Tympanic border cells are Wnt-responsive and can act as progenitors for postnatal mouse cochlear cells

The integrity of hair cells and supporting cells in the mammalian organ of Corti is required for hearing function. Because the postnatal sensory organ lacks the capacity to repair or regenerate, disorders causing cell loss in the cochlear sensory epithelium (SE) result in permanent hearing loss. In mice, cell proliferation in the developing organ of Corti terminates on embryonic day 14 (Ruben, 1967). Although various sensory and non-sensory cell types continue to mature until hearing function begins in the second postnatal week, these cells remain mitotically inactive. Recent evidence suggests, however, that the postnatal mammalian cochlea contains a relatively uncharacterized population of cells that retain a transient proliferative potential, as measured in vitro by cell-derived floating spheres (Malgrange et al., 2002; Oshima et al., 2007; Savary et al., 2008; Shi et al., 2012; Zhang et al., 2007) or by culturing the cells on specific feeder cells (Chai et al., 2012; Sinkkonen et al., 2011; White et al., 2006). Cultured cochlear cell-derived spheres are thought to be derived from stem/progenitor cells, whereas the sphere-derived cells can differentiate in vitro into cells exhibiting hair cell and supporting cell phenotypes (Oshima et al., 2007). However, the proliferative potential of this endogenous pool of cochlear progenitor cells rapidly declines during the first 3 postnatal weeks (Oshima et al., 2007; White et al., 2006). At present, there is limited understanding of both the origin of these cochlear progenitor cells and the mechanisms that regulate their proliferative capacity, although experimental evidence suggests that a subpopulation of cochlear supporting cells behave as progenitor cells in vitro (Sinkkonen et al., 2011; White et al., 2006).

The Wnt/β-catenin signaling pathway is essential in maintaining homeostasis of many tissues (Logan and Nusse, 2004). Active Wnt signaling marks endogenous stem cells in the gastrointestinal system (Barker et al., 2007; Ootani et al., 2009), integumentary system (Jaks et al., 2008) and the mammary gland (Zeng and Nusse, 2010). Expression of Axin2, a downstream target and feedback inhibitor of the Wnt pathway, reliably reports active canonical Wnt signaling in many tissues (Jho et al., 2002; Lustig et al., 2002). By manipulating the Wnt pathway with exogenous agonists, several investigators have reported proliferation and expansion of resident stem/progenitor cell populations (Kalani et al., 2008; Ootani et al., 2009; Zeng and Nusse, 2010). Although components of the Wnt pathway are expressed in the mammalian postnatal cochlea (Daudet et al., 2002; Shah et al., 2009), their roles in mediating Wnt signaling in inner ear progenitor cells are incompletely understood. Here, we tested whether active canonical Wnt signaling marks and promotes proliferation of cochlear progenitor cells. We found that the Wnt target gene Axin2 marks a previously poorly characterized cochlear cell population: tympanic border cells (TBCs). Both in vivo and in vitro, TBCs retain proliferative capacity, which is promoted by Wnt activation and suppressed by Wnt inhibition. By lineage tracing and culturing sorted cells, we found that Axin2-positive TBCs are able to generate new hair cells and supporting cells in vivo and in vitro.

Wild-type and Axin2^lacZ/+ mice (Jho et al., 2002; Lustig et al., 2002) in CD1 background, Pax2-Cre mice (provided by A. Groves, Baylor College of Medicine, Houston, TX, USA) (Ohyama and Groves, 2004), Actin-GFP mice (stock #007075) (Okabe et al., 1997), Actin-DsRed mice (stock #006051) (Vintersten et al., 2004), R26R^mTmG mice (stock #007576) (Muzumdar et al., 2007) and R26R^tdTomato mice (stock #007914) (Madisen et al., 2010) (all from the Jackson Laboratory); and Axin2^CreERT2 mice (van Amerongen et al., 2012) were used. For lineage tracing, pups were injected intraperitoneally with tamoxifen (0.05-1.00 mg/25 g dissolved in corn oil) (Sigma). Intraperitoneal injection of EdU (50 mg/kg; Invitrogen) was performed once per day for 2 days or twice per day for 3 days. The latter regimen, when combined with tamoxifen, caused ∼33% lethality. Cochleae were processed for cryosectioning and immunostaining as described below, and EdU detection was performed per product protocol using an Alexa Fluor 555 Imaging Kit. All protocols were approved by the Animal Care and Use Committee of Stanford University School of Medicine.

Tissues were fixed with 4% paraformaldehyde (Electron Microscopy Services) in phosphate-buffered solution (PBS, pH 7.4) for 30 minutes on ice and subsequently washed with 2 mM MgCl₂ (in PBS) before incubation with X-gal reagents at 37°C for 35 minutes. For cryosectioning, cochleae were similarly fixed and stained, then treated in a sucrose gradient (10-30% in PBS). Tissues were serially treated with sucrose/OCT compound (Sakura Finetek) mixture (1:1, 3:7, then 0:1) in a vacuum chamber for 1 hour at room temperature. Tissues were then sectioned at 10 μm and processed for immunohistochemistry. Decalcification with EDTA (0.5 M in PBS) was performed for cochleae from mice aged P7 or older.

