The Transcription Factor Sox2 Is Required to Maintain the Cell Type-Specific Properties and Innervation of Type II Vestibular Hair Cells in Adult Mice

Type II hair cells with Sox2 deletion lost Type II molecular markers and gained Type I markers

This study was conducted in young adult mice starting at 6 weeks of age. We studied the utricle, which is one of five vestibular organs in each inner ear (Fig. 1A) and is specialized for detecting linear head motions. In mice, the utricle is a disk-shaped organ whose sensory epithelium (macula) has two zones: the central zone, which is also called the striola, and the peripheral zone, which is further subdivided into the LES and MES. Both Type I and Type II hair cells are distributed throughout these zones (Fig. 1B; Desai et al., 2005; Kirkegaard and Nyengaard, 2005; Li et al., 2008; Pujol et al., 2014). Studies in chinchilla showed that most vestibular afferents innervating hair cells are dimorphic: they have both calyx and bouton terminals (Fernández et al., 1990; Goldberg et al., 1990). Dimorphic afferents are distributed throughout all zones of the utricle; the remaining afferent neurons are either calyx-only (confined to the striola) or bouton-only (confined to the extrastriola).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Type II hair cells in Sox2 CKO mice lost Sox2 protein after tamoxifen treatment. A, upper left, Schematic of the inner ear with vestibular sensory organs in blue. Lower right, Top-down view of the utricle showing its different zones: peripheral (including the LES and MES) and central (including the striola, S). The juxtastriolar zone is small and lies on either side of the striola (data not shown). B, Schematic cross-section of the utricular epithelium; extrastriola is shown on the left, and striola is shown on then right. Type I hair cells (gray), Type II hair cells (magenta), supporting cell nuclei (sc), and afferent calyx terminals (black) are shown. Tomato labeling driven by Atoh1-CreER^T2 is shown in magenta; Sox2 immunolabeling is shown in green. C, Low-magnification confocal image of Tomato-labeled hair cells (magenta) in the utricle of a Atoh1-CreER^TM:Rosa26^tdTomato mouse at one week post-tamoxifen. The approximate position of the striola (S) is outlined in white, and the LES and MES are indicated. D, Graph shows density of Tomato-positive hair cells (mean ± SD) that were Sox2-positive in Sox2 WT and CKO mice at one month post-tamoxifen. N = 3 Sox2 WT mice and 3 Sox2 CKO mice. E–H, In Sox2 WT mice at one month post-tamoxifen, Sox2 protein (green) was detected in the nuclei (blue) of Type II hair cells (hc; E, F, H) and supporting cells (sc; G, H). Arrows point to Sox2-positive Type II hair cells that were Tomato-positive (nucleus appears orange when the three channels overlap). Tomato-negative Type II hair cells were also Sox2-positive (arrowhead in F). I–L, In Sox2 CKO mice, Sox2 protein was lost from Tomato-positive hair cell nuclei (arrows) by one month post-tamoxifen. Tomato-negative hair cells retained Sox2 (arrowheads in I, J, L), as did supporting cells (sc; K, L). All images are top-down views of the utricle except for H, L, which show cross-sections. hc, hair cell layer; sc, supporting cell layer. Scale bar in C: 100 µm (C), 30 µm (E, I), and 7 µm (F–H, J–L).

To test the hypothesis that Sox2 is required to maintain the Type II hair cell phenotype, we deleted Sox2 from Type II hair cells using Sox2^loxP/loxP mice, which have loxP sites flanking the single Sox2 exon (Shaham et al., 2009). To target Type II hair cells, we employed transgenic Atoh1-CreER^TM mice (Chow et al., 2006), which in adult utricles have tamoxifen-inducible Cre activity primarily in Type II haircells (Bucks et al., 2017); details are described below. We generated Atoh1-CreER^TM:Rosa26^tdTomato:Sox2^loxP/loxP (Sox2 CKO) mice, in which tamoxifen drives expression of the fluorescent protein tdTomato (Tomato) for fate-mapping cells with Sox2 deletion. Atoh1-CreER^TM:Rosa26^Tomato:Sox2^+/+ littermates (Sox2 WT) served as negative controls in which Tomato was expressed in Type II hair cells and Sox2 remained intact following tamoxifen injection.

First, we characterized Tomato labeling in Sox2 WT mice. At four months after tamoxifen injection at six weeks of age, there were 989 ± 126 (mean ± SD) Tomato-positive hair cells per utricle (Fig. 1C). Numbers were similar at one, four, and eight months post-tamoxifen (Table 1), which was expected because there is little death of hair cells in adult mouse utricles (Bucks et al., 2017). As described previously (Bucks et al., 2017), Tomato-positive supporting cells were rare (1–3 cells per utricle), and Tomato expression was absent from vestibular neurons and glia. Labeled hair cells were fairly evenly distributed through all zones. 61% of labeled cells were in the LES, 22% in the MES, and 17% in the striola (Fig. 1C; Table 2, left side), these proportions reflect the relative areas of each zone. Most labeled hair cells (92%) were Type II (Table 2, right side). A large proportion of Type II hair cells in each region were Tomato-positive: 60% in the LES, 72% in the MES, and 60% in the striola. Less than 10 Tomato-positive hair cells and supporting cells per utricle were present in Sox2 WT mice when no tamoxifen was given (Bucks et al., 2017).

To verify the loss of Sox2 protein from Type II hair cells of Sox2 CKO mice, we immunolabeled utricles at one month post-tamoxifen (Fig. 1E–L). Sox2 WT mice had significantly higher density of Tomato-positive/Sox2-positive hair cells than Sox2 CKO mice (unpaired Student's t test p < 0.0001, t = 31.43, df = 4; Fig. 1D). On average, 98% of Tomato-positive Type II hair cells in Sox2 WT mice were Sox2-positive (Fig. 1E–H) compared with 2% in Sox2 CKO (Fig. 1I–L). Tomato-negative Type II hair cells retained Sox2 labeling in Sox2 CKO mice (Fig. 1J), as did supporting cells, which also normally express Sox2 (compare Fig. 1G,H and K,L). This was expected because Tomato-negative cells should lack CreER activity. We observed a similar loss of Sox2 at two weeks post-tamoxifen (data not shown), but this was not quantified.

