In vivo notch reactivation in differentiating cochlear hair cells induces sox2 and prox1 expression but does not disrupt hair cell maturation


  • Zhiyong Liu,

    1. Department of Developmental Neurobiology, St. Jude Children's Research Hospital, Memphis, Tennessee
    2. Integrated Program in Biomedical Sciences, University of Tennessee Health Science Center, Memphis, Tennessee
    Search for more papers by this author
  • Thomas Owen,

    1. Department of Developmental Neurobiology, St. Jude Children's Research Hospital, Memphis, Tennessee
    2. University of Bath, Bath, United Kingdom
    Search for more papers by this author
  • Jie Fang,

    1. Department of Developmental Neurobiology, St. Jude Children's Research Hospital, Memphis, Tennessee
    Search for more papers by this author
  • R. Sathish Srinivasan,

    1. Department of Genetics and Tumor Cell Biology, St. Jude Children's Research Hospital, Memphis, Tennessee
    Search for more papers by this author
  • Jian Zuo

    Corresponding author
    1. Department of Developmental Neurobiology, St. Jude Children's Research Hospital, Memphis, Tennessee
    • Department of Developmental Neurobiology, St. Jude Children's Research Hospital, 262 Danny Thomas Place, Memphis, TN 38105

    Search for more papers by this author


Background: Notch signaling is active in mouse cochlear prosensory progenitors but declines in differentiating sensory hair cells (HCs). Overactivation of the Notch1 intracellular domain (NICD) in progenitors blocks HC fate commitment and/or differentiation. However, it is not known whether reactivation of NICD in differentiating HCs also interrupts their developmental program and reactivates its downstream targets. Results: By analyzing Atoh1CreER+; Rosa26-NICDloxp/+ or Atoh1CreER+; Rosa26-NICDloxp/+; RBP-Jloxp/loxp mice, we demonstrated that ectopic NICD in differentiating HCs caused reactivation of Sox2 and Prox1 in an RBP-J-dependent manner. Interestingly, Prox1 reactivation was exclusive to outer HCs (OHCs). In addition, lineage tracing analysis of Prox1CreER/+; Rosa26-EYFPloxp/+ and Prox1CreEGFP/+; Rosa26-EYFPloxp/+ mice showed that nearly all HCs experiencing Prox1 expression were OHCs. Surprisingly, these HCs still matured normally with expression of prestin, wild-type-like morphology, and uptake of FM4-64FX dye at adult ages. Conclusions: Our results suggest that the developmental program of cochlear differentiating HCs is refractory to Notch reactivation and that Notch is an upstream regulator of Sox2 and Prox1 in cochlear development. In addition, our results support that Sox2 and Prox1 should not be the main blockers for terminal differentiation of HCs newly regenerated from postnatal cochlear SCs that still maintain Sox2 and Prox1 expression. Developmental Dynamics 241:684–696, 2012. © 2012 Wiley Periodicals, Inc.


Mice detect sound through the mechanosensory hair cells (HCs) that reside in the cochlea, a coiled auditory organ in the ventral part of the inner ear (Morsli et al.,1998). The cochlear auditory epithelium, also referred to as the organ of Corti, has three rows of outer hair cells (OHCs) and one row of inner hair cells (IHCs; Kelley,2006; Kelly and Chen,2009). Nonsensory supporting cells (SCs) surround these HCs. HCs and SCs are believed to derive from the same prosensory progenitors (Fekete et al.,1998). Cochlear development can be roughly divided into the early prosensory phase and the late cell fate determination phase (Hayashi et al.,2008a; Bermingham-McDonogh and Reh,2011). In the early prosensory phase (between embryonic day [E] 12 and E14.5), Notch signaling (primarily through Notch1/Jagged1) plays a critical role in specifying prosensory progenitors (Eddison et al.,2000; Kiernan et al.,2001,2006; Tsai et al.,2001; Brooker et al.,2006). Consistent with the ability to promote prosensory progenitors, overactivation of Notch1 in the prosensory phase causes formation of ectopic HCs (Daudet and Lewis,2005; Hartman et al.,2010; Pan et al.,2010). In the late cell fate determination phase (after E14.5), Notch signaling declines through “lateral inhibition” effects (Lanford et al.,1999) in progenitors that have committed to the HC fate but persists in progenitors that have committed to the SC fate (Zhang et al.,2000; Zheng et al.,2000b; Zine et al.,2000,2001; Kiernan et al.,2005; Takebayashi et al.,2007). Consistent with this pattern, loss of or decrease in Notch1 signaling on postnatal day 0 (P0) SCs can result in conversion of SCs to HCs (Yamamoto et al.,2006; Doetzlhofer et al.,2009). However, expression patterns of Notch1 intracellular domain (NICD) and Hes5 suggest that Notch signaling declines postnatally in cochlear SCs (Murata et al.,2006; Hartman et al.,2009).

Transfection of NICD into progenitors of ∼E13 cochlear explants blocks the activity of Atoh1, which is a crucial transcription factor for HC development (Bermingham et al.,1999), and consequently reduces the number of HCs (Dabdoub et al.,2008). These results strongly support that down-regulation of Notch activity is necessary for HC fate commitment and/or initial HC differentiation. However, it remains unknown whether persistent expression of NICD in differentiating HCs also prevents full differentiation or maturation of immature HCs and further turns on its normal downstream targets such as Sox2 and/or Prox1, as proposed previously (Dabdoub et al.,2008). This question is particularly interesting, because differentiating immature HCs could fail to be fully differentiated or even convert to SCs if Notch signaling persists and/or SC markers are ectopically expressed in these immature HCs. This scenario could also occur when HCs newly regenerated by Atoh1-mediated reprogramming/conversion of postnatal SCs maintain high levels of SC markers such as Sox2.

