|Home | About | Journals | Submit | Contact Us | Français|
Morphogenesis of sensory hair cells, in particular their mechanotransduction organelle, the stereociliary bundle, requires highly organized remodeling of the actin cytoskeleton. The roles of Rho family small GTPases during this process remain unknown. Here we show that deletion of Rac1 in the otic epithelium resulted in severe defects in cochlear epithelial morphogenesis. The mutant cochlea was severely shortened with a reduced number of auditory hair cells and cellular organization of the auditory sensory epithelium was abnormal. Rac1 mutant hair cells also displayed defects in planar cell polarity and morphogenesis of the stereociliary bundle, including bundle fragmentation or deformation, and mispositioning or absence of the kinocilium. We further demonstrate that a Rac-PAK signaling pathway mediates kinocilium-stereocilia interactions and is required for cohesion of the stereociliary bundle. Together, these results reveal a critical function of Rac1 in morphogenesis of the auditory sensory epithelium and stereociliary bundle.
In the mammalian inner ear, mechanosensory hair cells located within a specialized sensory epithelium in the cochlea, the organ of Corti (OC), are responsible for sound detection. These hair cells are interdigitated with non-sensory support cells, and each contain on their apical surface a V- shaped stereociliary bundle (or hair bundle) with intrinsic structural polarity: the actin-based stereocilia are organized into rows of graded heights, forming a staircase-like pattern. Furthermore, auditory hair bundles are uniformly oriented across the OC, with the V pointing toward the outer (lateral) border of the cochlear duct. These features are essential for the correct perception of sound and are established by an intricate gene network during development (Frolenkov et al., 2004; Petit and Richardson, 2009).
In the mouse, hair bundle development and maturation proceed in two perpendicular gradients, from the base to the apex and from the medial to lateral side of the cochlea over a period from late embryogenesis to the first two postnatal weeks (Frolenkov et al., 2004; Petit and Richardson, 2009). During the initial phase, a single tubulin-based kinocilium, derived from the primary cilium, migrates from the center to the lateral edge of the hair cell apex. Concomitant with this migration, microvilli around the kinocilium elongate to form stereocilia of graded heights. By the end of embryogenesis, nascent hair bundles exhibit a crescent shape with the kinocilium centered next to the tallest stereocilia. Next, during the first postnatal week, stereocilia undergo further row-specific, differential outgrowth, eventually forming a hair bundle with a staircase organization. Finally, the kinocilium retracts around postnatal day 10.
Tremendous progress has been made towards understanding the molecular mechanisms that regulate hair bundle morphogenesis. In particular, genetic analysis and cloning of deafness mutations in humans and mice have identified a number of structural and regulatory proteins of the actin cytoskeleton (El-Amraoui and Petit, 2005; Leibovici et al., 2008; Petit and Richardson, 2009). In addition, a Wnt/planar cell polarity (PCP) pathway, with both evolutionarily conserved (such as the Frizzled and Dishevelled proteins) and novel components (such as Scrb1 and PTK7), plays an important role in regulating cochlear elongation and hair bundle orientation (Rida and Chen, 2009).
Despite the importance of the actin cytoskeleton in hair cell morphogenesis, the roles of Rho GTPases, central regulators of actin dynamics, in this process remain poorly understood. Rho family members, including Rho, Rac and Cdc42, have been shown to influence cell migration, cell-cell adhesion, microtubule dynamics, cell proliferation, apoptosis, and gene transcription (Jaffe and Hall, 2005). Thus, Rho GTPases could potentially regulate many aspects of cochlear epithelial morphogenesis.
In this study, we show that the small GTPase Rac1 plays a critical role in the morphogenesis of the OC and the auditory hair bundles. Furthermore, we provide evidence that Rac1 regulates interactions between the kinocilium and stereocilia via PAK, which is required for the cohesion of the developing hair bundle. These results establish Rac1 as a key regulator of auditory hair cell morphogenesis.