We found no difference in numbers of Tomato-positive hair cells per utricle among Sox2 WT and CKO mice at one, four, and eight months times post-tamoxifen (Table 1). Two-way ANOVA confirmed no effect of either genotype (p = 0.2192, f = 1.58) or time post-tamoxifen (p = 0.2529, f = 1.44). This finding strongly suggested that Sox2 deletion from Type II hair cells did not cause them to die.

To test Sox2's requirement for maintaining the Type II hair cell phenotype in adult mice, we gave tamoxifen to six-week-old Sox2 WT and CKO mice and examined hair cells at three times post-tamoxifen: two weeks, one month, four to five months, and eight to nine months. We assessed molecular, morphologic, and ultrastructural criteria that are specific to the Type I or Type II hair cell phenotype in Tomato-positive hair cells.

First, we looked at expression of the calcium-binding protein calretinin (Calb2), which is a selective marker for Type II hair cells (Desai et al., 2005; Pujol et al., 2014; Fig. 2E). In Sox2 WT mice, antibodies to Calb2 labeled the cytoplasm and nucleus of 96 ± 0% of extrastriolar Type II hair cells (four months post-tamoxifen is shown; Fig. 2A1–A3) and 97 ± 0% of striolar Type II hair cells (Fig. 6C). Type I hair cells and supporting cells appeared Calb2-negative (Fig. 2A3). Calb2 IF was reduced in Sox2 CKO mice relative to Sox2 WTs (four months post-tamoxifen is shown; compare Figs. 2B1–B3 and 2A1–A3, 2D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Type II hair cells lost immunoreactivity for two Type II-specific markers following Sox2 deletion. A1, A2, Top-down views of the extrastriolar epithelium showing the same field of a Sox2 WT mouse at four months post-tamoxifen. In A1, A2, arrows point to Tomato-positive cells (magenta) that are immunoreactive for Calb2 (green). A3. Vertical slice from another region showing Calb2 labeling in Tomato-positive Type II hair cells (II) and a Tomato-negative Type II hair cell (arrowhead). Supporting cells (sc) and Type I hair cells (arrows) were Tomato-negative. B1, B2, Top-down views of the extrastriolar epithelium showing the same field of a Sox2 CKO mouse at four months post-tamoxifen. Tomato-positive hair cells (arrows) appeared Calb2-negative; Tomato-negative hair cells (arrowheads) retained Calb2 immunoreactivity, as expected. B3, Tomato-positive cells with very weak Calb2 immunolabeling. C, Cell population data showing distributions of Calb2 IF intensity in Type I hair cells (HCI) and Type II hair cells (HCII) from one representative mouse of each group (WT or Sox2 CKO mice at one, four, or eight months post-DT). Each dot represents one cell. Black bar shows the average while red bar shows ±1 SD. D, Graph shows mean ± 1 SD for Calb2 IF intensity in each group. Each dot represents the mean intensity level for all cells analyzed in a given animal; n = 3 mice per group except for eight months WT (two mice) and eight months CKO (six mice); **p < 0.01. E, Illustration of different cell type-specific markers in peripheral (extrastriolar; ES) and central (striolar; S) regions of the utricle. F, upper panel, Top-down views of the extrastriolar epithelium from a Sox2 WT mouse at one month post-tamoxifen. Tomato-positive (magenta) hair cells (arrows) were immunoreactive for Anxa4 (green), as were Tomato-negative Type II hair cells (arrowhead). Lower panel shows a vertical slice through the area indicated by the horizontal white line (left side, half-way down). Tomato-positive hair cells were Anxa4-positive while supporting cells (sc) and Type I hair cells (arrow) were not. G, Images, showing Sox2 CKO mice at eight months post-tamoxifen, are the same format as in F. In Sox2 CKO mice, Tomato-positive hair cells (arrows) had lost Anxa4 labeling; Tomato-negative hair cells (arrowheads) retained Anxa4 immunoreactivity, as expected. H, Graph shows mean number (±1 SD) of Tomato-positive hair cells that were Anxa4-positive. Each dot represents the average for one animal; n = 4 for the Sox2 WT at one month post-Tam and n = 3 mice for the two Sox2 CKO groups; *p < 0.05, ***p < 0.001. Scale bar in A1: 6 µm (A1–B3). Scale bar in F (lower left): 6 µm (F, G).

We measured Calb2 IF intensity in Tomato-positive cells in Sox2 WT and CKO mice. We sampled 19–25% of the extrastriolar epithelium (LES and MES), classifying each Tomato-positive cell as Type I or Type II and measuring Calb2 intensity in each cell, as described in Materials and Methods. Figure 2C shows cell population data from one representative animal at each time point. Each data point represents an individual hair cell, with cell numbers listed below each column on the x-axis. As expected for Sox2 WT mice, the average Calb2 intensity in Type I hair cells was much lower than Type II hair cells at all times. Interestingly, WT Type II hair cells had a large range of Calb2 intensities. In Sox2 CKO mice, there was a reduction in average Calb2 intensity in Type II hair cells at all times post-tamoxifen and an increase in numbers of Tomato-positive Type I hair cells.

We computed the average Calb2 intensity of Type II hair cells for each in a group (Fig. 2D). Two-way ANOVA of group data in Figure 2D showed a significant effect of genotype (p < 0.0001, f = 50.48) and time post-tamoxifen (p = 0.0109, f = 6.350), indicating that Calb2 IF had decreased in Sox2 CKO mice relative to WT by one month post-tamoxifen and did not decrease further over time. Although we only measured IF intensity in extrastriolar hair cells, we noted a similar loss of Calb2 IF in striolar hair cells (compare Fig. 6C and D).

We investigated whether extrastriolar Type II hair cells lost a second marker, annexin A4 (Anxa4), after Sox2 deletion. Anxa4 is only expressed in Type II hair cells (McInturff et al., 2018) and is localized to the cell's periphery (Fig. 2E), appearing as ring around the cell in top-down views of Sox2 WT utricles at one month post-tamoxifen (Fig. 2F). After Sox2 CKO, numbers of Anxa4-positive hair cells decreased significantly relative to WT controls (p = 0.0009, f = 22.34, by one-way ANOVA; Fig. 2F–H). This reduction was significant at one month post-tamoxifen and became significantly larger at eight months post-tamoxifen. These observations provided further evidence that mature Type II hair cells lost cell-specific properties following Sox2 deletion.

