|Home | About | Journals | Submit | Contact Us | Français|
Cochlear hair cells (HCs) are mechanosensory receptors that transduce sound into electrical signals. HC damage in non-mammalian vertebrates induces surrounding supporting cells (SCs) to divide, transdifferentiate and replace lost HCs; however, such spontaneous HC regeneration does not occur in the mammalian cochlea. Here, we acutely ablate the retinoblastoma protein (Rb), a crucial cell cycle regulator, in two subtypes of postmitotic SCs (pillar and Deiters’ cells) using an inducible Cre line, Prox1-CreERT2. Inactivation of Rb in these SCs results in cell cycle reentry of both pillar and Deiters’ cells, and completion of cell division with an increase in cell number of pillar cells. Interestingly, nuclei of Rb−/− mitotic pillar and Deiters’ cells migrate toward the HC layer and divide near the epithelial surface in a manner similar to the SCs in the regenerating avian auditory epithelium. In contrast to postmitotic Rb−/− HCs which abort cell division, postmitotic Rb−/− pillar cells can proliferate, maintain their SC fate and survive for more than a week. However, no newly formed HCs are detected and SC death followed by HC loss occurs. Our studies accomplish a crucial step toward functional HC regeneration in the mammalian cochlea in vivo, demonstrating the critical role of Rb in maintaining quiescence of postmitotic pillar and Deiters’ cells and highlighting the heterogeneity between these two cell types. Therefore, the combination of transient Rb inactivation and further manipulation of transcription factors (i.e., Atoh1 activation) in SCs may represent an effective therapeutic avenue for HC regeneration in the mammalian cochlea.
Damaged mammalian cochlear hair cells (HCs) cannot regenerate spontaneously; therefore, their loss causes permanent hearing deficits (Brigande and Heller, 2009). In non-mammalian vertebrates, replacement of auditory and vestibular HCs occurs even after sensory epithelia and hearing function have matured (Cotanche, 1987; Corwin and Cotanche, 1988; Ryals and Rubel, 1988) There are two proposed mechanisms of HC regeneration in non-mammalian vertebrates: direct transdifferentiation, where supporting cells (SCs), which lie underneath HCs, change cell fate to become HCs, and mitotic regeneration, where SCs divide and the daughter cells of that division further differentiate to replace both damaged HCs and previously differentiated SCs (Raphael, 1992; Stone and Cotanche, 2007). In the chicken auditory epithelium, when HC loss is severe, many of the regenerated HCs are formed from proliferating SCs (Roberson et al., 2004; Stone and Cotanche, 2007). In the zebrafish mechanosensory system, defects in SC proliferation prevent HC regeneration (Behra et al., 2009). Therefore, in non-mammalian vertebrates, SCs are the origin of HC regeneration and their proliferation is a crucial step.
HCs and SCs are believed to be derived from common progenitors during embryonic development (Fekete et al., 1998). In mice, these progenitors exit the cell cycle around embryonic days 12-14 (E12-14) and remain postmitotic thereafter (Ruben, 1967; Lee et al., 2006). Although the retinoblastoma protein (Rb) (Mantela et al., 2005; Sage et al., 2005; Sage et al., 2006; Weber et al., 2008), p27Kip1 (Chen and Segil, 1999; Lowenheim et al., 1999; Kanzaki et al., 2006; White et al., 2006), p19Ink4d (Chen et al., 2003), p21Cip1 (Laine et al., 2007), and cyclin D1 (Laine et al., 2009) have been implicated in regulating cell cycle in the inner ear, it remains unclear whether mammalian postmitotic SCs in the organ of Corti can proliferate in vivo.
Rb is a key cell cycle inhibitor which suppresses genes required for entering and progressing through the cell cycle. Rb is present in all cells in the E12.5 otocyst and its expression is prominent in HCs during embryonic and adult ages (Sage et al., 2005); however, its expression in postmitotic SCs remains unclear. Deletion of Rb in cochlear HC and SC progenitors produces supernumerary progenitors, which subsequently acquire features of differentiated HCs and SCs and later undergo cell death (Mantela et al., 2005; Sage et al., 2005; Sage et al., 2006); however, it is unclear whether such ectopic proliferation originates from progenitors or postmitotic differentiated HCs and SCs. The acute elimination of Rb in postmitotic HCs causes cell cycle reentry and mitosis; however, HCs die at different stages of the cell cycle before division is complete (Weber et al., 2008). The effect of inactivating Rb in postmitotic SCs remains unknown.
Here, we describe an inducible Cre line (Prox1-CreERT2) specifically expressed in two subtypes of postmitotic SCs, pillar cells (PCs) and Deiters’ cells (DCs). We induce the acute elimination of Rb in postnatal DCs and PCs and examine their capability to proliferate and transdifferentiate in vivo.