The Rac1 conditional allele, Foxg1-Cre mice, and Pax2-Cre mice were previously described (Hebert and McConnell, 2000; Glogauer et al., 2003; Ohyama and Groves, 2004). All strains were maintained on a mixed genetic background. LtapLp mice were obtained from the Jackson Laboratory and CD1 mice from Charles River. Foxg1-Cre; Rac1KO/+ males were bred with Rac1CO/CO females to generate Foxg1-Cre; Rac1KO/CO mutants and littermate controls. Pax2-Cre; Rac1CO/+ females were bred with Rac1CO/CO males to generate Pax2-Cre; Rac1CO/CO mutants and littermate controls. PCR genotyping of the Rac1 alleles was performed as described (Glogauer et al., 2003). Mice were genotyped for Cre using the following primers: 5’-AGAACCTGAAGATGTTCGCG-3’ and 5’GGCTATACGTAACAGGGTGT-3’. For timed pregnancies, the morning of the plug was designated as embryonic day 0.5 (E0.5), and the day of birth postnatal day 0 (P0). Animal care and use was in accordance with NIH guidelines and was approved by the Animal Care and Use Committee at the University of Virginia.
Temporal bones were dissected and fixed in 4% paraformaldehyde for an hour at room temperature or overnight at 4°C and washed in PBS. For whole-mounts, cochleae were dissected out of the temporal bones in PBS and the anlage of Reissner’s membrane removed to expose the sensory epithelium. For sectioning, fixed temporal bones were dissected in PBS, equilibrated overnight in 30% sucrose and then embedded and snap frozen in OCT (Tissue Tek). Temporal bones were cryosectioned at 14µm thickness. Sections or dissected cochleae were incubated in PBS + 5% heat inactivated goat serum + 0.1% Triton + 0.02% NaN3 (blocking solution) for 1 hour at room temperature, followed by overnight incubation with primary antibodies diluted in blocking solution at 4°C. After three washes in PBS + 0.1% Triton, samples were incubated with secondary antibodies diluted in blocking buffer for 1 hour at room temperature. Cochleae were flat-mounted in Mowiol with 5% N-propyl gallate. Images were collected using a Zeiss LSM 510 Meta confocal microscope and LSM Image Browser software, or using a DeltaVision deconvolution microscope and softWoRx software (Applied Precision). Images were then processed in Adobe Photoshop (Adobe Systems). For flat-mount images of the OC, Z-stacks of 5 or more microns were taken at 0.2 µm intervals, encompassing a volume from the tips of the hair bundle to the apical cytoplasm of the hair cell. When comparing protein localizations, the same exposure conditions were used for controls and mutants, and every plane of the Z-stack was carefully examined to ensure comparisons were made at equivalent axial levels. All comparisons were carried out at equivalent locations along the length of the cochlear duct. The following primary antibodies were used: anti-Rac1 (1:1000, BD Biosciences), anti-Myosin VI (1:1000, Proteus BioSciences), anti-Myosin VIIa (1:1000, Proteus BioSciences), anti-acetylated tubulin (1:500, Sigma), anti-α-tubulin (1:1000, Sigma), anti-β1/β2-tubulin (1:200, Sigma), anti-α-spectrin (1:100, Millipore), anti-E-cadherin (1:1000, Sigma), anti-β-catenin (1:1000, Cell Signaling), anti-Fz3 [1:200, (Wang et al., 2006b)], Pan-Espin antibody [1:500, (Sekerkova et al., 2006)], anti-Myosin XVa [1:150, (Belyantseva et al., 2005)], anti-Whirlin antibodies [1:300, (Belyantseva et al., 2005)], anti-harmonin (1:100, ProteinTech Group), anti-phosphoPAK1/2/3 (1:200, Biosource), and anti-p27kip1 (1:200, Neomarkers, used after antigen retrieval by boiling in 10mM citrate buffer, pH 6.0, for 10 min). Alexa–conjugated secondary antibodies (1:1000), Alexa- and rhodamine-conjugated phalloidin (1:100) and Hoechst 33342 (1:10,000) were from Invitrogen.
