Cochlear progenitor number is controlled through mesenchymal FGF receptor signaling

The sensory and supporting cells (SCs) of the organ of Corti are derived from a limited number of progenitors. The mechanisms that regulate the number of sensory progenitors are not known. Here, we show that Fibroblast Growth Factors (FGF) 9 and 20, which are expressed in the non-sensory (Fgf9) and sensory (Fgf20) epithelium during otic development, regulate the number of cochlear progenitors. We further demonstrate that Fgf receptor (Fgfr) 1 signaling within the developing sensory epithelium is required for the differentiation of outer hair cells and SCs, while mesenchymal FGFRs regulate the size of the sensory progenitor population and the overall cochlear length. In addition, ectopic FGFR activation in mesenchyme was sufficient to increase sensory progenitor proliferation and cochlear length. These data define a feedback mechanism, originating from epithelial FGF ligands and mediated through periotic mesenchyme that controls the number of sensory progenitors and the length of the cochlea. DOI: http://dx.doi.org/10.7554/eLife.05921.001


Introduction
The Organ of Corti contains mechanosensory hair cells (HC) and specialized supporting cells (SC) that are required for the transduction of sound (Wu and Kelley, 2012). The frequency spectrum of sound stimuli is tonotopically represented along the length of the mammalian cochlea (Fay and Popper, 2000). In mouse, the cochlea begins to grow from the ventral otic vesicle at embryonic day 11.5 (E11.5) and continues to grow and coil, forming approximately one and a half turns by birth. During its development, the length of the cochlea is limited by the number of progenitors that give rise to sensory HCs and SCs, and is further regulated through a process of convergent extension (Chen and Segil, 1999;Wang et al., 2005;Wu and Kelley, 2012). In mouse, sensory progenitors exit the cell cycle by E14.5 and begin to differentiate into HCs and SCs. Thus, the size of the progenitor population at this stage of development is the ultimate determinant of the size of the adult cochlea. Progenitor number is determined by proliferation, the timing of differentiation, and in some cases by aberrant cell death. Previous studies indicate that sensory progenitor growth requires mesenchymal signals (Phippard et al., 1999;Braunstein et al., 2008Braunstein et al., , 2009, however, the identity and source of the factors that control this activity are not known. Fibroblast Growth Factors (FGFs) have several stage-specific functions during inner ear development. FGF3 and FGF10 signal from hindbrain and head mesenchyme, respectively, to the overlying ectoderm to induce formation of the otic placode and vesicle (Urness et al., 2010). Later in development, FGF20 regulates differentiation of outer hair cells (OHC) and SCs, termed the lateral compartment of the cochlea . Phenotypic similarities with mice lacking Fgfr1 in the entire otic epithelium suggest that FGF20 signals directly to FGFR1, serving as a permissive factor for differentiation (Pirvola et al., 2002;Hayashi et al., 2008;Huh et al., 2012). FGF9 signaling regulates structural components of the vestibular system, but alone has no effect on cochlear development (Pirvola et al., 2004). During postmitotic stages, FGF8 signaling from the inner hair cell (IHC) to FGFR3 in SCs regulates pillar cell differentiation (Colvin et al., 1996;Mueller et al., 2002;Jacques et al., 2007).
Here, we identify another critical stage in inner ear development that requires FGF signaling. We show that Fgf9, expressed in the non-sensory epithelium, and Fgf20, expressed in the sensory epithelium, regulate the number of cochlear progenitors and the ultimate length of the cochlea through signaling to mesenchymal FGFRs. We find that in vivo FGF9/20 signaling to mesenchymal FGFR1 and FGFR2 is required for sensory progenitor proliferation and that mesenchymal FGFR signaling is sufficient to promote sensory progenitor proliferation and extend the length of the cochlear duct. In addition, we show that prosensory epithelial FGFR1 and FGF20 independently is required for differentiation of outer HCs and SCs.

