Wnt signaling induces proliferation of sensory precursors in the postnatal mouse cochlea

Isolated Lgr5⁺ Cells Behave as Progenitor Cells in Vitro.

We previously characterized the Lgr5^{EGFP-CreERT2/+} mouse and found it to have normal cochlear morphology and hearing thresholds (11, 13). In the cochleae of neonatal Lgr5^{EGFP-CreERT2/+} mice, GFP is coexpressed with Sox2 in the third Deiters’ cells, inner pillar cells, inner phalangeal cells, and lateral greater epithelial ridge cells (Fig. 1 A–C). When the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) was injected on postnatal day (P) 0–2 (50 mg/kg two times/d), no labeling was noted in the sensory epithelium (Fig. 1D), confirming previous findings that supporting cells are mitotically quiescent (6, 14). Next, we dissociated P0–3 cochleae from Lgr5^{EGFP-CreERT2/+} mice and isolated GFP⁺ cells via flow cytometry; GFP⁺ cells constituted ∼2.1% of viable cells (Fig. 1E). Immunostained Lgr5⁺ cells were 94% GFP⁺, 96% Sox2⁺, and 0% myo7a⁺ (Fig. 1 F and G), whereas Lgr5⁻ cells were 0% GFP⁺, 3% Sox2⁺, and 1% myo7a⁺ (Fig. S1 A and B). Quantitative RT-PCR showed that sorted Lgr5⁺ cells, in comparison with Lgr5⁻ cells, expressed higher levels of Lgr5 and Sox2 and lower levels of the hair cell marker Brn3.1 (Fig. 1H and Table S1) (15). These data indicate that sorted Lgr5⁺ supporting cells were highly pure.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Lgr5⁺ cochlear supporting cells act as progenitor cells in vitro. (A) Cryosection of P3 Lgr5^{EGFP-CreERT2/+} cochlea showed GFP expression in the third Deiters’ cells (DC), inner pillar cells (PC), inner phalangeal cells (IPC), and the lateral greater epithelial ridge (GER). IHC, inner hair cells; LER, lesser epithelial ridge; OHC, outer hair cells. (B) Schematic depicting cell types in the P0–3 cochlea. BC, Boettcher cells; CC, Claudius cells; HEC, Hensen’s cells; SG, spiral ganglia; TBC, tympanic border cells. (C) Sox2 is expressed in Lgr5⁺ cells and other supporting cell types. (D) Cochlea from P3 mouse injected with EdU (P0–2) demonstrated no EdU uptake in the sensory epithelium. (E) Lgr5^{EGFP-CreERT2/+} cochleae were dissociated, and GFP⁺ and GFP⁻ cells were isolated using flow cytometry. (F–H) Immunostaining and quantitative PCR showed that isolated GFP⁺ cells did not contain myo7a⁺ cells and robustly expressed Lgr5 and Sox2 but not brn3.1. (I and J) Lgr5⁺ cells isolated from Lgr5^{EGFP-CreERT2/+}; Actin-DsRed mice were mixed (1:1) with those from Lgr5^{EGFP-CreERT2/+} mice and cultured. After 10 d, the majority of cytokeratin-positive colonies formed were monochromatic. I and J show DsRed and DAPI labeling only. I′ and J′ show DsRed, cytokeratin (CK), and DAPI labeling. (K) Lgr5⁺ and Hes5⁺ cells formed more colonies than Lgr5⁻ cells. Forty percent of colonies from Lgr5⁺ cells contained myo7a⁺ cells. (L–N) Lgr5⁺ cells generated more myo7a⁺ cells than Lgr5⁻ or Hes5⁺ cells, most of which were outside (O/S) colonies. I/S, inside colonies. (O) In the presence of EdU, only a minority of myo7a⁺ cells were EdU⁺; double-positive cells were noted more commonly inside colonies. Data are represented as mean ± SD. **P < 0.01. In F, G, I–K, N, and O, n is shown in parentheses. (Scale bars, 25 μm.)

