Early and transient requirements for FGFR2b/1b ligands in cochlear sensory and neural cell subtype differentiation

Abstract

We probed the roles of FGFR2b/1b signaling in mid-gestation cochlear development by inducing dnFGFR2b, a ligand trap that sequesters FGF3 and FGF10. Analyses following E11.5-E18.5 induction showed that FGFR2b/1b ligands are required for normal hair cell numbers, to repress inner hair cell differentiation in the lateral organ of Corti and to promote outer hair cell differentiation. FGFR2b/1b ligands also repress outer support cell numbers and promote differentiation of outer pillar cells and Deiters’ cells. Finally, these ligands also promote normal cochlear ganglion size and morphology as well as differentiation of CALB2-positive spiral ganglion neuronal subtypes. Delaying the start of dnFGFR2b expression to successive days after E11.5 revealed a critical period of E11.5-E13.5 for FGFR2b/1b signaling in sensory development and of E11.5-E14.5 in ganglion development. Strikingly, even transient induction of dnFGFR2b from E11.5-E12.5 was sufficient to largely recapitulate the phenotypes seen following long-term induction, suggesting that some of the key FGFR2b/1b ligand-dependent events controlling cochlear cell subtype differentiation occur long before such differentiation is evident, and suggesting a new window within which to search for regulators of cochlear differentiation.

Summary statement

Inducible inhibition of FGFR2b/1b signaling revealed unexpected requirements for FGF3 and FGF10 ligands in differentiation of cochlear sensory and neural cell types and an early critical period for their roles.

1. Introduction

Auditory sensation is mediated by the cochlea, which in mammals consists of the spiraling cochlear duct epithelium and the associated spiral ganglion. The cochlear duct’s sensory region, the organ of Corti, contains two sensory hair cell types arranged in longitudinal and parallel rows from the cochlear base near the vestibular apparatus to the apical tip. The single medially positioned row of inner hair cells converts sound waves into electrical signals that are conveyed by spiral ganglion neurons to the brainstem. Inner hair cell electrical signals are amplified by sound-induced motion of three lateral rows of outer hair cells. Hair cells are interdigitated with a variety of supporting cell types, which provide physical support and recycle hair cell-derived ions and neurotransmitters. The cochlear duct also has important non-sensory domains. The organ of Corti is flanked medially by the inner sulcus, which forms from Kölliker’s organ, and laterally by the outer sulcus. The remainder of the non-sensory epithelium comprises Reissner’s membrane and the stria vascularis (annotated in E18.5 cross sections; Fig. 1C,C’,G,G’). Disrupted development or subsequent perturbations to any of these cochlear components in mammals can cause permanent deafness due to the lack of regenerative capacity.

Fig. 1.

Inducing dnFGFR2b from E11.5-E18.5 appears to cause a more severe cochlear phenotype than does germline Fgf10 deletion. (A–B) Lateral views of paint-filled control (R26^rtTA/+) and dnFGFR2b (R26^rtTA/+;dnFGFR2b/0) inner ears exposed to DOX from E11.5 to E15.5. (C–D) H&E-stained sagittal sections of cochleae induced from E11.5-E18.5. (E–F) Control (Fgf10^−/+) and Fgf10 null (Fgf10^−/−) E15.5 paint-filled inner ears. (G–H) E18.5 H&E-stained sagittal sections of control and Fgf10 null cochleae. Boxes in panels C, D, G, H indicate regions shown at higher magnification in C′, D′, G′, H’. Scale bar in A (500 μm) also applies to B, E, F. Scale bar in C (250 μm) also applies to D, G, H. Scale bar in C’ (50 μm) also applies to D′, G′, H’. Double-headed arrow in C′ and D′ indicates the medial-to-lateral orientation of the organ of Corti, which is altered in dnFGFR2b samples due to loss of Reissner’s membrane and the outer sulcus, and applies to all subsequent control and experimental panel pairs. The organ of Corti is boxed with a dashed white line where recognizable. (I) Average area measurements (±1 SD) of basal scala media, basal spiral ganglion, all scalae media and all spiral ganglia for 3 each of E18.5 control and dnFGFR2b cochleae and (J) for 3 each of E18.5 Fgf10^−/+ and Fgf10^−/− cochleae. Statistically significant differences indicated with asterisks. Adjusted P-values: Panel I, basal scala media, 0.0068; basal spiral ganglion, 0.4953; all scalae media, 0.0068; all spiral ganglion, 0.0208. Panel J basal scala media 0.0019, basal spiral ganglion, 0.0048; all scalae media, 0.0055; all spiral ganglia, 0.5866. Abbreviations: a, apex, b, base, cd, cochlear duct; i, inner hair cell; Ko, Kölliker’s organ; L, lateral, M, medial; o, outer hair cell; oC, organ of Corti; os, outer sulcus; Rm, Reissner’s membrane, sg, spiral ganglion; sm, scala media; sv, stria vascularis.

The cochlear duct epithelium and spiral ganglion neurons both originate from the otic placode, a region of head ectoderm. This tissue is induced to transform into the spherical otic vesicle, from which the cochleo-vestibular ganglion neuroblasts delaminate and aggregate as a single mass that is subsequently split into cochlear and vestibular divisions. Neural crest cells also contribute to the cochlear epithelium and ganglion, where they differentiate as melanocyte-like cells associated with the stria vascularis in the cochlear duct and as glia in the spiral ganglion. At E11.5 in mice, the cochlear duct becomes evident as a ventral outbud of the otic vesicle. The organ of Corti differentiates from a region of the cochlear duct termed the prosensory domain, which at E11.5 is characterized by a thickened SOX2-expressing region. As the cochlear duct elongates it coils, and between E13.5-E14.5 the prosensory domain exits the cell cycle in an apical-to-basal direction. Differentiation starts at E14.5 and proceeds in a basal-to-apical and medial-to-lateral progression such that by E18.5, although not functional yet, the basal organ of Corti is relatively mature. Hair cell markers are seen at E14.5 and at E15.5, inner hair cell-specific differentiation is induced in the medial prosensory domain by a yet unknown signal thought to come from Kölliker’s organ. The inner hair cell in turn, induces development of the supporting inner pillar cell, which is followed by differentiation of the outer compartment, comprising the outer hair cells and the remaining supporting cells; the outer pillar cell and three Deiters’ cells (for reviews, see Driver and Kelley, 2020; Kelley, 2022; Wu and Kelley, 2012).

The spiral ganglion also elongates and coils as it develops in parallel with the cochlear duct, becoming more distant from the epithelium as development proceeds. It develops two types of neurons; Type 1s (comprising 3 subtypes termed A, B and C) innervate the inner hair cells and Type 2s innervate the outer hair cells. Differentiation of these neuronal subtypes occurs during the same timeframe as differentiation of the cochlear duct. At E14.0, undifferentiated neuronal precursors split into two trajectories; Type 1A/Type 2 and Type 1B/C precursors, and at E16.0 these two lineages further separate with consolidation of final subtype identity at E18.0 and postnatally (Petitpre et al., 2022; Sanders and Kelley, 2022; Shrestha and Goodrich, 2019).

Many different signals control cochlear development, and prominent among these are the Fibroblast Growth Factors (FGFs). FGFs signal through FGF receptors (FGFRs) encoded by 4 genes, and molecular and genetic analyses suggest that the 15 canonical paracrine FGF ligands comprise two classes, those that signal principally through the “b” splice isoforms of FGFRs 1–3 and those signaling through the “c” isoforms of FGFRs 1–4 (Ornitz and Itoh, 2022). The “c”-type ligands, FGFs 8, 9 and 20, have well established roles in cochlear duct development (see schematic in Fig. S1). At E11.5, FGF9, expressed from the developing Reissner’s membrane, and FGF20 expressed within the prosensory domain, signal to surrounding otic mesenchyme, which in turn, signals back to the prosensory domain to proliferate (Huh et al., 2015). At about E13.5, FGF20 signals from the prosensory domain, promoting differentiation of the outer hair cells and the outer supporting cells (Hayashi et al., 2008; Huh et al., 2012) and at about E15.5, FGF8 expressed in the developing inner hair cell, promotes differentiation of the inner pillar cell (Hayashi et al., 2007; Jacques et al., 2007; Puligilla et al., 2007; Zelarayan et al., 2007).

The cochlea also expresses the “b”-type ligands, FGFs 3 and 10 (see schematic in Fig. S1), which signal through FGFR2b and FGFR1b, both of which are present in the developing cochlear duct and spiral ganglion (Pirvola et al., 2000, 2002). Fgf10 is expressed at high levels in the greater epithelial ridge, located medially to the developing prosensory domain, and extending into a region that ultimately becomes the lateral half of Kölliker’s organ. Fgf3 is expressed at much lower levels, overlapping with the most lateral portion of the Fgf10 domain, and beginning around E16.5, is also found on the lateral side of the organ of Corti (Pirvola et al., 2000; Urness et al., 2018; Wilkinson et al., 1989). We showed previously that E15.5 Fgf10^−/− (germline null) mutants have shortened, poorly coiled and narrow cochlear ducts, and at E18.5 they lack Reissner’s membrane and the outer sulcus, but that the organ of Corti, Kölliker’s organ and the stria vascularis are present with all expected cell types. Developmental studies showed that the Reissner’s membrane phenotype was first apparent at E12.5–13.5 and the outer sulcus phenotype at E15.5, both likely due to a failure to signal to the roof and lateral edge, respectively of the developing cochlear duct (Urness et al., 2015). In contrast, Tg(Pax2-cre)/Fgf3 conditional null mutants (cKOs), in which Fgf3 expression is ablated in the developing otic placode-derived tissues, bypassing an early requirement for FGF3 signaling from the hindbrain to the otic placode, have structurally and functionally normal inner ears (Hatch et al., 2007). Unfortunately, Tg (Pax2-cre)/Fgf3/Fgf10 double cKOs, in which both genes are deleted in otic placode derivatives, fail to develop a cochlea, thus negating the strategy of using a simple genetic deletion approach to assess their combinatorial roles during mid-gestation (Urness et al., 2018).