Next, we tested whether Type II hair cells acquired Type I-specific properties after Sox2 CKO. First, we labeled utricles with antibodies to secreted phosphoprotein 1 (Spp1; also called osteopontin). Spp1 is a selective marker for Type I hair cells in the extrastriola (McInturff et al., 2018; Fig. 2E). In the utricles of mature mice, Spp1 is expressed in type I hair cells but not type II hair cells (McInturff et al., 2018)(Fig. 2E). In Sox2 WT mice, we found most Spp1 labeling to be extrastriolar (Fig. 3A). As expected, type II hair cells (either Tomato-positive or negative) were Spp1-negative, and type I hair cells were Spp1-positive (Fig. 3B1–B3). After Sox2 CKO, numbers of Tomato-positive hair cells that were Spp1-positive per utricle increased significantly (Fig. 3C1–C3;G). Two-way ANOVA showed an effect of genotype (p = 0.0003, f = 25.10) but no effect of time post-tamoxifen (p = 0.2795, f = 1.42). Numbers did not increase in Sox2 CKO mice until four months post-tamoxifen, and after that time they did not increase significantly (Fig. 3G).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Type II hair cells gained immunoreactivity for two Type I-specific markers following Sox2 deletion. A, Low-magnification top-down view of Spp1 immunolabeling in adult utricle. S, striola; LES, lateral extrastriola; MES, medial extrastriola. B1, B2, Top-down views of Spp1-labeled extrastriolar epithelium showing the same field of a Sox2 WT mouse at four months post-tamoxifen. In B1, B2, arrows point to Tomato-positive cells (magenta) that lack Spp1 immunolabeling (green). Arrowheads indicate Tomato-negative cells that are Spp1-positive. B3, Vertical slice from another region showing lack of Spp1 labeling in Tomato-positive Type II hair cells (II; arrow) and Spp1 labeling in the neck of a Tomato-negative Type I hair cell (I; arrowhead). C1, C2, Top-down views of Spp1-labeled extrastriolar epithelium showing the same field of a Sox2 CKO mouse at four months post-tamoxifen. Tomato-positive hair cells (magenta, arrow) were Spp1-positive. C3, Spp1-positive/Tomato-positive hair cell (arrow) shown in a slice. D, Low-magnification top-down view of Ocm immunolabeling in adult utricle. S, striola. E1, E2, Top-down views of the striolar epithelium showing the same field of a Sox2 WT mouse at one month post-tamoxifen. Arrowheads point to the same Tomato-positive cells (magenta) that lack Ocm immunolabeling (green). F1, F2, Top-down views of the striolar epithelium showing the same field of a Sox2 CKO mouse at eight months post-tamoxifen. Arrowheads point to the same Tomato-positive cells (magenta) that are Ocm-positive (green). G, Graph shows number of extrastriolar Tomato-labeled hair cells per utricle that were Spp1-immunoreactive in Sox2 WT and CKO mice at various times post-tamoxifen. H, Graph shows number of striolar Tomato-labeled hair cells per utricle that were Ocm-immunoreactive in Sox2 WT mice at one month post-tamoxifen and in Sox2 CKO mice at one and eight months post-tamoxifen. For both graphs, data are expressed as mean ± SD. Each point is the average number for each mouse in the group; *p < 0.05, **p < 0.01. Scale bar in A: 150 µm (A, D), 12 µm (B1–C3), and 10 µm (E1–F2).

We also tested whether Type II hair cells in the striola gained a Type I-specific marker after Sox2 CKO by labeling utricles with antibodies to oncomodulin (Ocm), which is expressed in the cytoplasm and nucleus of striolar Type I hair cells (Simmons et al., 2010; see Fig. 2E,3D). In Sox2 WT mice at one month post-tamoxifen, very few Tomato-positive hair cells (Type II) were Ocm-positive (Fig. 3E1, E2, H). After Sox2 CKO, numbers of Tomato-positive cells that were Ocm-positive were significantly higher (Fig. 3F1, F2, H) than WT mice at both of the times we examined (p = 0.0037, f = 20.9, by one-way ANOVA).

These analyses showed that Sox2 deletion from Type II hair cells caused them to lose Type II-selective markers (Calb2 and Anxa4) and to gain Type I-selective markers Spp1 in the extrastriola and Ocm in the striola.

Type II hair cells adopted Type I morphology following Sox2 deletion

Next, we explored whether Sox2 deletion from Type II hair cells caused them to acquire Type I-like morphology. For this analysis, we focused on the extrastriola (MES and LES), where the cell nuclei are clearly organized into three layers and the shapes of the two hair cell types are consistently distinct from each other (Figs. 1B, 2E). The cellular lamination in the extrastriolar epithelium is evident in a low-magnification transmission electron micrograph (TEM) of a Sox2 WT mouse at nine months post-tamoxifen (Fig. 4A). By contrast, after Sox2 CKO, the epithelial lamination was lost in some places (Fig. 4B), presumably because many Type II nuclei had migrated to a more basal (Type I-appropriate) position as they underwent transdifferentiation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Changes in Type II hair cell position and shape following Sox2 deletion. All images were taken from the same approximate region in the LES. A, B, Low-magnification TEM images from a Sox2 WT mouse (A) and a Sox2 CKO mouse (B), both at nine months post-tamoxifen. Note that the WT epithelium has two hair cell layers (Type II near the lumen, and Type I more deeply located), while this lamination was less clear in some regions after Sox2 CKO. I, Type I hair cell; II, Type II hair cell; bm, basement membrane; sc, supporting cell. Scale bar in A: 7 µm (also applies to B). C, D, Confocal vertical re-slices of Tomato-positive hair cells a Sox2 WT mouse (C) and a Sox2 CKO mouse (D). In WTs, most Tomato-positive hair cells had Type II-like morphology: a thick cell body, short stereocilia (arrowheads), nuclei (nu) located close to the lumen, and basolateral processes (arrows). In Sox2 CKO mice, more Tomato-positive hair cells were Type I-like; they had flask-shaped cell bodies, long stereocilia (arrowheads), nuclei (nu) located deeper than the Type II hair cell nuclei, and no basolateral processes. Asterisk shows a cytoplasmic bleb from a hair bundle. Note that panel C contains two separate non-continuous panels separated by a thin line, showing four exemplary Tomato-positive Type II hair cells from a WT mouse. Scale bar in C: 6 µm (also applies to D). E, Graph shows number of extrastriolar Tomato-positive hair cells per utricle that had Type I-like morphology for Sox2 WT and CKO mice at various times post-tamoxifen. Data are expressed as mean ± 1 SD. Each point is the average number for each mouse in the group; for all groups, n = 3 mice; **p < 0.01, ****p < 0.0001.