293T cells were transfected in 6 well plates using the FuGene reagent (Roche) according to manufacturer’s instructions. Protein lysates were isolated 24 hours after transfection and analyzed by Western Blotting. Samples were run in duplicates on the same gel and immunoblotted separately with anti-Rac1 or anti-HA antibodies. pcDNA3.1-Rac1-3xHA, pcDNA3.1-Rac2-3xHA and pcDNA3.1-Rac3-3xHA plasmids were provided by Dr. Thomas Parsons (University of Virginia).
For determination of hair cell number, consecutive 100× images of the entire OC stained with phalloidin and/or myosin VI were obtained on a Zeiss AxioImager Z1 upright microscope and the number of hair cells counted. OC length was determined on 10× images of the OC. For quantitation of hair bundle defects, three mutants with a total of 622 hair cells, and four littermate controls with a total of 945 hair cells were analyzed from the basal region of the cochlea at E18.5. Phalloidin and acetylated tubulin stained hair cells were classified as having a normal, split, flat, or generally deformed bundle morphology, and also as having a normal, offcentered (relative to the hair bundle), or absent/poorly formed kinocilium. Mean ± s.d. is given. Statistical probability was measured using Student’s t-test. For quantitation of hair bundle orientation, hair cells from the basal region of three mutant and three littermate control cochleae were analyzed at E18.5. Orientation was measured on a total of 268 mutant and 268 control cells (60 IHC, 70 OHC1, 70 OHC2, and 68 OHC3, respectively) using Image J software. The orientation of a line drawn straight through the center of the bundle relative to a line drawn perpendicular to the pillar cells was used to determine bundle orientation. Bundles aligned with the perpendicular line and oriented towards the lateral border of the cochlear duct were assigned an orientation of 0°. Due to the kinocilium defects in the Rac1 mutant, orientation was estimated based upon phalloidin staining only. Mutant cells with deformed bundles lacking a discernable orientation were not included in the analysis.
For SEM analysis, inner ears from E18.5 mice were dissected in PBS and fixed in 4% paraformaldehyde with 2.5% glutaraldehyde overnight at 4°C. Cochleae were then dissected out of the temporal bone, postfixed in 1% osmium tetroxide, and dehydrated through a graded ethanol series. Specimens were then critical point dried, mounted on metal studs and sputter coated with gold. Samples were examined on a JEOL 6400 scanning electron microscope at 20 kV.
Cochlear explants from Rac1 mutants and littermate controls were established at E18.5. Briefly, cochleae were dissected in HEPES-buffered HBSS (Invitrogen) and the developing OC was exposed. Cochleae with attached stria vascularis were established on coverslips coated with Cell Tak (BD Biosciences). Explants were then maintained for four days in vitro (equivalent of P3) in DMEM/F-12 (Invitrogen) supplemented with N2 (Invitrogen) and penicillin (Sigma). To assess FM 1-43 uptake, samples were treated with 5µM FM 1-43 (Invitrogen) for 10 seconds as previously described (Lelli et al., 2009). For quantitation of hair bundle defects in Rac1 explants, a total of 499 Rac1 mutant hair cells and 325 control hair cells from three independent experiments were analyzed from the mid-basal region of the cochlea at the equivalent of P3. Phalloidin-stained hair cells were classified as having a normal, split, flat, or generally deformed bundle morphology. Mean ± s.d. is given. Statistical probability was measured using Student’s t-test. For PAK inhibition experiments, cochlear explants from CD1 mice were established either on E15.5, E16.5 or E17.5, and treated on the following day (equivalent of E16.5, E17.5 and E18.5, respectively) with vehicle (DMSO) or the PAK inhibitor IPA-3 (Calbiochem). Six hours after drug application, media were replaced with drug-free media. After four or five days in vitro (equivalent of P2), explants were fixed with 4% PFA and processed for acetylated tubulin and phalloidin staining. For quantitation of hair bundle defects in IPA-3 treated explants, a total of 514 hair cells (DMSO control), 280 hair cells (5µM IPA-3), and 533 hair cells (10µM IPA-3) from 3 independent experiments were analyzed from the mid-basal region of the cochlea and bundle morphology was scored as described above.