Results
Fgf9 and Fgf20 are expressed in the developing cochlea In a prior study, we showed that Fgf20 is required between E13.5-14.5 for differentiation of cochlear OHCs and SCs in the organ of Corti . However, Fgf20 is expressed in a portion of the otic vesicle sensory epithelium much earlier in development, beginning at E10.5 , but analysis of mice lacking Fgf20 did not reveal any function for Fgf20 at this stage of development. Since there are many examples of FGFs functioning redundantly during development, we hypothesized that redundancy could account for the lack of a phenotype in Fgf20 null inner ears between E10.5 and 12.5. Fgf9 is closely related to Fgf20 (Zhang et al., 2006;Itoh and Ornitz, 2008), and is also expressed in the otic epithelium at E10.5-12.5 (Pirvola et al., 2004); however, Fgf9 −/− mice have normal cochlear development and normal patterning of the organ of Corti (Pirvola et al., 2004). eLife digest Mammalian ears contain several structures that are involved in hearing. Within the inner ear is a spiral-shaped structure called the cochlea. This contains an array of cells called sensory hair cells that convert sound vibrations into electrical signals, which are then conveyed to the brain. Sounds of differing pitch are detected at different points along the cochlea, so its overall length helps to determine the range of sounds that an individual can hear.
In the embryo, sensory hair cells and their associated supporting cells develop from 'cochlear progenitor' cells. The final length of the cochlea is determined by the numbers of progenitor cells that commit to becoming either sensory hair cells or supporting cells. Two proteins called FGF9 and FGF20 are involved in the formation of the cochlea. FGF20 promotes the formation of the hair cells and supporting cells, but the precise roles of both proteins are not clear.
Here, Huh et al. studied FGF9 and FGF20 in the inner ear of mice at an early stage of development. The experiments show that these proteins work together to control the number of progenitor cells and the length of the cochlea. FGF20 is produced by the same tissue structure (called an 'epithelium') that gives rise to the hair cells and supporting cells. In contrast, FGF9 is produced in another epithelium tissue that produces the cells that line the fluid-filled tubes of the inner ear.
The experiments also show that both FGF9 and FGF20 act as signals to cells in an adjacent tissue called the mesenchyme, where they activate other proteins known as FGF receptors. These receptors, in turn, regulate an unknown molecule in the mesenchyme that influences the growth of progenitor cells and the length of the cochlea.
Sensory hair cells can be injured or lost by excessive sound exposure, some medications and as part of normal aging. These cells are not replaced, and so their loss is a major cause of permanent hearing loss. Understanding the signals that produce the progenitor cells will take us one step closer to being able to grow these cells in the laboratory for use in therapies to replace or repair damaged sensory hair cells.
We first examined the expression domain of Fgf9 relative to Sox2-expressing sensory progenitors (and Fgf20) using a new Fgf9-βGal reporter allele (Fgf9 lacZ ) in which a splice acceptor-lacZ gene was inserted into the first intron of Fgf9 (Skarnes et al., 2011). At E10.5, βGal activity was detected in the otic vesicle epithelium ( Figure 1A). Co-staining of βGal and Sox2 at E11.5 showed no overlap, indicating that Fgf9 is expressed in the non-sensory epithelium of the otic vesicle ( Figure 1B). Taken together with previous Fgf20 expression analysis at this stage   Figure 1C), Fgf9 and Fgf20 are both expressed in the otic vesicle, but in non-overlapping domains in the otic epithelium ( Figure 1D).
Decreased sensory progenitor number could also result from premature cell cycle exit. p27 kip1 is one of the cell cycle inhibitors that is expressed in sensory progenitors as they become postmitotic. Expression of p27 kip1 begins at E12.5 in the apex of the cochlea and progresses towards the base (Lee et al., 2006). By E14.5, the entire cochlear progenitor domain becomes p27 kip1 positive. Expression of p27 kip1 at E12.5 in the proximal cochlear duct was not detected in either control or Fgf9 −/− ;Fgf20 lacZ/lacZ embryos suggesting that there is no premature cell cycle exit in mice lacking Fgf9 and Fgf20 Epithelial Fgfr1 but not Fgfr2 is required for lateral compartment differentiation Next, we questioned which cell types are required for sensory progenitor proliferation and/or lateral compartment differentiation. Expression of both Fgfr1 and Fgfr2 have been reported in the otic epithelium and periotic mesenchyme between E10.5 and E12.5 (Pirvola et al., 2000(Pirvola et al., , 2002(Pirvola et al., , 2004Ono et al., 2014). Epithelial Fgfr1 has been conditionally inactivated in otic epithelium using Foxg1 Cre , Six1enh21 Cre , and Emx2 Cre (Pirvola et al., 2002;Ono et al., 2014). This results in a cochlear epithelium with reduced numbers of HCs, with OHC numbers being more severely affected than IHC numbers. In addition to the loss of differentiated HCs, a 40-50% decrease in cochlear length was reported when Fgfr1 was inactivated with Six1enh21 Cre , or Emx2 Cre (Ono et al., 2014).