In this study, we reactivated NICD in differentiating HCs in Atoh1CreER+; Rosa26-NICDloxp/+ or Atoh1CreER+; Rosa26NICD/EYFP mouse models. We found that Sox2 and Prox1 (Bermingham-McDonogh et al.,2006; Hume et al.,2007; Oesterle et al.,2008), which are normally not expressed in postnatal HCs, were reactivated in an RBP-J-dependent manner. Of interest, while Sox2 was expressed both in IHCs and OHCs with ectopic NICD, Prox1 was expressed exclusively in OHCs. Surprisingly, ectopic NICD starting at P0 and P1 did not affect the differentiation and maturation of HCs. Our data suggest that the intrinsic developmental program of mouse neonatal differentiating HCs is refractory to ectopic NICD, Sox2, and Prox1, and is thus less plastic than that of progenitor cells. Our observations indirectly support that Notch signaling and expression of Sox2 and Prox1 might not be the main factors blocking the complete differentiation of HCs newly converted from postnatal SCs, which has significant implications for HC regeneration studies in mammals (Brigande and Heller,2009; Bermingham-McDonogh and Reh,2011).


In Vivo Models to Induce Ectopic NICD Expression

Given the lateral inhibition effect elicited by Notch1 signaling (Lanford et al.,1999; Kiernan et al.,2005), we predicted that the normal maturation of differentiating HCs would be affected if Notch1 signaling was reactivated in them. To test this hypothesis, we used CreER/loxP technologies (Joyner and Zervas,2006) to specifically and reproducibly reactivate NICD in differentiating cochlear HCs at late embryonic (∼E16) and neonatal ages (P0 and P1). In the current study, owing to the dystocia associated with tamoxifen induction in mice at embryonic ages, we primarily focused on neonatal ages.

The Atoh1CreER+; Rosa26-NICDloxp/+ model was used to induce ectopic NICD in HCs. Atoh1CreER+ is a transgenic mouse line in which CreER is driven by the Atoh1 enhancer (Fig. 1A). Although the CreER driver is not the entire Atoh1 regulatory cis-element, when tamoxifen is given at postnatal day (P) 0 and P1, the Cre recombinase activity is specific to HCs and 80–90% cochlear HCs are targeted (Chow et al.,2006; Weber et al.,2008). Without tamoxifen induction, no nuclear Cre activity is detectable in the cochlea. Rosa26-NICDloxp/+ is a knockin mouse line (Murtaugh et al.,2003) in which NICD and enhanced green fluorescent protein (EGFP) are expressed within the same transcript driven by the ubiquitous Rosa26 promoter, but not translated unless the preceding stop fragment is removed by the Cre recombinase (Fig. 1A). After Cre-mediated excision, two different proteins are translated from the same transcript within the same cell: NICD by means of a cap-dependent mechanism and EGFP by means of Internal Ribosome Entry Site (IRES). The NICD subsequently translocates to the nucleus and further activates Notch1 signaling (Kopan and Ilagan,2009). In addition, because of the nuclear localization signal, EGFP is restricted in the nucleus (Fig. 1A). Thus, EGFP will faithfully reflect NICD cellular expression and also serve as a lineage tracer for NICD-expressing cells.

Figure 1.

Reactivation of Notch1 intracellular domain (NICD) in embryonic day (E) 16 hair cells. A: The scheme of NICD induction in differentiating hair cells (HCs). The two red arrows indicate two translation sites within the same cis element. B–C′: Double staining of Myosin-VI and enhanced green fluorescent protein (EGFP) at ∼E19 in embryos of Rosa26-NICDloxp/+ control mice (B–B′) and Atoh1CreER+; Rosa26-NICDloxp/+ experimental mice (C–C′) treated with tamoxifen at ∼E16. Images shown in (B,C) were taken at HC top surface layer and those in (B′ and C′) at the HC nucleus layer by confocal microscopy. OHCs, outer hair cells; IHCs, inner hair cells. Scale bar = 20 μm.

Differentiating Cochlear HCs Are Not Sensitive to NICD Reactivation

Atoh1CreER+; Rosa26-NICDloxp/+ (experimental group, n = 4) and Rosa26-NICDloxp/+ (control group, n = 4) littermate embryos were treated with tamoxifen at ∼E16 and analyzed on ∼E19 (Fig. 1B–C′). There were no EGFP+ cells in control embryos (Fig. 1B–B′). In experimental embryos, all EGFP+ cells exclusively resided in the HC layer and EGFP expression was restricted to the nucleus (Fig. 1C–C′), confirming the specificity of Cre activity in HCs. Notably, all EGFP+ cells expressed Myosin VI (Fig. 1C–C′), suggesting that HC differentiation is normal by E19 in the presence of ectopic NICD. However, dystocia (difficulty during birth) in mice made it difficult to further analyze the long-term effect of NICD on differentiating HCs at ∼E16 or ∼E16.5, if the slight delay between tamoxifen treatment and NICD expression was considered.