Of the three mammalian Rac family members, Rac1 is ubiquitously expressed whereas Rac2 and Rac3 are restricted to hematopoetic and neuronal cells, respectively (Haataja et al., 1997; Glogauer et al., 2003). Both Rac1 and Rac3 transcripts are present in the developing cochlea (Supplementary Figure 1A). To determine Rac protein distribution in the OC, we performed immunostaining using an anti-Rac1 antibody from BD Biosciences that recognizes all three Rac family members (Supplementary Figure 1B). At both E16.5 and E18.5, Rac proteins were present in both sensory hair cells and support cells (e.g., Deiter’s, pillar, inner phalangeal) along the entire cochlear spiral (Supplementary Figure 1C–F). In hair cells, Rac staining was particularly prominent in the nascent hair bundles, delineating the entire length of the stereocilia as well as the kinocilium. In support cells, we observed strong Rac staining in the actin-rich microvilli. Thus, their expression pattern during the first phase of stereocilium elongation (Frolenkov et al., 2004; Petit and Richardson, 2009) suggests a possible role for Rac proteins in the development of the auditory sensory epithelium and hair bundle. Since Rac3-deficient mice are viable, fertile and have no reported inner ear phenotypes (Cho et al., 2005), we focused our functional analysis on Rac1.
Rac1 knockout mice die around E8 during embryonic development (Sugihara et al., 1998). Therefore, to assess the in vivo function of Rac1 during inner ear development, we used a conditional Rac1 allele (Rac1CO) that contains loxP sites flanking exon 1 and the transcriptional start site of the Rac1 gene (Glogauer et al., 2003). Cre-mediated recombination between the loxP sites has been shown to result in deletion of this region thereby eliminating Rac1 protein expression (Glogauer et al., 2003; Chen et al., 2007). To inactivate Rac1 in the entire otic epithelium, we used the Foxg1-Cre strain, in which the Cre recombinase gene is targeted to the Foxg1 locus, resulting in Cre expression in the early otocyst from E8.75 (Hebert and McConnell, 2000). We also used a Rac1 null allele (Rac1KO) that was derived from the Rac1CO allele by germline Cre expression (Gu et al., 2003). To generate Rac1 mutants, we crossed Foxg1-Cre; Rac1KO/+ males to Rac1CO/CO females to obtain Foxg1-Cre; Rac1KO/CO progeny. Littermates were used as controls. Although no Foxg1-Cre; Rac1KO/CO mice survived after birth (most likely due to Cre expression outside the inner ear (Hebert and McConnell, 2000; Chen et al., 2007), Foxg1-Cre; Rac1KO/CO embryos were present in a Mendelian ratio up to E18.5 (155/682=23%, expected 25%).
At E18.5, the temporal bones from Foxg1-Cre; Rac1KO/CO (referred to as Rac1 mutant hereafter) embryos appeared highly dysmorphic, and this phenotype was fully penetrant in all mutants analyzed. At a gross level, the temporal bones were much smaller and abnormally shaped compared to littermate controls (Figure 1A). In particular, the otic capsule surrounding the vestibular apparatus had a narrow and pointed shape, suggestive of semicircular canal defects. In this study, we focused our analysis on the cochlear phenotypes. At E18.5, although a cochlea was present, it was dramatically reduced in length [controls: 4870µm ± 320 (n=6); mutants: 2040µm ±100 (n=4)] (Figure 1B). The cochlea spiral formed half to three quarters of one turn versus one and three quarters turns in the controls. Of note, the width of the cochlear duct was normal and we did not detect supernumerary rows of hair cells along the cochlea. We quantified the number of hair cells by myosin VI and/or phalloidin staining. The total number of hair cells present in the OC of Rac1 mutant embryos was significantly reduced compared to littermate controls [controls: 2437 ± 49 (n=5); mutants: 1032 ± 69 (n=5)]. This reduction in hair cell number directly correlated with the decreased length of the OC, indicating a relatively normal hair cell density.
