin, 1992).
Only a small number of mitotic chromosomes were iden-tified in the BrdU-labeled nuclei (Fig. 5b,c). In addition, there was little evidence for distinct pairs of labeled nuclei, which would indicate the presence of daughter cells after cell division (Rouse and Pickles, 1991; Raphael, 1992). These results suggest that the mitotic figures presumably originated from nuclei that had incorporated BrdU during the first injection in the series of 5, which occurred over an 8 h period prior to tissue fixation. Because pairs of BrdU-labeled nuclei were rare after the 8 h injection period, we suggest that the total duration of S+G2-phases is around 8 h. Obviously single BrdU-injection studies are necessary to determine the precise length of each phase in the cell cycle. The present results indicate that the mitotic chromosome characteristics of metaphase and anaphase are as clear with BrdU immunohistochemistry as with chromatin stains such as propidium iodide (Knapp, 1992). This suggests that BrdU immunohistochemistry might be used to determine the stages within the cell cycle.
The time course and anatomical features associated with hair cell regeneration following acoustic over-stimulation have been extensively described using scanning electron microscopy (Cotanche, 1987; Corwin and Cotanche, 1988; Girod et al., 1989; Marsh et al., 1990). Primitive hair cells, which are characterized by a small apical surface covered by numerous microvilli, have been identified as early as 48 h after the cessation of a 48 h pure tone exposure at 120 dB SPL. The number of regenerated hair cells continued to increase for the next 8 days (Cotanche, 1987). Given that the S-phase of DNA replication occurred 24–32 h after the onset of our 48 h exposure and that primitive hair cells are observed 48 h post-exposure, we suggest that it takes approximately 72-80 h for S-phase cells to divide and dif-ferentiate into hair cells. While the results from scanning electron microscopy suggest that cell proliferation might continue for 8 days, our results indicate that most of the DNA replication takes place over a period of 2 days. There are two plausible explanations for this discrepancy. First, we did not evaluate BrdU labeling beyond 4 days post-exposure. Therefore, we cannot exclude the possibility of a second mitotic wave that could account for hair cell pro-liferation between 4 and 10 days post-exposure. Second, it is conceivable that cell proliferation takes place for only 2 days, but that the period of cell migration and differen-tiation is more protracted than the period of DNA replica-tion.
Several other investigators have examined the time course of cell proliferation in the chick cochlea in response to acoustic trauma. Girod et al. (1989) used a prolonged [3H]thymidine pulse-labeling procedure in which animals were repeatedly injected with thymidine during and after an 18 h exposure and killed at various survival times. Labeled cells were first observed in the basilar papilla 15 h after the 18 h exposure (33 h after the onset of the expo-sure). The latency for onset of [3H]thymidine labeling from the onset of the exposure is comparable to our results (24-32 h). However, since the duration of [3H]thymidine admin-istration was varied between groups, the labeled cells from longer-survival groups are mixed in with those that had incorporated BrdU at various times up to 30 days ago. Therefore, one cannot directly determine the time at which cell proliferation ceases at different places within the basi-lar papilla. Raphael (1992) observed BrdU-labeled cells in the chick cochlear epithelium as early as 24 h after a 4 h exposure to a 120 dB SPL octave band of noise centered at 1.5 kHz. In this case, animals were given only a single injection of BrdU 10 min before the onset of the acoustic exposure, and killed at various times after exposure. The labeling potential of a single intramuscular injection of BrdU is presumably of the order of 90 min in avians, based on [3H]thymidine studies (Alvarez-Buylla et al., 1990). In addition, both [3H]thymidine and BrdU are cleared from the blood after an intravenous injection at approximately the same rate and time (Kriss and Revesz, 1962; Nowakowski and Rakic, 1974). We would assume that only the cells that had undergone S-phase during this time period would incorporate BrdU and therefore only these and their progeny would carry the potential for BrdU labeling. In ani-mals killed after long survival times, most BrdU-labeled cells were arranged in pairs (Raphael, 1992), implying that a subpopulation of cells had incorporated BrdU around the time of exposure and subsequently divided at a later stage. Raphael’s results, however, raise some questions about the onset of cell proliferation. If BrdU can be incorporated into DNA only during the first 90 min after injection (Alvarez-Buylla et al., 1990), as suggested by [3H]thymidine studies, then hair cell loss and the entry of cells into the S-phase of the cell cycle must begin near the onset of the exposure. Since we found no evidence of hair cell loss until 12 h of exposure and since it probably takes several hours for cells to reach the S-phase, we suggest that the potential for BrdU labeling must be maintained for at least 12 h after the injec-tion of BrdU.