Next, we studied neonatal HCs at P0 and P1 (Fig. 2). At this stage, neonatal HCs are far from fully differentiated and NICD can be used to potentially interrupt their differentiation program. We injected Atoh1CreER+; Rosa26-NICDloxp/+ (experimental group, n = 3) and Rosa26-NICDloxp/+ (control group, n = 3) neonatal mice with tamoxifen at P0 and P1, thereby bypassing the dystocia problem and allowing analysis of mice at adult ages. There was no EGFP expression in the control mice (Fig. 2A). However, HCs expressing low levels of EGFP at P2 (Fig. 2B) and high levels of EGFP at P6 (Fig. 2C) were found along all cochlear turns of Atoh1CreER+; Rosa26-NICDloxp/+ mice. There was no substantial difference in the number of EGFP+ HCs among different turns at P6 (Fig. 2D). Only ∼15% of OHCs were EGFP+, much lower than that (80–90%) of Atoh1CreER active HCs reported in our previous studies (Chow et al.,2006; Weber et al.,2008), which is likely due to the low EGFP translation efficiency mediated by the IRES but not due to a problem with the Atoh1CreER+ transgenic line. This explanation was further supported by the much higher percentage (80–90%) of enhanced yellow fluorescent protein-positive (EYFP+) HCs in the Atoh1CreER+;Rosa26EYFP/NICD model in which EYFP was translated in a cap-dependent manner (discussed later). In addition, 93% of EGFP+ HCs were OHCs and 7% were IHCs (Fig. 2E), likely due to the overall higher level of expression of Atoh1 (or higher Cre activity) in neonatal OHCs than IHCs, or combinations of Rosa26 promoter activity and IRES-mediated translation efficiency favoring OHCs. Because EGFP was expressed in HCs only, EGFP+ cells could be defined as HC lineages.

Figure 2.

Reactivation of Notch1 intracellular domain (NICD) in neonatal hair cells at postnatal day (P) 0 and P1. A–C: Images of enhanced green fluorescent protein (EGFP) and Myosin VI double immunostaining in control samples at P6 (A), and experimental samples at P2 (B) and P6 (C). D: Quantification of EGFP+ hair cells (HCs) in ∼200 μm different cochlear turns. E: Distribution of EGFP expression in OHCs and IHCs. OHCs, outer hair cells; IHCs, inner hair cells. F–F″: EGFP and Sox2 double immunostaining of samples from Atoh1CreER+; Rosa26-NICDloxp/+ mice at P3. G–G″: EGFP and Prox1 double immunostaining of samples from Atoh1CreER+; Rosa26-NICDloxp/+ mice at P4.5. F″ and G″ are merged images of (F) and (F″), and (G) and (G″), respectively. Scale bar = 20 μm.

As shown in previous studies (Bermingham-McDonogh et al.,2006; Hume et al.,2007), Sox2+ or Prox1+ HCs were not found in Rosa26-NICDloxp/+ control mice at P6. For experimental mice (n = 3), samples were analyzed at P2, P3, P4, P4.5, P5, and P6. Ectopic Sox2 expression was first detected at ∼ P3 (Fig. 2F–F″), and ectopic Prox1 expression began at ∼P4.5 (Fig. 2G–G″). The delayed ectopic Prox1 expression suggests that Sox2 is needed to turn on Prox1, consistent with the fact that turning on Prox1 at embryonic cochlea depends on Sox2 (Dabdoub et al.,2008). At P6, all Sox2 and Prox1 expressing HCs were EGFP+ (Fig. 3A–A′″). Of interest, though all EGFP+ HCs (both IHCs and OHCs) were Sox2+, ectopic Prox1 expression was present exclusively in EGFP+ OHCs (Fig. 3A–A′″).

Figure 3.

Ectopic Notch1 intracellular domain (NICD) causes re-expression of Sox2 and Prox1. A–A′″: All EGFP+ HCs (enhanced green fluorescent protein-positive hair cells) reactivated Sox2 expression, whereas Prox1 was expressed exclusively in OHCs but not in IHCs (arrows in A′″). B–C′: Sox2 (B′) and Prox1 (B″) expression disappeared in EGFP+ HCs (B) with ectopic NICD but without RBP-J, but their expression in supporting cells (SCs) was intact (C,C′). OHC, outer hair cell; IHC, inner hair cell. Scale bar = 20 μm.

We also analyzed Atoh1CreER+; Rosa26-NICDloxp/+; RBP-Jloxp/loxp mice in which NICD was ectopically induced but the key Notch signaling effector RBP-J was deleted. As expected, 276 of 281 (98.2%) EGFP+ cochlear HCs (n = 3) were Sox2-negative and Prox1-negative (Fig. 3B–B″). The remaining 5 (1.8%) EGFP+ HCs expressed Sox2/Prox1 likely because of the occasional mosaic Cre-mediated deletion (i.e., reactivating NICD, but without deleting RBP-J). Cochlear SCs had a normal Sox2 and Prox1 expression pattern at P6 (Fig. 3C–C′), serving as internal positive controls for Sox2 and Prox1 immunostaining. Taken together, these results confirm that ectopic Sox2 and Prox1 expression in HCs are induced by the reactivated RBP-J-dependent Notch1 signaling.

Although HCs expressed Sox2 (or Prox1), they still maintained Myosin VI at P6 (Fig. 4A–A′″), again suggesting that the fate of HCs is maintained even in the presence of ectopic NICD. In addition, when Atoh1CreER+; Rosa26-NICDloxp/+ pups (n = 3) were injected with FM 4-64FX dye (red color) at P6 and analyzed 12 h later, all cochlear HCs, including EGFP+ HCs with ectopic NICD, took up the FM 4-64FX dye (Fig. 4B–B″). These results support that the mechanotransduction (MET) channel of EGFP+ HCs is intact.

Figure 4.