The gene-targeted RbloxP/loxP mouse line (Marino et al., 2000) was a generous gift from A. Berns through the National Cancer Institute Mouse Models of Human Cancers Consortium. Prox1-CreERT2 mice (Srinivasan et al., 2007) were kindly provided by G. Oliver at St. Jude Children’s Research Hospital. ROSA26-LacZ reporter (Zambrowicz et al., 1997) and ROSA26-enhanced yellow fluorescent protein (EYFP) reporter (Srinivas et al., 2001) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). The CAG-CAT-enhanced green fluorescent protein (EGFP) reporter line (Nakamura et al., 2006) was kindly provided by J. Robbins from the University of Cincinnati. Genotyping for RbloxP/loxP, Prox1-CreERT2 and ROSA26R-LacZ lines and the administration of tamoxifen at postnatal days 0 and 1 (P0-P1) were described previously (Srinivasan et al., 2007; Weber et al., 2008; Yu and Zuo, 2009). Genotyping of CAG-CAT-EGFP and ROSA26-EYFP lines was performed as described previously (Srinivas et al., 2001; Nakamura et al., 2006).
5-bromo-2-deoxyuridine (BrdU) injection, immunostaining and microscopic analysis were performed using BrdU labeling and detection kit I (Roche Diagnostics, Indianapolis, IN) as previously described (Weber et al., 2008). 5-ethynyl-2-deoxyuridine (EdU) staining was performed using Click-iT EdU imaging kits (Invitrogen, San Diego, CA) (Salic and Mitchison, 2008; Kaiser et al., 2009) following manufacturer instructions. X-gal staining of the cochlea using β-Gal Staining Set (Roche Diagnostics) was also previously described (Chow et al., 2006). Cochlear whole mounts and cyrosections were immunostained with rabbit anti-myosin VIIa (Myo7a) (1:200 dilution, Proteus Bioscience, Ramona, CA), rabbit anti-Prox1 (1:400 dilution, Millipore, Temecula, CA), goat anti-Sox2 (1:250 dilution, Santa Cruz, Santa Cruz, CA), Alexa 647-conjugated rabbit anti-myosin VI (Myo6) (1:20 dilution, Proteus Biosciences), Alexa 488-conjugated rabbit anti-phospho-histone H3 (pH3) (1:20 dilution, Cell Signaling Technology, Danvers, MA), Alexa 488-conjugated rabbit anti-GFP (1:50 dilution, Invitrogen), Hoechst 33342 (1:2,000 dilution, Invitrogen) and 4′,6-Diamidino-2-phenylindole dihydrochloride (DAPI) (1:8,000 dilution, Sigma, St. Louis, MO). Fluorescence images were obtained by using a Zeiss LSM 510 confocal microscope (Carl Zeiss, Jena, Germany).
To perform fluorescence in situ hybridization (FISH) staining, Prox1-CreERT2;RbloxP/loxP mice were given tamoxifen injections at P0 and P1, then BrdU injections at P4 (one injection every 2 hrs, for a total of five injections) with the same dosage as previously described (Weber et al., 2008), and were sacrificed 36-48 hrs after the first BrdU injection. Cochleae were dissected and immersed in methanol/acetic acid (3:1) fixative for 6 hours at 4°C and then cryosectioned. Slides were then denatured with 70% formamide in 2X SSC at 70°C and hybridized with a digoxigenin dUTP labelled bacterial artificial chromosome (BAC) clone that is specific for the Gapdh locus (RP24-489C24). Specific hybridization signals were detected with FITC-coupled anti-digoxigenin antibodies; the slides were then stained with Alexa 488-conjugated mouse anti-BrdU antibody (1:100 dilution, Invitrogen) and counterstained with DAPI.
To determine whether Rb−/− SCs are able to progress through more than one round of S phase, BrdU was injected at P3 in the same manner as described above, followed by one injection of EdU at P5. Mice were then sacrificed 12 hrs after EdU injection. Cochleae were fixed in methanol/acetic acid (3:1) for 6 hours at 4°C followed by whole mount dissection. The whole mounts were stained with EdU using Click-iT EdU imaging kits (Invitrogen) following manufacturer instructions, then stained with BrdU using Alexa 488-conjugated mouse anti-BrdU antibody (1:100 dilution, Invitrogen) and counterstained with DAPI.