Because Fgfr2 often exhibits redundancy with Fgfr1, it is important to consider potential Fgfr2 function in the inner ear prosensory epithelium. However, Fgfr2 is required for formation of the otic vesicle and Foxg1 Cre , which is active before and during the otic vesicle stage (Hébert and McConnell, 2000), could not be used to investigate the role of Fgfr2 at later stages of otic vesicle development. In addition, due to overall activity of Foxg1 Cre in the otic vesicle, cell type specificity of Fgfr1 was still unknown. To study whether Fgfr1 and/or Fgfr2 function cell autonomously or non-cell autonomously in the Fgf20 + domain of the prosensory epithelium, we generated an Fgf20 Cre allele (Figure 4-figure supplement 2A) to allow conditional gene targeting of the Fgf20 lineage. To assess Cre activity, Fgf20 Cre/+ ;ROSA mTmG/+ mice were generated. Cre activity was detected at E10.5 in a subset of the Sox2 + prosensory domain, in a pattern identical to that of Fgf20 lacZ embryos (Figure 4-figure  supplement 2B). At P0, all of the components of the organ of Corti were positive for the Fgf20 Cre/+ ; ROSA mTmG/+ lineage tracer, indicating that Fgf20 Cre is active in prosensory progenitors or their lineage (Figure 4-figure supplement 2B).
The length of the cochleae from E18.5 Fgfr1 −/f ::Fgf20 Cre/+ and Fgfr1 −/f ;Fgfr2 −/f ::Fgf20 Cre/+ embryos was decreased by 19% and 25%, respectively, compared to controls (p < 0.0001, Figure 4G). However, the length of the cochleae from Fgfr2 −/f ::Fgf20 Cre/+ was comparable (p > 0.5) to controls. Together, these data, and those presented above, showed that epithelial Fgfr1, but not Fgfr2, is required for lateral compartment differentiation, and has a modest effect on cochlear duct length of a similar magnitude to the 10% reduction in cochlear length seen in Fgf20 lacZ/lacZ mice . This reduction in cochlear length could be due to reduced numbers of progenitors or to other effects of FGFR1 signaling on cochlear duct elongation at later stages of development. Whole mount Sox2 staining at E14.5 of Fgfr1 −/f ;Fgfr2 −/f ::Fgf20 Cre/+ cochleae showed a similarly sized sensory progenitor domain as compared to controls indicating that the Sox2 + progenitor population was not affected by inactivation of epithelial Fgfr1 and Fgfr2 ( Figure 4D). In addition, proliferation of Fgfr1 −/f ; Fgfr2 −/f ::Fgf20 Cre/+ cochleae was comparable (p > 0.5) to controls at E12.5 ( Figure 4E,H).
To determine whether the effect of loss of mesenchymal FGFRs on cochlear length originates early in development, we examined the size of the Sox2 + progenitor domain at the time that HCs commit to differentiate, and cell proliferation within the Sox2 + domain before the onset of differentiation. The size of the Sox2 + progenitor domain, visualized by whole mount Sox2 staining of E14.5 cochleae was decreased in Fgfr1 −/f ;Fgfr2 −/f ::Twist2 Cre/+ embryos compared to control embryos ( Figure 5D). In addition, proliferation of Sox2 + progenitors from Fgfr1 −/f ;Fgfr2 −/f ::Twist2 Cre/+ cochleae was significantly (p < 0.01) decreased compared to control cochleae at E12.5 ( Figure 5E,H). Together, these data show that mesenchymal FGFR signaling is a necessary determinant of cochlear length and sensory progenitor proliferation, but not for cochlear pattern formation or differentiation.
To determine whether the FGF signaling pathway is affected in periotic mesenchyme, whole mount RNA in situ hybridization was used to localize expression of Etv4 and Etv5, two transcription factors that are commonly regulated by FGF signaling (Raible and Brand, 2001;Firnberg and Neubüser, 2002;Brent and Tabin, 2004;Mao et al., 2009;Zhang et al., 2009). Compared to double heterozygous control and Fgf9 −/+ ;Fgf20 lacZ/lacZ inner ears, Fgf9 −/− ;Fgf20 lacZ/lacZ inner ears showed decreased expression of Etv4 and Etv5 in mesenchyme surrounding the cochlear duct ( Figure 6-figure supplement 1A,B). The only known mesenchymal signaling pathway to regulate sensory progenitor proliferation is a Tbx1/Pou3f4 dependent retinoic acid (RA) signaling cascade (Braunstein et al., 2008(Braunstein et al., , 2009). However, Tbx1 and Pou3f4 expression, using RNA in situ hybridization in the embryos lacking Fgf9 and Fgf20 (Fgf9 −/− ;Fgf20 lacZ/lacZ ), did not reveal a change in expression of these transcription factors compared to doble heterozygous control and Fgf9 −/+ ; Fgf20 lacZ/lacZ embryos ( Figure 6-figure supplement 1C,D), suggesting that FGF signaling may function independent of RA signaling.