We used confocal microscopy to analyze the shape of Tomato-positive hair cells and classify them as either Type I or Type II (described in Materials and Methods). In Sox2 WT mice, as anticipated based on marker labeling, most Tomato-positive hair cells were Type II-like in cell shape, and they did not change their morphology over time post-tamoxifen (Fig. 4C,E). After Sox2 CKO, however, the numbers of Tomato-positive hair cells with a Type I morphology increased significantly (Fig. 4D,E). Two-way ANOVA showed effects of genotype (p < 0.0001, f = 410.30) and time post-tamoxifen (p < 0.0001, f = 24.75), with a significant difference in Type I-like hair cells between Sox2 WT and CKO mice as early as one month post-tamoxifen and with significant differences between Sox2 CKO mice at all times post-tamoxifen.

Sox2 deletion from mature Type II hair cells altered the morphology of the afferent terminals they contacted

We were curious whether the transition of Type II hair cells toward the Type I phenotype after Sox2 deletion induced the primary afferent neurons contacting them to shift their terminal morphology from bouton to calyx. First, we examined effects in the extrastriola (MES and LES), where the vast majority of afferent nerves are dimorphic, i.e., they have both afferent calyx and bouton nerve terminals (Fernández et al., 1990; Goldberg et al., 1990). We immunolabeled utricles for Tubb3, which is abundant in all afferent terminals. We scored Tomato-positive cells as “with a calyx” if they had Tubb3 label that surrounded all of the cell body at the level of the nucleus; in most cells, Tubb3 label extended up the cell's neck (Fig. 5C). In Sox2 WT mice, very few Tomato-positive hair cells had a calyx at four months post-tamoxifen (Fig. 5A1–A4) and later (Fig. 5D), which was expected because most Tomato-positive hair cells were Type II. Sox2 CKO induced a significant increase in the numbers of Tomato-positive hair cells with a calyx (Fig. 5B1–B4,D). Two-way ANOVA revealed significant effects of genotype (p < 0.0002, f = 28.32) and time post-tamoxifen (p = 0.0014, f = 11.93; Fig. 5D). Numbers of Tomato-positive hair cells with a calyx did not increase significantly in Sox2 CKO mice relative to Sox2 WTs until eight months post-tamoxifen, and they increased significantly in Sox2 CKO mice between one and eight months, and between four and eight months, post-tamoxifen.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Extrastriolar Type II hair cells gained a full or partial calyx afferent following Sox2 deletion. A1, A3, In Sox2 WT mice at four months post-tamoxifen, Tomato-positive hair cells (magenta, arrows) were Type II-like and lacked a calyx (labeled with antibodies to Tubb3, green). By contrast, Type I hair cells (arrowheads), which were Tomato-negative, had a calyx. A2, A4, Same fields as A1, A3 but just the Tubb3 channel. B1, B3, In Sox2 CKO mice at four months post-tamoxifen, some Tomato-positive hair cells (magenta, arrows) were Type I-like, with most of their cell body enwrapped by a calyx. B2, B4, Same fields as B1, B3 but just the Tubb3 channel. C, Depiction of the appearance of cells in Sox2 WT mice, with dimorphic afferent nerve (Aff_D, green) surrounding a Type I hair cell as a calyx and contacting a Tomato-positive Type II hair cell as bouton terminals. D, Graph shows the number of extrastriolar Tomato-labeled hair cells per utricle (mean ± SD) with a calyx in Sox2 WT and CKO mice at different times post-tamoxifen. Each point is the average number for each mouse in the group. N = 3 mice for all groups; **p < 0.01, ***p < 0.001. E, Image from a Sox2 CKO mouse at four months post-tamoxifen showing a Tomato-positive cell with a partial calyx (arrow). xy, top-down view, while xz and yz are slices of the same field in each plane. F, Drawing of Tomato-positive hair cell partially enveloped by calyx nerve afferent (green), similar to what is shown for a Sox2 CKO mouse in panel E. G–O, TEM images of cells from the extrastriolar region. All afferent nerve terminals are pseudo-colored in blue. G–I, Exemplary Type I and Type II hair cells from Sox2 WT mice. Calyx afferents (c) fully enwrap the Type I hair cell body, up to the neck (arrows in I). Bouton afferents (b and arrowheads in G) only contact Type II hair cells. J, K, N, O, Examples of presumed converting Type II hair cells (II-c) with a partial calyx in Sox2 CKO mice at one or nine months (m) post-tamoxifen. Arrowheads point to places where the calyx is lacking; in one case (N), the hair cell cytoplasm bulges between the two pieces of calyx material. L, M, Examples of presumed Type I hair cell precursors (I-prec) with a partial calyx in developing mouse utricles at P18 (L) and P14 (M). Arrowheads point to places where the calyx is lacking. I, Type I hair cell; II, Type II hair cell; sc, supporting cell; bm, basal membrane; bp, basolateral process. Scale bar in A1: 6 µm (A1–B4); in E: 6 µm; in G: 5 µm (G, I, J, L, M, O) and 1 µm (N); and in H: 3 µm (H, K).

In the extrastriolar regions of Sox2 CKO mice, we also observed Tomato-positive hair cells in which continuous Tubb3 labeling was present over only a portion (∼20–70%) of the cell body at the level of the nucleus (Fig. 5E; illustrated in Fig. 5F). These structures resembled partial calyces, which are common in developing mammals but are rarely seen in adults (Favre and Sans, 1979; Rüsch et al., 1998; Warchol et al., 2019). Tubb3 antibodies also labeled boutons, and some boutons were aligned on WT hair cells so as to resemble partial calyces. Therefore, we could not reliably compare partial calyces in the extrastriola between Sox2 WT and CKO mice in a rigorous, quantitative manner.