To determine the percentage of SCs that showed Cre activity, cochlear whole mounts of Prox1-CreERT2; ROSA26-LacZ mice were stained with X-gal and Myo7a to label Cre-positive cells and HCs respectively, and observed using an Olympus BX60 microscope attached with an Olympus DP71 digital camera (Olympus Optical Co., Tokyo, Japan). The length of the entire cochlear whole mount along the basilar membrane was measured by ImageJ (http://rsbweb.nih.gov/ij/) and divided into three pieces of equal length designated basal, middle and apical turns. The number of HCs and X-gal positive SCs in each piece was counted. It was difficult to determine the SC subtype of lacZ-positive cells by bright field and fluorescence images at different focal planes (bright field focused on SCs and fluorescence focused on HCs); therefore we did not distinguish between DCs and PCs in our analysis of this reporter line. To localize Cre activity in SC subtypes accurately, cochlear whole mounts of Prox1-CreERT2;ROSA26-EYFP and Prox1-CreERT2:CAG-CAT-EGFP mice were co-stained for GFP (Cre activity) and Myo7a (HCs) or Sox2/Prox1 (DCs and PCs), and examined using confocal microscopy. To quantify the total number of SCs, we used DAPI and Myo7a to label nuclei and HCs in cochlear whole mounts, and then counted DCs and PCs based on their precise location relative to inner and outer HCs in confocal 3D reconstructed images when the organ of Corti is still organized at P4 and P6. The length of the entire cochlear whole mount was measured by ImageJ or LSM Image Browser. Starting from the cochlear hook region, the area at 25% of the entire cochlear length was considered representative of the basal turn, 50% as the middle turn and 75% as the apical turn (Fig. 1D). We quantified cell numbers in 200 μm segments of the apical, middle and basal turns. Since the murine cochlea elongates substantially at neonatal ages and Rb−/− SCs proliferate at the same time, whereas HC numbers do not change before P9, we normalized the number of Cre-positive cells, BrdU-positive cells, Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)-positive cells and the total number of DCs and PCs to the number of HCs in the same 200 μm segment (i.e., expressed as a percentage of HCs) to allow for comparison between age, cochlear turn and genotype.
The TUNEL assay was performed using in situ cell death detection kit (Roche Diagnostics). After fixation of cochleae with 4% paraformaldehyde overnight at 4°C, they were dissected for whole mount preparation, followed by incubation in freshly prepared 0.5% Triton X-100 in 0.1% sodium citrate for 15 min at room temperature. Whole mount samples were incubated in TUNEL reaction mixture containing dUTP-FITC and TdT for 60 min at 37°C in the dark. Whole mount samples were then washed and counterstained with Myo7a and imaged using confocal microscopy as described above.
ABR was measured with a Tucker-Davis Technologies (Alachua, FL) system III as described previously (Wu et al., 2004).
All statistical analyses were completed using GraphPad Prism 5.0 software package (San Diego, CA).
In order to eliminate Rb in postmitotic DCs and PCs, we characterized Prox1-CreERT2 mice, in which CreERT2 was inserted into the Prox1 locus following an internal ribosome entry site (IRES) (Srinivasan et al., 2007). Prox1, a homeodomain transcription factor, is initially expressed in the cochlea at E14.5 in developing HCs and SCs (Bermingham-McDonogh et al., 2006). After E18, Prox1 is no longer detectable in cochlear HCs but continues to be expressed in two SC subtypes, DCs and PCs. The expression of Prox1 is down-regulated in these postmitotic cells between birth and P20, in a gradient from the base to the apex of the cochlea. Since the Cre-coding sequence is followed by a Neo-cassette, we only used mice that were heterozygous for the Prox1-CreERT2 allele to avoid potential variation of Prox1 protein expression levels caused by the Neo cassette (Bermingham-McDonogh et al., 2006; Kirjavainen et al., 2008). Prox1-CreERT2/+ mice showed normal Prox1 expression (Fig. 1A), normal morphology of the organ of Corti and normal hearing (data not shown).
To induce Cre recombinase activity restricted to postmitotic DCs and PCs and to directly compare with the previously reported effects of acute Rb inactivation specifically in postmitotic HCs, we injected tamoxifen once daily at P0 and P1. This procedure has been successfully used in the neonatal mouse cochlea, with no abnormal cochlear morphology or hearing defects in tamoxifen induced control or reporter mice at various postnatal ages (Chow et al., 2006; Weber et al., 2008). To visualize Cre activity, we crossed Prox1-CreERT2 mice with three independent reporter lines: Rosa26-EYFP reporter mice in which EYFP is driven by the Rosa26 promoter in cells with Cre recombinase activity (Srinivas et al., 2001), Rosa26-LacZ reporter mice in which lacZ encoding β-galactosidase (β-gal) is expressed in cells with Cre recombinase activity (Yu and Zuo, 2009) and CAG-CAT-EGFP reporter mice in which EGFP is driven by the CMV/β-actin promoter in cells with Cre activity (Kawamoto et al., 2000).
As revealed by analyzing whole mount preparations and cryosections at P6, Cre activity in the organ of Corti was only detected in DCs and PCs along the length of the entire cochlea when tamoxifen was injected at P0 and P1 (Figs. 1B-F, S1A-C). The precise architecture of the organ of Corti allows us to determine whether Cre-positive cells are DCs or PCs based on their location relative to inner and outer HCs. The double labeling of YFP and Prox1 in Prox1-CreERT2;Rosa-EYFP whole mounts at P4 further confirmed that Cre activity was indeed restricted exclusively to DCs and PCs in the organ of Corti (Fig. 1F). No Cre activity was observed in the organ of Corti of reporter mice without tamoxifen injection or in Prox1-CreERT2 negative mice with tamoxifen injection (data not shown). The absence of Cre activity in HCs was confirmed by counter-staining with the HC marker, Myo7a (Figs. 1C-E, S1B-C).