As previously described (Jan et al., 2011), cochleae from P0-P2 Axin2^lacZ/+ mice were isolated, with the stria vascularis and spiral ganglia removed before incubation in 0.125% trypsin (Invitrogen; in PBS for 8 minutes) and then in trypsin inhibitor/DNase1 cocktail (1:1; 10 mg/ml; Worthington Biochem). Following trituration, cells were passed through a 40 μm filter and labeled with 3-carboxyumbelliferyl β-D-galactopyranoside (CUG, 1:50 for 25 minutes; Marker Gene Technologies) and propidium iodide (1 μg/ml; Sigma). Wild-type cochleae were used to determine background labeling levels in each sort. Using a BD Aria FACS cytometer (BD Biosciences), we consistently achieved over 90% cell viability and over 93% purity for sorted cells as measured via re-sort analysis. For staining, sorted cells were plated on fibronectin (Sigma)-coated slides (2 hours at room temperature) before fixation and immunohistochemistry.

Total RNA isolation was carried out using a RNeasy Mini Extraction kit (Qiagen), followed by cDNA synthesis using SuperScript III First-Strand Synthesis System kits (Invitrogen). Primer pairs were designed using the online Primer3 software (http://frodo.wi.mit.edu/primer3/) as follows: Sox2 forward, 5′-ATGAACGGCTGGAGCAACGGCA-3′; Sox2 reverse, 5′-TCACATGTGCGACAGGGGCAGT-3′; Axin2 forward, 5′-ATGTTAGAGAGTGAGCGGCAGA-3′; Axin2 reverse, 5′-CTTCAGCATCCTCCTGTATGGA-3′; Sp5 forward, 5′-CCGTCGTACCCTTACGAGTTCT-3′; Sp5 reverse, 5′-ATCTGGCTCTGGTACTGTGCAA-3′; p27^Kip1 forward, 5′-AAGCACTGCAGAGACATGGAAG-3′; p27^Kip1 reverse, 5′-GTAGAAGAATCGTCGGTTGCAG-3′; Brn3.1 forward, 5′-ACCCAAATTCTCCAGCCTACAC-3′; Brn3.1 reverse, 5′-GGCGAGATGTGCTCAAGTAAGT-3′; Prestin forward, 5′-TACCTCACGGAGCCGCTGGT-3′; Prestin reverse, 5′-GCAGTAATCAGTCCGTAGTCC-3′; β-actin forward, 5′-ACGGCCAGGTCATCACTATTG-3′; β-actin reverse, 5′-AGGGGCCGGACTCATCGTA-3′. qPCR reactions were performed with SYBR Green PCR Master Mix on a 7900HT-Fast Real time PCR system (both Applied Biosystems). The ΔΔC_T method with β-actin as the endogenous reference was used (Livak and Schmittgen, 2001). Reactions were carried out in triplicate.

Feeder cells were collected by isolating embryonic 18-day-old chicken utricles and removing the SE via thermolysin treatment (0.5 mg/ml in HBSS; Sigma) (Oshima et al., 2010; Warchol, 1999). Dissociated cells were cultured in DMEM/F12 with 10% FBS and propagated cells were grown to 90% confluence and pre-treated with mitomycin C (10 μg/ml, Sigma) prior to use as feeder cells. Adding 100 μl of 70 sorted cells/μl (i.e. 7000 cells/well) onto a 0.97 cm² surface area was determined to foster clonal colony formation. For colony expansion assays, Actin-DsRed-positive cells were cultured on feeder cells and colonies were quantified after 7 days and then passaged (0.125% trypsin for 3 minutes at 37°C). Quantification of propidium iodide-labeled cells found ∼70% cell death with this passaging protocol.

Cochleae cultured in defined conditions were lysed and homogenized. A BCA Protein Assay kit was used to determine protein levels and a FluoReporter lacZ/Galactosidase Quantification kit (both Molecular Probes) was used to determine β-galactosidase activity. Manufacturer’s protocols were followed.

