Deletion of the Ebf1, a mouse deafness gene, causes a dramatic increase in hair cells and support cells of the organ of Corti

The inner ear contains the sensory structures necessary for our sense of both hearing and balance. A single sensory structure mediates hearing, the organ of Corti, which is the sensory region of the cochlear duct. It consists of one row of inner hair cells (iHCs) and three rows of outer HCs (oHCs), all with attendant support cells (SCs) and an extremely regular arrangement along the ventral cochlear duct. The inner border cell (iBC) and inner phalangeal cell (iPhC) are associated with the iHC. iBC projections wrap around the medial face of iHCs, and iPhC projections wrap around the lateral face. Each oHC is associated with a single Deiters' cell (DC). In between these HC/SC groups there are two pillar cells (PCs); the one closest to the iHC is called the inner PC (iPC) and the one closest to the oHCs is called the outer PC (oPC). These PCs form the fluid-filled tunnel of Corti without which the animal cannot hear (Colvin et al., 1996; Hayashi et al., 2007). The development, position and arrangement of these HCs and SCs are dependent on various signaling molecules from multiple families of growth factors expressed in medial (neural) to lateral (abneural) gradients (Driver and Kelley, 2020; Groves and Fekete, 2012). For example, there is a strong expression of BMP4 at the lateral side, which likely restricts the sensory domain in this direction (Ohyama et al., 2010). However, what exactly restricts the development of sensory cells medially is not known. This study investigating the role of EBF1 in developing cochleae suggests that the expression of this transcription factor establishes the medial border of the sensory domain.

In our analysis of the epigenetic regulation of cochlear development, we discovered an enrichment of EBF transcription factor-binding motifs in the open chromatin of SOX2-EGFP prosensory cells (Wilkerson et al., 2019). Other transcription factors known to be important in sensory development were also found. This suggests that a member of the EBF family of transcription factors is likely to play a role in the development of the organ of Corti. We looked at single cell RNA-seq obtained from developing cochleae at the same ages as the ATAC-seq samples, and discovered that EBF1 is likely the EBF family member involved. Indeed, the expression of Ebf1 in the inner ear was first described by Davis and Reed (1996) and later revealed in the Allen Brain Atlas.

EBF1 is a member of the COE (Collier, OLF1 and EBF1-4) family of transcription factors. The EBFs are helix-loop-helix proteins containing an unusual Zn coordinating domain that binds to DNA. Ebf genes are expressed in many parts of the nervous system and frequently co-expressed with other family members. Mutations in a single Ebf gene can often be compensated for by another family member; thus, loss of function of a single Ebf gene reveals a phenotype if it is the only member of the family expressed. This appears to be the case in the developing cochlear epithelium, as there is little to no expression of Ebf2, Ebf3 and Ebf4.

In this study, we examine the effect of deleting Ebf1 from the otic epithelium on the developing cochlea using a tissue-restricted Cre or a tamoxifen-inducible Cre. We find that there is a dramatic increase of both iHCs and oHCs, and their attendant SCs in the mice with the conditional deletions. The mutant mice also have ectopic patches of sensory epithelia that develop in Kölliker's organ, suggesting that EBF1 is suppressing sensory development in this medial region. The expanded sensory domain is innervated, but the outer spiral bundles do not develop normally in the Ebf1 mutant mouse. The ectopic HCs we see in the non-sensory Kölliker's organ are also innervated and are surrounded by SCs. Interestingly these extra iHCs and oHCs are retained in the adult cochlea; however, the mice are severely hearing impaired.

Recent work from our lab identified enrichment of EBF-binding motifs in the open chromatin of cochlear prosensory cells collected at key developmental time points (Wilkerson et al., 2019). To investigate Ebf gene expression in the developing cochlear sensory epithelium, we performed single cell RNA-seq on SOX2-EGFP⁺ cells FAC-sorted from E12, E14 and E16 cochlear ducts. Clustering the cells after sequencing using Seurat allowed us to identify the various cell types present in the sample. We were able to capture the major cell types (prosensory cells, iHCs, oHCs, SCs, neurons, glia, Kölliker's organ cells, cycling cells, roof cells, mesenchymal cells and outer sulcus cells) present during these three developmental time points (Fig. 1A). Our analysis of the SOX2-EGFP⁺ single cells revealed that Ebf1 is expressed in the developing cochlear epithelium whereas Ebf2, Ebf3 and Ebf4 show little to no expression (Fig. 1C). More specifically, Ebf1 is strongly expressed in prosensory cells, iHCs, oHCs, SCs, neurons, Kölliker's organ cells, cycling cells and mesenchymal cells (Fig. 1A,C). In situ hybridization of the E14.5 inner ear confirmed our single cell results. Ebf1 transcript expression was detected in the ventral cochlear duct, spiral ganglion and mesenchyme (Fig. 1B).

Our transcript analyses indicated that Ebf1 is the primary contributor to EBF protein expression in the developing cochlear sensory epithelium (Fig. 1A-C). To further explore the timing and localization of EBF expression, immunolabeling was performed using a pan-EBF antibody. At E12.5, EBF protein is expressed in a band that overlaps with the medial edge of the expression domain for SOX2 (Fig. 1D,D′), a well-known prosensory marker (Kiernan et al., 2005b). At E14.5, EBF protein expression extends medially into Kölliker's organ and overlaps with the SOX2 domain (Fig. 1E-E″). By E16.5, MYO7a⁺ HCs can be seen in the base and middle of the cochlear sensory epithelium (Sahly et al., 1997). At this stage, EBF protein is most strongly expressed in the population of SOX2⁺ Kölliker's organ cells immediately medial to the iHC row (Fig. 1F-F″) and shows comparatively weak expression in HCs and SCs of the sensory domain (bracket in Fig. 1F″). Although the developing cochlea maintains strong expression of EBF proteins in Kölliker's organ by E18.5 (Fig. 1G-G″), expression is diminished in the sensory cells, particularly in iHCs (bracket in Fig. 1G″). We continue to see EBF immunolabeling in the DCs and iPhCs of the adult organ of Corti (see below).