To study the behavior of Lgr5⁺ cells, we cocultured 5,000 Lgr5⁺ cells with mitomycin-inactivated feeder cells derived from embryonic chicken utricle mesenchyme (Fig. S2). Inner ear-derived mesenchymal tissues have been shown to foster differentiation of cochlear supporting cells (7, 9). These mesenchymal cells do not express hair cell or supporting cell markers (16). After 10 d in serum-free medium, Lgr5⁺ cells formed epithelial colonies (consisting of at least five DAPI⁺ cells), which were immunostained with the pan-cytokeratin antibody (Fig. 1 I and J). When Lgr5⁺ cells from Lgr5^{EGFP-CreERT2/+};Actin-DsRed mice were mixed (1:1) with those from Lgr5^{EGFP-CreERT2/+} animals, 99% of colonies were monochromatic, suggesting that they were clonally derived from single cells (Fig. 1 I and J and Fig. S2B). To determine if Lgr5 serves as a marker for sensory progenitor cell enrichment, we compared it with Hes5, which is a Notch target gene expressed in most cochlear supporting cell types (Fig. S3 A and B) (17). We isolated and cultured Hes5⁺ cells from cochleae of P0–3 Hes5-GFP transgenic mice (18) and found that the colony counts from Hes5⁺ and Lgr5⁺ cells were comparable and significantly higher than in Lgr5⁻ cells (Fig. 1K).

Myo7a is a well-established specific marker of hair cells (9, 16). To investigate Lgr5⁺ cells’ potential for hair cell differentiation, we immunostained and characterized two classes of myo7a⁺ cells: those residing inside or outside epithelial colonies. On average, the majority of myo7a⁺ cells (308 of 389) resided outside colonies (Fig. 1 L–N). In comparison with Hes5⁺ and Lgr5⁻ cells, Lgr5⁺ cells generated significantly more myo7a⁺ cells. With EdU supplementation, 11.0% of myo7a⁺ cells within epithelial colonies were labeled. We observed a correlation between myo7a⁺ cells residing outside colonies and decreased EdU labeling (Fig. 1O), suggesting that the majority of hair cells were generated via direct differentiation. In support of their hair cell identity, myo7a⁺ cells also expressed other hair cell markers: calretinin, parvalbumin, and myo6 (Fig. S4 A–C). In addition to myo7a⁺ cells, many Sox2⁺ and GFP⁻ cells were found both inside and outside colonies (Fig. S4 C and G). Seventy-one percent of myo7a⁺ cells exhibited polarized protrusions expressing F-actin and the actin-bundling protein espin (Fig. S4 D–F) (19). These protrusions occasionally are V-shaped and resemble stereocilia of nascent hair cells. These results show that Lgr5⁺ cells can act as sensory progenitors and led us to investigate their in vivo behavior.

Lgr5⁺ Cells Behave as Precursors to Hair Cells and Supporting Cells in Vivo.

Prior in situ hybridization experiments had shown that Lgr5 expression is restricted to supporting cell subtypes (13). Cochleae from P3 Lgr5^{EGFP-CreERT2/+} also showed this expression pattern with no apical-to-basal gradient (Fig. 2A). However, we noticed rare pairings of GFP⁺, myo7a⁺ cells (two pairs in eight cochleae), with one located at the outer hair cell and the other at the third Deiters’ cell position (Fig. 2B). These pairs of double-positive cells also were rare at P5 (two pairs in four cochleae), and none were found at P9 (Fig. 2 D–F). In light of Lgr5⁺ cells’ ability to differentiate into hair cells in vitro, we postulate that GFP⁺, myo7a⁺ cells may represent differentiating hair cells. To probe this possibility further, we performed lineage-tracing experiments using Lgr5^{EGFP-CreERT2/+}; R26R^tdTomato/+ mice (20). Tamoxifen administration at P3 activated tdTomato labeling of Lgr5⁺ cells at P5 (Fig. 2 C–E and Fig. S5 A and B). In the P5 cochleae, traced cells included rare GFP⁺, myo7a⁺ cells but also occasional GFP⁻ hair cells and other supporting cells (Fig. S5 A and B). It is likely that GFP⁺, myo7a⁺ cells at both the hair cell and the third Deiters’ cell positions from the P3 cochlea contributed, at least in part, to tdTomato⁺, myo7a⁺ cells in the P5 and P9 cochleae.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Lineage tracing of Lgr5⁺ cells in the postnatal cochlea. (A and B) P3 Lgr5^{EGFP-CreERT2/+} cochlea showed GFP signals in supporting cell subtypes. GFP was rarely detected in myo7a⁺ cells and always was adjacent to a myo7a⁺, GFP⁺ cell at the third Deiters' cell position (arrowheads in B). (B′--B′′) Side views of reconstructed confocal images from B. B′ corresponds to the ′ position and B′′ to the ′′ position in B. (C) Tamoxifen was administered to P3 Lgr5^{EGFP-CreERT2/+}; R26R^tdTomato/+ mice, and cochleae were examined 2 and 6 d later. (D and E) Two days later, traced cells included the third Deiters' cells, inner pillar cells, inner phalangeal cells, and greater epithelial ridge cells. Traced GFP⁺, myo7a⁺ cells were rarely noted, and all were associated with myo7a⁺ cells at the the third Deiters' cell position (arrow in E′). (E′–E′′′′) Reconstructed z-stack images from E, showing merged labeling (E′), GFP and myo7a (E′′), tdTomato and myo7a (E′′′), and myo7a only (E′′′′). HC, hair cell. (F) Counts of GFP⁺ cells per cochlea. n is stated in parentheses. (G) At P9, traced cells included Hensen's cells, the second Deiters' cells (arrowheads), and hair cells (arrows). (H) GFP and myo7a labeling only. H′–H′′′′ are reconstructed z-stack images from G, showing merged labeling (H′), GFP and myo7a (H′′), tdTomato and myo7a (H′′′), and myo7a only (H′′′′). (I) Corn oil controls showed no tdTomato labeling. (J and K) Traced cells among outer and inner hair cells (OHC and IHC). (J′ and J′′) Side view and 3D reconstruction of confocal images from J′′. (K′–K′′′′) Reconstructed z-stack images from K, showing merged labeling (K′), GFP and myo7a (K′′), tdTomato and myo7a (K′′′), and myo7a only (K′′′′). (L) Counts of labeled hair cells and supporting cells from Lgr5^{EGFP-CreERT2/+}; R26R^tdTomato/+ mice. n is shown in parentheses. (Scale bars, 25 μm.)