The time course of cell reproduction in the chick inner ear is remarkably fast compared to that in other sensory organs. Hitchcock et al. (1992) demonstrated clusters of dividing cells in the neural retina of the goldfish 7 days after the trans-scleral lesion, which was followed by com-plete regeneration by 8 weeks. Similarly, Raymond (1991) found BrdU-labeled cells two weeks after destroying the retina with injections of ouabain, which blocks Na,K-ATPase. In the olfactory epithelium, thymidine pulse studies showed the peak of cell proliferation occurred 4 days after the olfactory nerves were sectioned in mice (Camara and Harding, 1984) or 12 days after bulbectomy was performed in rats (Carr and Farbman, 1992). Although the methods for destroying sensory receptor cells in these studies were different from the method employed by us, we do not believe that these methodological factors can account for the large difference in latency or the time course of cell proliferation between the cochlear epithelium and other sen-sory organs. One factor that could conceivably contribute to the latency and rate of cell proliferation is the metabolic rate, which is much higher in birds than in mammals or amphibians.
Spatial patterns of DNA replication
The first site of cell proliferation occurred in the inferior portion of the crescent-shaped lesion within the sensory epithelium (Fig. 2b). BrdU-labeled cells were subsequently observed in more superior locations of the papilla, result-ing in almost complete labeling of the crescent-shaped lesion (Fig. 2c). At later stages (Fig. 2f), DNA replication was observed only near the superior border of the lesion. The general patterns of DNA replication observed in the present study is consistent with the progression of hair cell loss from the inferior edge of the papilla toward the supe-rior edge as the duration of exposure increases (Cotanche and Dopyera, 1990). The spatio-temporal pattern of cell
proliferation following acoustic overstimulation differs from the patterns of hair cell proliferation observed during the normal embryological development of the chick cochlea (Katayama and Corwin, 1989). During embryonic devel-opment, several hot zones of cell proliferation run trans-versely across the basilar papilla; these results suggest that there are so-called cellular genic foci in the chick embry-onic cochlea. However, post-embryonic cell proliferation due to acoustic trauma appears to be guided by the pattern of hair cell loss rather than by intrinsic positional cues and/or hot zones of cell proliferation. Moreover, support-ing cells in the immediate region of hair cell loss seem to be the primary precursors of the regenerated hair cells and supporting cells.
In some of the cochleas, BrdU-labeled cells were observed outside of the sensory epithelium. From cross-sec-tions, the labeled cells were characterized as fibroblasts, based on their morphological features and location beneath the sensory epithelium. Labeled cuboidal/hyaline cells in non-sensory areas adjacent to the basilar papilla have also been reported (Girod et al., 1989; Warchol and Corwin, 1992; Raphael, 1992). Girod et al. (1989) suggested that the neuroepithelial cells were the primary precursors for regenerated hair cells. However, the number of labeled neu-roepithelial cells was relatively small and they were present prior to the most active period of DNA replication. There-fore, cells outside of the sensory epithelium can prolifer-ate, but may not be a major source of regenerated hair cells. It is conceivable that cells outside the sensory epithelium are destroyed by acoustic exposure before hair cells and supporting cells in the epithelium and that surviving, non-epithelial cells proliferate to replace lost cells.
We observed BrdU-labeled cells just above the nerve fiber bundle after the peak of DNA replication (Fig. 5a). A thin strip of surviving hair cells separated these cells from a monolayer of cells located near the inferior edge of the papilla. It is conceivable that some interaction occurs between the precursor cells and auditory nerve fibers; for example, a diffusible chemical or contact of neural processes that stimulate the precursor cells to enter the cell cycle. These nerve-derived signals may be delayed com-pared to the trigger signals mediated by direct cell-cell com-munication within the epithelium. In post-embryonic devel-opment of Drosophila, the selective approach of axons from photoreceptor neurons stimulate G1-phase neural precursor cells to enter S-phase of the cell cycle (Selleck et al., 1992). Possible roles for growth factors in the up-regulation of cell production were suggested, in the olfactory epithelium (Carr and Farbman, 1992) or in the neural retina (Park and Hollenberg, 1989). In the present study, BrdU-labeled fibroblasts were observed primarily in the region of hair cell loss (Fig. 4c). Increased [3H]thymidine labeling in fibroblasts associated with hair cell regeneration has also been reported before (Girod et al., 1989; Ryals and West-brook, 1990). Although cell proliferation in connective tissue was interpreted as due to a high basal turnover rate (Girod et al., 1989), it could be due to the influence of sol-uble trophic factors.
ACKNOWLEDGEMENTS
We thank Drs Christopher Cohan, Masahiro Sokabe and Mr Erik TinHan for their comments on the manuscript, and Ms Karen Miller for technical assistance. This study was supported by NSF(BNS9007822), NIH/NIDCD(DC01685) to R.J.S. and the Sound Technology Promotion Foundation to E.H.