To explore these structures further, we examined utricular cross-sections using TEM (Fig. 5G–O). In these images, afferent neural structures in contact with hair cells are highlighted in blue. In Sox2 WT mice, sectioned Type II hair cells had discrete afferent bouton terminals contacting the basal half of their cell body (Fig. 5G), and in some sections, displayed no neural elements (Fig. 5H). Type I hair cells had a calyx terminal that surrounded the entire cell body from the base to almost the top of the neck with no breaks in calyx structure (Fig. 5G,I). By contrast, in Sox2 CKO mice, many Type II-like hair cells (with a thick neck and apically located nucleus) were partial wrapped by continuous afferent nerve resembling a calyx (Fig. 5J,K). These cells with both Type I and Type II properties (“hybrids”) had various forms of afferent terminals. In some, the partial calyx wrapped the apical portion of the hair cell body, and there were small breaks in the basal portion that were in contact with supporting cells (Fig. 5J,O) or afferent bouton terminals (Fig. 5O). We also noted cases in which a patch of hair cell cytoplasm seemed to be pinched between two stretches of presumably new calyx (Fig. 5N). In other hair cells, the calyx wrapped only the cell's base (Fig. 5K). We detected cells with partial calyces as early as one month (Fig. 5N,O) and as late as nine months (Fig. 5J,K) post-tamoxifen. It is important to note that hybrid cells were never seen in adult Sox2 WT mice. However, we readily detected them using TEM in C57Bl6/J mice at postnatal day (P)14–P18 (Fig. 5L,M), and they were described in prior studies for mice at earlier postnatal stages (Rüsch et al., 1998; Warchol et al., 2019). We conclude that these hybrid cells are converting Type II hair cells (II-c), in the process of acquiring Type I-like properties in response to Sox2 deletion.

Next, we analyzed calyces on Tomato-positive hair cells in the striola. We used antibodies to Calb2 to selectively label calyx-only afferents, which receive inputs from only Type I hair cells and are confined to the striola (Fig. 6A,B; Desai et al., 2005; Li et al., 2008; Hoffman et al., 2018). Some Calb2-positive afferents wrap more than one Type I hair cell and are called complex calyx afferents (illustrated in Figs. 1B, 6G). This analysis did not quantify calyces that originated from dimorphic afferents in the striola.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Type II hair cells in the striola joined calyx-only afferents following Sox2 deletion: Evidence with confocal microscopy. A, B, Calb2 (green) labeling of calyx-only afferents in the striola from Sox2 WT and CKO mice at four months post-tamoxifen. Calb2 is highly expressed in calyx-only afferents, which are confined to the striola (S; outlined in white). C–F, Images from the striola showing calyx-only afferents labeled for Calb2 (C, E) or Calb1 (D, F) from Sox2 WT or CKO utricles at four months post-tamoxifen. Arrows point to Tomato-positive hair cells (magenta) that were surrounded by a Calb2 or a Calb1-labeled calyx in mice with Sox2 CKO (D, F) but not in Sox2 WT (C, E) mice. Insets in C, E show Tomato and Calb2 nuclear labeling in separate channels for the cells indicated by arrows. G, Depiction of the appearance of cells in Sox2 WT mice with a calyx-only afferent nerve (Aff_c, green) surrounding Type I hair cells. This example is a complex calyx afferent. The tomato-positive Type II hair cell is not contacted by the calyx-only afferent. H, Graph shows the number of Tomato-positive (Tom+) striolar hair cells per utricle (mean ± SD) that were fully or mostly wrapped by a Calb2-labeled calyx. Each point is the average number for each mouse in the group. For Sox2 WT, n = 5 mice at one month, 6 mice at four months, and 3 mice at eight months post-tamoxifen. For Sox2 CKO, n = 6 mice at one month, 7 mice at four months, and 5 mice at eight months post-tamoxifen; *p < 0.05, **p < 0.01, ***p < 0.0001, ****p < 0.0001. I, Sox2 CKO utricles labeled with antibodies to Calb1 (top) or Calb2 (bottom), showing complex calyces extending neurites (arrowheads) to enwrap nearby Tomato-positive hair cells. J, Drawing of Tomato-positive hair cell partially enveloped by calyx nerve afferent (green), similar to what is shown for Sox2 CKO mice in panel I. K, Graph shows the number of Tomato-positive (Tom+) striolar hair cells per utricle (mean ± SD) that were partially wrapped by a Calb2-labeled calyx. Each point is the average number for each mouse in the group. For Sox2 WT, n = 4 mice at each time post-tamoxifen; for Sox2 CKO, n = 3 mice at each time post-tamoxifen. Scale bar in B: 25 µm (A, B); and in D: 4 µm (C–F) and 6 µm (I).

In Sox2 WT mice, very few Tomato-positive striolar hair cells had a Calb2-positive calyx (Fig. 6C, four months post-tamoxifen is shown, H), and this was consistent over time post-tamoxifen (Fig. 6H). This finding was anticipated for normal Type II hair cells. By contrast, many Tomato-positive striolar hair cells in Sox2 CKO mice were fully wrapped by a Calb2-positive calyx (Fig. 6E, four months post-tamoxifen is shown, H). Two-way ANOVA revealed significant effects of genotype (p < 0.0001, f = 68.31) and time post-tamoxifen (p = 0.0267, f = 4.18). Specifically, numbers of Tomato-positive hair cells with a calyx did not increase significantly in the striola of Sox2 CKO mice relative to Sox2 WTs until four months post-tamoxifen, and they did not increase significantly in Sox2 CKO mice between four and eight months post-tamoxifen. We also noted that Tomato-positive hair cells in the striola of Sox2 CKO mice had reduced nuclear Calb2 IF, a Type II-specific property (compare Fig. 6C, arrow, and E, arrows; see also Fig. 6C,E, insets). This finding is similar to our observations in the extrastriola of Sox2 CKO mice (Fig. 2A1–D).

We analyzed utricles using a second calyx marker, calbindin (Calb1), which in rodents is expressed only in calyces located in the striola and in a small region lying adjacent to the striola called the juxtastriola (Cunningham et al., 2002; Leonard and Kevetter, 2002; Prins et al., 2020). Our findings confirmed that many converting Tomato-positive Type II hair cells in the striola had gained a full calyx after Sox2 deletion (compare Fig. 6D and F, four months post-tamoxifen is shown).