To quantify the amount of Cre activity in Prox1-CreERT2 mice and to reduce variability, we measured the length of the entire cochlear duct from the hook, and used the region at 75% of the total length to represent the apical turn, 50% for the middle turn and 25% for the basal turn (Fig. 1E). We first normalized the number of Cre-positive cells to the total number of SCs (both DCs and PCs) from corresponding cochlear turns (i.e., expressed as percentages of total SCs) (Fig. 1G-I; Fig. S2A-C). Since the mouse cochlea elongates dramatically at neonatal ages and in our Rb−/− model HC numbers did not change, whereas SC numbers did change significantly before P9 (see below), we normalized the number of Cre-positive cells (Fig. 1J-L; Fig. S1D; Fig. S2D-F) to the number of HCs in the same 200 μm segment (i.e., expressed as a percentage of HC number) at the same location to allow for comparison between age, cochlear turn, genotype and subsequent analyses (i.e., BrdU and TUNEL).
Quantitative analysis of EYFP-positive (Fig. 1G-L), lacZ-positive (Fig. S1D) and EGFP-positive cells (Fig. S2A-F) at P6 showed: 1) that only a fraction of DCs and PCs were Cre-reporter positive (i.e., approximately 13% of total DC and PC numbers in the apical turn); 2) that there is a gradient of decreasing Cre activity from apical to basal turns, as determined by linear regression analysis (r2 = 0.87 for EYFP reporter; r2 = 0.80 for lacZ reporter; r2 = 0.72 for EGFP reporter); 3) that there were more Cre-reporter positive DCs than PCs in every cochlear turn and 4) that there were more Cre-reporter positive DCs in the apical turn than in middle or basal turns; however, Cre-reporter activity in PCs exhibited no statistical difference among cochlear turns. Based on these results, we focused most of our subsequent analysis on the apical turn.
The mosaic Cre-reporter activity among Prox1-positive cells and differential Cre activity between DCs and PCs may be caused by reduced translation efficiency of IRES in the Prox1-CreERT2 locus, heterogeneous activity of the Prox1 promoter driving CreERT2 expression, mosaic activity of the ROSA26 or CMV/β-actin promoter in reporter mice and/or differential uptake of tamoxifen by DCs and PCs in the developing cochlea at P0-P1 (Hildinger et al., 1998; Zhou et al., 1998). Regardless, Prox1-CreERT2 represents a powerful and unique genetic tool for gene manipulation specifically in two postnatal cochlear SC subtypes.
To investigate whether the acute deletion of Rb in postnatal DCs and PCs results in proliferation, we crossed a gene-targeted RbloxP/loxP mouse line, in which exon 19 of the Rb gene was flanked by two loxP sites (Marino et al., 2000), with Prox1-CreERT2 mice. After injection of tamoxifen at P0 and P1, Prox1-CreERT2;RbloxP/loxP mice (designated Rb−/−) received intraperitoneal injections of BrdU (one injection every 2 hrs, for a total of five injections) or EdU (one injection daily) (Salic and Mitchison, 2008) at P4 or P6. BrdU-positive DCs and PCs were frequently detected in the Rb−/− cochlea (Fig. 2A-C), but never in the cochlea of control mice (littermates that were either Cre-negative or RbloxP/+) (Fig. 2D), consistent with the observation that Cre activity was exclusively detected in DCs and PCs in Prox1-CreERT2 reporter cochleae (Figs. (Figs.1,1, S1, S2).
The number of BrdU-positive cells in the Rb−/− organ of Corti after a 10 hr BrdU pulse at P4 was counted by analyzing whole mount preparations (Fig. 2E-G). We identified 7.00 ± 0.40, 2.79 ± 0.51 and 0.40 ± 0.25 BrdU-positive cells in apical, middle and basal turns, respectively. When normalized to HC number for comparison with other analyses, the number of BrdU-positive cells displayed a gradient from apical (5.73 ± 0.42 %) to middle (2.27 ± 0.40 %) to basal (0.37 ± 0.23 %) turns (Fig. 2H; linear regression r2 = 0.90), in agreement with the gradient of the Cre activity detected in Prox1-CreERT2 reporter cochleae (Figs. (Figs.1J;1J; S1D, S2D).
Because of lack of specificity for mouse Rb antibodies, low level of Rb expression in SCs (Mantela et al., 2005; Sage et al., 2005) and the brief 10 hr BrdU pulse, we could not determine the actual percentage of Rb−/− cells that reentered the cell cycle; however, comparison of BrdU-positive cells in the apical turn of Prox1-CreERT2;RbloxP/loxP cochleae (5.73% in Fig. 2H) with Cre-positive cells in the apical turn of Prox1-CreERT2 reporter cochleae (~18% in Fig. 1J, ~14% in S1D and ~15% in S2D) suggests that ~30-40% of cells with Cre activity in the apical turn reentered the cell cycle during the 10 hr BrdU pulse at P4. This is comparable with our previous result that ~40% of Cre-positive HCs rapidly reentered the cell cycle using similar procedures of Rb inactivation and BrdU labeling (Weber et al., 2008).