Hair cells with Notch1 intracellular domain (NICD) expression are normal at postnatal day (P) 6. A–A′″: Triple staining of enhanced green fluorescent protein (EGFP), Sox2, and Myosin VI. Images were visualized at confocal YZ plane. Arrows point to the same EGFP+/ Sox2+/ Myosin VI+ hair cell (HC). B–B″: The FM 4-64FX dye was taken up by all HCs, including EGFP+ HCs with ectopic NICD, when injected at P6 and analyzed 12 hr later. Arrows label the same EGFP+ HC. OHCs, outer hair cells; IHCs, inner hair cells. Scale bar = 20 μm.

Lineage Tracing of Cells Experiencing Prox1 Expression During Normal Cochlear Development

Ectopic Prox1 expression was exclusive to OHCs in Atoh1CreER+; Rosa26-NICDloxp/+ mice (Fig. 3). Furthermore, among EGFP+ OHCs, only 79.8% were Prox1+/Sox2+ and the remaining 20.2% were Prox1-negative/Sox2+. We speculated that this exclusive expression pattern of Prox1 in OHCs could be due to either reactivation of the intrinsic program controlling Prox1 expression during cochlear embryonic development or a new, modified program confining ectopic Prox1 to differentiating OHCs (with ectopic NICD in neonatal ages). If the former were true, Prox1 expression would also occur exclusively in OHCs during normal cochlear development; if the latter were true, other new factors would be activated for exclusive expression of ectopic Prox1 in OHCs.

To distinguish between the two scenarios, we first performed lineage tracing analysis of Prox1CreER/+; Rosa26-EYFPloxp/+ embryos at ∼E19 that were treated once with tamoxifen at ∼E16 (the day between onset and end of Prox1 expression in HCs; Bermingham-McDonogh et al.,2006). OHCs and SCs (pillar and Deiters' cells), but not IHCs, were traced with EYFP (Fig. 5). However, a single induction of tamoxifen might be insufficient to trace all Prox1 expressing cells at ∼E16.

Figure 5.

Tamoxifen-dependent lineage tracing of Prox1 expressing cells at embryonic day (E) 16. A,B: Whole-mount image of Myosin VI (A) and enhanced yellow fluorescent protein (EYFP; B) in cochleae from ∼E19 Prox1CreER/+; Rosa26-EYFPloxp/+ embryo treated with tamoxifen once at ∼E16. C–E: High-magnification image of the white rectangular area in (B). The numbered cells (1–3 in C) were also viewed in the confocal YZ plane in (D, E). 1: OHC; 2: OPC; 3: DC. F: Percentage of different EYFP+ cells. OHC, outer hair cell; IHC, inner hair cell; OPC, outer pillar cell; IPC, inner pillar cell; DC, Deiters' cell; TMX, tamoxifen. Scale bars = 200 μm in A,B, 20 μm in C–E.

To label all cochlear cells experiencing Prox1 expression (regardless of its spatial and temporal pattern), we further analyzed another independent line, Prox1CreEGFP/+; Rosa26-EYFPloxp/+ (n = 4) at P23 (Fig. 6A–C″), in which the Prox1 coding region was replaced by the fusion protein CreEGFP, so that no tamoxifen induction was required to trigger Cre-mediated recombination (Srinivasan et al.,2010). Therefore, EYFP+ cells (driven by the Rosa26 promoter after CreEGFP mediated excision) were defined as Prox1-expressing cells during cochlear development. Because Prox1 is turned on before birth in cochlear cells (Bermingham-McDonogh et al.,2006; Fritzsch et al.,2010), all cells experiencing Prox1 expression would have enough time to accumulate EYFP expression by P23. As shown in Figure 6D, among the EYFP+ HCs, ∼96.1% were OHCs and 3.9% were IHCs. After normalizing the number of EYFP+ OHCs (or IHCs) to the total number of OHCs (or IHCs) in the same confocal scanning region, we found that ∼30% of OHCs were EYFP+ and ∼3.8% of IHCs were EYFP+ (Fig. 6E).

Figure 6.

Tamoxifen-independent lineage tracing of Prox1 expressing cells during the entire cochlear development. A: Illustration of the first tamoxifen-independent lineage tracing strategy. B: Double staining of enhanced yellow fluorescent protein (EYFP) and Calbindin (another hair cell [HC] marker) at postnatal day (P) 23 cochlear samples. C–C″: High-magnification image of the white square area in (B). Arrows point to the same EYFP+ IHC and arrow heads point to the same EYFP+ OHC. The HCs without EYFP expression were wrapped by the EYFP+ SCs. D: Distribution of EYFP expression in OHCs and IHCs. E: The percentage of EYFP+ OHCs and IHCs. LER, lesser epithelium ridge; GER, greater epithelium ridge; OHCs, outer hair cells; IHCs, inner hair cells; PCs, pillar cells. Scale bars = 200 μm in B, 20 μm in C″.

As expected, almost all PCs and DCs were EYFP+ (or Prox1+). Intriguingly, cells in the greater epithelium ridge (GER) or lesser epithelium ridge (LER) were also EYFP+, likely due to the high detecting sensitivity with Prox1CreEGFP/+ mice. These results were consistent with the finding of weak Prox1 expression in LER/GER cells in a previous study (Kirjavainen et al.,2008). Taken together, the lineage tracing results support that almost all (∼96.1%) Prox1 expressing HCs are OHCs during cochlear development; NICD reactivation in postnatal OHCs leads to reuse of the intrinsic program controlling Prox1 activation in embryonic cochlear development; and other unknown signaling pathways are responsible for activation of Prox1 in OHCs and/or inhibition of Prox1 in IHCs during normal development.