In Sox2 CKO samples as early as one month post-tamoxifen, we found evidence with both Calb1 labeling (Fig. 6I, top panel) and Calb2 labeling (Fig. 6I, bottom panel) that afferent fibers deriving from a complex calyx (comprised of two or more continuous calyces) had partially wrapped adjacent Tomato-positive hair cells, suggesting that they were “recruiting” them to join the complex calyx afferent (Fig. 6J). We never detected Tomato-positive cells with these partial calyces in Sox2 WT mice, but we saw them in Sox2 CKO mice at all times post-tamoxifen (Fig. 6K).

We used TEM to further explore these changes in the striola. In Sox2 WT mice, complex calyx afferents were readily detected (Fig. 7A). In all cases, the cell body of every Type I hair cell in the complex was fully wrapped by neural material; no partial calyces were seen. As a reference for how Type II hair cells might be recruited by a nearby calyx afferent to create a complex calyx after Sox2 deletion in adulthood, we studied neonatal utricles in which striolar calyces were developing (Warchol et al., 2019). Figure 7B shows a calyx afferent from a striolar Type I hair cell at P6 that extends to partially wrap an adjacent Type I hair cell precursor. We found similar developing calyceal structures in Sox2 CKO mice at two weeks post-tamoxifen (Fig. 7C,D). Neural material had branched from one calyx to partially wrap an adjacent hair cell that was Type II-like based on its afferent boutons. Similar structures were seen as late as nine months post-tamoxifen (Fig. 7E). These findings demonstrated that new calyces were formed on striolar Type II hair cells from branches of adjacent calyces after Sox2 CKO, and calyx formation were occurring at early and late times after tamoxifen treatment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Type II hair cells in the striola joined calyx-only afferents following Sox2 deletion: evidence with TEM. All panels show TEM images, with calyx afferents pseudo-colored in blue. A, An exemplary complex calyx wrapping two Type I hair cells (I) from a Sox2 WT mouse. B, A developing complex calyx in a P6 mouse utricle. Arrows point to region of a Type I hair cell precursor (I-prec) that is not wrapped by calyx. C–E, Two converting Type II hair cells (II-c) with a partial calyx derived from a neighboring Type I hair cell at two weeks (C) and nine months (E) after Sox2 CKO. Arrowheads point to regions of the converting Type II hair cells not wrapped by calyx. D, The boxed area in C is shown at higher magnification; arrows indicate the point where the full calyx on a Type I hair cell (left) branches to partially wrap the adjacent converting Type II hair cell (right). I, Type I hair cell. Scale bar in A: 5 µm (A, C) 4 µm (B), and 3 µm (E).

Based on these observations, we hypothesized that the addition of Type I hair cells to the central region of the utricle (striola plus juxtastriola) might increase its size. To investigate this, we labeled utricles with antibodies to Calb1, which as described above, marks both striolar and juxtastriolar regions (Cunningham et al., 2002; Prins et al., 2020). The area of the Calb1-positive region increased significantly in Sox2 CKO mice relative to Sox2 WT mice by four months post-tamoxifen and remained elevated at eight months post-tamoxifen (Fig. 8A–C). Two-way ANOVA revealed a significant effect of genotype (p < 0.0001, f = 52.30) but no significant effect of time post-tamoxifen (p = 0.0732, f = 3.219) This expansion was also evident in utricles that were labeled with antibodies to Calb2, a marker of calyces in the striola (Fig. 6A,B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

The aea of the central Calb1-positive region increased as Type II hair cells converted to Type I hair cells and joined, or formed, new calyces after Sox2 deletion. A, B, Calbindin (Calb1) immunolabeling (cyan) reveals that the central zone (CZ) of the utricle (striola plus juxtastriola) increases in width in Sox2 CKO relative to Sox2 WT utricles (eight months post-tamoxifen is shown). C, Graph shows the area (µm²) of the central zone (mean ± SD) in Sox2 WT and CKO mice at one, four, and eight months post-tamoxifen. Each point is the average number for each mouse in the group; for all groups, n = 3 mice except for Sox2 WT mice at one month post-Tam, where n = 4; *p < 0.05, **p < 0.01, ***p < 0.001. Scale bar in B (lower right corner): 100 µm (A, B).

Type II hair cells with Sox2 deletion adopt nuclear properties of Type I hair cells

To further explore the extent of Type II-to-Type I hair cell conversion after Sox2 deletion, we examined the nucleus, which has a different appearance in Type I versus Type II hair cells. We measured nuclear size (area) in the same Tomato-positive extrastriolar hair cells as for the Calb2 intensity analysis in Figure 2. We noted that, in Sox2 WT mice, the nuclear area of Type II hair cells was larger than that of Type I hair cells (Fig. 9A1–B2,L), and it did not change over time post-tamoxifen (Fig. 9L). By contrast, the nuclear area of Type II hair cells was smaller after Sox2 deletion, more closely resembling Type I hair cells (compare Fig. 9A1,A2 and C1,C2,L). As expected, the area of Type I hair cell nuclei did not differ between Sox2 WT and CKO mice (4 months post-tamoxifen is shown) (compare Fig. 9B1,B2 and D1,D2,L). Two-way ANOVA revealed a significant effect of genotype (p < 0.0001, f = 57.65) but not time post-tamoxifen (p = 0.8762, f = 0.1334) for the nuclear area of Type II hair cells (excluding Type I hair cell data in Fig. 9L). Specifically, the nuclear areas of Type II hair cells with Sox2 CKO mice were smaller than Sox2 WT mice at one and four months post-tamoxifen, but not at eight months post-tamoxifen. This latter finding suggests nuclear area increased over time after Sox2 deletion to more closely resemble normal Type II hair cells. However, this interpretation is refuted by the lack of significant difference between Sox2 CKO groups over time.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Converting hair cells acquired additional Type I-like features: changes in nuclear size and heterochromatin distribution. A1–D2, Confocal images of the Type I and Type II hair cell nuclear layers (Fig. 1B for reference) in the extrastriola of Sox2 WT and CKO mice at four months post-tamoxifen. A1, B1, C1, D1, Tomato (magenta), Calb2 (green), and DAPI (blue) labeling. A2, B2, C2, D2, DAPI labeling (white) only for each corresponding field. Arrows point to the same nuclei for each set of images (e.g., A1, A2). E, F, TEM images of exemplary Type I and Type II hair cells from Sox2 WT utricles. Red arrowheads in E, G point to small clumps of heterochromatin in each Type I nucleus, and red arrows in E, F point to areas in Type II nuclei lacking heterochromatin clumps. H–K, TEM images of four exemplary converting Type II HCs (II-c) at two weeks post-tamoxifen (H) and at nine months post-tamoxifen (I–K). Each hair cell has a partial calyx (pseudo-colored in blue). Red arrowheads in H, K point to examples of small clumps of heterochromatin in each cell's nucleus. HCI, Type I hair cells; HCII, Type II hair cells; mo, months. Scale bar in B1: 5 µm (A1–D2); in E: 5 µm (E–K). L, Graph shows nuclear area (mean ± SD) for Tomato-labeled Type I and Type II hair cells for Sox2 WT and CKO mice at various times post-tamoxifen. Each point is the average number for each mouse in the group; for all groups, n = 3 mice except for Sox2 WT at eight months post-Tamoxifen, n = 2 and for Sox2 CKO at eight months post-Tam, n = 6; **p < 0.01, ***p < 0.001.