Discussion
Sensory progenitor proliferation and differentiation are temporally distinct events in cochlear development. In mice, sensory progenitors exit from the cell cycle beginning at the apical end of the cochlea at ∼E12.5 and ending at the base at ∼E14.5. In contrast, differentiation begins in the mid-base at ∼E14.5 and then extends to the base and apex (Wu and Kelley, 2012). Under physiological conditions, once progenitors exit the cell cycle, they do not reenter the cell cycle throughout the life of the organism. Previous studies suggested that during development both epithelial and mesenchymal signals are required to regulate cochlear progenitor proliferation and differentiation Doetzlhofer et al., 2004). However, the mechanisms that control cochlear sensory progenitor proliferation are not known. In this study, we found that epithelial FGF9 and FGF20 signaling to mesenchymal FGFR1 and FGFR2 is required for normal levels of cochlear sensory progenitor proliferation and that inactivation of either the ligands or the mesenchymal receptors results in a shortened cochlea. We also demonstrated that activation of mesenchymal FGFR signaling is sufficient to increase sensory progenitor proliferation and extend cochlear length.
Fgf9 is expressed in non-sensory epithelia of the cochlea and loss of Fgf9 results in defects in periotic mesenchymal cell proliferation, causing a hypoplastic otic capsule (Pirvola et al., 2004). Based on known expression patterns in mesenchyme, Fgfr1 and Fgfr2 were considered the most likely targets of FGF9 signaling (Pirvola et al., 2004). The critical time window for FGF9 signaling was determined to occur before E14.5. By contrast, Fgf20 is expressed in the sensory epithelium and loss of Fgf20 results in failure of the lateral compartment of the organ of Corti to fully differentiate . Based on expression patterns and phenotypic similarities with epithelial Fgfr1 conditional gene inactivation, FGFR1 was identified as the epithelial target receptor (Pirvola et al., 2002;Huh et al., 2012).
The effects of epithelial FGFR1 signaling on the length of the cochlear duct exhibit variability among studies. Ono et al. (2014) report a 40-50% decrease in cochlear length in Fgfr1 f/f :: Six1enh21 Cre , and Fgfr1 f/f ::Emx2 Cre conditional knockout mice, by contrast, Fgf20 lacZ/lacZ , Fgf9 −/+ ; Fgf20 lacZ/lacZ , and Fgfr1 −/f ::Foxg1 Cre/+ mice that we studied , and this study) showed only a 10-25% decrease in cochlear length. It is clear that in both studies defects in epithelial differentiation is likely to result in some decrease in cochlear length. It is also possible that differences in genetic background could contribute to differences in these two studies.
FGF9 and FGF20 are members of the same FGF subfamily and share similar biochemical properties (Zhang et al., 2006;Ornitz and Itoh, 2015). Redundancy between these FGFs has also been demonstrated in kidney development, where both ligands are required for nephron progenitor maintenance (Barak et al., 2012). Interestingly, in both cases, the expression patterns of these two FGFs do not overlap, but nevertheless they appear to signal to a common target tissue, periotic mesenchyme in the developing inner ear and CAP mesenchyme in the developing kidney. For the evolution of the kidney and inner ear, it is possible that additive expression of these FGFs from distinct sources was required to take advantage of their unique receptor specificities or unique interactions with the extracellular matrix.
Tbx1 is a transcription factor that is expressed in both sensory epithelium and mesenchyme (Vitelli et al., 2003;Raft et al., 2004). Deletion of Tbx1 in mesenchymal cells resulted in defects in cochlear epithelial proliferation indicating a non-cell autonomous requirement for Tbx1 for cochlear epithelial development (Xu et al., 2007). In addition, the Pou domain containing transcription factor, Pou3f4, also known as Brn4, is expressed in mesenchymal cells in the developing inner ear (Phippard et al., 1999). Deletion of Pou3f4 resulted in reduction of cochlear length and defects in derivatives of the otic mesenchyme including the spiral limbus, scala tympani, and strial fibrocytes (Phippard et al., 1999). Furthermore, decreasing gene dosages of Tbx1 and Pou3f4 resulted in a significant decrease in sensory epithelial proliferation and cochlear length indicating that Tbx1 and Pou3f4 genetically interact. The RA catabolizing genes Cyp26a1 and Cyp26c1, both targets of Tbx1 and Pou3f4, were decreased in these mice, suggesting that increased RA signaling could directly or indirectly suppress sensory progenitor proliferation (Braunstein et al., 2008(Braunstein et al., , 2009. Analysis of Fgf9 and Fgf20 double mutant mice showed no change in the expression of Tbx1 and Pou3f4 in mesenchyme surrounding the otic vesicle, suggesting that mesenchymal FGF signaling does not directly affect transcription factors that regulate RA signaling ( Figure 6-figure supplement 1C,D). On the other hand, Etv4 and Etv5 function as downstream targets of FGF signaling in other systems including the limb (Mao et al., 2009;Zhang et al., 2009), and Etv4 and Etv5 expression were decreased in Fgf9/Fgf20 mutant ears. Future studies will be needed to determine whether FGF signaling including ETV4 and ETV5 regulates RA signaling downstream of Tbx1/Pou3f4 or act in parallel to the Tbx1/Pou3f4/RA signaling pathway to regulate sensory progenitor proliferation. Whether the cellular target of RA signaling is in the periotic mesenchyme or the sensory progenitor epithelium also remains to be determined. It is also possible that the number of nearby mesenchymal cells may influence sensory progenitor proliferation. However, considering that loss of Fgf9 resulted in decreased mesenchymal cell proliferation (Pirvola et al., 2004) but did not affect HC formation or cochlear length (Figure 2), alternative mechanisms may need to be considered.
The reactivation of developmental signaling pathways may be important for regeneration. Recent publications showed that inhibition of Notch signaling could induce transdifferentiation of SCs to HCs in a damaged cochlea (Korrapati et al., 2013;Mizutari et al., 2013). In addition, Wnt/β-catenin signaling can induce SC proliferation in neonatal mice (Chai et al., 2012;Shi et al., 2012). One intermediate goal of regenerative biology for the inner ear would be to generate large numbers of sensory progenitor cells that could be differentiated into functional HCs and SCs and then be reintroduced into the damaged inner ear. The studies presented here suggest that in efforts to grow inner ear sensory progenitor cells in vitro, that FGF-induced mesenchyme may be necessary. The identification of mesenchymal factors that are regulated by FGF or RA could also be used to support the growth of sensory progenitor cells.