To investigate the role of EBF1 in cochlear development, we generated a cKO mouse in which the Slc26a9 promoter directs Cre-mediated excision of floxed Ebf1 exons 6-16 (Vilagos et al., 2012). SLC26A9 is a chloride transporter present at E9.5 in the otic vesicle that gives rise to the inner ear epithelium (Urness et al., 2020). Unlike the Ebf1 global knockout (KO), which is postnatal lethal (Nelson et al., 2019), Slc26a9^Ebf1-cKO mice are fertile and survive into adulthood. The cochleae of mice with the conditional deletion lose EBF immunolabeling in the cochlear epithelium but maintain EBF expression in the mesenchyme (Fig. S1), likely due to Ebf3 and Ebf4, which show low expression levels in the otic epithelium in our single cell analyses (Fig. 1C).

To determine the effects of loss of EBF1 in the developing cochlea, we analyzed P1 cochlear wholemounts of control and Slc26a9^Ebf1-cKO mice. We found a dramatic expansion of the sensory epithelium in the mutant mice (Fig. 2). Unlike control littermates with the expected single row of iHCs (vGLUT3⁺ and MYO7a⁺) and three rows of oHCs (vGLUT3^- and MYO7a⁺; Fig. 2E-G), mice with the conditional deletion possess three to five rows of iHCs medial to approximately six rows of oHCs (Fig. 2K-M). Slc26a9^Ebf1-cKO cochlear ducts exhibit approximately threefold and twofold increases, respectively, in iHC and oHC counts by P1 (Fig. 2Q). Slc26a9^Ebf1-cKO neonates also demonstrate a delayed base-to-apex sweep of vGLUT3 expression relative to littermate controls (Fig. 2A,B versus C,D), suggesting a developmental delay in cochlear maturation. However, the lengths of P1 Slc26a9^Ebf1-cKO and littermate control cochlear ducts are not significantly different (P>0.05, Welch's t-test, 4598.1±493.6 µm for nine Slc26a9^Ebf1-cKO mice and 4653.2±349.7 µm for nine littermate controls from four litters; one duct per mouse).

To determine whether the extent of HC expansion varies along the length of the cochlear duct in the mutant mice, we quantified HC density in the base, middle and apex of cochlear wholemounts. Compared with littermate controls, iHC and oHC densities are significantly greater in all regions of the Slc26a9^Ebf1-cKO samples (Fig. 2R,S and Table S1). Although the mutants show an overall increase in HC density across all regions, the increase was greatest at the base for both iHCs and oHCs (Table S1). The increases in HCs are accompanied by defects in either stereocilia bundle or HC orientation. The supernumerary iHCs in the mutant are arranged on either side of the space presumably formed by the developing tunnel of Corti (Fig. 2K). Although we did not quantify the support cell increases, there was a clear expansion of the SOX2⁺ domain (Fig. 2H-J versus N-P)

In addition to an excess of sensory cells within the sensory domain, the Slc26a9^Ebf1-cKO mice also possess ectopic sensory patches throughout their Kölliker's organs (arrowheads in Fig. 2C,D). These ectopic patches resemble miniature vestibular organs, consisting of vGLUT3⁺ HCs resting on a bed of SCs that do not express the markers characteristic of sensory domain SCs (magnified inset capturing the ectopic patch outlined in Fig. 2D and Fig. S2). Slc26a9^Ebf1-cKO supernumerary and ectopic patch HCs have functioning mechanoelectrical transduction (MET) channels, as demonstrated by their uptake of FM1-43 dye (Fig. S3).

Supernumerary HCs are neighbored by supernumerary SCs (Fig. 2N-P), and mutant mice exhibit more nuclear layers in Kölliker's organ than littermate controls (Fig. 3A,A′ versus B,B′). We measured the areas of the developing sensory domain and Kölliker's organ in the base of mid-modiolar sections collected from P1 Slc26a9^Ebf1-cKO and littermate control mice. Anti-SOX2 antibody was used to visualize the developing sensory domain in the base of mid-modiolar sections collected from P1 Slc26a9^Ebf1-cKO and littermate control mice (Fig. 3). As expected, we found that the area of SOX2 expression was significantly greater in Slc26a9^Ebf1-cKO than littermate control sections (P<0.05, 2534±682 µm² for Slc26a9^Ebf1-cKO and 1751±331 µm² for littermate controls; Table S1). Similarly, we used a marker unique to Kölliker's organ, PRDM16, to assess changes in the overall size of this transient structure (Ebeid et al., 2022). We found that the area of PRDM16⁺ expression was significantly greater in Slc26a9^Ebf1-cKO than in littermate control sections (P<0.05, 4391±940 µm² for Slc26a9^Ebf1-cKO and 3146±398 µm² for littermate controls; Table S1). Kölliker's organ appears to increase in size, along with the sensory domain, in the mutant mice.