We overcame this difficulty by inhibiting signaling by both FGF3 and FGF10 through both FGFR2b and FGFR1b between E11.5 and E18.5 by using doxycycline (DOX)-inducible expression of a globally expressed and secreted form of the extracellular domain of FGFR2b (dnFGFR2b), which functions as a dominant-negative ligand trap that is reversible following DOX withdrawal (Danopoulos et al., 2013; Parsa et al., 2008; Urness et al., 2018). Previous use of this strategy revealed that FGFR2b/1b ligands are required for early patterning and morphogenesis of the otic vesicle (Urness et al., 2018). Here we extend our analysis to cell type differentiation within the mid-gestation cochlear duct and spiral ganglion. We find that FGFR2b/1b ligands are required for normal hair cell numbers, to repress inner hair cell differentiation at the lateral edge of the organ of Corti and to promote outer hair cell differentiation. FGFR2b/1b ligands are also necessary to repress outer supporting cell numbers and for differentiation of the outer compartment supporting cell types; the outer pillar cells and Deiters’ cells. Finally, these ligands are also required for normal cochlear ganglion size and morphology as well as for differentiation of CALB2-positive spiral ganglion neuron subtypes. Delaying the start of dnFGFR2b expression to successive days after E11.5 revealed that the critical period for FGFR2b/1b signaling in sensory development is E11.5-E13.5 and for ganglion development is E11.5-E14.5. Strikingly, even transient induction of dnFGFR2b from E11.5-E12.5 was sufficient to largely recapitulate the phenotypes seen following long-term induction, suggesting that some of the key FGFR2b/1b ligand-dependent events controlling cochlear sensory and neural subtype differentiation occur long before such differentiation is evident.

2. Results

2.1. Inducing dnFGFR2b from E11.5-E18.5 causes qualitatively more severe cochlear phenotypes than does germline Fgf10 deletion

To block all signaling through FGFR2b and FGFR1b during cochlear elongation, patterning and differentiation, we generated pregnant females carrying both control (Rosa26^rtTA/+; abbreviated as R26^rtTA/+) and experimental (Rosa26^rtTA/+;tet-O(s)dnFGFR2b)/0; abbreviated as R26^rtTA/+;dnFGFR2b/0) embryos, and induced rapid and sustained ubiquitous expression of dnFGFR2b by injecting dams with DOX at E11.5 and substituting DOX-chow for normal chow as described (Urness et al., 2018). To assess overall inner ear morphology, embryos were harvested at E15.5 and the inner ears were paint-filled. As shown previously (Urness et al., 2018), relative to controls (Fig. 1A), dnFGFR2b-expressing cochlear ducts were narrow, short and poorly coiled (Fig. 1B; n = 9). These phenotypes appeared somewhat more severe than those of Fgf10^−/− mutants (Fig. 1F; n = 24 and Urness et al., 2015). To assess gross cochlear duct morphology as late as dnFGFR2b-expressing embryos survive, we prepared sagittal head sections at E18.5 and stained the tissue with hematoxylin and eosin (H&E). Control cochleae were normal with all expected structures readily identified (Fig. 1C–C’). As expected from and similarly to Fgf10^−/− mutants (Fig. 1H–H’ and Urness et al., 2015), dnFGFR2b-expressing cochlear ducts were abnormal, lacking Reissner’s membrane and the outer sulcus (Fig. 1D–D’; n = 3). However, they had additional abnormalities; only two cochlear cross sections were evident, and it was difficult to clearly discern distinct boundaries and cell types of Kölliker’s organ, the organ of Corti and the stria vascularis. Also, the spiral ganglion appeared elongated with discontinuous clumps of cells.

To quantify the extent of cochlear duct and spiral ganglion reductions, we measured the cross-sectional areas of the basal cochlear duct scala media (interior cochlear duct cavity) and basal spiral ganglion as well as the combined areas of all scalae media and all spiral ganglia in central sections from control and dnFGFR2b samples (n = 3 each). We found significant reductions in both the basal and total scala media areas and in the total spiral ganglion, but not in the basal only spiral ganglion area in dnFGFR2b samples (Fig. 1I). We only measured basal scala media areas in our previous study of Fgf10^−/− mutants (Urness et al., 2015), so we remeasured control and Fgf10^−/− samples (n = 3 each) in the same way as for dnFGFR2b samples. As expected, we found that the basal and total scala media areas were significantly reduced. Unexpectedly, however, the basal spiral ganglion area was significantly reduced while the total spiral ganglion area was not (Fig. 1J). This may reflect germline vs. timed inhibition of FGF10. However, due to the large amount of time between processing the Fgf10 mutant and dnFGFR2b samples, and the potential for differential tissue shrinkage, we did not make statistical comparisons between the two different perturbations. Together, these results suggest a mid-late gestation role for FGF3 in combination with FGF10 in promoting normal development of the cochlear duct and spiral ganglion.

2.2. Cochlear non-sensory development following E11.5-E18.5 dnFGFR2b induction is similar to that of E18.5 Fgf10^−/− mutants

To further characterize the dnFGFR2b cochlear phenotype, we applied RNA and antibody probes to inner ear sections. As expected, a full-length Fgfr2-cDNA probe revealed prominent expression of the endogenous transcript in the outer sulcus of control samples and weaker expression in the stria vascularis and the epithelial layer of Reissner’s membrane, but was nearly ubiquitous in the dnFGFR2b cochlea, reflecting global induction of dnFGFR2b (Fig. 2A–A’ and B–B’). In contrast, the Fgfr2-3′ UTR probe, which cannot detect the transgene, marked the three expected regions in controls but was virtually absent from dnFGFR2b samples (Fig. 2C–C’ and D–D’). Similar results were obtained with the probe for Bmp4, another outer sulcus marker (Fig. 2E–E’ and F–F’). Similarly, markers of Reissner’s membrane (Fgf9-epithelial and Aldh1a2-mesenchymal) were evident in controls and largely absent from dnFGFR2b samples (Fig. 2G–J’). While the dnFGFR2b stria vascularis appeared shorter than that of controls, markers of each of the three cell layers (Kcnq1-marginal, Trp2-intermediate and Cldn11-basal) were apparent in dnFGFR2b cochleae (Fig. 2K–P’). Each of these findings is like that described for Fgf10^−/− cochleae, which have a deletion of the non-contiguous tissues, the outer sulcus and Reissner’s membrane (Urness et al., 2015). These results reflect efficient inhibition of FGF10 activity by dnFGFR2b, confirming that these non-sensory tissues require FGF10 activity after E11.5 and suggesting that FGF3 has little to no role in their development.

Fig. 2.

In situ hybridization analysis shows the efficacy of dnFGFR2b induction and the resultant loss of Reissner’s membrane and the outer sulcus, phenocopying the effects of Fgf10 deletion on cochlear non-sensory tissues. Embryos were exposed in situ to DOX from E11.5 to E18.5 and sagittal head sections were hybridized with the indicated DIG-labeled probes for Fgfr2 (either the cDNA, which detects both the endogenous and dnFGFR2b transcripts or the 3′UTR, which detects only the endogenous transcript) or various non-sensory markers. (A–P) Views of the entire cochlea with probe indicated in the upper left and (A′-P′) an enlargement of the basal cross section. Positively hybridizing tissues are labeled. Genotypes are indicated to the left of each row. Scale bar for panel A (250 μm) also applies to panels B–P; similarly, scale bar for panel A’ (100 μm) applies to B′-P’. Abbreviations: os, outer sulcus; Rm^ep, Reissner’s membrane epithelial layer; Rm^mes, Reissners’ membrane mesodermal layer; sv, stria vascularis; sv^bas, sv basal cells; sv^int, sv intermediate cells; sv^mar, sv marginal cells.