Tissues were fixed for 30-60 minutes in 4% paraformaldehyde (in PBS, pH 7.4) prior to incubation with primary antibodies overnight at 4°C. We used antibodies against the following markers: myosin 7a (1:400; Proteus Bioscience), Ki67 (1:400; ThermoScientific), phosphohistone 3 (1:400; Millipore), pancytokeratin (1:200; Sigma), parvalbumin (1:1000; Sigma), calretinin (1:1000; Millipore), fibronectin (1:400; BD Biosciences), β-galactosidase (1:10,000; Promega), jagged 1 (1:800; Santa Cruz Biotechnology), Sox2 (1:400; Santa Cruz Biotechnology), espin (1:1000; a gift from A. Hudspeth, Rockefeller University, New York, NY, USA), Prox1 (1:1000; Millipore), E-cadherin, vimentin (both 1:400; Sigma), p27^Kip1 (1:400; Fisher Scientific), GFP (1:1000; Abcam) and Brn4 (1:500; a gift from B. Crenshaw, University of Pennsylvania, Philadelphia, PA, USA). The secondary antibodies were conjugated with FITC, TRITC, or Cy5 (1:500; Invitrogen). Images were acquired using epifluorescent or confocal microscopy (Axioplan 2, Zeiss, Germany) and analyzed with Photoshop CS4 (Adobe Systems). Three-dimensional reconstruction was performed using Volocity software (v5.3.0; Improvision).

Cochleae were cultured as previously described (Chai et al., 2011). Briefly, organs were cultured in growth factor-enriched, serum-free media (GM) consisting of DMEM/F12, N2 (1:100), B27 (1:50; all from Invitrogen), bFGF (1 ng/ml), IGF-1 (50 ng/ml), EGF (20 ng/ml), heparan sulfate (50 ng/ml) and ampicillin (50 μg/ml; all from Sigma) (Oshima et al., 2007) in four-well Petri dishes (Greiner Bio-one) with media replenished every 1-2 days. R-spondin 1 (1 μg/ml; R&D Systems), purified Wnt3a (200 ng/ml), Fz8CRD (25 μg/ml) and EdU (25 ng/ml; Invitrogen) were added. The synthesis and purification of Wnt3a and Fz8CRD have been previously described (DeAlmeida et al., 2007; Willert et al., 2003).

Statistical analyses were conducted using Excel (Microsoft) and Origin softwares (OriginLab). Two-tailed, unpaired Student’s t-test and one-way ANOVA were used to determine statistical significance. P<0.05 was considered significant.

We used an Axin2^lacZ/+ reporter mouse (Jho et al., 2002; Lustig et al., 2002) to detect active canonical Wnt signaling in the postnatal cochlea. As a feedback inhibitor, Axin2 is expressed in cells with active Wnt signaling (Jho et al., 2002). We established that Axin2^lacZ/+ mice had no detectable phenotypes in the auditory system, as the mice had normal cochlear morphology and function (Fig. 1A; supplementary material Fig. S1A). Analysis of cryosections and whole mounts of neonatal cochlea revealed that TBCs underneath the SE on the basilar membrane express LacZ (Fig. 1A,B; supplementary material Fig. S1E). In situ hybridization demonstrated robust Axin2 expression in the TBCs (supplementary material Fig. S1B) (Chai et al., 2011), corroborating the Axin2^lacZ reporter gene activity data. TBCs extend from the lateral cochlear wall to the habenula perforata (Fig. 1B; supplementary material Fig. S1C,D). To further characterize the TBCs, we examined expression of markers of proliferation, hair cells, supporting cells, epithelial cells and mesenchymal cells. In contrast to the P1 SE cells, TBCs were proliferative (Ki67 positive) and did not express markers of hair cells (myosin 7a) or supporting cells (jagged 1 and Sox2) (Fig. 1C-E). Furthermore, unlike the SE cells, which expressed E-cadherin and cytokeratin, TBCs were devoid of these epithelial cell markers (Fig. 1F,G). The mesenchymal markers vimentin and fibronectin were also absent in the TBCs (Fig. 1H,I), but they expressed the transcription factor Brn4 at P0 (supplementary material Fig. S2A-C) (Ahn et al., 2009; Phippard et al., 1999). In combination, an absence of markers seen in mature cells of the cochlear SE and the expression of proliferative markers led us to conclude that TBCs are poorly differentiated, actively cycling and distinct from the Axin2-negative mitotically inactive SE.

To investigate whether there is a lineage relationship between the SE cells and TBCs, we first examined cochleae from the Pax2-Cre;R26R^mTmG/+ (Rosa26-mTmG) transgenic mice (Muzumdar et al., 2007; Ohyama and Groves, 2004). Pax2 is expressed in the otic placode and developing otocyst and Pax2-positive cells have been shown to contribute to the cochlear ductal epithelium (Basch et al., 2011; Ohyama and Groves, 2004). However, their relationship with TBCs is unclear. Crossing the Pax2-Cre mouse strain with the R26R^mTmG double reporter strain allowed us to examine the identity of cells derived from Pax2 expressing cells. The presence of Cre recombinase initiates expression of mGFP reporter while switching off the mTomato signal. In the P1 cochlea, mGFP-positive cells occupied the majority of the SE, including hair cells and supporting cells, whereas TBCs were not labeled (Fig. 1J; supplementary material Fig. S1F; Table 1). In contradistinction, Brn4 expression is undetectable from the sensory epithelium and robust in the periotic mesenchymal cells in the embryonic period, and was transiently present in TBCs in the early postnatal period (supplementary material Fig. S2A-C) (Ahn et al., 2009; Phippard et al., 1999). We performed lineage tracing of Brn4-expressing cells using the B4OE-Cre;R26R^tdTomato/+ mouse strain, and found a subset of TBCs traced in the P1 cochleae (not shown). These results show that TBCs originate from Brn4-expressing cells outside the cochlear ductal epithelium.