Based on immunostaining of hair cell and support cell markers, differentiation of hair cells and support cells in Rac1 mutants was essentially normal (Figure 2, Supplementary Figure 2A–F and data not shown). This suggests that Rac1 is not required for cell fate determination in the OC. In addition, three rows of outer hair cells and one row of inner hair cells were mostly present throughout the mutant cochlea. However, cellular disorganization in the Rac1 mutant OC can be detected as early as E15.5; by E18.5, the invariant mosaic pattern of hair cells interdigitated with support cells with regular spacing was abnormal in Rac1 mutants (Figure 2A–B, Supplementary Figure 2A–H). In many cases, pairs of hair cells appeared to be in contact with each other (Figure 2B, arrows). Similarly, the apical extensions of two support cells often appeared to be in contact (Figure 2B, arrowhead). Moreover, both the shape and the size of hair cells were irregular in Rac1 mutants. The apical surface of many hair cells often appeared oblong, in contrast to a uniformly round appearance in controls (Figure 2B).
Cellular disorganization was also apparent in cross-sections of the OC along the apical-basal polarity axis (Figure 2C–F). In control cross-sections, the hair cell nuclei were aligned at a uniform distance from the luminal surface of the epithelium. By contrast, in Rac1 mutants the hair cell nuclei were often found at variable distances from the luminal surface (Figure 2D, arrow). Moreover, because of disrupted spacing of the cellular mosaic, more than four hair cells were frequently present in cross-sections of Rac1 mutants, in contrast to control cross-sections, which invariably have four hair cells. We also observed irregular spacing between support cell nuclei in Rac1 mutants (Figure 2D, arrowheads). The shape of the nuclei of both hair cells and support cells was also irregular. Finally, the height of the hair cells and support cells in Rac1 mutants appeared shorter than controls, and as a result the overall thickness of the OC was thinner in Rac1 mutants compared to controls. Together, these data suggest that Rac1 is critical for cellular patterning in the OC.
To determine if cellular disorganization in the Rac1 mutant OC resulted from defective cell-cell adhesion (Jaffe and Hall, 2005), we examined the localization of E-cadherin and β-catenin, components of the adherens junction (Supplementary Figure 3). At E18.5, E-cadherin and β-catenin are localized on the basolateral membranes of both hair cells and support cells. In Rac1 mutants, there are no overt changes in the distribution of either E-cadherin or β-catenin, suggesting that Rac1 is not required for the recruitment of these junctional proteins to cellular contacts.
Immunostaining of Rac1 mutant hair bundles at E18.5 revealed both bundle misorientation and structural deformation (Figure 5). The orientation defects were observed in all hair cell rows and in both structurally normal and abnormal bundles. Overall, the angle measurements of misoriented bundles were mostly within 50 degrees, with outer hair cell rows 2 and 3 more affected than other rows (Supplementary Figure 4).
Bundle orientation, a readout for hair cell PCP, is regulated by a PCP pathway with both conserved and novel components (Rida and Chen, 2009). The first sign of planar polarization of hair cells is the migration of the kinocilium from the center to the lateral edge of the hair cell apex, which starts at the base of the cochlea at E16.5 (Montcouquiol et al., 2003). To determine if Rac1 regulates the initial establishment of hair cell PCP, we scored the location of the kinocilium in the basal region of the cochlea at E16.5. In the Rac1 mutant, most kinocilia were positioned towards the hair cell lateral edge, similar to controls (Supplementary Figure 2I–L). Thus, establishment of PCP appears to be normal in the Rac1 mutant.