To better characterize the increase in SCs of the sensory domain, we used several markers to identify SC subtype-specific changes in patterning within the base of P1 mid-modiolar sections. In addition to anti-SOX2 antibody, anti-jagged1 (JAG1) antibody was used to examine overall SC expansion. At P1, JAG1 is expressed in sensory domain SCs and lateral Kölliker's organ cells, and SOX2 expression overlaps with JAG1 and extends more medially into Kölliker's organ. The expression domain of JAG1 is expanded in cochlear sections from the mutant mice compared with littermate controls (Fig. 3C,C′ versus D,D′). To determine which SC subtypes contribute to the observed SC expansion in the mutant mice, we used specific markers for iBCs, iPhCs, iPCs, oPCs and DCs. When compared with littermate controls, the mutant mice exhibited an increase in anti-FABP7 labeled iBCs/iPhCs (arrows in Fig. 3E,E′ versus F,F′) and anti-PROX1 labeled PCs and DCs (Fig. 3G,G′ versus H,H′). Further analysis with anti-p75 and anti-CD44 antibodies, which label iPCs and oPCs (Hertzano et al., 2010; Knipper et al., 1996), respectively, indicates that increases in oPC and DC numbers likely account for the observed increases in PROX1⁺ nuclei. Although multiple CD44⁺ oPCs can be seen in Slc26a9^Ebf1-cKO cochlear sections, only one p75⁺ (NGFR) iPC is typically seen (arrowheads in Fig. 3J-J‴). Cells lateral to the oHCs do not demonstrate changes in their numbers. FABP7 is expressed by Hensen's cells (HeCs) in addition to iBCs and iPhCs, and CD44 is expressed by Claudius cells (CCs) in addition to oPCs. Compared with littermate control cochlear sections, HeC and CC populations are unchanged in Slc26a9^Ebf1-cKO cochlear samples (Fig. 3E,E′,I,I‴ versus F,F′,J,J‴).

To further investigate the effects of Ebf1 deletion in patterning of medial SCs, we performed immunolabeling on P1 cochlear wholemounts. Anti-FABP7 labeling revealed that mice with a conditional deletion of Ebf1 possess more than the typical two rows of iBCs/iPhCs (arrows in Fig. 4A-A‴ versus B-B‴). In the mutant cochlea, FABP7⁺ iBCs/iPhCs reside next to the medial and/or lateral faces of supernumerary HCs that resemble iHCs in their cell shape and position along the medial-lateral axis (arrows in Fig. 4B-B‴). The occasional iHC can be observed in the oHC region (Figs 2K and 4B-B‴). Even among the oHCs, this iHC continues to direct patterning of a FABP7⁺ iBCs/iPhCs (Fig. 4B-B‴). Like iBCs/iPhCs, Slc26a9^Ebf1-cKO cochlear ducts possess more CD44⁺ oPCs than littermate controls. Although controls possess a single continuous row of oPCs lateral to the row of iPCs (Fig. 4C-C″,G-G‴), Slc26a9^Ebf1-cKO samples exhibit at least two discontinuous rows of oPCs lateral to their iPCs (Fig. 4D-D‴,H-H‴). Although the Ebf1 deletion induces an increase in many types of SCs, it does not appear to increase iPC numbers. p75⁺ iPCs are present in a single row in both littermate control and Slc26a9^Ebf1-cKO cochlear ducts (Fig. 4C-C‴,E-E‴ versus D-D‴,F-F‴).

The increase in sensory cell numbers after Ebf1 deletion suggests that the proliferation of the sensory progenitors may be prolonged in the mutant mice. To assess whether loss of EBF1 leads to prosensory cell proliferation beyond the typical E12-E15 window (Ruben, 1967), we performed EdU labeling on pregnant dams carrying Slc26a9^Ebf1-cKO and control littermates. EdU was administered twice a day for 3 days starting on E15.5, and cochlear wholemounts were analyzed at P1 (Fig. 5G). Although HCs and attendant SCs fail to incorporate EdU in control cochleae (Fig. 5A-B′), many HCs and SCs in the sensory domains of Slc26a9^Ebf1-cKO are EdU⁺ (open arrowheads in Fig. 5C-D′). Quantitative analysis of EdU⁺ iHC and oHC density revealed that the cochleae of the mutant mice possess significantly more EdU⁺ HCs than samples from littermate controls (Fig. 5E,F and Table S1). The increase in EdU⁺ HCs is present in the Slc26a9^Ebf1-cKO samples throughout the extent of the cochlea, with more EdU⁺ oHCs in the apex followed by the base and middle of Slc26a9^Ebf1-cKO cochlear ducts (Table S1). These results suggest that prolonged cell proliferation in the prosensory cells likely contributes to the sensory expansion seen after Ebf1 deletion.

The precise coordination of cell cycle exit and sensory cell differentiation are essential for cochlear development (Benito-Gonzalez and Doetzlhofer, 2014; Driver and Kelley, 2020). To assess whether delayed cell cycle exit in Slc26a9^Ebf1-cKO mice is accompanied by delayed differentiation, we measured the sweep of MYO7a expression at E16.5 (Fig. S4), well after the onset of HC differentiation (Chen et al., 2002). The distance from the apex to the first MYO7a⁺ cell is significantly greater in Slc26a9^Ebf1-cKO cochlear ducts than in control cochlear ducts. This difference is not due to impaired formation of the SOX2⁺ prosensory domain, which suggests that the sweep of differentiation is delayed in Slc26a9^Ebf1-cKO samples (Fig. S4A″,B″,C). At E16.5, it is worth mentioning that the lengths of Slc26a9^Ebf1-cKO and control cochlear ducts are significantly different, suggesting that convergent extension is delayed (P<0.01, 3724±120 µm for Slc26a9^Ebf1-cKO and 4463±287 µm for littermate controls; Table S1). By P1, Slc26a9^Ebf1-cKO and littermate control ducts are similar in length.