Because the SE continues to mature during the first 3 weeks of postnatal development, we further investigated TBCs during this period. When the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) was administered for 2 days, robust label uptake was seen in TBCs in the P3 wild-type cochleae (6.8±1.2 EdU-positive TBCs per turn), less frequent uptake in the P9 cochleae (1.8±0.9) and there was rare uptake at P23 (0.7±1.2); at all three ages, there was no uptake into the SE cells (Fig. 2A-D). These data show that TBCs proliferate until the second week of postnatal development, in contrast to the quiescent SE cells. Interestingly, the number of TBCs declined during the first 3 postnatal weeks, and they cease to proliferate between the first and second postnatal week, while the anatomy of this region gradually transitions from a three- to four-cell layered structure at P3 to a one- to two-cell layer by the second postnatal week (Fig. 2A-C,E).

Concomitant with the decline in proliferation and quantity of TBCs, mRNA expression levels of the Wnt target genes Axin2 and Sp5 in the cochlea decreased, whereas expression of the outer hair cell gene Prestin (Slc26a5) was elevated. Decreased expression of Sox2, a supporting cell marker, correlated with the natural maturation of the organ of Corti (Fig. 2F). The hair cell-specific gene Brn3.1 is important for hair cell specification (Erkman et al., 1996) and its expression levels did not change significantly. In Axin2^lacZ/+ cochleae, LacZ was expressed at P3 and declined over the subsequent 2-3 weeks (Fig. 2G-J; supplementary material Fig. S1C,D).

These experiments show that TBCs beneath the organ of Corti display a robust but transient proliferative potential during the first postnatal week in vivo, after which proliferation rapidly decreases. This decline in proliferation is accompanied by decreased expression levels of Wnt target genes as well as increased expression of mature cell markers. These observations led to the hypothesis that high levels of Wnt signaling might be indicative of progenitor cell features, and we therefore investigated whether Axin2-positive cells can behave as progenitor cells for the SE.

The architecture of the organ of Corti is dynamic during the first 3 postnatal weeks, and SE cell types are distinguished by their anatomic locations and specific markers (Fig. 3A-C). Using lineage-tracing experiments, we tested the hypothesis that TBCs can act as progenitor cells for the cochlear SE in vivo during this period. We generated an Axin2^CreERT2 mouse strain (van Amerongen et al., 2012) and crossed it with the R26R^mTmG reporter strain to examine cells produced by Axin2-positive TBCs. A single injection of tamoxifen on P1 or P3 activated Axin2-driven Cre recombinase to initiate mGFP reporter activity while switching off the mTomato signal 2 days post injection (DPI) (Fig. 3D; supplementary material Fig. S3A,B; Movies 1, 2). mGFP expression was first observed in the TBCs at 2 DPI and was not detected in cochleae of mice that received drug vehicle alone (Fig. 3D,G). At 5 and 7 DPI, mGFP-positive cells occupied the SE and constituted distinct supporting cell types, which are specified by their anatomical locations (Fig. 3E,H,I; supplementary material Fig. S3C-F). No apical-basal gradient of traced cells was observed (not shown). Quantitative analyses showed that the number of mGFP-positive cells within the population of sensory hair cells, supporting cells, and greater and lesser epithelial ridges (GER and LER) significantly increased from P3 to P8 (P<0.05, Table 1). These cells remained integrated in the P15 cochleae (at 12 DPI) and included Deiters’ and pillar cells (Fig. 3F), which are defined by the expression of Sox2, Prox1, E-cadherin and jagged 1 (Fig. 1D-F; supplementary material Fig. S6F). mGFP-positive cells were also seen among Hensen’s (Sox2 and E-cadherin positive) (Fig. 3J; supplementary material Fig. S3C,H,I), Claudius, Boettcher (both E-cadherin-positive) (Fig. 3I,J; supplementary material Fig. S3D) and GER cells (Sox2, E-cadherin and jagged 1 positive) (Fig. 3E,F,H-L; supplementary material Fig. S3C-F,H,I). These findings suggest that TBCs can migrate into the organ of Corti and adjacent tissues and differentiate into multiple cochlear supporting cell types. On occasion, we found outer and inner hair cells that were mGFP positive (Fig. 3J-L; supplementary material Fig. S3E,F; Movies 1, 2). Hair cells are distinguished by their anatomical locations, expression of myosin 7a and stereociliary hair bundles on their apical surfaces (Fig. 3J-L; supplementary material Fig. S3E,F; Movies 1, 2). Outer hair cell stereocilia are organized in a V-shaped pattern, which is observed in mGFP-positive outer hair cells (Fig. 3L; supplementary material Movies 1, 2). The finding that both inner and outer hair cells were mGFP-positive further supports the idea that TBCs have the potential to generate multiple cell types in the SE.