When the tracing period was extended to P9, significantly more traced myo7a⁺ hair cells were found. Traced cells were found in a subset of outer and inner hair cells and supporting cells (first and second Deiters’ cells and outer pillar cells; Fig. 2 J and K, Fig. S5 C and D, and Movies S1 and S2). Parallel lineage-tracing experiments using Lgr5^{EGFP-CreERT2/+}; R26R^lacZ/+ mice (21) similarly found traced (lacZ⁺) cells among hair cells and supporting cells (Fig. S5 E–H). In both tracing paradigms, corn oil alone did not alter GFP expression or induce Cre-reporter activity (Fig. 2I and Fig. S5E). In addition, traced cells increased significantly from 2 to 6 d after tamoxifen administration (Fig. 2L). The increase in traced cells can represent fully differentiated cells derived from Lgr5⁺ cells, but we cannot rule out the possibility of a delayed accumulation of Cre reporter activity. In sum, these results support the notion that Lgr5⁺ cells in the neonatal cochlea behave as precursors to hair cells and supporting cells in vivo.

Active Wnt Signaling Induces Proliferation in Lgr5⁺ Cells.

In self-renewing organs, overexpression of β-catenin/Wnt signaling expands Lgr5⁺ cell populations and causes tumor formation (22, 23). To understand how Wnt signaling regulates the mitotically quiescent Lgr5⁺ cochlear cells, we characterized the Lgr5^{EGFP-CreERT2/+}; Catnb^{flox(exon3)/+} mice (24), in which tamoxifen-responsive Cre recombinase initiates overexpression of β-catenin in Lgr5⁺ cells. Following tamoxifen injection at P0–1, we observed the formation of multiple GFP⁺ foci at P8 adjacent to inner hair cells and lateral to outer hair cells (Fig. 3A). Because Lgr5 expression can be an indication of active Wnt signaling in the cochlea (13), these foci represent expanded clusters of Wnt-activated cells that normally become down-regulated and are most detectable in the third Deiters’ cells at this age (13).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Wnt signaling induces proliferation of Lgr5⁺ cells. (A–C) Tamoxifen was given to P0–1 Lgr5^{EGFP-CreERT2/+}; Catnb^{flox(exon3)/+} mice. Foci of GFP⁺ cells were noted 7 d later, abutting the inner hair cells and laterally in the lesser epithelial ridge. The medial foci (arrowheads) shifted adjacent to inner hair cells and disappeared by P21. (D) Pulse–chase experiments showed BrdU uptake (arrowheads) among formed foci. (D′ and D′′) High magnification and side view of foci. (E) Corn oil injection did not induce foci formation. (F) Foci counts per cochlea. (G and H) Tamoxifen was administered to P0–1 Lgr5^{EGFP-CreERT2/+}; Catnb^{flox(exon3)/+}; R26R^tdTomato/+ mice. All GFP⁺ foci expressed tdTomato and Sox2 but not myo7a. G′–G′′′′ Side view of reconstructed images from G. (H′–H′′′′) Individual foci with tdTomato, GFP, Sox2, and Hoescht (H′), Sox2 only (H′′), tdTomato and Hoescht (H′′′), and GFP and Hoescht labeling (H′′′′). (I–O) Purified Lgr5+ cells were cultured (10 d) with Wnt3a or R-spondin1. Treatment cells with Wnt3a or R-spondin1 increased cytokeratin-positive colonies, myo7a⁺ cells inside colonies, and EdU⁺, myo7a⁺ cells, but the total counts of myo7a⁺ cells were unaffected. *P < 0.05; **P < 0.01. In F and M–O, n is shown in parentheses. (Scale bars, 25 μm in A–E, G, H, and J–L; 10 μm in H′–H′′′′.)