Cochlear HCs With Ectopic NICD Undergo Normal Maturation at Adult Ages

Next, we investigated whether cochlear HCs could complete their normal terminal differentiation process under conditions of ectopic NICD. As reported for adult retinal cells (Jadhav et al.,2006), EGFP (from NICD-IRES-EGFP) was difficult to visualize in adult cochlear HCs, too (data not shown). Therefore, we bred Atoh1CreER+; Rosa26-NICDloxp/loxp with Rosa26-EYFPloxp/loxp and obtained Atoh1CreER+; Rosa26EYFP/NICD mice as the experimental group, and Rosa26EYFP/NICD littermates were controls. In experimental mice, on tamoxifen induction at P0 and P1, most HCs should express EYFP that was translated directly by means of the efficient Cap-dependent mechanism. We used EYFP to guarantee the occurrence of Cre-mediated recombination in samples analyzed at adult ages, and to identify HCs with ectopic NICD expression at a single-cell resolution.

Of note, in the experimental Atoh1CreER+; Rosa26EYFP/NICD mice, there were two independent floxed “STOP” fragments in the homologous Rosa26 loci for Cre to delete, which is different from previous Atoh1CreER+; Rosa26-NICDloxp/+ model in which only one floxed “STOP” fragment was present. Despite the higher (two vs. one “STOP” fragment to delete) burden for Cre, there were still many EYFP+ HCs (80–90%) along the entire cochleae. This confirms the high efficiency of Atoh1CreER+ and further supports our explanation for why only 15% of HCs were EGFP+ in Atoh1CreER+; Rosa26-NICDloxp/+. Only a fraction of HCs expressing NICD-IRES-EGFP in Atoh1CreER+; Rosa26-NICDloxp/+ mice would have detectable EGFP expression because of the lower efficiency of IRES-mediated than the Cap-dependent translation. Thus, it is likely that only HCs with very high Rosa26 promoter activity were able to drive sufficient expression of EGFP.

Following tamoxifen injection at P0 and P1, Atoh1CreER+; Rosa26EYFP/NICD mice were analyzed at 6 weeks of age (Figs. 7, 8). Similar to the observations at P6 (Fig. 3), ectopic Sox2 (Fig. 7) and Prox1 (Fig. 8) expression were found in adult EYFP+ HCs. Again, ectopic Sox2 expression was found in both IHCs and OHCs but ectopic Prox1 expression was exclusive to OHCs. Triple staining of EYFP, Prox1, and Sox2 was not performed because EYFP endogenous fluorescence (excitation wavelength ∼514 nm) overlapped weakly with the red channel (∼579 nm). Because EYFP and NICD expression might not be coupled perfectly due to the mosaic Cre-mediated recombination in the two Rosa26 alleles, a few Sox2+ or Prox1+ HCs were EYFP-negative (data not shown). Of interest, Sox2+ or Prox1+ HCs always had a much higher expression of EYFP than the neighboring Sox2-negative or Prox1-negative HCs did. This is likely because of the heterogeneity in Rosa26 promoter activity among different HCs, and suggests that ectopic Prox1 and Sox2 expression can only be induced in HCs in which the Rosa26 promoter activity is strong. In other words, NICD can induce Sox2 and Prox1 expression only when its level is above the working threshold.

Figure 7.

Sox2 expression in adult hair cells (HCs) with ectopic Notch1 signaling. A–D: Whole-mount image of Sox2 and enhanced yellow fluorescent protein (EYFP) in cochlear basal turns. (B–D) High-magnification image of the square area in (A). Arrows point to an EYFP and Sox2 double-positive OHC. E–H: Whole-mount image of Sox2 and EYFP in cochlear middle turns. F–H: High-magnification image of the square area in (E). Arrows point to an EYFP and Sox2 double-positive OHC and arrow heads to an EYFP and Sox2 double-positive IHC. I–L: Whole-mount image of Sox2 and EYFP in cochlear apical turns. J–L: High-magnification image of the square area in (I). Arrows point to an EYFP and Sox2 double-positive OHC. OHC, outer hair cell; IHC, inner hair cell; SG, spiral ganglion. 25%, 50% and 75% represent basal, middle and apical parts, respectively, as described in the Experimental Procedures section. Scale bars = 200 μm in A,E,I, 20 μm in B–D,F–H,J–L.

Figure 8.

Prox1 expression in adult outer hair cells with ectopic Notch1 signaling. A–C: Whole-mount image of Prox1 and enhanced yellow fluorescent protein (EYFP) in cochlear basal turns. Arrows point to an EYFP and Prox1 double-positive OHC. D–F: Whole-mount image of Prox1 and EYFP in cochlear middle turns. Arrows point to an EYFP and Prox1 double-positive OHC. G–I: Whole-mount image of Prox1 and EYFP in cochlear apical turns. Arrows point to an EYFP and Prox1 double-positive OHC. Note that in all turns, all EYFP and Prox1 double-positive cells were OHCs. OHC, outer hair cell; IHC, inner hair cell. Scale bar = 20 μm.