To examine the possibility that delayed Cre activity could lead to reporter activity within the SE, we administered tamoxifen at P8 and found mGFP expression limited to TBCs 2 DPI (supplementary material Fig. S3J), confirming tight regulation of the Axin2^CreERT2 activity. Administration of EdU (50 mg/kg twice daily, P0-2) revealed EdU-positive cells among the TBCs (15.5±3.8%), but none in the organ of Corti at P3 (Fig. 3M; Table 2). Among cells in the lesser (LER) and greater epithelial ridges (GER), 5.0±0.5% and 0.02±0.02% were EdU positive, respectively (n=3). P8 and P15 cochleae revealed that the EdU-labeled cells occasionally contributed to organ of Corti cells (Fig. 3N,O). At P8, EdU labeled significantly more cells: 29.1±6.3% TBCs, 8.6±1.5% LER cells and 2.0±0.5% GER cells [P<0.001 for all (n=4), Table 2]. Among four cochleae, we detected three EdU-positive cells in the organ of Corti. Concurrent EdU and tracing experiments identified occasional EdU, mGFP double-positive SE cells at P8, but also untraced EdU-positive cells within the SE (supplementary material Fig. S3M-O). Together, these data suggest that Axin2-positive TBCs are multipotent progenitor cells in the postnatal mouse cochlea.

To test the in vitro potential of Axin2-positive TBCs, we used flow cytometry to isolate cochlear Axin2-expressing cells for further analyses (Jan et al., 2011). Using cochleae from P0-2 Axin2^lacZ/+ mice, we labeled and isolated the most fluorescent cells (top 16.1±0.6%), expressing high levels of the Wnt target genes Axin2 and Sp5, and the fewest fluorescent cells (bottom 23.2±2.3%), which expressed hair cell (Brn3.1) and supporting cell (p27^Kip1 and Sox2) genes (Fig. 4A,B,H; supplementary material Fig. S4A). We designed parameters to isolate the least fluorescent cells to minimize contamination and allow comparisons between the two groups based on the levels of Axin2 expression. These two groups are henceforth referred to as Axin2^hi and Axin2^lo cells, respectively. To confirm cell purity, we immunostained cells immediately post-sort and found that the Axin2^hi cells contained over 93% β-galactosidase-positive cells, 0% myosin 7a-positive hair cells, 0% Prox1-positive supporting cells, 75% Brn4-positive cells and <0.1% other supporting cell types (Sox2-, Jag1- or p27^Kip1-positive; n=3 sorts with 2500-5000 DAPI-positive cells analyzed) (Fig. 4C-G; supplementary material Fig. S4B-F).

We first compared the proliferative capacity of Axin2^hi and Axin2^lo cells using a colony formation assay: cells were seeded in serum-free, N2/B27-supplemented media onto feeder cells harvested from embryonic chicken utricles (supplementary material Fig. S4G). The use of embryonic otic mesenchymal tissues and deprivation of growth factors has been shown to promote differentiation of inner ear progenitor cells (Chai et al., 2012; Doetzlhofer et al., 2004; Oshima et al., 2010; Sinkkonen et al., 2011; White et al., 2006). After 10 days in vitro, all newly generated colonies were immunopositive for the epithelial marker cytokeratin. Moreover, Axin2^hi cells formed fourfold more colonies than the Axin2^lo cells (88.2±19.4 versus 22.0±6.2 colonies per 7000 plated cells, P<0.0001) (Fig. 4I,J). To determine whether these colonies arose from individual cells, Axin2^hi cells were isolated from cochleae of Axin2^lacZ/+ mice that were transgenic for GFP driven by the ubiquitous Actin promoter and mixed (1:1) with those from GFP-negative Axin2^lacZ/+ cochleae (supplementary material Fig. S4G). When this mixture of cells was cultured in the above conditions, we found that 94% of the colonies were monochromatic and thus clonal (Fig. 4K-M). These data demonstrate that isolated TBCs from the postnatal cochlea have colony-forming capacity.