Although most foci were adjacent to inner hair cells, a few also were observed in the region of pillar cells and lateral to outer hair cells (Fig. S6 G–I). BrdU injected at P4 was incorporated into these foci, all of which were Sox2⁺ and myo7a⁻ (Fig. 3 D, G, and H), suggesting proliferation without further differentiation. When Lgr5⁺ cells were traced simultaneously using the R26R-tdTomato Cre reporter allele, all GFP⁺ foci expressed tdTomato, implying that they arose from Lgr5⁺ cells overexpressing β-catenin. Interestingly, we observed a decrease in both the number and size of foci from P8 to P15, and no foci were detected at P21 (Fig. 3 A–C and F and Fig. S6 E and F). This result highlights the transient nature of these foci and suggests that cells constituting these foci underwent cell death. A subset of Lgr5⁺ cells did not degenerate, because many tdTomato⁺ cells remained at P21 (Fig. S6F). These results support our hypothesis that active Wnt signaling stimulates the proliferation of Lgr5⁺ cells. To test this theory further, we next examined the effects of Wnt agonists on isolated Lgr5⁺ cells.

Isolated Lgr5⁺ cells formed more colonies when cultured with Wnt3a (25) or R-spondin1 (Fig. 3 I–L and N) (23), but colony size was variable and was not affected by either treatment. Myo7a⁺ cell counts were comparable in drug-treated cells and vehicle-only controls; however, EdU⁺, myo7a⁺ cells doubled when treated with either Wnt agonist (P < 0.01 and <0.001, respectively) (Fig. 3 M and O). In support of the notion that more hair cells differentiated following a mitotic division, Wnt agonists increased the number of myo7a⁺ cells residing inside colonies and decreased the number residing outside colonies (Fig. 3O).

Wnt Inhibition Decreases Proliferation in Lgr5⁺ Cells.

IWP-2, a small-molecule Wnt inhibitor (26), significantly decreased Lgr5⁺ cells’ colony-forming capacity and EdU uptake (P < 0.001 for both) (Fig. 4 B–F). In comparison with vehicle-treated controls, IWP-2, reduced EdU, myo7a double-labeled cells, myo7a⁺ cells inside colonies, and EdU-labeled colonies. As a positive control, we treated Lgr5⁺ cells with the DNA polymerase inhibitor aphidocilin and found similar results. The number of myo7a⁺ cells outside colonies did not change significantly with aphidocilin but increased slightly with IWP-2 treatment (Fig. 4G). These results show that Wnt inhibition diminishes the proliferative capacity of Lgr5⁺ cells, whose ability to differentiate into hair cells does not require proliferation. Similarly, IWP-2 and aphidocilin decreased the colony-forming capacity of and EdU uptake in isolated Lgr5⁻ cells (Fig. S7 C and D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Wnt inhibition reduced proliferation of Lgr5⁺ cells in vitro. (A) Isolated Lgr5⁺ cells were cultured (10 d) with IWP-2 or aphidocilin. (B–G) IWP-2 or aphidocilin decreased colony formation and EdU⁺ colonies. Drug treatment also reduced EdU⁺, myo7a⁺ cells and myo7a⁺ cells inside colonies but not outside colonies. In E–G, n is shown in parentheses. **P < 0.01. (Scale bar, 25 μm in B–D.) (H) Hypothetical model of Wnt-dependent proliferation and differentiation of Lgr5⁺ cells.