2.3. FGFR2b/1b ligands are required for normal hair cell numbers, to suppress inner hair cell differentiation in the outer compartment and for outer hair cell differentiation

Next, we examined hair cell differentiation, continuing to use cross-sectioned material because the lack of Reissner’s membrane and the small cochlear duct diameter of dnFGFR2b samples precluded preparation and flattening of intact whole mount tissue. Control tissue stained with antibodies directed against the pan-hair cell markers MYO7A and POU4F3 showed the expected 4 hair cells along the medial-to-lateral axis basally (Fig. 3A), but dnFGFR2b samples had a smaller and variable number of hair cells in each cross section, most commonly 2 or 3 (Fig. 3B). We counted MYO7A-positive cells in basal cross sections of 10 controls co-stained either with POU4F3 or with another marker, and each had the expected 4 hair cells. In contrast, only 1 of 10 similarly stained dnFGFR2b basal cross sections had 4 hair cells; 1 had 3 hair cells, 3 had 2 hair cells and 3, for which we analyzed multiple sections from the same cochlea, had a variable number of hair cells. In one case there were either 1, 2 or 3 hair cells, in another we found either 2 or 3 hair cells, and in the last, 2 or 4 hair cells. Thus, dnFGFR2b-expressing embryos had inter- and intra-sample variability with respect to total hair cell number per basal cochlear cross section and the number was often less than normal.

Fig. 3.

Inducing dnFGFR2b from E11.5 reduces hair cell number, induces ectopic inner hair cell-like differentiation and represses outer hair cell differentiation. (A-BB) Basal cross sections of cochlear ducts from control and dnFGFR2b embryos exposed to DOX and immunostained or hybridized with probes indicated above each panel pair. Induction intervals are annotated in the upper right of each panel pair. (CC-FF) E18.5 control (either Fgf10^+/+ or Fgf10^−/+) and Fgf10^−/− basal cochlear duct sections immunostained with antibodies as indicated. Scale bar in A (50 μm) applies to all panels. Symbols: (C,D) arrowhead, ihc innervation; caret, ipc (E,F,AA, BB) arrow, innervation of ihc (S,U) bracket, Bcl11b puncta in ohc region (AA’) asterisks, examples of utricular hcs. Abbreviations not previously defined: hc, hair cell; i-l, inner hair cell-like.

In some cases of MYO7A staining, we noticed a laterally positioned hair cell that was somewhat flask-shaped, appearing more like an inner hair cell than an outer hair cell. Therefore, we co-stained sections with antibodies for MYO7A and the inner hair cell marker, CALB2, which also labels fibers from spiral ganglion neurons on the pillar side of the inner hair cell (arrowhead) and basal inner pillar cells (caret) (Dechesne et al., 1994). Controls showed a single double-labeled cell, the inner hair cell, and 3 cells labeled with MYO7A only, i.e. the outer hair cells (Fig. 3C, n = 4). In contrast, we found that 3 of 4 dnFGFR2b samples showed 2 double-labeled hair cells, (Fig. 3D). Additional inner hair cell markers confirmed the subtype identity of the unusual lateral hair cells. Immunostaining for vGLUT3 (SLC17A8) showed control samples had only one labeled epithelial cell (Fig. 3E, n = 3), whereas 2 of 3 dnFGFR2b samples had 2 labeled cells (Fig. 3F). Co-staining for TUJ1 (TUBB3), the pan-axonal marker, indicated a single innervated inner hair cell in controls (Fig. 3E), but in dnFGFR2b-expressing samples, both the medial and lateral vGLUT3-positive cells, were innervated (Fig. 3F, arrows). In situ hybridization (ISH) using an Fgf8 probe to detect the inner hair cell showed a single positive cell in all 3 controls (Fig. 3G), whereas 3 of 4 dnFGFR2b samples had 2 Fgf8-positive cells (Fig. 3H). We also probed Fgf8 expression at E15.5, the first stage at which it is normally detected in the cochlear base (Hayashi et al., 2007; Jacques et al., 2007). We found a single basal Fgf8-expressing cell in control cross sections (Fig. 3I, n = 4), whereas 3 of 6 dnFGFR2b cross sections had 2 Fgf8-positive cells (Fig. 3J). Finally, since Tbx2 is expressed in inner hair cells and is necessary and sufficient for their differentiation (Bi et al., 2022, 2024; Garcia-Anoveros et al., 2022; Kaiser et al., 2022), we conducted fluorescence ISH using both Fgf8 and Tbx2 probes on control and dnFGFR2b samples induced from E11.5-E18.5. Both controls had a single double-positive cell in each cross section (the inner hair cell; Fig. 3K), whereas 2 of 3 dnFGFR2b samples had more than one double-positive cell and in one sample there were three double-positive cells (Fig. 3L). Furthermore, similar labeling of samples induced from E11.5-E15.5 showed two normal controls with one double-positive inner hair cell per cross section (Fig. 3M), whereas 2 of 3 dnFGFR2b samples showed two double-positive cells (Fig. 3N). Together, these data suggest that FGFR2b/1b ligands promote cochlear hair cell differentiation and that they suppress lateral inner hair cell differentiation from the earliest stages of hair cell subtype differentiation. We refer to the ectopic laterally positioned hair cells expressing inner hair cell markers as “inner hair cell-like” (i-l in subsequent figures) as their full transcriptome and function were not determined.

To assess outer hair cell differentiation, we first stained control and dnFGFR2b samples with antibodies directed against BCL11B, an outer hair cell marker, and CALB2, the inner hair cell marker. As expected, basal cross sections from controls showed robust expression of CALB2 in the inner hair cell and BCL11B in 3 outer hair cell nuclei (Fig. 3O, n = 4). Surprisingly, although each dnFGFR2b sample expressed CALB2 in 2 cells per cross section; one located medially, (the presumptive normal inner hair cell), and the other located more laterally (i-l), none expressed BCL11B (Fig. 3P, n = 5). BCL11B expression normally commences at E15.5, becoming readily detectable at E16.5 (Jiang, 2022; Wiwatpanit et al., 2018) and we found that E16.5 controls co-stained for BCL11B and CALB2 had the expected one inner hair cell and three outer hair cells per basal cross section (Fig. 3Q, n = 2), whereas those expressing dnFGFR2b had two CALB2-positive hair cells, but no BCL11B-positive outer hair cells (Fig. 3R, n = 2). These data were confirmed and extended by examining the same fluorescence ISH samples mentioned previously but comparing Fgf8 and Bcl11b signals. Controls induced from E11.5-E18.5 or E11.5-E15.5 had Bcl11b signals in outer hair cells adjacent to the Fgf8-expressing inner hair cells (Fig. 3S and U; n = 2 at each stage), whereas the corresponding dnFGFR2b samples had 2 or 3 Fgf8-expressing inner hair cells, but Bcl11b was not detected (Fig. 3T and V; n = 3 at each stage). Next we assessed INSM1, an outer hair cell-specific transcriptional repressor required to maintain outer hair cell identity (Li et al., 2023; Wiwatpanit et al., 2018) that is expressed in the cochlear base starting at E15.5, progressing apically and diminishing in the base by E18.5 (Li et al., 2023; Lorenzen et al., 2015). Basal sections of control and experimental cochleae induced to express dnFGFR2b from E11.5-E15.5 or E11.5-E16.5 were co-stained with antibodies to INSM1 and POU4F3, the pan-hair cell marker. Controls had three double-labeled hair cells per cross section, indicative of outer hair cell identity (Fig. 3W and Y), whereas dnFGFR2b samples (n = 3 for each induction) had 1–3 POU4F3-labeled hair cells, none of which expressed INSM1 (Fig. 3X and Z), suggesting that FGFR2b ligands promote outer hair cell differentiation.

Finally, we asked whether any dnFGFR2b-expressing hair cells had taken on a vestibular identity by co-staining for SPP1 and MYO7A. As expected, E18.5 control utricular Type 1 hair cells expressed SPP1 (Fig. 3AA’; n = 2; asterisks indicate a few examples) but control cochlear hair cells did not (McInturff et al., 2018; Swanson et al., 1989), though the inner hair cell was innervated as expected by SPP1-expressing spiral ganglion neurons (Fig. 3AA, arrow) (Kim et al., 2014; Swanson et al., 1989). The dnFGFR2b cochlear ducts had 2–4 MYO7A-positive hair cells and both the medial (presumptive endogenous inner hair cell) and lateral inner hair cell-like cell (i-l) were innervated by SPP1-positive spiral ganglion neurons (Fig. 3BB, arrow, n = 4). Similarly to controls, however, none of the dnFGFR2b-expressing cochlear hair cells themselves expressed SPP1. Together, these data suggest that FGFR2b/1b signaling is required for outer hair cell differentiation from the onset, but not for specifying vestibular vs. cochlear hair cell differentiation.

Fgf10^−/− cochleae have hair cells that are normal with respect to total number per basal cross section and MYO7A expression (Urness et al., 2015). To determine whether they have any perturbations of hair cell subtype differentiation, we stained E18.5 control and Fgf10^−/− basal cochlear cross sections with MYO7A or BCL11B and CALB2. Controls (n = 2) and mutants (n = 3) each showed 4 MYO7A-positive hair cells per cross section, one of which was also CALB2-positive (the inner hair cell, Fig. 3CC,DD). Similarly, controls (n = 2) and mutants (n = 5) each showed a single CALB2-positive inner hair cell and three BCL11B-positive outer hair cells (Fig. 3EE,FF). Thus, unlike dnFGFR2b-expressing embryos, Fgf10^−/− mutants have no abnormalities of hair cell subtype differentiation. Therefore, such abnormalities seen in dnFGFR2b-expressing cochleae must be caused by the inhibition of both FGF10 and FGF3 activity.