The cochlear SE comprises two main cell types: supporting cells and hair cells. After 10 days, colonies derived from the Axin2^hi cells uniformly expressed E-cadherin and cytokeratin, both epithelial markers expressed in the SE (Fig. 5A-D). This observation suggests that the isolated TBCs are able to adopt epithelial characteristics in vitro, similar to their behavior in vivo. In the postnatal organ of Corti, Sox2 is a marker of supporting cells and Prox1 is expressed in two subtypes of supporting cells (Deiters’ cells and pillar cells) (Fig. 1E; supplementary material Fig. S6F) (Bermingham-McDonogh et al., 2006; Oesterle et al., 2008). Colonies derived from Axin2^hi cells expressed Sox2 and Prox1 (77.5±3.2% and 18.2±0.6%, respectively) with Prox1-positive cells always co-expressing Sox2 (Fig. 5A-B). In addition, jagged 1 was also highly expressed (74.3±3.5%) in TBC-derived colonies (Fig. 5D). Control experiments ascertain that chicken mesenchymal cells do not express used markers: cytokeratin, Prox1, jagged 1, myosin 7a, Sox2 or E-cadherin (supplementary material Fig. S4H-K). As TBCs do not express Sox2, Prox1, jagged 1, E-cadherin or cytokeratin in vivo (Fig. 1D-G), the finding that their progeny can express these markers in vitro implies that isolated TBCs have an enhanced potential to acquire an epithelial phenotype and display markers that are specific for SE supporting cells.

Within the cochlear SE, hair cells are marked by antibodies to myosin 7a (Hasson et al., 1997) and parvalbumin (Hackney et al., 2005). The apical surfaces of hair cells contain highly ordered stereocilia bundles and cuticular plates, which are enriched in filamentous actin and espin, an actin-bundling protein that is essential for hearing function (Zheng et al., 2000). Myosin 7a-positive hair cell-like cells were infrequently noted among Axin2^hi colonies (6.9±1.4%); when present, they were always juxtaposed to Sox2-positive cells (Fig. 5A) and detected in 9.4±0.9% of Sox2-positive colonies. Although purified Axin2^lo cells initially expressed high levels of the hair cell marker Brn3.1 (Fig. 4B,H), myosin 7a-positive hair cell-like cells were more frequently observed in the Axin2^hi colonies after 10 days in vitro (P<0.005) (Fig. 5F). When EdU was present during the first 3 days of the culture period, some of the hair cell-like cells were EdU labeled, suggesting that Axin2^hi cells generated new sensory cells via mitotic divisions (Fig. 5E; supplementary material Fig. S5A), although most myosin 7a-positive cells were EdU negative (supplementary material Fig. S5B). The hair cell-like cells also co-expressed parvalbumin and exhibited stereocilia-like structures, as indicated by a polarized expression pattern of filamentous actin and espin (Fig. 5G-I). Interestingly, we observed that a subset of cells within each type of colony remained proliferative (9.2 Ki67-positive cells per Axin2^hi colony versus 3.9 per Axin2^lo colony; Fig. 5M). Overall, significantly more Ki67-positive colonies were derived from the Axin2^hi than from the Axin2^lo cell populations (P<0.001) (Fig. 5J-K). Labeling for the M-phase marker phosphohistone 3 further demonstrated that a subset of cells within Axin2^hi colonies was proliferating (Fig. 5L). In summary, these findings demonstrate the heterogeneity of colonies originating from the Axin2-positive TBCs; some progeny differentiated into cell types resembling supporting cells and hair cells of the cochlear SE, whereas others remained proliferative. We next investigated whether proliferation of these colony-forming cells is regulated by Wnt signaling.

Because Axin2 is a marker of active canonical Wnt signaling, we tested whether addition of Wnt proteins would enhance the proliferative capacity of Axin2-positive TBCs. Wnt3a treatment increased the number of Ki67-positive Axin2^hi colonies, whereas addition of the Wnt inhibitor Fz8CRD, a soluble analog of Frizzled receptors that can inhibit the activity of Wnt proteins (DeAlmeida et al., 2007), suppressed the formation of Ki67-positive Axin2^hi colonies (Fig. 6A-E; supplementary material Table S1). Ki67-positive colonies were rare in the Axin2^lo group and their numbers were not affected by Wnt3a/Fz8CRD treatment, suggesting that the Axin2^hi, and not Axin2^lo, cells were competent to respond to Wnt proteins. Individual Axin2^hi colonies, but not Axin2^lo colonies, demonstrated an increase in Ki67-positive cells and in Axin2 expression in response to Wnt3a treatment (Fig. 6A-G; supplementary material Table S1). However, the number of colonies with myosin 7a-positive hair cells did not change significantly with Wnt3a/Fz8CRD treatment (supplementary material Fig. S5C). These data suggest that Wnt proteins promote proliferation and expansion of the Axin2^hi cells. To further examine this possibility, we cultured purified Axin2^hi and Axin2^lo cells from Actin-DsRed-positive Axin2^lacZ/+ cochleae in serum-free N2/B27 media containing Wnt3a proteins without other factors. We measured proliferative capacity of the cells by quantifying DsRed-positive colonies and found that the Axin2^hi cell lineage had a significantly higher colony-forming capacity than Axin2^lo cell lineage (P<0.01, one-way ANOVA). Wnt3a treatment further increased the number of colonies over multiple passages (P<0.05) (Fig. 6H-J). However, Axin2^lo colonies failed to respond to Wnt3a treatment. After two cycles of expansion, the number of Axin2^hi colonies increased 319-fold in comparison with those from the Axin2^lo cell lineage, and this difference further increased to 924-fold with Wnt3a treatment. Similar to their first generation precursors, these expanded colonies from the Axin2^hi cell population displayed an epithelial phenotype, expressed cytokeratin, and expressed markers for hair cells (myosin 7a) and supporting cells (Sox2) (Fig. 6K,L). All Sox2-positive and myosin 7a-positive cells formed within colonies were Actin-DsRed positive, indicating that they are derived from mouse tissues and not chicken mesenchymal cells (supplementary material Fig. S5D,E). With expansion via passaging, the percentage of Sox2-positive colonies (∼60%) was maintained with Wnt3a supplementation (supplementary material Fig. S5F), whereas that of myosin 7a-positive colonies declined (first generation=8.9±0.4%, second generation=5.0±1.2%, third generation=1.2±1.5%; none detected in subsequent generations). These findings lend credence to the notion that Wnt proteins act as growth factors to purified Axin2^hi TBCs from the cochlea.