Interestingly, we found more PCs than DCs labeled with BrdU in the apical turn at P4 (Fig. 2I), while more DCs appeared to have Cre activity in the apical turn (Fig. 1H-I and K-L; Fig. S2B-C and E-F). At P6 the number of BrdU positive DCs increased to the level of BrdU positive PCs (Fig. 2I). It is possible that DCs and PCs continue to reenter the cell cycle after P6; however, we could not distinguish between DCs and PCs after P6 because of the severely disorganized structure of organ of Corti and the lack of specific markers for each cell type.
Although no Myo7a-positive cells in the Rb−/− cochlea were labeled by BrdU, we occasionally observed BrdU-positive nuclei in the HC layer, above the normal position of PCs and DCs (Fig. 2B-C); therefore, it appears that the nuclei of some proliferating SCs migrated upward from their normal position into the HC layer, a pattern similar to the avian HC regeneration model (Tsue et al., 1994; Stone and Cotanche, 2007).
To determine whether Rb−/− SCs can progress through the cell cycle and enter mitosis (M) phase, we performed immunostaining with pH3, a marker of M phase nuclei. We detected PCs with strong pH3 labeling in Rb−/− cochlear whole mounts at P4 (Fig. 3A) and P8 (data not shown), as well as pH3-positive DCs at P6 (Fig. 3B). BrdU-positive SCs in the Rb−/− cochlea also showed condensed individual chromosomes (Fig. 3C, E, G-I), a morphologic feature of M phase cells (Weber et al., 2008). Such cells were usually observed at the luminal surface of the sensory epithelium (Supplemental Video 1). Among ten BrdU-positive SCs in M phase, eight were located in the HC layer and two were located between the HC nuclear layer and the SC nuclear layer. In addition, six of the eight SCs in the HC layer were even more elevated than HC nuclei (i.e., very close to the luminal surface of the sensory epithelium). None of the mitotic cells were labeled with Myo7a (Fig. 3A-F).
To test whether Rb−/− SCs could complete the cell cycle, we used BrdU labeling in combination with FISH, which allows discrimination between cells with unduplicated (G1 or early S phase) or duplicated (late S, G2 or M phase) chromosomes. We used a genomic probe containing the ubiquitously expressed and early duplicated gene, Gapdh. Among BrdU-positive PCs, 2 of 9 cells showed G1 or early S phase features (i.e., two FISH signals; Fig. 4A) at P6, whereas 7 cells showed late S, G2 or M phase features (i.e., four FISH signals, data not shown). All BrdU-positive DCs (5 cells) showed late S, G2 or M phase features (i.e., four FISH signals; Fig. 4B). Observation of BrdU plus unduplicated chromosomes (two FISH signals) indicates that these cells have reentered and completed the cell cycle; whereas, cells that were labeled with BrdU and duplicated chromosomes (four FISH signals) have either not completed the cell cycle or reentered the cell cycle a second time. These data alone, however, cannot exclude the possibility that some of Rb−/− SCs (particularly DCs) were arrested in S phase or at the G2/M check point.
To confirm that Rb−/− SCs progress through more than one round of S phase, we injected BrdU at P3 and EdU at P5. BrdU/EdU double-positive PCs were repetitively identified in Rb−/− cochlear whole mounts (asterisks in Fig. 4C). Interestingly, pairs of BrdU and/or EdU labeled PCs were observed (asterisks in Fig. 4C, D), which also suggests completion of the cell cycle. In the cochlear apical turns of Rb−/− mice we identified 32 BrdU single positive PCs, 7 BrdU single postive DCs, and 8 BrdU/EdU double positive PCs. The number of cells going through consecutive rounds of the cell cycle was likely underestimated due to the decreased window of time that EdU was present (the toxicity of EdU prevented multiple injections). Nevertheless, these results suggest that postmitotic Rb−/− PCs have entered S phase a second time and thus have the potential to divide multiple times in vivo; however, there is no evidence that DCs can complete the cell cycle. To confirm that there is no cross-reaction between BrdU and EdU immunostaining procedures, we performed BrdU/EdU double staining on Rb−/− cochlear whole mounts treated with either BrdU alone or EdU alone and did not detect any cross-reactivity (Fig. 4 D-E).