Next, we used TEM to explore the chromatin structure of converting hair cells (identified as Type II-like cells with a partial calyx) after Sox2 deletion. Heterochromatin is transcriptionally inactive (for review, see Schueler and Sullivan, 2006; Eymery et al., 2009). It is electron-dense and therefore appears dark in TEM images. In adult Sox2 WT mice, heterochromatin was distributed in large or small clumps throughout the nucleus of Type I hair cells (Fig. 9E,G). By contrast, in Type II hair cells, clumps of heterochromatin were usually limited to the nuclear periphery, and chromatin in the center of the nucleus was a more uniform mix of euchromatin and heterochromatin (Fig. 9E,F). These different distributions of heterochromatin in vestibular Type I and Type II hair cells were previously reported in neonatal mice (Rüsch et al., 1998). As early as two weeks after tamoxifen in Sox2 CKO mice, converting Type II hair cells (i.e., those with a partial calyx) had clumps of heterochromatin that more closely resembled Type I hair cells (Fig. 9H). Similar cells were seen at nine months after tamoxifen in Sox2 CKO mice (Fig. 9I–K). These observations indicated that Sox2 deletion from Type II hair cells caused them to acquire nuclear features typical of Type I hair cells.

Sox2 deletion from Type II hair cells induced the stereocilia and synapses to resemble Type I hair cells

To further explore the extent of conversion to a Type I-like phenotype, we examined the stereocilia, the sensory organelles of hair cells, following Sox2 deletion. Prior studies showed that the stereocilia of Type I hair cells in mice are considerably thicker than those of Type II hair cells (Rüsch et al., 1998; Li et al., 2008), and we found this to be true in Sox2 WT controls using TEM (unpaired Student's t test, p < 0.0001, t = 7.002, df = 14; Fig. 10A,B,C,E,F). By contrast, the stereocilia of converting Type II hair cells (with a partial calyx) in Sox2 CKO mice at one month post-tamoxifen were considerably thicker than Sox2 WT Type II hair cells and more closely resembled Type I hair cell stereocilia (Fig. 10A,D,G,H). Two-way ANOVA of Sox2 CKO mice showed significant differences across cell types (Type I, Type II, and converting Type II; p < 0.0001, f = 30.45) and time post-tamoxifen (p = 0.0037, f = 9.61). In both the one month and the five to nine months post-tamoxifen groups, the stereocilia of Type I hair cells were significantly thicker than those of Type II hair cells, but they were not significantly thicker than converting Type II hair cells. In both groups, the stereocilia of converting hair cells were significantly thicker than Type II hair cells. There was no significant change in the thickness of stereocilia in converting hair cell types between one and five to nine months.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

Stereocilia width and synaptic ribbons in Type II hair cells become more Type I-like after Sox2 deletion. A, Graph shows stereocilia width (nm; mean ± SD) for Type I and Type II hair cells from Sox2 WT mice, and for Type I, Type II, and Type II-converting (II-c) hair cells from Sox2 CKO mice at one month or five to nine months post-tamoxifen. For this analysis, we examined: seven Type I hair cells and nine Type II hair cells from five Sox2 WT mice; nine Type I hair cells, nine Type II hair cells, and nine converting Type II hair cells from three Sox2 CKO mice at one month; and five Type I hair cells, five Type II hair cells, and six converting Type II hair cells from six Sox2 CKO mice at five to nine months; *p < 0.05, ***p < 0.001, ****p < 0.0001. B, C, TEM images show the typical appearance of stereocilia in Sox2 WT mice, with Type I hair cell stereocilia (B) considerably thicker than Type II hair cell stereocilia (C). Images were taken from the extrastriola. D, Stereocilia on a converting Type II hair cell (II-c, left, identified by its partial calyx; not shown) after Sox2 CKO appear to be as thick or thicker than those on the adjacent Type I hair cell on the right (identified by its full calyx; not shown). E–H, Higher-magnification TEM images of stereocilia from Sox2 WT (E, F) and CKO (G, H) hair cells. I, TEM images of two hair cells from a Sox2 WT mouse, a Type I hair cell (I) in the upper left and a Type II hair cell (II) in the lower right. The black arrow points to a bar-type ribbon in the Type I hair cell with the afferent calyx located below the hair cell plasma membrane. The black arrowhead points to a round ribbon in a Type II hair cell with a bouton (b) located to the left of the hair cell plasma membrane. A vestibular efferent (eff) sits slightly above the bouton. J, K, Higher magnifications of the boxed areas in I showing a bar-type ribbon in the Type I hair cell (J) and a round synapse in the Type II hair cell (K). L, Cluster-type ribbon in a Type I hair cell. Arrowheads point to the central core of the ribbon, and arrows point to the calyx plasma membrane, which in this section lacks a postsynaptic specialization. M–P, TEM images of converting hair cells (II-c) from Sox2 CKO mice at nine months post-tamoxifen. Arrows in M, N point to bar-like ribbons located close to partial calyx in two different converting cells. O, Low-magnification TEM image showing a converting Type II hair cell contacted by three partial calyces (pc). Arrows point to two enlarged ovoid floating ribbons. P, Enlargement of region in O that is indicated by arrows. Type I, Type I hair cell; II, Type II hair cell; c, calyx afferent; b, bouton afferent; eff, efferent presynapse; sc, supporting cell. Scale bars in B: 3 µm (B–D); in E: 250 nm (E–H); and in I: 200 nm (I, M, N), 100 nm (J–L, P), and 400 nm (O).