2.4. E11.5-E18.5 dnFGFR2b induction causes ectopic inner pillar cell differentiation and suppresses differentiation of outer pillar cells, lateral Kölliker’s organ and Deiters’ cells

The striking appearance in dnFGFR2b-expressing cochlear ducts of lateral hair cells expressing inner hair cell markers, particularly Fgf8, the product of which signaling through FGFR3, is necessary and sufficient for inner pillar cell differentiation (Hayashi et al., 2007; Jacques et al., 2007; Puligilla et al., 2007; Zelarayan et al., 2007), prompted us to evaluate supporting cells in cross sections from dnFGFR2b-expressing samples. We first co-stained E11.5-E18.5 DOX-induced samples for NGFR (p75^NTR, marking the inner pillar cell) and either CALB2 or CD44 (marking the outer pillar cell in addition to lateral Kölliker’s organ, the outer sulcus and the stria vascularis). The control stained for NGFR and CALB2 had a single inner hair cell and an inner pillar cell (Fig. 4A). dnFGFR2b samples (n = 3) had 2 or 3 CALB2-positive cells, either widely separated (Fig. 4B) or closely spaced (Fig. 4C), and each had an adjacent NGFR-positive cell. The presumptive ectopic inner pillar cell was found either lateral (Fig. 4B) or medial (Fig. 4C) to the inner hair cell-like cell. Control samples co-stained for NGFR and CD44 showed a single inner pillar cell and a single outer pillar cell. In addition, the lateral half of Kölliker’s organ and the entire outer sulcus and stria vascularis also expressed CD44 (Fig. 4D, n = 4). Four of 5 dnFGFR2b samples had two NGFR-positive cells and again these could be either separated by a few cells (Fig. 4E) or closely spaced (Fig. 4F). One of 5 had only a single (endogenous) inner pillar cell (not shown). None of the dnFGFR2b samples had an outer pillar cell (Fig. 4E and F). As expected from our studies of Fgf10^−/− mutants (Urness et al., 2015), all 5 dnFGFR2b samples lacked the outer sulcus and retained some stria vascularis tissue, but in addition, we found that all 5 had a greatly reduced CD44-positive domain in lateral Kölliker’s organ (Fig. 4E and F). As expected, E18.5 controls co-stained with SOX2 (a nuclear marker of all supporting cells and lateral Kölliker’s organ at this stage) and CD44 showed co-labeling of lateral Kölliker’s organ and the outer pillar cell (Fig. 4G, n = 4). In contrast, dnFGFR2b-expressing cochleae had a reduced region of SOX2/CD44 co-staining in lateral Kölliker’s organ and no evidence of co-staining in the outer compartment, suggesting that Kölliker’s organ is reduced and the outer pillar cell is missing (Fig. 4H and I; n = 2). Also as expected, controls co-stained with NGFR and PROX1 showed a single inner pillar cell and 5 PROX1-positive nuclei marking both pillar cells and all 3 Deiters’ cells (Fig. 4J, n = 3). In contrast, 4 of 5 dnFGFR2b samples had two NGFR-positive cells (all with PROX1-positive nuclei) and the total number of PROX1-positive supporting cells varied from 5 to 8 per cross section (Fig. 4K and L). To assess Deiters’ cell differentiation, we co-stained for S100A1, which also labels the inner hair cell, inner phalangeal cell and the outer pillar cell (Hayashi et al., 2007; Kaiser et al., 2022; Urness et al., 2015) and for PROX1. As expected, controls all showed a single inner hair cell, inner phalangeal cell, outer pillar cell and 3 Deiters’ cells (Fig. 4M, n = 3). The dnFGFR2b phenotype varied. We found S100A1-positive/PROX1-negative cell duplications (presumptive inner hair cell-like cells) in all 3 samples (Fig. 4N–P). In addition, in one case there was no S100A1 in PROX1-positive cells (Fig. 4N), in another case there was weak S100A1 expression in only 2 of at least 11 PROX1-positive cells (Fig. 4O) and one sample had weak S100A1 expression in 2 of 4 PROX1-positive cells (Fig. 4P). The PROX1-positive/S100A1-negative cells adjacent to inner hair cells are likely inner pillar cells, but many of the PROX1-positive cells had not differentiated as Deiters’ cells.

Fig. 4.

E11.5-E18.5 induction of dnFGFR2b causes ectopic inner pillar cell differentiation and suppresses differentiation of outer pillar cells, lateral Kölliker’s organ and Deiters’ cells. (A–P) Basal cochlear duct sections from embryos exposed to DOX from E11.5-E18.5 and immunostained as indicated. Scale bar in panel A (50 μm) also applies to panels B–P. Medial-to-lateral orientation axis (double headed arrow) in A-C are relative to the organ of Corti and apply to all corresponding control and experimental panels. (J–P) PROX1-positive supporting cells are demarked with a dot. (Q,R) Cochlear ducts from embryos exposed to DOX from E11.5-E15.5 and subjected to RNAscope ISH using an Fgf20 probe. Scale bar in panel Q (200 μm) also applies to panel R. Abbreviations not used in previous figures: Dc, Deiters’ cells; ipc, inner pillar cell; iphc, inner phalangeal cell, opc, outer pillar cell.

Since FGF20 signaling through FGFR1 is required for both outer hair cell and outer supporting cell differentiation (Hayashi et al., 2008; Huh et al., 2012), we asked whether Fgf20 expression was reduced by E11.5-E15.5 dnFGFR2b induction. However, both control (n = 2) and experimental (n = 3) genotypes expressed Fgf20 (Fig. 4Q and R). If anything, Fgf20 expression was more robust in the dnFGFR2b samples, perhaps reflecting a delay in the start of basal-to-apical down-regulation expected at this stage (Hayashi et al., 2008). By E18.5, however, there was little Fgf20 expression in either control or dnFGFR2b-expressing cochlear ducts (Fig. S2). This shows that FGFR2b/1b ligands do not regulate outer compartment differentiation by simple downregulation of FGF20 signaling.

2.5. Inducing dnFGFR2b has no major effect on prosensory domain formation

Since dnFGFR2-expressing cochleae had such severe disruptions of both sensory and supporting cell differentiation, we asked whether these phenotypes were reflected in abnormal development of the prosensory domain, from which both cell types arise in a basal to apical progression. We first co-stained E11.5-E18.5 DOX-induced samples with SOX2, which is necessary and sufficient for prosensory development and marks the prosensory domain (Kiernan et al., 2005; Mak et al., 2009; Pan et al., 2013), becoming restricted by E18.5 in the base to lateral Kölliker’s organ and supporting cells, and POU4F3, the pan-hair cell marker that is expressed as soon as hair cells differentiate. All controls had the expected staining pattern with the base showing the fully mature pattern (Fig. 5A, n = 4). dnFGFR2b samples were like controls with respect to SOX2 expression but had reduced numbers of basal hair cells (0–3 per cross section), which were either separated at the medial and lateral edges of the presumptive organ of Corti or located immediately adjacent to one another (Fig. 5B and C; n = 4). Controls treated from E11.5-E15.5 had the characteristic pattern of multiple layers of SOX2-positive cells and 4 developing POU4F3-positive hair cells in the base. At this stage the nascent hair cells still expressed SOX2 (Hume et al., 2007) (Fig. 5D, n = 2). Similarly to controls, the E15.5 basal dnFGFR2b cochleae had a thickened SOX2-positive domain with multiple layers of cells and had SOX2/POU4F3-expressing hair cells, but their number per cross section varied from 1 to 3 (Fig. 5E and F; n = 3). Thus, SOX2 expression in the dnFGFR2b prosensory domain seemed relatively normal even when the reduction in hair cell differentiation was already apparent at E15.5.

Fig. 5.

dnFGFR2b has no major effect on prosensory domain development. (A–C) Basal cochlear duct cross sections of control and experimental embryos induced from E11.5-E18.5 or (D–F) E11.5-E15.5 and immunostained for SOX2 and POU4F3. (G,H) Cochlear duct sections from embryos induced from E11.5-E13.5, (I,J) E11.5-E14.0 or (K,L) E11.5-E14.5 and stained for SOX2 and CDKN1B. The individual stains and the merge are shown for both apical and basal sections. DAPI only is shown in low-magnification views (G–L). Dotted lines demark areas of cell cycle exit (CDKN1B-positive). Scale bar in A (50 μm) applies to B-F. Scale bar in G (200 μm) applies to H-L. Scale bar in Ga (50 μm) applies to Ha-La and Gb-Lb. Abbreviations not used previously: m, middle cochlear cross section, psd, prosensory domain; sc, support cells.