To accurately quantify and compare the number of SCs in the organ of Corti at different ages, we again normalized the number of SCs to the number of HCs. The total number of SCs (both DCs and PCs) in a 200 μm segment of the apical turn was 139 ± 1% at both P4 and P6 in control cochlea, 150 ± 3% and 156 ± 5% in the Rb−/− cochlea at P4 and P6, respectively (Fig. 4F). In other words, the total number of SCs in the Rb−/− apical turn at P6 was 1.12 fold increased or 12% greater than control (Fig. 4F). Although not significant, there was a trend toward an increase in total SC number at P4. Moreover, when DCs and PCs were analyzed separately, the number of PCs, again normalized to HC number, was significantly increased compared to control at P6 (1.21 fold increase or 21% greater, Fig. 4H); whereas the number of DCs normalized to HC number at P4 and P6 were not significantly different from control (Fig. 4G). These results are consistent with our observation that there are more BrdU-positive PCs than DCs at P4 (Fig. 2I). Considering that only ~3.5% of PCs are Cre-positive in the apical turns of Rb−/− mice (Fig. 1K; Fig. S2E), such a dramatic increase (~21%) in the number of PCs at P6 (Fig. 4H) suggests that these cells undergo multiple rounds of cell divisions from P2 to P6.
In summary, incorporation of BrdU together with detection of unduplicated chromosomes, a second S phase marker and an increase in cell number, provides strong evidence that Rb−/− PCs can complete the cell cycle, as well as initiate multiple rounds of DNA replication, although we cannot exclude the possibility that some of Rb−/− SCs are arrested in S phase or at the G2/M check point.
Loss of Rb can induce apoptosis in certain tissues and cells, including postmitotic HCs (Weber et al., 2008). At P7, we first detected dying cells in the organ of Corti using TUNEL staining (a late marker of apoptosis) (Fig. 5A). These dying cells are likely DCs whose nuclei have migrated into the HC layer since TUNEL-positive cells were not Myo7a positive and HCs surrounding the TUNEL-positive cells appeared intact. This could explain why we did not observe proliferation or increase in cell number for DCs (Fig. 4G). Alternatively, we may not have detected proliferation of DCs due to lower chance of cell cycle reentry of DCs than that of PCs (Fig. 2I). We are unable to determine when PC death begins as the architecture in the organ of Corti is severely disrupted after P7, and thus we can no longer distinguish PCs from DCs. At P9, the average number of TUNEL-positive SCs, when normalized to HC number, in the apical turn of Rb−/− cochleae was 4.0 ± 0.9%, while no TUNEL-positive cells were observed in the control (p < 0.05; n=3, two-tailed Student’s t test; data not shown). TUNEL staining is prone to underestimating the actual number of dying cells and its sensitivity is dependent on fixation and pretreatment variables (Lucassen et al., 1995; Negoescu et al., 1996); thus, the number of TUNEL-labeled cells is likely a low estimate of the actual number of dying cells. Interestingly, cell death of Rb−/− SCs appeared at a later time than cell death of Rb−/− HCs in our previous model (Weber et al., 2008).
HC numbers were not significantly changed at P9 and no ectopic HCs were observed by Myo7a (Fig. 5A, B) or Myo6 staining (data not shown); however, the HC layer was disorganized, with cell bodies squeezed together, likely caused by the translocation of SC nuclei from underneath (Fig. 5A, B). At P12 scattered HC loss was clearly evident in the Rb−/− cochlea (Fig. 5C) and was more severe at 7 weeks of age (Fig. 5D). Interestingly, HC loss was more prominent in the apical turn than in middle and basal turns at P12 and P21 (Fig. S3A, B), consistent with the gradient of Cre activity (Figs. 1G-L, S1D, S2A-F) and BrdU positive SCs (Fig. 2E-H) at earlier ages. It is therefore likely that HC loss is secondary to SC proliferation and subsequent loss. Lack of trophic and mechanical support from SCs, non-cell autonomous effects of Rb deletion in SCs or secondary effects of excessive proliferation within the precisely orchestrated organ of Corti could all result in subsequent HC loss. Furthermore, ABR thresholds of Rb−/− mice at 5 weeks of age were significantly increased from 6-32 kHz compared to controls (Fig. 5E). Neither HC loss, SC loss nor hearing deficits have been observed in control mice or reporter lines treated with identical amounts of tamoxifen at the same ages and with the same strain backgrounds (Chow et al., 2006; Weber et al., 2008).
To determine the cell fate of proliferating Rb−/− DCs and PCs, we counterstained EdU-labeled Rb−/− cochleae with Prox1 or Sox2, two well characterized SC markers. All EdU-positive cells were also labeled with Prox1 or Sox2 at P4 (Fig. 6A-F; data on DCs not shown). Together, these results demonstrate that proliferating cells still maintained their SC fate during cell cycle reentry.