We also used TEM to explore whether synaptic ribbons shifted from Type II-like to Type I-like following Sox2 deletion. Ribbons are the presynaptic element in the vestibular hair cell-afferent nerve synapse (for review, see Moser et al., 2006). Ribbon morphology in each hair cell type is similar in some respects: both are composed of an electron-dense central region surrounded by synaptic vesicles. Our analysis of hundreds of hair cells from all zones of adult Sox2 WT mouse utricles demonstrated that ribbon synapses in Type I hair cells were either (1) single bar-shaped ribbons located close to the plasma membrane (Fig. 10I, top box, J) or (2) multiple ribbons that were clustered together and not apposed to a differentiated postsynaptic membrane (Fig. 10L). By contrast, ribbons in adult Type II hair cells in Sox2 WT mice were round or ovoid (Fig. 10I, bottom, box, K; Pujol et al., 2014). We never detected bar-type or clustered ribbons in Sox2 WT Type II hair cells in adult mice. However, investigators have described these forms of ribbons in Type II hair cells in chinchilla (Lysakowski and Goldberg, 1997).

Following Sox2 deletion from Type II hair cells, we used TEM to examine synapses in hair cells with partial calyces, which we assumed to be converting into Type I hair cells. In these cells, synaptic ribbons were either ovoid, as in normal Type II hair cells (data not shown), or bar-shaped, as in normal Type I hair cells (Fig. 10M,N). Furthermore, ribbons in converting hair cells were located near afferent nerve terminals that appeared larger than the typical bouton afferent, resembling a growing calyx afferent (Fig. 10M,N). We also detected some converting cells with highly enlarged round or ovoid ribbons that were located in the cell's most basal portion and sometimes relatively far from the plasma membrane, or “floating” (Fig. 10O,P). These ribbons may have been undergoing re-modeling as the Type II hair cell basolateral process retracted and the calyx grew around the converting hair cell. Altogether, our observations of synapses support the interpretation that Sox2 deletion from Type II hair cells causes them to adopt Type I-like synaptic ribbons.

Spatiotemporal progression of Type II-to-Type I conversion after Sox2 deletion

In an effort to better understand the steps by which Type II hair cells converted into Type I-like hair cells after Sox2 deletion, we performed three additional analyses. First, we compared two phenotypic features, Calb2 IF intensity (Fig. 2) and nuclear area (Fig. 9), in Sox2 WT and CKO mice at different times following tamoxifen. We measured each feature in the same Tomato-labeled nucleus in either the Type I layer or the Type II layer (Fig. 11A). In Sox2 WT mice, as expected, Type I hair cells at each time post-tamoxifen (gray-scale circles) had significantly smaller nuclei and lower Calb2 intensity than Type II hair cell nuclei (orange-red-scale circles). By contrast, Type II hair cells in Sox2 CKO mice (blue-green scale circles) had reduced Calb2 intensity and smaller nuclei. A Pearson correlation test showed that the two variables were significantly associated (r = 0.6929, p = 0.0007). Interestingly, though, not all mice with Sox2 deletion changed to the same degree over time post-tamoxifen. For example, some Sox2 CKO mice had, on average, larger nuclei at eight months post-Tamoxifen than did Sox2 CKO mice at four months post-tamoxifen (also see Fig. 9).

• Download figure
• Open in new tab
• Download powerpoint

Figure 11.

Spatiotemporal patterns of Type I-to-Type II conversion after Sox2 deletion. A, Graph shows mean values per animal for nuclear area and Calb2 IF intensity for Tomato-labeled Type I and Type II hair cells from Sox2 WT and Tomato-labeled Type II-layer hair cells Sox2 CKO mice at various times post-tamoxifen. B, Graph shows percentage of (1) Tomato-positive cells in the extrastriola that were Spp1-positive (red bars), had a Type I morphology (blue bars), or had a calyx (green bars); and (2) Tomato-positive cells in the striola with a calyx (magenta bars). All data are from Sox2 WT or CKO mice at eight months post-tamoxifen. C, Maps of Tomato-positive Type I-like hair cells from a representative Sox2 WT and CKO utricle at four months post-tamoxifen. Striolar hair cells (located in the Calb2 zone) are represented by blue dots, while extrastriolar hair cells are shown as white dots.

Next, we examined the percentage of Tomato-labeled Type II hair cells with Sox2 deletion that acquired various Type I-specific features by eight months post-tamoxifen (Fig. 11B; Table 3). Because a small proportion of Type I hair cells in Sox2 WT controls expressed Cre, 8–10% of Tomato-positive extrastriolar hair cells were Spp1-positive, had Type I-like morphology, and/or were contacted by a calyx-type afferent as defined by Tubb3 labeling. After Sox2 deletion, a higher percentage of Tomato-positive extrastriolar hair cells had Type I-like features: 33% were Spp1-positive, 28% had a Type I-like morphology, and 26% had a calyx. The variation in these percentages suggested that each feature changed to a different extent within each hair cell. Interestingly, a higher percentage of Tomato-positive hair cells in the striola (60%) acquired a full calyx than in the extrastriola (26%).

Finally, we assessed spatial patterns of Type II-to-Type I conversion in Sox2 CKO mice by generating maps of Tomato-positive Type I-like hair cells from one representative utricle from Sox2 WT and CKO mice at four months post-tamoxifen (Fig. 11C), defining Tomato-positive hair cells as Type I-like if (1) in the extrastriola, they had Type I morphology (described in association with Fig. 4) and (2) in the striola, they had a Calb2-positive calyx (described in association with Fig. 6). Tomato-positive Type I-like hair cells were distributed throughout the utricle, with higher numbers in Sox2 CKO mice relative to Sox2 WT mice. We estimated that, at eight to nine months post-tamoxifen, the conversion of Type II hair cells to Type I hair cells increased the total number of Type I hair cells by 13% in the extrastriola (Table 4) and 21% in the striola (Table 5), and decreased the total number of Type II hair cells by the same amounts.