To address the possibility that a delay in cell cycle exit in the prosensory domain might account for the reduced hair cell number in dnFGFR2b samples, we induced control and dnFGFR2b embryos from E11.5 to E13.5, E14.0 or E14.5 and co-stained cochlear cross sections with antibodies directed against SOX2 to label the prosensory domain and CDKN1B (p27^KIP1), to label cells that had exited the cell cycle and were competent to differentiate. At E13.5, control (n = 2) and dnFGFR2b (n = 4) samples each had 2 cochlear cross sections, apical and basal (Fig. 5G and H). SOX2 expression was readily apparent in discrete domains of both cross sections in both control and dnFGFR2b samples (Fig. 5Ga,Gb,Ha,Hb). As expected, controls had a discrete zone of CDKN1B partially overlapping SOX2 in the apical cross section but had less evident CDKN1B expression in the basal cross section (Fig. 5Ga,Gb). Three of 4 dnFGFR2b samples had no obvious CDKN1B in either cross section (not shown), but the other expressed CDKN1B in both cross sections (Fig. 5Ha,Hb). At E14.0 and E14.5, the controls had lengthened and coiled enough to have three cross sections (Fig. 5I and K), while the dnFGFR2b samples still had only two (Fig. 5J and L). Nevertheless, both genotypes expressed CDKN1B within the SOX2-positive domain at all levels of the cochleae (Fig. 5Ia,Ja,Ka,La,Ib,Jb,Kb,Lb; n = 3 at each stage). This suggests that while there may be a slight delay in initiation of cell cycle exit in some dnFGFR2b samples, there are no major differences between the two genotypes after E14.0 with respect to cell cycle exit and establishment of the prosensory domain.

2.6. FGFR2b/1b ligands are required for spiral ganglion morphogenesis and for differentiation of CALB2-positive type 1 neurons, but not for survival or proliferation of ganglion cells

To assess spiral ganglion development, we prepared control and dnFGFR2b samples induced from E11.5-E18.5 and stained cochlear cross sections with antibodies directed against RBFOX3 (NEUN) or PROX1 to reveal spiral ganglion neurons, derived from the otic placode, and SOX10 or SOX2 to reveal glia, derived from the neural crest. Like the control ganglia (Fig. 6A and C), the dnFGFR2b-expressing samples each contained both neurons and glia, however, similarly to the histology (Fig. 1), the dnFGFR2b-expressing ganglia appeared smaller and were clumpy (Fig. 6B, n = 2; 6D n = 3). Surprisingly, dnFGFR2b ganglia co-stained with NGFR and CALB2 antibodies, which co-label most spiral ganglion neurons in controls (Fig. 6E), had an extreme depletion of CALB2-positive neurons (Fig. 6F; n = 2), suggesting a requirement for FGFR2b/1b ligands in Type 1A and/or Type 1B spiral ganglion neuronal subtype differentiation. This depletion was noted in all other staining combinations involving CALB2 and cochlear duct markers (not shown). To determine whether the paucity of CALB2-positive spiral ganglion neurons also applied to Fgf10^−/− mutants we co-stained them for NGFR and CALB2 (n = 2) but saw no obvious differences relative to littermate controls (Fig. 6G and H). This suggests that as with the sensory epithelial phenotypes, the dnFGFR2b spiral ganglion neuron abnormalities arise from inhibition of both FGF10 and FGF3.

Fig. 6.

E11.5-E18.5 induction of dnFGFR2b disrupts spiral ganglion morphogenesis and differentiation of CALB2-positive neurons but not through effects on cell death or proliferation. (A–F) Basal spiral ganglion cross sections of control and experimental embryos induced from E11.5-E18.5 and immunostained with antibodies as indicated above each pair. (G,H) E18.5 control (either Fgf10^+/+ or Fgf10^−/+) and Fgf10^−/− basal spiral ganglion sections immunostained as indicated. (I,J) Oblique coronal sections of cochleae from embryos induced from E11.5-E13.5 and stained with cCASP3 and TUJ1 antibodies. (K,L) Sagittal cochlear sections from embryos induced from E11.5-E15.5 or (M,N) E11.5-E16.5 and stained with cCASP3 and GATA3 antibodies. Dotted lines encircle the spiral ganglion. (Q,R) Oblique coronal sections of otic region of embryos induced from E11.5-E12.5 and stained with pHH3 and TUJ1 antibodies. Dotted lines demark quantified regions. Abbreviation not used previously: mes, mesenchyme; sg. Scale bar in panel A (50 μm) applies to panels B–H; scale bar in Panel I (200 μm) applies to panels J–N; scale bar in panel Q (100 μm) applies to panel R. (O) Quantification of average sg area, (P) dying cell density and (S) mitotic cell density in 3 regions. Multiple unpaired t-tests of control vs dnFGFR2b samples were adjusted using the Holm-Šidák method. Asterisk in O: AdjP = 0.0035. Asterisk in S: AdjP = 0.042.

To determine whether the reduced spiral ganglion size in E11.5–18.5 induced dnFGFR2b samples (Fig. 1I) was due to an increase in cell death we stained control and dnFGFR2b samples induced from E11.5-E13.5 with antibodies directed against cleaved-Caspase3 (cCasp3) and TUJ1 (Fig. 6I and J; n = 3 each), and those induced from E11.5-E15.5 and E11.5-E16.5 with cCasp3 and GATA3 (Fig. 6K–N; n = 3 each). A small reduction in the experimental spiral ganglion area only reached statistical significance in E11.5-E16.5 samples (Fig. 6O). Very few cCasp3-positive cells were detected in either sample type at any stage. Indeed, although the dnFGFR2b samples tended to have a higher spiral ganglion dying cell density than the controls, the differences were not statistically significant at any stage (Fig. 6P). Thus, FGFR2b ligands may not play a major role in spiral ganglion cell survival.

To assess whether reduced proliferation might account for the ganglion size differences we prepared sections from control and dnFGFR2b samples induced from E11.5-E12.5 and co-stained them with antibodies directed against phospho-Histone H3 (pHH3), the mitotic marker, and TUJ1 to label the spiral ganglion neurons (Fig. 6Q and R; n = 3 each). We measured the areas of the spiral ganglion and cochlear duct epithelium, and counted mitotic cells in each area, as well as in a fixed box size in mesenchyme adjacent to the cochlear duct. There was no significant reduction in mitotic cell numbers in either the spiral ganglion or mesenchyme of dnFGFR2b samples, but we did find a significant reduction in the cochlear duct (Fig. 6S). Thus, FGFR2b/1b ligands may promote proliferation of the epithelium between E11.5 and E12.5, but not of the ganglion or mesenchyme.

2.7. The critical periods for FGFR2b/1b ligands in controlling cochlear epithelial and ganglion development fall between E11.5-E13.5 and E11.5-E14.5, respectively

To determine when during the E11.5-E18.5 interval FGFR2b/1b ligands are required for cochlear duct and spiral ganglion development, we induced dnFGFR2b on progressively later days of development from E12.5 to E15.5, sustained the induction through E18.5, and immunostained cochlear cross sections with many of the same antibody combinations used for the E11.5-E18.5 inductions. E12.5-E18.5 Rosa26^rtTA/+ samples served as controls for all induction intervals, except where noted, and an E11.5-E18.5 control was substituted. Controls stained with the cochlear duct marker combinations MYO7A/POU4F3, MYO7A/CALB2, NGFR/CALB2, NGFR/CD44, NGFR/PROX1 or SOX2/CD44 all appeared normal with respect to total hair cell number, inner hair cell, inner pillar cell, outer pillar cell and other supporting cell identities (Fig. 7A₁,B₁,C₁,D₁,E₁,F₁). Samples expressing dnFGFR2b between E12.5 and E18.5 had very similar staining patterns to those induced from E11.5-E18.5 (Figs. 3 and 4), with a reduction in total hair cell number, duplication of inner hair cell and inner pillar cell identities, loss of outer pillar cell identity and variability (not shown) in the total number of supporting cells (Fig. 7A₂,B₂,C₂,D₂,E₂,F₂). The stain for CD44 highlighted loss of the outer sulcus and a reduction in Kölliker’s organ (Fig. 7D₂,F₂). Strikingly, although E13.5-E18.5 inductions caused loss of the outer sulcus, and Kölliker’s organ did not appear entirely normal (Fig. 7D₃,F₃), organ of Corti differentiation (hair cell fate and number) was unperturbed (Fig. 7A₃,B₃,C₃,D₃,E₃,F₃). Similar results were seen for the E14.5-E18.5 inductions (Fig. 7A₄,B₄,C₄,D₄,E₄,F₄), and even the outer sulcus (Fig. 7D₅,F₅) and Reissner’s membrane (not shown) appeared normal in samples induced from E15.5-E18.5. These results suggest that there is a relatively short critical period, from E11.5-E13.5, during which FGFR2b/1b ligands are required for cell identity in the organ of Corti, and that Kölliker’s organ and the outer sulcus require FGFR2b/1b ligands over a longer duration.

Fig. 7.

Critical periods for FGFR2b/1b ligands in controlling cochlear epithelial and ganglion development fall between E11.5-E13.5 and E11.5-E14.5, respectively. Embryos were exposed to DOX starting at E12.5, E13.5, E14.5 or E15.5 and cochleae were collected at E18.5. Sagittal head sections were immunostained as indicated in each row. Genotypes and DOX treatment are shown above each column. Only E12.5-E18.5 controls are shown unless an E11.5-E18.5 control was substituted as indicated. Scale bar in A₁ (50 μm) applies to all panels.