Our results on the developing cochlea show dramatic sensory expansion in mice with a conditional deletion in Ebf1; to assess whether this phenotype persists in adult mice, we performed immunolabeling on mid-modiolar cochlear sections collected from P60 Slc26a9^Ebf1-cKO and littermate control mice. Much like Slc26a9^Ebf1-cKO neonatal cochleae (Fig. S1B,B′), adult Slc26a9^Ebf1-cKO cochleae continue to express EBFs in their mesenchymal cells and show little to no EBF expression in the cochlear sensory epithelium (Fig. 6B,B′). Control littermates exhibit EBF expression in their SCs in addition to their mesenchymal cells (Fig. 6A,A′). To examine the effects of Ebf1 deletion in SC survival and maturation in the adult mice, we looked at expression of several HC and SC markers. We found that in adult mutant mice, the increase in SC numbers persisted and, strikingly, the additional SCs appear to give rise to multiple tunnels of Corti (asterisks in Fig. 6D″,D‴,F,F‴). To assess HC survival and maturation, we performed labeling using anti-vGLUT3 and anti-prestin antibodies. We found that there was a loss in oHCs in the base of a subset of adult Slc26a9^Ebf1-cKO cochleae; however, supernumerary HCs appear to survive throughout the rest of the duct. Adult supernumerary vGLUT3⁺ iHCs are typically found medial to tunnels of Corti, although the occasional vGLUT3⁺ iHC can be seen in the PC region (open arrowhead in Fig. 6F,F′). Neurofilament (NF-M) labeling indicates that supernumerary iHCs and oHCs are innervated (Fig. 6F‴). Interestingly, Slc26a9^Ebf1-cKO mice demonstrate significantly elevated auditory brainstem response (ABR) thresholds for all tested stimulus frequencies, except 60 kHz. These results suggest that adult Slc26a9^Ebf1-cKO mice are deaf (Fig. 6G,H).

HCs are known to secrete neurotrophins that direct their innervation by spiral ganglion neurons (Fariñas et al., 2001; Fritzsch et al., 1997). To determine whether supernumerary HCs are innervated, we performed immunolabeling with anti-NF-M antibodies. P1 littermate control mice demonstrate the typical innervation pattern. Numerous neuronal projections terminate on iHCs (arrowheads in Fig. 7A″), and the type II afferent projections turn towards the base, innervating multiple oHCs in the same row and forming three spiral bundles (Fig. 7A,A′) (Liberman et al., 1990). Although fibers continue to terminate on Slc26a9^Ebf1-cKO iHCs (arrowheads in Fig. 7B″), the spiral bundles are completely lost in the oHC region and the innervation appears disorganized (Fig. 7B,B′). In addition to fibers extending to supernumerary HCs in sensory domain, HCs in the ectopic sensory patches of Kölliker's organ are also innervated (Fig. 7C-C″).

The Slc26a9^Ebf1-cKO mice show the effects of the loss of the Ebf1 gene at a very early stage of cochlear development (E9.5). To determine the role of EBF1 at later stages of development, we used an inducible Cre-mouse line: Sox2^CreER. This second Cre-line was crossed to Ebf1^fl/fl mice, and then Ebf1 deletion was induced with two tamoxifen treatments: one at E11 and another at E12. Owing to tamoxifen toxicity (Ved et al., 2019), we opted to analyze these litters at E18 (Fig. 8O). iHCs were distinguished from oHCs based on their position relative to the developing tunnel of Corti. At E18, Sox2^Ebf1-cKO mice show extra iHCs and oHCs in an apically-biased manner. In the apex, iHC density and oHC density are both significantly greater in Sox2^Ebf1-cKO than Sox2^CreER-negative control samples. In the middle, iHC density is greater in Sox2^Ebf1-cKO than that in Sox2^CreER-negative controls. Sox2^Ebf1-cKO cochlear ducts show no difference in oHC density in the middle and no differences in HC density in the base (Fig. 8M,N and Table S1). In Sox2^Ebf1-cKO cochlear ducts, most of the extra iHCs are in a second row medial to the main row, and instances of oHCs in a fourth row could be observed. Alignment of oHC rows is imperfect in Sox2^Ebf1-cKO samples (Fig. 8G-I). Loss of EBF1 in the developing sensory epithelium at E11/12 does not significantly affect cochlear duct length (P>0.05, Welch's t-test, 3592.3±581.0 µm for seven Sox2^Ebf1-cKO and 4042.6±476.9 µm for 13 Sox2^CreER-negative controls, five litters; one duct per mouse). To discriminate between the effects of Ebf1 deletion and Sox2 haploinsufficiency on HC formation (Atkinson et al., 2018), we compared HC density in Sox2^CreER Ebf1^fl/fl mice not treated with tamoxifen with that of Sox2^CreER-negative littermate controls treated with tamoxifen (13.0±0.8 iHCs and 51.8±2.8 oHCs per 100 µm in the base, 16.2±0.4 iHCs and 53.0±2.8 oHCs per 100 µm in the middle, 15.7±0.5 iHCs and 49.3±2.3 oHCs per 100 µm in the apex for three Sox2^CreER Ebf1^fl/fl mice not treated with tamoxifen; one duct per mouse). No significant differences were observed in HC density of Sox2^CreER Ebf1^fl/fl mice not treated with tamoxifen and the Sox2^CreER-negative control samples (P>0.05, two-way ANOVA with Tukey's multiple comparison's test).