Generation of Fgf20 Cre mutant mice
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Washington University Division of Comparative Medicine Animal Studies Committee (Protocol Number 20130201). All efforts were made to minimize animal suffering.

Histology
For frozen sections, embryos were fixed with 4% paraformaldehyde overnight and washed with PBS. Samples were soaked in 30% sucrose and embedded in OCT compound (Tissue-Tek, Torrance, CA). Samples were sectioned (12 μm) and stored at −80˚C for immunohistochemistry.

Quantification of HC and SC numbers
Either phalloidin or Prox1 immunostaining were used to identify HCs and SCs, respectively. To measure the density of HCs and SCs, at least 300 μm regions of the base (10%), middle (40%), and apex (70%) of the cochleae were counted and normalized to 100 μm along the length of the cochlear duct. Inner and OHCs were identified by location and morphology of phalloidin staining. Cell counting was performed using Image J software.

Proliferation and cell death analyses
To analyze progenitor proliferation and cell death, frozen sections were prepared from the entire ventral inner ear of E11.5 or E12.5 embryos. Alternate sections were subjected to staining for pHH3 and Sox2 (for proliferation) or activated-Caspase 3 and Sox2 (for cell death, data not shown). For EdU labeling, pregnant females were injected with 50 μg/g (body weight) of EdU according to the manufacture's recommendation. Embryos were collected 2 hr after EdU injection. EdU was detected with the Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen, Carlsbad, CA) according to manufacture's instructions. The total area of Sox + cells was measured using Image J software and pHH3 + or activated-Caspase 3 + cells within the Sox2 + domain were counted. Counting was normalized to 10,000 μm 2 of Sox2 + prosensory epithelium.

Statistics
Numbers of samples are indicated for each experiment. All data are presented as mean ± standard deviation (sd). The p value for difference between two samples was calculated using a two-tailed Student's t-test or one-way ANOVA where appropriate. p < 0.05 was considered as significant.