The spiral ganglion was assessed following sequentially later dnFGFR2b inductions using RBFOX3/SOX10, NGFR/PROX1 and NGFR/CALB2, all of which showed the expected staining patterns in controls (Fig. 7G₁,H₁,I₁). The dnFGFR2b samples induced from E12.5-E18.5 or E13.5-E18.5 had a basal spiral ganglion morphology that was not as clumpy as that of E11.5-E18.5 samples (Fig. 6B–D,F) and most notably, the number of CALB2-positive spiral ganglion neurons was minimal (Fig. 7G_2–3,H_2–3,I_2–3). However, the spiral ganglion neurons appeared relatively normal in samples expressing dnFGFR2b from E14.5–18.5 or E15.5-E18.5 (Fig. 7G_4–5,H_4–5,I_4–5). These data show that the spiral ganglion requires FGFR2b/1b signaling over a somewhat longer period (E11.5-E14.5) for normal morphology and CALB2-positive neuronal subtype differentiation than for organ of Corti differentiation.

2.8. Early transient exposure to dnFGFR2b is sufficient to cause cochlear epithelial and ganglion perturbations

Since exposure to dnFGFR2b from E11.5-E12.5 appeared critical in generating a variety of cochlear abnormalities, we asked whether such an early and limited exposure was sufficient to cause any of the phenotypes. To this end, we injected pregnant females with DOX at E11.5 and supplied DOX chow transiently, for only 24 h, before replacing it with standard chow. Embryos were harvested at E15.5 for inner ear paint-filling and at E18.5 for immunostaining of cochlear cross sections. Relative to controls (Fig. 8A) and similarly to ears induced continuously from E11.5-E15.5 (Fig. 1B), the transiently induced dnFGFR2b samples have a shortened, narrow and less well coiled cochlea (Fig. 8B; n = 5); however, the narrowing of the cochlear duct did not appear as extreme as in the continuously inhibited ears (Fig. 1B). Immunostaining of E18.5 basal cochlear cross sections from 4 transiently induced dnFGFR2b cochleae showed even more inter- and intra-sample variability than that of continuously induced samples. Based on MYO7A or POU4F3 staining, the total hair cell number/basal cross section was generally greater than that for the continuously induced samples, with a range of 1–7 per cross section (Fig. 8C–F), most sections having 4 or 5 hair cells. Virtually all dnFGFR2b-exposed hair cells expressed some CALB2, with the strongest expression in the bona fide inner hair cell and in a lateral, flask-shaped inner hair cell-like cell found in several sections of 3 of the 4 dnFGFR2b cochleae (Fig. 8D,G,H). Lateral inner hair cell-like cells expressing vGLUT3 or CALB1 were seen in only one of the transient dnFGFR2b-expressing samples (Fig. 8E,L), but in all cases of lateral inner hair cells, that cell was innervated like a normal inner hair cell (Fig. 8F). None of the dnFGFR2b organs of Corti expressed the outer hair cell marker, BCL11B (Fig. 8G). Thus, induction of dnFGFR2b for only 24 h does not block overall hair cell differentiation as extensively as with continuous induction, but it is sufficient to induce an ectopic inner hair cell-like cell in the outer compartment and block outer hair cell differentiation.

Fig. 8.

Transient exposure to dnFGFR2b from E11.5-E12.5 is sufficient to cause cochlear epithelial and ganglion perturbations. (A,B) Lateral views of paint-filled inner ears from embryos exposed to DOX from E11.5 to E12.5 and then allowed to develop normally until E15.5. Genotypes are indicated in the upper right. Scale bar in panel A (500 μm) also applies to panel B. (C–O) Sagittal head sections prepared from embryos exposed to DOX from E11.5 to E12.5 and then allowed to develop normally until E18.5 were immunostained as indicated atop each cochlear cross-section pair. Genotypes are indicated to the left of each row. The scale bar in panel C (50 μm) also applies to panels D–O. Unlabeled symbols: (F) arrowheads, innervation of inner and inner-like hair cells; (G,H) asterisks, lateral hair cells with some CALB2 expression; caret, inner pillar cell; (J,K) dots, supporting cells. (P) Model depicting proposed FGF3 and FGF10 signaling occurring at E11.5-E12.5 and leading, after continuous induction, to the phenotypes seen at E18.5, most of which are also characteristic of transient induction from E11.5-E12.5. Abbreviations not used previously: mKo, medial Kölliker’s organ; lKo, lateral Kölliker’s organ.

Sections in which a lateral CALB2-positive cell was detected tended to show ectopic NGFR staining in the adjacent cell, indicative of inner pillar cell differentiation, (Fig. 8H and I), but these ectopic NGFR-positive cells appeared smaller and were stained more weakly than those detected following continuous DOX induction (Fig. 4B,C,E,F). Nevertheless, all 4 dnFGFR2b samples had sections with ectopic NGFR in cells lateral to the endogenous inner pillar cell, suggesting that the ectopic inner hair cell-like cells induced inner pillar cell differentiation. Outer pillar cell differentiation as assessed by CD44 expression was quite different in transiently induced samples than in continuously induced samples. All 4 dnFGFR2b samples had one CD44-positive cell per section in the position expected for an outer pillar cell (Fig. 8I and L) and all samples had at least one section with a total of 2–3 CD44-positive cells (Fig. 8I). Thus, transient dnFGFR2b expression induces ectopic outer pillar cell differentiation, whereas continuous dnFGFR2b prevents it. To assess Deiters’ cell differentiation, we compared PROX1 and S100A1 expression. Similarly to continuous induction, all transiently induced dnFGFR2b samples had some basal sections with excess PROX1-positive cells. The total number of PROX1 cells/section ranged from 5 to 10, with most having 5–8 (Fig. 8C,J,K). In this case, however, the excess PROX1-positive cells did not appear to express S100A1, whereas cells in the hair cell layer were S100A1-positive, i.e. somewhat inner hair cell-like (Fig. 8K). This, together with the abundance of CALB2-positive cells, suggests that developing hair cells are differentially sensitive to the duration of FGFR2b/1b signaling inhibition.

Markers of Kölliker’s organ behaved similarly in all dnFGFR2b samples following transient or continuous induction. The lateral CD44-positive domain was reduced, but the medial CALB1-positive domain was intact (Fig. 8L). This suggests that transient dnFGFR2b expression is sufficient to reduce lateral Kölliker’s organ. In addition, and as for the continuous induction, all transiently induced dnFGFR2b samples lost CD44 staining between the organ of Corti and the stria vascularis, showing an early requirement for FGFR2b/1b signaling to initiate outer sulcus development, consistent with developmental studies of Fgf10^−/− mutants (Urness et al., 2015).

Finally, we assessed spiral ganglion differentiation in control and experimental samples and found that as with continuous induction, transient induction of dnFGFR2b reduced spiral ganglion size and led to a clumpy morphology (Fig. 8M–O). Also, the dnFGFR2b ganglia were comprised of both NEUN-positive neurons and SOX10-positive glia (Fig. 8M) and the dnFGFR2b spiral ganglion neurons expressed both NGFR and PROX1 (Fig. 8N) but were severely depleted of CALB2-positive cells (Fig. 8O). Thus, only a brief, early inhibition of FGFR2b ligands is sufficient to cause many of the epithelial phenotypes and all the ganglion phenotypes seen following continuous inhibition.

3. Discussion

By employing an inducible ligand trap we identified several new requirements for FGFR2b/1b ligands in cochlear sensory and neural cell subtype differentiation, uncovering a surprising involvement of FGF3, particularly with respect to spiral ganglion development, and establishing early critical periods for FGFR2b/1b signaling in controlling cochlear cell fates. These findings provide new avenues for manipulating cell fate during investigations of hearing restoration.

3.1. Differentiation of the outer compartment and lateral Kölliker’s organ requires FGFR2b/1b ligands signaling from the inner compartment

Inducing dnFGFR2b at E11.5 and maintaining expression through E18.5 had profound effects on the organ of Corti not seen in E18.5 Fgf10^−/− mutants, pointing to new requirements for FGF10, together with FGF3, in hair cell and supporting cell subtype differentiation. These included a reduction in total hair cell numbers. Since normal inner hair cells were always detected as expected at the medial boundary of the organ of Corti, the deficit in total hair cell numbers must come from the outer compartment. The dnFGFR2b ears also had ectopic, laterally located inner hair cell-like cells and showed a complete failure of outer hair cell differentiation. Supporting cells were also affected, such that there were additional PROX1-positive supporting cells, including ectopic inner pillar-like cells located adjacent to the ectopic inner hair cell-like cells, which produce Fgf8, encoding the signal required for inner pillar cell differentiation. In addition, there was no outer pillar cell differentiation and reduced Deiters’ cell differentiation. Finally, the most lateral region of Kölliker’s organ was reduced.

Since hair cells and supporting cells differentiate from a common prosensory domain, we entertained the idea that FGFR2b/1b signaling controls the ratio of hair cell to supporting cell differentiation, but there did not seem to be a one-to-one replacement of hair cells with supporting cells, instead, the increase in PROX1-positive supporting cells was highly variable, but usually larger than the decrease in POU4F3- or MYO7A-positive hair cells. Thus, it seems likely that FGFR2b/1b signaling normally acts independently on hair cell and supporting cell differentiation. It promotes hair cell differentiation generally, but represses the response of outer compartment hair cells to inner hair cell-inducing signals, and suppresses supporting cell production generally, but promotes outer compartment supporting cell subtype differentiation.