A hallmark feature of PCP proteins is that they become asymmetrically localized during signaling. In the OC, for example, Dishevelled-2 appears to be enriched on the lateral side of the hair cell membranes, whereas Vangl2 and Frizzled-3 (Fz3) are located on the medial side (Wang et al., 2005; Montcouquiol et al., 2006; Wang et al., 2006a; Wang et al., 2006b; Etheridge et al., 2008). Asymmetric membrane localization of PCP proteins requires cell-cell interactions (Adler, 2002; Rida and Chen, 2009). Because the normal pattern of cell-cell interaction is disrupted in the Rac1 mutant OC (Figure 2 and Supplementary Figure 2A–H), we examined whether the localization of the PCP protein Fz3 was affected. In controls, at E17.5 in the mid-basal region of the cochlea, Fz3 staining is already medially enriched, forming a characteristic line abutting the medial border of hair cell rows (Figure 3A). By E18.5, asymmetric Fz3 staining is even more apparent, forming a tight crescent on the medial side of hair cell membranes, as previously reported (Montcouquiol et al., 2006; Wang et al., 2006b) (Figure 3C). Fz3 was also present on cell membranes around the pillar cells. By contrast, in E17.5 Rac1 mutants, although Fz3 still appeared to be partially localized to cell membranes, it was no longer uniformly enriched on the medial side (Figure 3B). By E18.5, the membrane crescent of Fz3 was not detectable in the mutant (Figure 3D). However, membrane staining around the pillar cells remained normal. These results suggest that Rac1 is required for asymmetric distribution of Fz3 in the developing OC. Together with the fact that the Rac1 mutation has no apparent effect on kinocilium migration, an early event during planar polarization, these results are consistent with a role of Rac1 in the maintenance, rather than the initial establishment of PCP in the developing OC.
Previously Rac1 was identified as a downstream effector of PCP signaling regulating convergent extension during gastrulation (Habas et al., 2003). To test if Rac1 activity is regulated by PCP signaling in the OC, we focused on p21-activated kinases (PAK), which are well known downstream effectors of Rac that regulate cytoskeletal dynamics (Bokoch, 2003). We first examined the pattern of PAK activation in the wild type OC, using a phospho-PAK (pPAK) antibody that specifically detects a phosphoepitope present in activated PAK. At E16.5, during the initial establishment of PCP, pPAK appeared to be enriched on the lateral membranes of hair cells in the basal to mid-basal region but not yet in the more apical regions of the cochlea (Figure 4A–B). Concomitant with the wave of hair cell differentiation along the cochlear duct, the asymmetric membrane localization of pPAK spread toward the cochlear apex by E17.5. In the mid-basal region at E17.5, the pPAK crescent correlated perfectly with hair bundle orientation (Figure 4C–D). Next, we examined pPAK localization in the Lp mice, mutant for the conserved core PCP gene, vangl2 (Montcouquiol et al., 2003). In E16.5 Lp mutants, while pPAK was still asymmetrically localized on apical membranes, it was misoriented relative to the medial-lateral polarity axis (data not shown). Similarly, in E17.5 Lp mutants, the pPAK crescent was present but misoriented, which correlated perfectly with the misoriented hair bundle (Figure 4E–F). These results suggest that the pPAK localization is a novel readout for PCP in hair cells, and that the PCP pathway may spatially regulate PAK activity.
We closely examined hair bundle structural defects using immunostaining (Figure 5) and scanning electron microscopy (SEM) (Figure 6). Normal hair bundles have intrinsic structural polarity: stereocilia have graded heights, forming a crescent with the kinocilium centered next to the tallest stereocilia (Figure 5A and and6A).6A). By contrast, in Rac1 mutants, many hair bundles exhibited a variety of structural defects: stereocilia formed either flat, straight rows (flat bundles) (Figure 6D–F), or multiple clumps of stereocilia within the same cell (split bundles) (Figure 6G–H), or had other abnormal shapes (generally deformed bundles) (Figure 6I–J). In flat bundles, stereocilia either appeared uniform in height across the row (Figure 6D), or were organized in a wavy line or even a slight inverted crescent shape, with the taller stereocilia at the edges of the bundle (Figure 6E–F). Hair bundles defects were accompanied by abnormalities of the kinocilium. In many flat bundles, the kinocilium was located at one end of the bundle (off-center), rather than in the center (Figure 5B, 6B, 6D–F). Within the split bundles, the kinocilium was usually found centered within one group of stereocilia (Figure 5B, 6B, 6G–H). Moreover, the kinocilium in a fraction of mutant cells was either absent or poorly formed, as assessed by acetylated tubulin staining (Figure 5B). Of note, in Rac1 mutants, although the positioning of the kinocilium relative to the stereociliary bundle was often abnormal, the migration of the kinocilium to the lateral edge of the hair cell apex was essentially normal. Overall, 64% of Rac1 mutant hair cells scored from the basal region of the cochlea displayed some form of hair bundle disorganization (Figure 5C–D). Together these observations demonstrate a requirement of Rac1 during the early phase of bundle morphogenesis.