All adult HCs had normal morphology and expressed Myosin VIIa (another early HC marker; Fig. 9A), and all OHCs expressed the terminal differentiation marker Prestin (Zheng et al.,2000a; Liberman et al.,2002; Fig. 9B). During cochlear development, Prestin is not expressed in OHCs, especially the apical turns, by P3 (Legendre et al.,2008), when Sox2 had already been reactivated (Fig. 2F–F″). This suggests that reactivation of NICD, Sox2, and Prox1 does not interfere with expression of HC markers at various stages. The perfect arrangement of three rows of OHCs and one row of IHCs further suggested that no ectopic cell proliferation or cell death occurred. Furthermore, to determine whether these HCs maintain the characteristics of mechanotransduction at a single-cell resolution, Atoh1CreER+; Rosa26-NICDloxp/+ mice (n = 3) were injected with tamoxifen at P0 and P1 and with FM 4-64FX dye and furosemide at P30 and cochlear samples were analyzed at P31. The FM 4-64FX dye was taken up by all HCs, including HCs with ectopic NICD (identified by ectopic Sox2 expression; Fig. 9C–E). These results support that the MET channel of adult HCs with ectopic NICD is intact.

Figure 9.

Normal morphology and function of adult hair cells (HCs) with ectopic Notch1 signaling. A: Whole-mount, high-magnification image of Myosin VIIa immunofluorescence in the white rectangular area in the inset. B: Whole-mount, high-magnification image of the immunofluorescence of the OHC marker prestin in the white rectangular area in the inset. C–E: The FM 4-64FX dye was taken up by all HCs, including Sox2+ OHCs (D–D′) and IHCs (E) with ectopic Notch1 signaling at postnatal day (P) 31 when injected at P30. D–D′: Confocal YZ plane view through the dashed line in (C″). E: High-magnification image of the square area in (C″). Arrows in (D-E) label HCs with ectopic NICD. OHC, outer hair cell; IHC, inner hair cell. Scale bars = 20 μm in A,D,E, 10 μm in B, 200 μm in C″.


By expressing NICD in differentiating HCs, we show that ectopic Notch signaling led to re-activation of Sox2 (in both OHCs and IHCs) and Prox1 (only in OHCs). Surprisingly, HCs did not display any detectable phenotypes; they completed their terminal differentiation (summarized model in Fig. 10). These unexpected results suggest that unlike their progenitors at ∼E13 (Dabdoub et al.,2008), differentiating HCs are insensitive to NICD, Sox2, and Prox1 reactivation.

Figure 10.

Model showing how hair cell development is affected by ectopic Notch1 intracellular domain (NICD). Overactivation of NICD in ∼ embryonic day (E) 13 progenitor cells blocks hair cell (HC) development (Dabdoub et al.,2008). Ectopic reactivation of NICD in neonatal HCs at postnatal day (P) 0 and P1 results in re-expression of Sox2 (in both IHC and OHC) and Prox1 (only OHC). Of note, NICD+ HCs have functions similar to those of wild-type HCs. OHC, outer hair cell; IHC, inner hair cell.

Multiple Activities of Notch Signaling During Mouse Cochlear Development

Notch signaling plays dual but contrasting roles during cochlear development (Daudet and Lewis,2005). In the prosensory phase, it promotes formation of prosensory areas in which the progenitors reside (Hartman et al.,2010; Pan et al.,2010). However, Notch signaling might not be necessarily required in specifying prosensory progenitors (Basch et al.,2011; Yamamoto et al.,2011). This is supported by observations that other signaling pathways, fibroblast growth factor receptor (Fgfr; Pirvola et al.,2002; Wright and Mansour,2003; Hayashi et al.,2008b) and Wnt (Stevens et al.,2003; Ohyama et al.,2006; Jayasena et al.,2008) signals, can contribute to specification of prosensory regions. In the subsequent phase when progenitors undergo cell fate specification, the progenitors committed to HCs or new differentiating HCs (Notch-signaling cells) prevent their neighboring progenitor cells (Notch-receiving cells) from adopting the HC fate. This occurs through the well-known lateral inhibition effect of Notch signaling (Lanford et al.,1999; Kelley,2006). Briefly, Notch signaling activates its target genes (i.e., Hes or Hesr family genes; Li et al.,2008; Tateya et al.,2011) and prevents Atoh1 expression in the Notch-receiving cells that will commit to SCs (Cotanche and Kaiser,2010). However, it is not clear how Notch-signaling cells (HC lineage) and Notch-receiving cells (SC lineage) are gradually distinguished from the common undifferentiated progenitor pool.

Consistent with the lateral inhibition effect of Notch1, loss of Notch signaling in the SC lineage leads to conversion of SCs into HCs (Lanford et al.,1999; Kiernan et al.,2005; Yamamoto et al.,2006; Doetzlhofer et al.,2009). In addition, neonatal SCs change to HCs when cultured in vitro (White et al.,2006). These observations support that SCs at late embryonic ages or neonatal ages are plastic and their SC fate is not fully established. Alternatively, these SCs are indeed undifferentiated progenitors; however, currently no markers are available to distinguish progenitor cells and differentiating SCs. Therefore, we predicted that in the presence of ectopic NICD in differentiating HCs, full maturation of these HCs would be affected or these HCs would be even converted into SCs, similar to the reversible cell fates reported in different pancreatic (Thorel et al.,2010), and, recently, testis cell types (Matson et al.,2011). However, our study showed that HCs were insensitive to ectopic NICD and the resulting Sox2 and/or Prox1 expression, which suggests that HC fate and its normal developmental program are irreversible after birth. Because NICD overactivation prevents cochlear progenitors (∼E13) from becoming HCs (Dabdoub et al.,2008), our results are consistent with the idea that developmental stages determine potential cell fate change (Eminli et al.,2009).