Since cultured cochleae treated with antibodies directed against FGF20 (a “c”-type ligand) and Fgf20^−/− mutants also display abnormalities/loss of outer compartment hair cells and supporting cells starting prior to E14.5 (Hayashi et al., 2008; Huh et al., 2012), we examined Fgf20 expression following dnFGFR2b induction, but given the much more severe sensory phenotype caused by these methods of reducing FGF20 signaling, it was perhaps unsurprising that Fgf20 expression was not down-regulated in dnFGFR2b-expressing cochleae. Instead, it appeared upregulated at E15.5, though not at E18.5. Since normal mid-gestation cultured cochleae treated with FGF9, which has similar activity to FGF20, have no abnormalities of hair or supporting cell differentiation, we suggest that FGF20 may be functioning on processes upstream or independent of those of FGFR2b/1b ligands. Like dnFGFR2b-induced samples, Insm1^−/− mutants also have sporadic inner hair cell-like cells in the outer compartment (Wiwatpanit et al., 2018) and indeed we found no INSM1 expression following dnFGFR2b induction. However, this is not necessarily a specific effect on Insm1 expression because the other outer hair cell marker we tested, BCL11b, was also abrogated. Instead, we suggest that FGFR2b/1b signaling is required after hair cell specification and the initiation of inner hair cell subtype differentiation to initiate outer hair cell differentiation.

The normal expression patterns of Fgf3 and Fgf10 in the cochlear duct epithelium at E11.5-E12.5 bear on the ultimate dnFGFR2b cochlear duct phenotypes (Fig. 8P and S1). Fgf10 is highly expressed in the lateral region of the greater epithelial ridge, immediately adjacent to the developing organ of Corti, whereas Fgf3 is expressed at much lower levels overlapping with Fgf10 in the most lateral portion of its domain. Apparently, cells in the greater epithelial ridge domain that express both Fgfs require this expression to fully differentiate and express CD44, suggesting an autocrine/juxtacrine signaling pathway. The lack of any effect of dnFGFR2b expression on inner hair cell and inner pillar cell development suggests that their progenitors, which should in theory be exposed to the highest levels of FGF10 and FGF3, are already specified by a different pathway and either do not express the appropriate receptors at E11.5 or are enriched in negative regulators of signaling. What little data exist on Fgfr2b or Fgfr1b isoform-specific expression around E11.5 show Fgfr2b expression in non-sensory regions of the cochlear duct and Fgfr1b in sensory regions, with both transcripts found in the mesenchyme (Pirvola et al., 2000, 2002) and the negative regulators, Dusp6, Spry1, and Spry2 also in the developing sensory regions (Urness et al., 2008; Zhang et al., 2001). Thus, either explanation is conceivable. In contrast, progenitor cells of the outer compartment, which presumably experience lower levels of FGF3 and FGF10 than those of the inner compartment, respond to dnFGFR2b expression by becoming supporting cells in preference to hair cells and then fail to fully differentiate into appropriate supporting cell subtypes. The more severe block to differentiation of the outer pillar cell relative to Deiters’ cells suggests a greater dependence on FGFR2b/1b signaling in cells located closer to the ligands to complete differentiation. Together, these observations suggest that during normal development, there is a graded FGFR2b/1b paracrine signaling pathway functioning in the outer compartment to regulate appropriate hair cell and supporting cell subtype differentiation and that efforts to promote outer hair cell differentiation following loss of these cells may require careful spatial and temporal manipulation of FGFR2b/1b signaling levels.

3.2. FGFR2b/1b ligands are required for normal spiral ganglion size and morphology as well as for type 1 spiral ganglion neuronal subtype differentiation

We found that FGFR2b/1b ligands are required for normal spiral ganglion size and morphology, as well as for differentiation of CALB2-positive Type 1 spiral ganglion neurons. The critical role of FGF3 in the spiral ganglion phenotypes of dnFGFR2b embryos is supported by the relative lack of spiral ganglion phenotypes in Fgf10^−/− mutants (this study and Pauley et al., 2003) and may be more in accordance with the spiral ganglion phenotypes described for germline Fgfr2b^−/− mutants, which have a rudimentary otic ganglion, high levels of apoptosis in the E13 spiral ganglion and a failure of spiral ganglion neurons to innervate the cochlear duct (Pirvola et al., 2000). This suggests that FGF10 and FGF3 do have roles in spiral ganglion cell survival, but that this function may be most critical before E11.5, when we started dnFGFR2b inductions. Although Fgf3 is expressed in the otic ganglion at E9.5 (Hatch et al., 2007; McKay et al., 1996; Urness et al., 2018), and loss of FGF3 signaling (together with that of FGF10, which is expressed at high levels throughout spiral ganglion neuronal development) likely contribute to the Fgfr2b^−/− phenotype, Fgf3 is rapidly shut down and not detectably expressed in the spiral ganglion during the dnFGFR2b induction periods we employed, begging the question of how FGF3 could contribute to spiral ganglion development during mid-gestation. One possibility is that FGF3 protein is very stable in tissue surrounding the spiral ganglion after expression of its mRNA ceases, so inhibiting its function (together with FGF10) starting at E11.5 results in the observed phenotypes. Another possibility is that since the ganglion at E11.5 is so physically close to the cochlear duct, where Fgf3 is clearly expressed at E11.5 and beyond, it may be that epithelial FGF3 plays a role in spiral ganglion development (Fig. 8P). This issue could be addressed by restricting dnFGFR2b expression to ganglion vs epithelial locations.

We were unable to discern significant changes in spiral ganglion proliferation or death at the stages examined following dnFGFR2b induction. This does not rule out the possibility of small, cumulative effects on these processes, which could reduce the ultimate spiral ganglion size, but it does suggest that other explanations should be considered and investigated. For example, FGFs can play roles in cell adhesion, and this has been proposed to account for FGF9 function in the development of Reissner’s membrane (Pirvola et al., 2004). It is possible that expression of dnFGFR2b disrupts adhesion of the spiral ganglion neurons to the mesenchyme and as the cochlear duct and spiral ganglion elongate during development, the inhibited spiral ganglion is literally pulled apart. Thus, it could be profitable for future studies to examine adhesive interactions in the developing dnFGFR2b spiral ganglion.

The depletion of CALB2-positive spiral ganglion neurons in the dnFGFR2b samples suggests that FGFR2b/1b ligands play a role in spiral ganglion neuronal subtype specification and further studies are needed to establish the relative levels and developmental appearance of other spiral ganglion neuronal subtypes in dnFGFR2b embryos. Nevertheless, these results are the first to establish a specific signal linked to spiral ganglion neuronal subtype differentiation.

3.3. FGFR2b/1b ligands are required for cell subtype differentiation at a very early stage of cochlear development

Due to the inducibility of dnFGFR2b, we were able to show that starting dnFGFR2b expression at stages later than E11.5 recapitulated the inhibitions starting at E11.5, at least up to a point, at which development of the cochlear duct sensory epithelium and spiral ganglion neurons became insensitive to dnFGFR2b. Indeed, inhibiting signaling transiently from E11.5-E12.5 while not identical to the sustained inhibition, recapitulated the key phenotypes of ectopic lateral inner hair cell-like and inner pillar cell-like differentiation and failure to differentiate CALB2-positive spiral ganglion neurons. These observations define a critical window of E11.5-E12.5 in which FGFR2b/1b signaling functions to control cochlear cell subtype differentiation. This window is 3 days prior to the onset of overt inner hair cell differentiation (Kolla et al., 2020; Wiwatpanit et al., 2018) and 2 days prior to the first spiral ganglion subtype differentiation (Petitpre et al., 2022; Sanders and Kelley, 2022). Thus, future studies aimed at understanding the effectors of FGFR2b/1b signaling responsible for determining cochlear cell subtype differentiation should be aimed at this interval and the development of organoid models of cochlear development should account for the levels and timing of FGFR2b/1b signaling.

4. Materials and methods

4.1. Mouse models and induction of dnFGFR2b

Mice were housed, bred and euthanized according to protocols approved by the University of Utah Institutional Animal Care and Use Committee. All strains were maintained on mixed backgrounds mainly comprising C57BL/6 and 129S6 (The Jackson Laboratory or Taconic). The germline-recombined Rosa26^rtTA allele [derived from Gt(ROSA) 26^{Sortm1(rtTA, EGFP)Nagy}; MGI:3583817] (Belteki et al., 2005; Parsa et al., 2008), which expresses the reverse tetracycline transactivator ubiquitously, and the Tg(tetO-dnFgfr2b) allele [Tg(tetO-Fgfr2b/Igh)1.3Jaw; MGI:5582625] (Hokuto et al., 2003), further abbreviated to dnFGFR2b in the text and figures, were obtained and genotyped as described (Urness et al., 2018). dnFGFR2b expression was induced by intraperitoneal injection of pregnant females with 0.1 ml/10 g body weight of 0.15 mg/ml doxycycline hyclate (Sigma-Aldrich) prepared in PBS, followed by provision of DOX-chow (200 mg/kg, Custom Animal Diets, LLC #AD3006S or Bio-Serv #S3888) ad libitum for the designated time periods. For the transient (24-h) induction, we did not assess the time course of dnFGFR2b depletion following resumption of standard chow but based on studies of the limb (Danopoulos et al., 2013) it is unlikely to persist beyond 24 h after chow exchange. Fgf10 null mutants were generated from intercrosses of heterozygotes [Fgf10^tm1.1Sms/+; MGI:3526181) and genotyped as described previously (Urness et al., 2015).