EBF1 is best known for its role in B cell lineage development, but it also plays a role in nervous tissues. EBF1 has been shown to be necessary for striatal medium spiny neuron development and differentiation (Lobo et al., 2008). It has also been shown to play a role in dopaminergic neuron development with loss of these cells in the substantia nigra of Ebf1^-/- mice (Yin et al., 2009). In chick neural tube, early neurogenic genes such as Ngn2 and NeuroM have been shown to activate Ebf1, allowing neuronal differentiation and migration to take place (Garcia-Dominguez et al., 2003). The olfactory system is the most closely related to the auditory system, and, in this system, Ebf1 has roles in the regulation of genes in the odorant signaling cascade (Davis and Reed, 1996). Davis and Reed have described Ebf1 expression in many areas of the nervous system, including the inner ear. In this study, we have systematically described the expression of Ebf1 in the developing cochlea using immunohistochemistry, in situ hybridization and single cell RNA-seq. Although many studies suggest that Ebf1 plays a role after exit from the cell cycle, we observed expression in areas of the cochlear epithelium that are still in the cell cycle. However, our data are suggestive of a regulation of exit from the cell cycle, because we see a dramatic increase in cells that develop into sensory cells when Ebf1 is deleted at the otocyst stage.

The most obvious phenotype in the Slc26a9^Ebf1-cKO is the more than doubling of the sensory epithelium. This result suggests that Ebf1 plays a role in either regulating proliferation of prosensory cells or in limiting the extent of the developing sensory epithelial domain. Sensory epithelium patterning in the cochlea can be subdivided into two key processes: prosensory domain establishment and sensory cell differentiation (Brown and Groves, 2020). The prosensory domain is defined between E12 and E15 as prosensory cells undergo cell cycle exit in a wave that starts at the apex of the cochlea and sweeps to the base (Ruben, 1967). HC and SC differentiation proceeds in the opposite direction, starting at the base around E14 and reaching the apex during neonatal stages (Chen et al., 2002).

The development of the sensory epithelium is highly regulated and dependent on many signaling pathways (Groves and Fekete, 2012). Changes in several pathways give rise to cochleae with extra HCs. BMP4 is produced by the cells on the lateral edge of the developing organ of Corti, the development of which is dependent on low levels of BMP4 but inhibited by high levels. BMP4 is responsible, at least in part, for the sharp border at the lateral edge of the sensory epithelium (Ohyama et al., 2010). There is a gradient of WNTs and FGF10 from Kölliker's organ on the medial side of the sensory domain. The sensory domain itself is defined by expression of Fgf20, which is initially expressed in base and sweeps towards the apex ahead of MYO7a expression (Hayashi et al., 2008). The Notch pathway is also involved in sensory epithelial patterning. During early development, it is responsible for inducing a sensory fate via signaling from JAG1 (Kiernan et al., 2006) expressed in Kölliker's organ and overlapping the region where iHCs will form (lateral induction). Later in development, the Notch pathway suppresses HC fate because of the early expression of Notch ligands by HCs, thus allowing the cells next to each HC to develop as SCs (lateral inhibition). There are numerous mutations in Notch, its ligands or effectors that will give extra HCs (Basch et al., 2016a,b; Kiernan et al., 2005a).

How does EBF1 fit with all these signaling mechanisms? It is clear from our data that there is a delay in cell cycle exit. Other labs have shown that prolonged prosensory cell proliferation leads to increases in both HCs and SCs (Chen and Segil, 1999; Jacques et al., 2012; Kiernan et al., 2005a; Tateya et al., 2011). A doubling in the width of the sensory epithelium could be accounted for by progenitors going through one more cell cycle. This, however, does not account for the more than doubling of the iHCs. We propose that EBF1 plays an early role in regulating exit from the cell cycle and a later role in maintaining the medial border of the sensory epithelium such that the cells closest to the sensory epithelium do not develop as HCs or SCs. The suppression of sensory epithelium formation also accounts for the ectopic sensory patches we see in the cochlea from Slc26a9^Ebf1-cKO animals. This two-stage requirement for EBF1 activity is consistent with our finding that when Ebf1 is eliminated at E11/12, we only find extra iHCs, whereas oHC numbers are unaffected.

In addition to extra HCs in the Ebf1 conditional deletions, we also observed extra SCs. The extra HCs in mice with either conditional deletion (Slc26a9^Ebf1-cKO or Sox2^Ebf1-cKO) appear to pattern their immediately adjacent SCs, possibly via a secreted factor. Such a factor may be Fgf8, which is uniquely expressed only by iHCs. Mutations that affect the FGF signaling pathway have been shown to lead to both defects in PC differentiation as well as extra oHCs (Hayashi et al., 2007; Shim et al., 2005). The extra iHCs in the Ebf1 cKOs should lead to a higher concentration of FGF8; previous studies have shown that overexpression of Fgf8 in explant cultures will induce extra p75⁺ iPCs (Jacques et al., 2007). One might have anticipated that the mutant would have extra iPCs rather than the extra oPCs that we observe in the Slc26a9^Ebf1-cKO mice. Mice with a Fgfr3 mutation, which causes an overactive kinase activity of the receptor, also have extra oPCs (Mansour et al., 2013). In the Fgfr3 mutants, it is thought that the receptor is activated by FGF10, which is abundantly expressed in Kölliker's organ.