Genetic Interaction Among NICD, Sox2, and Prox1 in the Cochlea

Gain- and loss-of-function in vitro studies support the signaling hierarchy from NICD to Sox2 to Prox1 (Dabdoub et al.,2008; Jeon et al.,2011). Consistent with these results, our in vivo study also demonstrated that NICD was sufficient to reactivate Sox2 expression but not to induce Prox1, because ectopic Prox1 expression was exclusive to OHCs. Also, ectopic Sox2 appears approximately 1.5 days earlier than that of ectopic Prox1. It suggests that Sox2 is required to turn on Prox1 in neonatal HCs, as the case in embryonic cochlear development in which Prox1 expression is dependent on Sox2 (Dabdoub et al.,2008). Our results also indicate that decrease of Sox2 and Prox1 in HC lineage in the normal cochlear development is caused by the reduction of Notch (summarized model in Fig. 10).

In addition, RBP-J was required for NICD to reactivate Sox2 and Prox1 in neonatal HCs (Fig. 3), supporting that Notch signaling in the cochlea is dependent on RBP-J (Kopan and Ilagan,2009). Last, it has been suggested that Notch and/or Fgfr signaling maintains Prox1 expression in neonatal SCs (Doetzlhofer et al.,2009). However, the factors or signaling pathways that maintain Sox2 expression in SCs in neonatal and adult ages remain to be elucidated because Notch is no longer active after neonatal stages (Murata et al.,2006).

Lineage Tracing Analysis of Prox1 Expressing Cells During Cochlear Development

To gain insights into the exclusive ectopic Prox1 expression in OHCs when NICD was reactivated, we used Prox1CreER/+; Rosa26-EYFPloxp/+ and Prox1CreEGFP/+; Rosa26-EYFPloxp/+ models for lineage tracing of cochlear cells experiencing Prox1 expression during development, irrespective of their different temporal and spatial patterns. Because it has a constitutive Cre expression, many more HCs were labeled with EYFP in Prox1CreEGFP/+; Rosa26-EYFPloxp/+ mice (among which 96.1% were OHCs and 3.9% were IHCs) than in Prox1CreER/+; Rosa26-EYFPloxp/+ mice. This observation supports that Prox1 is primarily expressed in OHCs during development, which is controlled by yet-unknown intrinsic developmental programs or mechanisms. However, Prox1 does not seem to be required to bias OHC fate commitment (Fritzsch et al.,2010). Indeed, it is suggested that Fgf20-mediated Fgfr signals are required to control development of the cochlear lateral cell populations including OHCs (Hayashi et al.,2008b; Huh et al.,2012).

Why did we not observe IHCs (even at very limited numbers) with ectopic Prox1, given that Prox1 lineage tracing study revealed that a small fraction (3.9%) of IHCs also experienced Prox1 expression? We proposed the following explanations: (1) it is possible that the small fraction of IHCs was not included in the EGFP+ HCs, which only represented ∼15% of the total HCs (Fig. 2D); (2) it is possible that reactivation of NICD in neonatal IHCs turn on additional targets that could repress Prox1; (3) it is also possible that the Prox1 locus is permanently silenced (i.e., by epigenetic mechanisms) in neonatal IHCs.

Why only 79.8% (not 100% as Sox2+) of EGFP+ OHCs were Prox1+ in Atoh1CreER+; Rosa26-NICDloxp/+ mice might be explained by our lineage tracing results that only 30% of OHCs expressed Prox1 during normal cochlear development (Fig. 6E). In addition, 79.8% of EGFP+ OHCs in Atoh1CreER+; Rosa26-NICDloxp/+ mice were Prox1+, which is substantially more than the expected 30% (percentage of OHCs experiencing Prox1 expression during normal development). It suggests that OHC subpopulations that express Prox1 during normal development were somehow enriched in the EGFP+ OHCs in the Atoh1CreER+; Rosa26-NICDloxp/+ mice after tamoxifen injection at P0 and P1.

Effects of Ectopic Prox1 in Neonatal Cochlear HCs

A previous in vitro mouse cochlear explant study showed that ectopic Prox1 expression by virus caused death of HCs (more at ∼E15 than at ∼P1; Kirjavainen et al.,2008). Our in vivo data show that HCs survived to adult ages in the presence of ectopic Prox1. Besides possible differences between the in vivo and in vitro environments, the contrasting results might be due to the different levels of Prox1 used in the studies. The virus-mediated approach may lead to a higher level of Prox1 expression in vitro than in our study in vivo, because the ectopic NICD in our study was driven by the Rosa26 promoter, which is normally not strong.

Implications on Mammalian Hair Cell Regeneration

Nonmammalian vertebrates such as birds can regenerate HCs after damage by means of the conversion of SCs into HCs (Stone and Cotanche,2007). Overexpression of Atoh1 may cause conversion of mammalian SC lineages to HCs (Izumikawa et al.,2005; Gubbels et al.,2008). However, after SCs are converted into HC lineages, it is not known whether some SC markers (i.e., Sox2 and Prox1) continue to be expressed in these newly regenerated HCs.

Our findings show that reactivation of NICD, Sox2, and Prox1 in differentiating HCs did not block the differentiation process toward the fully mature state. It supports the notion that once HC fate is committed, NICD, Sox2, and Prox1 reactivation does not have significant impact on differentiating HCs. In agreement with this, some adult or fully mature mouse utricle HCs have been shown to still maintain the expression of Sox2 (Hume et al.,2007). Lineage tracing studies show that Atoh1-expressing cells can develop into HCs and SCs, whereas Gfi1 expressing cells exclusively differentiate into HCs in the cochlea (Matei et al.,2005; Yang et al.,2010a,b). Therefore, in our opinion, HC fate commitment occurs first at ∼E15.5 at the middle-basal turns when Gfi1 expression starts, and this wave further migrates to the basal and apical turns, after which other HC markers such as Myosin VI or Myosin VIIa are turned on.