4.2. Paint filling and histology of embryonic inner ears

Filling of E15.5 embryonic inner ears with latex paint and imaging were performed as described (Urness et al., 2015). E18.5 heads for histologic staining were bisected in the sagittal plane, fixed, paraffin embedded, sectioned and stained with Hematoxylin and Eosin as described previously, except that decalcification was omitted (Mansour et al., 2009). Areas of the scalae media and sg were measured and averaged from photographs of three central sections of each of three control and experimental samples. The data were graphed and analyzed using GraphPad Prism10 software (multiple t-tests with a Holm-Šídák correction).

4.3. Immunostaining of tissue sections

In general, embryo heads were bisected in the sagittal plane, fixed, embedded in sucrose/gelatin, cryosectioned at 8 μm, immunostained and imaged as described previously (Urness et al., 2015, 2018, 2020). Embryos used for CDKN1B/SOX2 staining were sectioned in the transverse plane, E13.5 embryos for cell death analysis and E12.5 embryos for proliferation analysis were sectioned in a plane between coronal and transverse. The antibodies, their sources and working dilutions are listed in Table S1. Optimal detection of BCL11B, CALB1, CDKN1B, vGLUT3 and INSM1 required pretreatment of rehydrated slides for 12–14 min in boiling 10 mM citrate (pH 6.0) followed by gradual cooling at room temperature for 30 min. Slides were then rinsed in PBS and blocked in 10 % of the appropriate normal serum. Mouse On Mouse Blocking Reagent (Vector Labs, MKB-2213) was applied to sections prior to incubation with primary antibodies generated in mouse. Fluorescent secondary antibodies, which were raised in goat or donkey, were obtained from Invitrogen, applied at 1:400, and washed. DAPI was included in the mounting medium (Vector Laboratories, H1500–10).

4.4. Quantification of mitotic or dying cells

To quantify mitotic cells after induction of dnFGFR2b for 24 h (E11.5-E12.5), embryos were fixed and sectioned as indicated above. The sections from three each of control and experimental embryos were arrayed sequentially with the entire otic region on a single slide. Slides were stained with antibodies directed against phospho-Histone H3 and TUJ1. Each section was photographed, and the photographs were aligned to identify 10 alternating sections spanning the cochlear duct and ganglion. The files with each section to be quantified were then coded and placed in a random order. A person blinded to the code counted mitotic cells in the cochlear duct, the ganglion and in a fixed region of mesenchyme adjacent to the cochlear duct and calculated the number of mitotic cells in each area. The data were then unblinded and the average number of mitotic cells per area was calculated for each sample. GraphPad Prism 10 software was used to graph the data and determine the statistical significance of differences between control and experimental samples (multiple t-tests with the Holm-Šídák correction for multiple comparisons).

Quantification of dying cells in the spiral ganglion after inducing dnFGFR2b expression from E11.5-E13.5 was similar to the mitotic analysis (n = 3 control and 3 experimental samples) except that dying cells were detected using anti-cleaved-Caspase-3 paired with TUJ1 to define the ganglion and we did not analyze the epithelium or mesenchyme, which appeared to have almost no dying cells. Dying cells/area was averaged from 6 matched and alternating sections for each sample. To quantify dying cells in the spiral ganglion after inducing dnFGFR2b expression from E11.5-E15.5 or E16.5, we analyzed sagittal sections and employed antibodies directed against cCasp3 and GATA3, a marker of spiral ganglion neurons. We attempted to blind the E15.5 samples, but this was not very effective, and we did not blind the E16.5 samples because the genotypes were clear from the section morphology. In both cases, we selected 3 alternating sections from the most central region of the cochlear duct to count and average mitotic cells per ganglion area. The statistical analysis of the data was the same as for the mitotic analysis.

4.5. mRNA detection by classic in situ hybridization

Paraffin-embedded half heads were sectioned in the sagittal plane at 8 μm and hybridized with digoxigenin-labeled antisense cRNA probes, which were detected with alkaline phosphatase-conjugated anti-digoxigenin antibodies as described previously (Urness et al., 2008, 2010). Probes were prepared either by transcription of cDNA-containing plasmids (Table S2) or from PCR-amplified fragments of mouse genomic DNA (Table S3) as described previously (Urness et al., 2015).

4.6. mRNA detection by RNAscope in situ hybridization

For multiplex RNAscope ISH, E15.5 or E18.5 embryo half heads were prepared as described for classic ISH and sectioned at 6 μm. Appropriate control and experimental sections were positioned in duplicate on the same positively charged glass slide (Genesee, 29–108). RNAscope ISH was conducted by the University of Utah Biorepository and Molecular Pathology Core using a Leica Bond Rx autostainer (Leica Microsystems). Briefly, the tissue samples on slides were baked at 72 °C for 30 min, then deparaffinized using Bond Dewax solution (Leica Biosystems, AR9222) followed by sequential hydration in ethanol solutions. Tissue was treated with protease (RNAscope 2.5 LSx Protease) at 40 °C for 15 min then subjected to heat retrieval at 95 °C for 15 min, pH 9 using Bond ER2 (Leica Biosystems, AR 9040). RNA transcripts were localized using the RNAscope 2.0 RNAscope^™ LS Multiplex Fluorescent Reagent Kit for BDZ 11 (ACDBio 322800) and the RNAscope^® LS 4-Plex Ancillary kit (ACDBio 322830) according to ACDBio instructions. Target oligonucleotide probes (all from ACDBio) were Mm-Fgf8 (#313418-C1), Mm-Fgf20 C2 (#526398-C2), Mm-Bcl11b (#413278–C3), Mm–Tbx2-C4 (#448998-C4). Additional slides from the same samples were hybridized at the same time with the RNAscope Multiplex Positive Control Probe-Mm (containing Polar2a C1, Ppib C2, Ubc C3 and Hprt-1 C4; ACDBio, 321818), or the RNAscope Multiplex Negative Control Probe (containing DapB from B. Subtilis in all channels; ACDBio, 321838). Slides were hybridized with target or control probes at 40 °C for 2 h, followed by signal amplification oligonucleotides at 40 °C for 15–30 min). The markers were stained with the corresponding Opal dyes with two dispenses each with a total incubation period lasting for 30 min (Opal 520- FP1487001KT; Opal 570-FP1488001KT; Opal 620-FP1495001KT; Opal 690-FP1497001KT). Finally, tissue sections were counterstained with two applications of spectral DAPI for 5 min each. Slides were rinsed with water and mounted using ProLong Diamond Antifade Mountant (Life Technology, P36961).

4.7. RNAscope imaging

RNAscope slide imaging was performed by the University of Utah Cell Imaging Shared Resource. Epifluorescence slide scanning of the entire sample area was performed either on a Zeiss AxioScan Z1 slide scanner microscope, equipped with X-Cite 120LED Boost (Excelitas Technologies) for epifluorescence excitation, or on a Zeiss AxioScan 7 slide scanner microscope, equipped with X-Cite Xylis XT720L 385 (Excelitas Technologies) for epifluorescence excitation. Scans were performed with a 20x/0.8NA air objective (Zeiss, Plan-Apochromat). The AxioScan Z1 filter cubes could distinguish up to nine OPAL dyes, including DAPI, OP480, OP540, OP620 and OP690 for this hybridization and the AxioScan 7 filter cubes could distinguish up to seven OPAL dyes, including DAPI, OP480, OP620 and OP690. An OP570 cube was used to detect OP540. In each case the light source was set to 100 % power and camera exposure times/er channel were adjusted to obtain sufficiently good brightness and SNR levels. The region of interest (the cochlea) was identified for each of the samples on the slide and confocal microscopy of that region was performed on a Leica SP8 X WLL system. A white light laser (2nd generation) provided freely selectable excitation wavelengths in the range 470–670 nm. In addition, a 405 nm laser and an argon laser with lines at 458, 476, 488, 496, 514 was available. The system was equipped with 5 detectors for confocal scanning (PMT1, HyD2, PMT3, HyD4, PMT5). The SP8 detector head layout allowed for freely adjustable detection wavelengths windows. These were chosen as DAPI (Ex = 405 nm, Em = 412–450), OP480 (Ex = 458 nm, Em = 465–513), OP540 (Ex = 523 nm, Em = 530–580), OP620 (Ex = 588 nm, Em = 595–645), OP690 (Ex = 670 nm, Em = 680–750) Confocal scanning was performed with an HC PL APO 40x/1.30 OIL immersion objective (Leica Microsystems). Tile scanning and stitching was performed in the LAS X Navigator module. Each tile was acquired with the following settings: line sequential mode, 2048 × 2048 pixel, zoom 1 = 290.62μm FOV (resulting in a pixel size of 142um), 200Hz line speed in unidirectional mode (resulting in a pixel dwell time of 600 ns). Laser powers and detector gains were adjusted to obtain sufficiently good brightness and SNR levels.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.ydbio.2025.08.017.

Data availability

Data will be made available on request.