The number of ectopic sensory patches in Kölliker's organ is highly variable; however, every mutant cochlea examined had at least one. These are located throughout the cochlea and at various positions along the medial axis of Kölliker's organ. We have never seen complete conversion of Kölliker's organ to sensory epithelium, which suggests that there is more than one inhibitor of sensory development present. We also do not know what kind of HCs are in these ectopic patches. The surrounding cells express SOX2, suggesting that they are SCs, but they do not express any of the specific markers for medial SCs (iBCs, iPhCs and PCs). The HCs do express Slc17a8, suggesting that these are either iHCs or vestibular HCs. They also show rapid uptake of FM1-43, indicating active MET channels. We have only examined adult cochleae from the Ebf1 mutants as sections, so we cannot say at this stage whether these ectopic HCs are retained. It appears that the inner sulcus does form at least in the base.

In the Slc26a9^Ebf1-cKO animal, there are many defects in the sensory epithelium. We see extra iHCs, oHCs, iPhCs, oPCs and DCs, as well as defective neural connections, any of which would dramatically change the response to sound of the animal. In the adults, we found that the extra HCs were retained, and the animal showed dramatic hearing loss. There are multiple potential reasons for this result, including the sheer weight of the extra cells causing mechanical disturbances to the propagation of sound waves along the basilar membrane. We also noticed defects in the tectorial membrane, which appeared thinner and more disorganized in the mutant animals, and defects in innervation.

In summary, we have identified EBF1 as a key factor for patterning the mammalian cochlea and deletion of Ebf1 leads to deafness in mice. We also found expression of Ebf1 in the vestibular system but see no vestibular phenotype in the cKO mice. Remaining questions relate to the mechanism of action of EBF1 in cochlear development. From what is known about this family, it could be acting as a transcriptional activator, repressor or chromatin remodeler (Ramírez et al., 2010). It may also have different mechanisms that are context dependent at the various stages of development. We are currently looking into the pathways that are mis-regulated when Ebf1 is deleted in the developing inner ear. In a recent study, the expression of Ebf1 was shown to be regulated by SOXC family. In a KO of Sox4 and Sox11 in the inner ear, the expression of Ebf1 was dramatically downregulated along with other prosensory factors, such as Fgf20 (Hayashi et al., 2008; Wang et al., 2023). In addition, EBF1 appears to be regulated by FGF; Ebf1 is downregulated in the otic mesenchyme but appears to be upregulated in the epithelium at E12.5 in Fgf9/20 double KO mice (Ebeid and Huh, 2020). The importance of Ebf1 in cochlear development is underscored by a recent publication that also analyzed a KO and a cKO for Ebf1 (Kagoshima et al., 2024). The results from this study are similar to ours, although different mutant mice are used. Their findings agree with those we detail in this report and further support an important role for EBF1 in development of the organ of Corti and suggest its mutation might cause deafness in humans.

Mice were housed in either the University of Washington Department of Comparative Medicine or in the Medical University of South Carolina Division of Lab Animal Resources. All experiments were approved by the Institutional Animal Care and Use Committee of the University of Washington, as well as by the Medical University of South Carolina, and were performed in accordance with the standards outlined by the National Institutes of Health (NIH). Mice were euthanized in accordance with IACUC approved procedures and in line with NIH policies.

Ebf1^fl/fl mice were obtained from Jackson labs on a C57BL/6 background (strain 028104). Sox2^CreER mice obtained from Jackson labs on a mixed background of C57BL/6 and 129S4/SVJae (strain 017593), and the Slc26a9^P2A-Cre mice were obtained on a C57BL/6 background (Urness et al., 2020). To generate Slc26a9- and Sox2-conditional Ebf1 knockout mice, Slc26a9^P2A-Cre and Sox2^CreER mice were crossed with Ebf1^fl/fl females to generate Cre⁺ Ebf1^fl/+ progeny. Male Cre⁺ Ebf1^fl/+ mice identified by genotyping (see below) were then crossed with Ebf1^fl/fl mice to generate litters containing Cre⁺ Ebf1^fl/fl mice. Cre⁺ Ebf1^fl/fl mice identified by genotyping were bred with Ebf1^fl/fl mice to generate Cre⁺ Ebf1^fl/fl and Ebf1^fl/fl littermates for our phenotypic analyses. To induce Cre recombination, timed-pregnant dams carrying Sox2^CreER Ebf1^fl/fl and Ebf1^fl/fl embryos received tamoxifen diluted in corn oil (100 mg/kg; Sigma-Aldrich T5648) via oral gavage at E11 and at E12. Sox2-EGFP knock-in mice (Arnold et al., 2011) from Jackson labs (strain 017592) were bred to generate time-pregnant litters. Sox2-EGFP⁺ embryos were identified by epifluorescence and their cells were purified by FAC-sorting for single cell analysis. Stages were verified by Theiler's criteria. Tail tips or ear punches were collected for genotyping.

Mice positive for Sox-CreER were genotyped with the following primer pair: TCCTTAGCGCCGTAAATCAA and TGCCAGGATCAGGGTTAAAG. Mice positive for Slc26a9^P2A-Cre were genotyped using the following primer trio: GGAGGAACACAGTTCACAGT, GTGTCTGGTGTGGCTGATGACC and ATGGGTTCACCAGAGTCTCATC. Ebf1^fl/fl mice were genotyped using two distinct primer pairs: (1) TGTGGCAACCGAAATGAG and CCTGTGAGCGACACAAAGC, in addition to (2) ACGACTTCTTCAAGTCCGCC and TCTTGTAGTTGCCGTCGTCC. The first primer pair was designed to identify the presence of the wild-type Ebf1 allele (Vilagos et al., 2012). The second primer pair was used to reveal the presence of the EGFP associated with the fusion protein encoded in the floxed Ebf1 allele (Vilagos et al., 2012). Mice were identified as Ebf1^fl/fl if they tested negative for wild-type Ebf1 allele and positive for EGFP.