We further investigated the effects of endogenous Wnt signals on the Axin2-positive TBCs in cochlear explants, where architecture of the organ of Corti is preserved. We manipulated the Wnt pathway with Wnt3a, R-spondin 1 and Fz8CRD. R-spondin 1 is an alternative Wnt agonist that acts synergistically with endogenous Wnt proteins (Kim et al., 2006; Ootani et al., 2009). Wnt3a or R-spondin 1 treatment robustly enhanced TBC proliferation (Fig. 7A-C; supplementary material Fig. S6A-C,F; Movie 3). Quantitative analyses of Ki67- or EdU-positive cells indicated that Wnt activation induced a doubling of the baseline proliferation, and that this mitogenic effect was evident for at least 5 days in vitro (Fig. 7D; supplementary material Table S2). Like isolated Axin2^hi cells, TBCs were competent to respond to exogenous Wnt agonists by showing a marked increase in proliferation, which was not observed in the SE. In addition, proliferation among TBCs was highly dependent on Wnt signaling as Fz8CRD effectively reduced EdU-positive and Ki67-positive cells and suppressed Axin2 expression (Fig. 7E-I; supplementary material Fig. S6D,E; Table S2). We used ELISA to quantify lacZ (Axin2) expression in cochlear cultures and found it to be significantly enhanced by Wnt3a treatment (P<0.01) and inhibited by Fz8CRD treatment (P<0.005) (Fig. 7J). Based on these observations, we conclude that the proliferative capacity of Axin2-positive TBCs is dependent on Wnt signaling.

Sensory hair cells are mechanoreceptors that are required for auditory function; their irreversible loss leads to permanent hearing loss. Previously thought to lack regenerative capacity, the early postnatal cochlea was recently shown to harbor cells with colony-forming capacity and the potential to generate new hair cell-like cells in vitro (Chai et al., 2012; Oshima et al., 2007; Savary et al., 2008; Shi et al., 2012; Sinkkonen et al., 2011; White et al., 2006; Zhang et al., 2007). In defined culture conditions, supporting cells isolated from SE exhibit both of these qualities and therefore are prime candidates as progenitor cells (Sinkkonen et al., 2011; White et al., 2006). In vivo, we and others have found no evidence of label incorporation in the organ of Corti immediately after EdU injection, suggesting that proliferation is either rare or non-existent. Conversely, robust proliferation is seen among TBCs, a distinct, yet poorly characterized, cell population beneath the SE. Also known as mesothelial cells, TBCs were first described in mammals by Claudius (Claudius, 1856), and subsequently in more detail in other species, including humans (Bhatt et al., 2001; Cabezudo, 1978; Keithley et al., 1994). Because of their close association with the basilar membrane, TBCs have been suggested to secrete extracellular matrix proteins (Amma et al., 2003). Others have proposed a biomechanical role for the TBCs, as their number varies along the cat cochlea (Cabezudo, 1978). However, their exact physiological function(s) remain unclear.

Our study provides the first pieces of in vitro and in vivo evidence that Axin2-positive TBCs can behave as progenitor cells in the postnatal cochlea. Wnt/β-catenin signaling is crucial for the maintenance of stem cell niches, thereby promoting stem cell self-renewal in many mammalian systems (Barker et al., 2007; Kalani et al., 2008; Lie et al., 2005; Zeng and Nusse, 2010). The expression of Axin2 is a marker for active Wnt signaling in many cell types (Kalani et al., 2008; Lustig et al., 2002; Zeng and Nusse, 2010). Using Axin2^lacZ mice to isolate Wnt-responsive cells in the postnatal cochlea, we discovered that Axin2-positive TBCs have colony-forming capacity. Furthermore, they can differentiate to multiple cell types, including supporting cells and hair cells both in vitro and in vivo. Therefore, our data suggest that TBCs can act as progenitor cells in the neonatal cochlea.