In the developing organ of Corti, HCs prevent SCs from also becoming HCs by Notch-mediated lateral inhibition (Lanford et al., 1999). When HCs are lost, the release of this lateral inhibition may induce SCs to proliferate and further regenerate HCs. While HCs can be effectively damaged in the adult mouse cochlea, with antibiotics in combination with diuretics or acoustic injury (Oesterle et al., 2008; Oesterle and Campbell, 2009), it has been difficult, if not impossible, to damage HCs in the neonatal mouse cochlea in vivo. To test whether in vivo HC loss secondary to SC proliferation and loss would further induce proliferation and transdifferentiation of Rb−/− SCs into new HCs, we injected EdU, one dose daily from P8 to P14, in Rb−/− mice and analyzed the cochlea at P14. EdU-positive SCs were found co-labeled with the SC marker, Sox2, in the area where HCs were lost (Fig. 6G-I). No EdU/Myo6 double positive cells were observed, suggesting that Rb−/− SCs did not transdifferentiate into HCs even after the loss of neighboring HCs and such failure to transdifferentiate into HCs is likely due to intrinsic defects of proliferating Rb−/− SCs. However, we did not damage HCs before inactivating Rb in SCs, which may be the reason why these new cells did not differentiate into HCs.
In this study we characterized a new, inducible Cre recombinase mouse line, Prox1-CreERT2, using three independent reporter lines. When induced at P0 and P1, Prox1-CreERT2 displays Cre activity specifically in DCs and PCs, but not in other cell types of the organ of Corti. Despite the mosaic and differential pattern among DCs and PCs, the apical to basal gradient of Cre activity is in agreement with the observation that the expression of Prox1 is down-regulated in a basal to apical gradient of the cochlea after birth (Bermingham-McDonogh et al., 2006; Kirjavainen et al., 2008). DCs and PCs are located directly below HCs, which largely resembles the location of SCs that regenerate HCs in non-mammalian vertebrates (Stone and Cotanche, 2007); therefore, the Prox1-CreERT2 line allows specific inactivation of genes in a subset of postmitotic cochlear SCs, which are good candidates for the source of HC regeneration. Compared to viral transfection of cochlear cells (Minoda et al., 2007), the use of CreER restricts gene alteration to specific cell types in a reproducible manner. Our model, however, is limited in that Prox1-CreERT2 can likely be used to manipulate genes only in neonatal SCs because of the widespread expression of Prox1 in HCs and SCs during embryonic development (before E18) and down-regulation of Prox1 in SCs from birth until its disappearance at P20 (Bermingham-McDonogh et al., 2006; Kirjavainen et al., 2008).
Although Rb has not been convincingly detected in postnatal SCs, we provide several lines of evidence that postmitotic SCs can reenter the cell cycle after Rb ablation in vivo. BrdU signals were restricted to DCs and PCs in the Rb−/− organ of Corti, consistent with the Cre-activated cell types detected in the Prox1-CreERT2 reporter cochlea. In addition, there was a similar apical to basal gradient observed in both BrdU-positive cells and Cre-activated cells. In sharp contrast to Rb inactivation in postmitotic HCs, Rb−/− PCs complete the cell cycle and display a significant increase in cell number by P6. These results clarify that Rb is expressed in postmitotic SCs and, as in other postmitotic cells (Sage et al., 2003), is required to maintain their quiescent state. These findings are also in agreement with the in vitro findings that SCs from the early postnatal mouse cochlea retain the ability to divide (White et al., 2006). Furthermore, they support the possibility that the supernumerary SCs found in previous Rb inactivation mouse models might also result from proliferation of postmitotic SCs, in addition to proliferation of progenitor cells (Mantela et al., 2005; Sage et al., 2005; Chen, 2006; Sage et al., 2006). It is unknown whether cell death of Rb−/− DCs and PCs in our model is primarily caused by Rb ablation or by the lack of space and nutrition needed for rapidly proliferating cells in the precise architecture of the organ of Corti.
The fact that Rb−/− HCs cannot complete the cell cycle and proliferate, whereas, Rb−/− PCs can, may indicate that Rb plays essential roles in other phases of the cell cycle in postmitotic HCs but not in postmitotic PCs. Alternatively, PCs may be less mature or differentiated, and thus more able to proliferate than HCs at this postnatal age. It remains to be further determined whether an intrinsic mechanism exists for proliferating Rb−/− PCs to abort their proliferation after several rounds of division.
Based on our results that PCs were able to complete the cell cycle and proliferate for multiple divisions, whereas only a small portion of Rb−/− DCs were detected reentering the cell cycle, it is tempting to hypothesize that PCs are more “stem cell-like” and can thus more easily reenter the cell cycle and proliferate. Such striking heterogeneity among various types of SCs is also supported by in vitro studies of isolated SCs (White et al., 2006) and the different pathways for differentiation and maintenance of DCs versus PCs (Doetzlhofer et al., 2009). In addition, a recent study found differential expression of cyclin D1 between neonatal PCs and DCs, which may contribute to the ability of PCs to proliferate (Laine et al., 2009). Moreover, in p27Kip1 germline knockout mice, postnatal PCs, but not DCs, were found to reenter the cell cycle at P6 (Chen and Segil, 1999). Our results that postnatal Rb−/− DCs and PCs can reenter the cell cycle provide strong evidence that Rb acts downstream of p27Kip1 in PCs and suggest different upstream regulation of Rb in DCs, although compensational mechanisms may exist. It is also possible that DC proliferation could have occurred after P6 when the organ of Corti became disorganized. Indeed, the late increase of BrdU labeling of DCs at P6 compared to that at P4 (Fig. 2I) is consistent with this notion. Differences in degradation of Rb, length of G0-G1 phase or uptake of BrdU may contribute to the delayed detection of cell cycle reentry in DCs.