Thus, with the aim of regenerating fully differentiated or functional HCs from SCs, is it necessary to inactivate SC markers (i.e., Sox2 and Prox1) and overactivate Atoh1 simultaneously? Our results suggest that Sox2 and Prox1 will not interrupt the program of HC differentiation, assuming that newly generated HCs (from SCs) follow the same developmental program as wild-type HCs.



Atoh1CreER+, Prox1CreER/+, and Prox1CreEGFP/+ mouse lines were used as described previously (Chow et al.,2006; Srinivasan et al.,2007,2010; Weber et al.,2008). Rosa26-EYFPloxp/loxp (stock number: 006148) and Rosa26-NICDloxp/loxp (stock number: 008159) lines were purchased from The Jackson Laboratory. The RBP-Jloxp/loxp mice (Han et al.,2002) were from RIKEN BioResource Center in Japan (RBRC01071). Mice were crossed at 5 PM and the next morning was designated as E0.5 when vaginal plugs were found. Tamoxifen (Sigma, T5648) was first resolved in coin oil (Sigma, C8267) at 37°C. Pregnant female mice were given tamoxifen (intraperitoneal, 100 μg/g body weight; Rawlins et al.,2009) once when embryos were at ∼E16. Neonatal mice were given tamoxifen (3 mg/40 g body weight) at P0 and P1. All animal work conducted for this study was approved by the Institutional Animal Care and Use Committee at St. Jude Children's Research Hospital and performed according to NIH guidelines.

Histology and Immunofluorescence

Samples of the inner ear were processed by our routine protocols described previously (Liu et al.,2010; Yu et al.,2010). All images were examined using a Zeiss LSM 700 confocal microscope. The following primary antibodies were used: anti-Myosin-VI (rabbit, 1:200, 25-6791, Proteus Bioscience, Ramona, CA), anti-Myosin-VIIa (rabbit, 1:200, 25-6790, Proteus Bioscience), anti-Prox1 (rabbit, 1:500, AB5475, Millipore, Billerica, MA), anti-GFP (rabbit, 1:50, A-21311, Invitrogen, Carlsbad, CA) or anti-GFP (chicken, 1:1,000, ab13970, Abcam, Cambridge, UK), anti-calbindin (rabbit, 1:500, AB1778, Millipore), anti-Prestin (goat, 1:200, sc-22692, Santa Cruz Biotechnology, Santa Cruz, CA), and anti-Sox2 (goat, 1:1,000, sc-17320, Santa Cruz Biotechnology). The following secondary antibodies were used: donkey anti-rabbit Alexa Fluor 647 (1:1,000, A31573, Invitrogen), donkey anti-chicken DyLt 488 (1:200, 703-486-155, Jackson ImmunoResearch, West Grove, PA), donkey anti-goat Alexa Fluor 568 (1:1,000, A11057, Invitrogen), goat anti-rabbit Alexa Fluor 568 (1:1,000, A11036, Invitrogen), goat anti-mouse Alexa Fluor 647 (1:1,000, A21236, Invitrogen) and goat anti-chicken Alexa Fluor 488 (1:1,000, A11039, Invitrogen).

Cell Counting

The entire cochlea was carefully divided into basal, middle, and apical turns. With the preliminary low-magnification confocal image, the length of each turn was first measured by drawing a line in the middle of the OHCs and IHCs. By defining the basal hook part as 0% and the most apical portion as 100%, the 25%, 50%, and 75% areas were used to represent basal, middle and apical turns, respectively. Confocal Z stack scanning was then performed at 1-μm intervals.

Mechanosensory Transduction (MET) Channel Measurement

The red fluorescent FM 4-64FX dye (F34653) was purchased from Invitrogen. FX means that this analog is fixable with aldehyde-based fixatives. The dye has been used in two recent studies to label HCs (Nagiel et al.,2008; Roux et al.,2011). FM 4-64FX dye injection was performed in neonatal mice with the procedures described previously (Meyers et al.,2003). For adult cochleae, we additionally used loop diuretic furosemide (400 mg/kg, Hospira Inc., RL-1206), which in conjunction with kanamycin can damage adult mouse OHCs (Oesterle et al.,2008). Furosemide (intraperitoneal) was given 30 min after the FM 4-64FX injection (5 mg/kg body weight, subcutaneous) to potentially increase the accessibility of FM 4-64FX to adult cochlea. In addition, we tried FM 4-64FX injection alone to adult control and experimental group mice and FM 4-64FX dye were absent in HCs. It suggests that adult brain barrier or blood-cochlea barrier prevents FM 4-64FX dye from penetrating into cochlear duct.

Statistical Analyses

Cell counts were compared by a one-way ANOVA, followed by a Student's t-test with a Bonferroni correction. GraphPad Prism 5.0 was used for all statistical analyses.


We thank J. Woods, G. Redd, C. Davis, and D. Wash for determining the embryonic ages and S. Connell, J. Peters, and Y. Ouyang for help in confocal imaging. The RBP-Jloxp/loxp mice were kindly provided by Dr. Tasuku Honjo through RIKEN. This work was supported in part by Travel Awards from Academic Programs of St. Jude Children's Research Hospital, University of Tennessee Health Science Center, and Society of Developmental Biology (Z.L). J.Z. is a recipient of The Hartwell Individual Biomedical Research Award.