As demonstrated for somatic stem cells in other organ systems (Barker et al., 2010; Ootani et al., 2009; Willert et al., 2003; Zeng and Nusse, 2010), and recently in the embryonic and postnatal cochlea (Chai et al., 2012; Jacques et al., 2012; Shi et al., 2012), Wnt agonists act as growth factors for TBCs. However, Wnt inhibition decreases the self-renewal capacity of TBCs, but it does not affect the frequency of hair cell differentiation, which may operate in a Wnt-independent manner. Interestingly, we frequently observed supporting cells among the progeny of TBCs in vitro as well as in vivo, suggesting that TBCs more readily differentiate into supporting cells than into sensory hair cells in both microenvironments. The mechanisms regulating this differentiation are currently unknown and warrant further investigation.

Our lineage tracing and thymidine analog labeling results further suggest that the postnatal organ of Corti remains dynamic, contradicting the previous assumption that hair cell formation is complete by E16.5 in mice (Kelley, 2007) and failed detection of proliferated cells in the postnatal organ of Corti (Ruben, 1967). Instead, our findings correspond with data from other mammalian species suggesting that postnatal hair cell addition occurs (Kaltenbach and Falzarano, 1994; Mu et al., 1997). Given that our results are surprising, it is important to point out that the incorporation of proliferating cells into the SE is rare (Table 2). Moreover, detecting proliferating cells resulted from six injections of EdU over a 3-day period, in contrast to fewer rounds of potentially more cytotoxic thymidine analogs used previously by others (Lee et al., 2006; Ruben, 1967). One candidate mechanism for such cell addition is migration of TBCs across the basilar membrane. In the embryonic inner ear, gaps in extracellular matrix proteins may mediate the process of delamination when neuroblasts migrate out of the otic epithelium (Davies, 2011). Moreover, neuroepithelial derivatives migrate into the otic epithelium through gaps in the basal lamina (Freyer et al., 2011). These gaps are characterized by the absence of fibronectin, which is also observed in the basilar membrane in the neonatal cochlea (Fig. 1I). Together, our data support the notion that TBCs may also serve as a pool of progenitors for SE cells in the postnatal cochlea.

Although the current study examines the Axin2-positive TBCs as a single cell population, it is more likely that TBCs are a heterogeneous population that consists of cells with different potentials for proliferation and differentiation. Lineage tracing and thymidine analog labeling experiments showed that the progeny of Axin2-positive TBCs in the P1-3 cochleae appear in the SE in older cochleae, while others remain in the TBC region. Likewise, purified Axin2^hi cells generated colonies with differentiated SE cells and Ki67-positive proliferative cells in vitro. In vivo, the low number of EdU-positive traced cells may have resulted from the stochastic nature of lineage tracing and the limited efficiency of the EdU labeling assay, which may be better addressed by using mosaic analysis with double markers (MADM) in future experiments (Zong et al., 2005).

Our results raise several important questions, including how is cell fate of individual TBCs determined, what is the mechanism underlying the decrease in TBCs during the first 3 weeks of postnatal development, and what are the roles of the Axin2-positive cells that have been demonstrated to surround the embryonic cochlear duct (Chai et al., 2011)? Do TBCs from older animals retain such progenitor cell potentials? It is plausible to consider applying Wnt agonists to TBCs in the adult cochlea in an attempt to rekindle such progenitor cell potentials. Stimulation of progenitor cell capacity in the mature cochlea is of clinical interest, especially in the context of regenerating the damaged organ of Corti. Although it has been reported that TBCs show increased proliferation post-injury through ³H-thymidine uptake in adult mammals (Roberson and Rubel, 1994; Sobkowicz et al., 1992), no regeneration of sensory cells was noted. The current study raises the possibility that enhancing Wnt signaling may promote regeneration of the damaged inner ear, an approach that has shown promise in the context of bone regeneration (Minear et al., 2010).

In summary, we have characterized tympanic border cells as Wnt-dependent progenitor cells and have demonstrated their potential to become sensory epithelial cells both in vitro and in vivo. We postulate that TBCs may potentially serve as a reservoir of cells that function as a fail-safe mechanism for the intricate organization of the organ of Corti during early postnatal development. This may explain the proliferative nature of these cells and their subsequent decrease in proliferation and decline in number as the organ matures. Nonetheless, there exists a remnant quiescent population of TBCs in the adult cochleae that may serve as potential targets of regenerative therapeutics. However, any future therapy will require our full understanding of the mechanisms that help guide TBCs to migrate and differentiate into functional sensory cells.

We thank A. Ricci, L. Cunningham, O. Bermingham-McDonogh and members of our laboratories for fruitful discussions; K. Ahn, P. White and T. Hayashi for excellent technical assistance; A. Groves for sharing the Pax2-Cre mouse; and B. Crenshaw for sharing the B4OE-Cre mouse and anti-Brn4 antibodies.