The hybridization was performed as described previously (Hayashi et al., 2007; Wilkerson et al., 2021). Briefly E14.5 heads were fixed in a modified Carnoy's solution overnight at 4°C. Heads were then dehydrated using increasing concentrations of ethanol and left in 100% ethanol overnight at 4°C. Heads were washed in xylene twice before embedding in low melt paraffin (Thermo Fisher Scientific, 23-021-401) and sectioning at 8 mm. Mouse Ebf1 cDNA was obtained from the Allen Institute (Seattle, WA, USA). Digoxigenin (DIG)-labeled probes for Ebf1 were generated using SP6 RNA polymerase (Promega P1085) and the following nested PCR primers: CTCAGTCACCACAAGCATGAAT, CACTTCATTCTCCCCTTCCATA and GCGATTTAGGTGACACTATAGCATGGAGTCTTGTTTATAGTGGC. In situs were performed using the DIG-labeled Ebf1 probes. The in situ product was visualized using NBT/BCIP (Sigma-Aldrich B1911) and anti-DIG conjugated to secondary antibody (Roche, 11093274910). Tissue sections were imaged using a Zeiss Axioplan 2 microscope.

SOX2-GFPhigh⁺ cochlear duct cells were isolated as described previously (Wilkerson et al., 2019) from 10-12 cochleae gathered from one to two litters per stage. Libraries were prepared using the Chromium Single Cell 3′ Library & Gel Bead Kit v3 (10x Genomics) according to the manufacturer's instructions. Reads were aligned to mm10 and filtered using 10X Genomics Cell Ranger, as described previously (Wilkerson et al., 2021). Briefly, E12, E14 and E16 samples were batch-corrected and normalized in Seurat 4 (Butler et al., 2018; Satija et al., 2015). Principal component, UMAP and clustering analysis (McInnes and Healy, 2018 preprint) was performed in Seurat; trajectory analysis was performed in Monocle 3 (Cao et al., 2019). Cell types were identified based on cluster-specific expression of known markers (see Fig. 1). Differential expression analysis was performed on raw counts using the FindMarkers function in Seurat.

For embryonic stages, heads were fixed overnight at 4°C. P1 and adult temporal bones were fixed in 4% paraformaldehyde (PFA) either for 30 min at room temperature or overnight at 4°C. After fixation, heads and temporal bones were washed in PBS three times for 10 min on a nutator at room temperature. For wholemount tissue experiments, cochlear ducts were dissected from embryonic heads and P1 temporal bones. For tissue section experiments, the heads and temporal bones were then incubated in a sucrose series consisting of 5%, 10% and 15% sucrose washes. Adult temporal bones and embryonic heads were incubated in an additional 25% sucrose and 30% sucrose wash, respectively. After incubating in OCT (Tissue-Tek 4583) at 4°C, heads and temporal bones were embedded in OCT and cryosectioned at 12 µm.

Wholemount tissue was permeabilized with 0.5% TX-100 for experiments visualizing cell surface markers and 2% Triton X-100 for experiments visualizing nuclear or cytoplasmic markers. Before blocking, adult sections to be stained with rabbit anti-SOX2 antibody were treated with 1% SDS for 5 min. Sections and wholemount tissue were incubated in primary antibodies diluted in block overnight at room temperature. Sections and wholemount tissue were blocked in 10% serum (donkey or fetal bovine) with 0.1% and 0.5% Triton X-100, respectively. Wholemount tissue was washed for 15 min or 1.5 h three times in 0.1% and 0.5% Triton X-100 for sections and wholemount tissue, respectively. Wholemount tissue was subsequently incubated in block for another 1 h. Sections and wholemount tissue were incubated in secondary antibodies diluted in block for 50 min or overnight at room temperature. Antibody details are provided in Table S2. All samples were imaged on a Zeiss LSM 880 confocal microscope.

To identify the presence of mechanoelectrical transduction (MET) channels in HCs, we treated freshly dissected P6 cochlear ducts with a styryl dye, FM1-43 FX (4 µM; Thermo Fisher Scientific F35355), for 30 s at room temperature. After removing the FM1-43 dye, the cochlear ducts were washed with HBSS three times. The cochlear wholemounts were imaged on a Zeiss LSM 880 confocal microscope.

To trace proliferation at embryonic stages, we administered 5-ethynyl-2′-deoxyuridine (EdU; 50 mg/kg; Invitrogen A10044) reconstituted in sterile PBS to pregnant dams via intraperitoneal injection twice a day (separated by 8 h) starting on E15 and ending on E17. Click-iT Alexa-Fluor-647 (Invitrogen C10430) was used to detect incorporated EdU in P1 cochlear ducts using the manufacturer's protocol. The EdU-labeled cochlear wholemounts were imaged on a Zeiss LSM 880 confocal microscope.