SCs play key roles in providing trophic and mechanical support for HCs. During postnatal development, efferent fibers also initiate their transition from inner to outer HC areas through SCs. To our knowledge, there are no prior reports of selective SC ablation prior to HC damage in the neonatal mouse cochlea. It remains possible that rapid proliferation of PCs (by either Rb inactivation or other manipulations) would disrupt the precise architecture of the organ of Corti and deprive necessary support for HCs, thus leading to secondary loss of HCs.
Proliferating SCs in our study appear to undergo cell cycle and migratory changes that are similar to those in the regenerating non-mammalian sensory epithelium (Stone and Cotanche, 2007). In the mature chicken vestibular sensory epithelium, SCs reenter S phase of the cell cycle at the basal level of the sensory epithelium. Their nuclei then migrate upwards toward the lumen during G2 and early M phase, dividing near the surface of the sensory epithelium (Tsue et al., 1994). In the regenerating chicken auditory epithelium, SC mitotic figures have also been observed near and parallel to the luminal surface (Raphael, 1992). In our mouse model, a majority of M phase SCs were observed near and in parallel to the luminal surface of the sensory epithelium, indicating that Rb ablation is likely sufficient to induce not only S phase reentry, but also nuclear migration and mitosis in postmitotic mammalian SCs. Our findings, therefore, suggest that mammalian SCs follow mechanisms similar to those of avian SCs to induce cell division and nuclear migration, which is the first stage of HC regeneration. The failure of newly generated SCs to transdifferentiate into HCs may be due to their inability to reestablish a G0 state since these cells now lack Rb; therefore transient Rb inactivation in SCs may represent a better strategy for HC regeneration. In addition, future in vitro experiments using Rb−/− cochlear explants and ablation of selected HCs using a laser or ototoxic drugs can be used to experimentally determine if newly generated SCs can differentiate into HCs in the context of initial HC loss.
Isolated SCs from the postnatal mouse cochlea can down-regulate the cell cycle inhibitor, p27Kip1, proliferate and transdifferentiate into HCs in vitro (White et al., 2006). Thus, postmitotic mammalian cochlear SCs may be good candidates for HC regeneration; however, in our in vivo study, there was no evidence of SC transdifferentiation into HCs. This is consistent with in vivo studies in the adult guinea pig cochlea where forced expression of Skp2, a ubiquitin ligase that enhances the degradation of p27Kip1, led to proliferation of non-sensory cells outside the organ of Corti, but was not sufficient to generate HCs (Minoda et al., 2007). There are many potential differences between in vitro and in vivo model systems to explain these results. For example, isolated postnatal SCs are cultured with embryonic mesenchymal cells and several growth factors supplied by the media (White et al., 2006). In addition, lateral inhibition via the Notch pathway is disrupted or eliminated in isolated SCs. In our in vivo model, proliferating SCs still express Prox1 and Sox2, and maintain their SC cell fate. Our results thus demonstrate that whereas SC proliferation is a necessary component of HC regeneration, additional factors are needed to restore their intrinsic ability to transdifferentiate into HCs in the mammalian cochlea.
Several potential factors may facilitate transdifferentiation into HCs. Atoh1 is a basic helix-loop-helix transcription factor that is both necessary and sufficient for the differentiation of HCs in the mammalian cochlea (Bermingham et al., 1999; Zheng and Gao, 2000; Chen et al., 2002; Shou et al., 2003; Izumikawa et al., 2005; Gubbels et al., 2008). Inhibition of Ids, Prox1 and Sox2 may also up-regulate Atoh1 in Rb−/− proliferating SCs (Jones et al., 2006; Khidr and Chen, 2006; Kirjavainen et al., 2008). Therefore, a combination of transient Rb inactivation and Atoh1 activation or inactivation of Ids, Prox1 or Sox2 may represent an effective measure to regenerate cochlear HCs.
We thank Drs. S. Baker, M. Dyer, G. Oliver, M. Roussel and R. Srinivasan for advice and for sharing reagents; Drs. J. Fang, M. Mellado-Lagarde, K. Steigelman and other members of the Zuo lab for discussion and critical comments; S. Connell, L. Zhang, Y. Ouyang and J. Peters for expertise in confocal imaging. This work was supported in part by grants from the National Institutes of Health (DC006471, DC008800, DC010310 and CA21765), Office of Naval Research (N000140911014) and the American Lebanese Syrian Associated Charities of St. Jude Children’s Research Hospital. J. Zuo is a recipient of The Hartwell Individual Biomedical Research Award.