To quantify HCs in Slc26a9^Ebf1-cKO (Slc26a9^P2A-CreEbf1^fl/fl) and control (Ebf1^fl/fl) littermates, anti-MYO7a⁺ HCs were counted using the point tool and ROI Manager in ImageJ/Fiji. iHCs were distinguished from oHCs based on their comparatively strong expression of vGLUT3 (SLC17A8) protein and cell shape. Just as the oHCs taper from three rows in the base to two rows in the hook region of Ebf1^fl/fl control cochleae, a reduction in the number of oHC rows can be seen in the hooks of Slc26a9^Ebf1-cKO cochleae. Counts were made starting in the base of each cochlea up to the point in the apex where anti-vGLUT3⁺ iHCs were no longer visible. To measure cochlear duct length, we used the segmented line tool to generate a line composed of 50 µm segments along the space between the iHCs and oHCs. We chose to use this space rather than the row of iPC nuclei (as was done for the Sox2-conditional model) because iPCs cannot be identified based on the unique shape of their nuclei in Slc26a9^Ebf1-cKO mice. The segmented line tool was also used to measure the sweep of differentiation in embryonic cochlear wholemounts. We measured from the apex of the cochlear epithelium to the first SOX2⁺ cell in the apex and along medial edge of the strongly SOX2⁺ prosensory domain to the first MYO7a⁺ cell in the apex. For all regional analyses in the cochlea, HC density was measured in ∼400 μm samples in the base, middle and apex. Basal measurements were made starting immediately above the hook. Apical measurements were made in the most apical region of the duct where a space was still evident between the iHCs and oHCs. Middle measurements were made at the approximate midpoint for the duct. For EdU⁺ counting in HCs, EdU⁺ iHCs were distinguished from EdU⁺ oHCs based on vGLUT3 protein expression and cell shape.

To quantify HCs in Sox2^Ebf1-cKO (Sox2^CreER Ebf1^fl/fl treated with tamoxifen) mice, littermate controls (Ebf1^fl/fl treated with tamoxifen) and Sox2^CreER Ebf1^fl/fl mice not treated with tamoxifen, anti-MYO7a⁺ HCs were counted using the point tool and ROI Manager in ImageJ/Fiji. The hook region was excluded from counting and length measurements. Counts were made starting in the base of each cochlea, immediately above the hook, up to the point in the apex where iHCs were missing and intermittent (i.e. ∼10% distance from the apex in E18). To quantitate HC density, the length of the counted region was measured using the segmented line tool to measure between iHCs and oHCs, along the zipper-like row of anti-SOX2⁺ PC nuclei. To measure total cochlear duct length, the segmented line tool was used to draw a line along the PC region. For regional analysis of iHC density in the cochlea, HC density was measured in ∼300 μm samples in the base, middle and apex. Basal measurements were made starting immediately above the hook (90-100% distance from the apex). Apical measurements were made from the point in the apex where iHCs were missing and intermittent, extending toward the base (10-20% distance from the apex). Middle measurements were made at ∼40-50% distance from the apex. For regional analysis of oHC density in the cochlea, HC density was measured in ∼100 μm samples in the base, middle and apex. oHC differentiation slightly lags iHC differentiation at E18, so apical oHC counts were made at ∼20-30% distance from the apex in regions containing three or more rows of oHCs. Basal and middle oHC counts were made in approximately the same regions as the iHC counts in each image.

Mice were anesthetized by intraperitoneal injection of ketamine (90 mg/kg) and xylazine (10 mg/kg). Subcutaneous electrodes were placed at the cranial vertex for recording, behind the pinna on the stimulus side for reference and behind the contralateral pinna for grounding. Mice were placed in a sound-attenuating chamber and then calibrated tone pips (at 6, 12, 18, 24, 32 45 and 60 kHz; 500 responses at 21/s) and broadband click stimuli (300 responses, at 42.6/s) were presented and the responses recorded from 0-90 dB SPL (5 dB steps) using an RZ6 system (Tucker Davis Technologies). ABR thresholds were determined manually by experienced readers as the minimum sound level, eliciting identifiable ABR waveforms compared with the 0 dB level.

The number of animals analyzed per genotype and/or treatment group is listed under sample size in the figure legend for each experiment as well as in Table S1. Biological replicates were analyzed in expression experiments. Male and female mice were used for all experiments.

Graphs were created in GraphPad Prism. Mean total HC counts, HC and EdU⁺ HC densities, and sweep of differentiation are shown as bar graphs with standard deviation error bars. Mean ABR thresholds are shown as a line graph with standard deviation error bars. Significance is indicated as follows: *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

For duct length, total HC counts, PRDM16 and SOX2 expression areas, and sweep of differentiation, statistically significant differences between genotypes were identified using Welch's t-tests. Two-way ANOVA with Holm-Šídák's multiple comparisons test was used to test statistically significant differences in HC and EdU⁺ HC density between Slc26a9^Ebf1-cKO and littermate control samples for each region analyzed (base, middle and apex). To test for statistically significant differences between regions in these genotypes, two-way repeated measures ANOVA was performed with the same multiple-comparisons correction. Two-way ANOVA with Holm-Šídák's multiple comparisons test was also used to determine statistically significant differences in ABR thresholds between adult Slc26a9^Ebf1-cKO and control littermates for each stimulus frequency. Two-way ANOVA with Tukey's multiple comparisons test was used to test for statistically significant differences in HC density between samples from Sox2^Ebf1-cKO mice, littermate controls treated with tamoxifen and Sox2^CreER Ebf1^fl/fl mice not treated with tamoxifen. To test for statistically significant differences between regions within these genotypes, two-way repeated measures ANOVA was performed with the same multiple-comparisons correction (Table S1).

We thank Dr Suzanne Mansour for suggesting the Slc26a9^P2A-Cre mice, members of the Reh lab for helpful critiques on experiments, and Drs Reh, Pavlou and Kaplan for comments on this manuscript. We thank Connor Finkbeiner for help with single cell RNA-seq bioinformatics and Erica Lee for technical assistance.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202816.reviewer-comments.pdf