To investigate a potential role of Rac1 in regulating stereocilium elongation, we tested if Rac1 regulates the localization of known actin regulatory proteins that control stereocilium elongation. In particular, the espin actin-bundling protein, mutated in jerker mice, can drive stereocilium elongation when overexpressed and is localized throughout the stereocilia (Rzadzinska et al., 2005; Sekerkova et al., 2006). Myosin XVa and whirlin, mutations in which result in abnormally short stereocilia, are required for the differential elongation of the stereocilia and are localized to the tips of stereocilia (Belyantseva et al., 2005). Myosin VIIa, which is mutated in shaker-1 mice and human patients with Usher Syndrome 1B, is localized to the cuticular plate, the tips of the stereocilia and the kinocilium (Lefevre et al., 2008). We observed no overt changes in the localizations of these proteins in Rac1 mutants at E18.5 (Supplementary Figure 5), suggesting that Rac1 is not required for the localization of these key hair bundle proteins to the nascent hair bundle.
In Foxg1-Cre mice, the Cre recombinase is targeted to the Foxg1 locus, replacing the Foxg1 coding sequences thereby resulting in Foxg1 haploinsufficiency (Hebert and McConnell, 2000). The Foxg1 gene itself also regulates inner ear morphogenesis (Pauley et al., 2006). To exclude the possibility that the Foxg1-Cre; Rac1KO/CO mutant phenotype is due in part to Foxg1 haploinsufficiency, we used an independent Cre deletor strain to inactivate Rac1 in the developing inner ear. Pax2-Cre is a BAC transgenic line that expresses Cre throughout the early otocyst (E8.75) (Ohyama and Groves, 2004). We generated Pax2-Cre; Rac1CO/CO embryos and analyzed their inner ears at E17.5 [We were unable to recover mutants at E18.5, probably due to Cre expression outside the inner ear (Ohyama and Groves, 2004)]. In these embryos, we observed the full spectrum of phenotypes of Foxg1-Cre; Rac1KO/CO embryos, including abnormally shaped temporal bones, a shortened cochlear duct, cellular organization defects within the OC and hair bundle structural and orientation defects (Supplementary Figure 6). Together these data strongly support the conclusion that the phenotypes observed in Foxg1-Cre; Rac1KO/CO embryos are specific effects of Rac1 deletion.
Staircase formation and functional maturation of the hair bundle occurs in the first postnatal week in mice (El-Amraoui and Petit, 2005; Lelli et al., 2009). Because Rac1 mutant embryos die at birth, we were not able to assess hair bundle maturation of these mutants in vivo. To further examine maturation of Rac1 mutant hair cells, we established cochlear explant cultures from Foxg1-Cre; Rac1KO/CO mutant mice at E18.5 and allowed them to grow for 4 days in vitro (equivalent of P3). We then examined the stereociliary bundles by immnostaining and SEM analysis. In both mutant and control cultures, extra rows of outer hair cells were frequently observed, as has been observed previously in cochlear explants (Montcouquiol and Kelley, 2003). Nevertheless, in control cultures the hair bundles had normal morphology and orientation (Figure 7A, D, F, Supplementary Figure 7C). In mutant explants, on the other hand, 86% of hair bundles appeared dysmorphic. The severity of bundle disorganization varied among cultures, examples of which are shown in Figure 7B (moderate) and Figure 7C (severe). In severe cases, the majority of cells had fragmented (split) bundles detached from the kinocilium. Flat and other types of abnormal bundles were observed less frequently (Figure 7G–K, Supplementary Figure 7C). Consistent with the normal localization of bundle elongation proteins in Rac1 mutants (Supplementary Figure 5), the stereocilia in mutant explants still exhibited graded heights in spite of the fragmentation defects (Figure 7G–K), suggesting that formation of the staircase can proceed in the absence of Rac1.
To further evaluate functional maturation of Rac1 mutant hair cells, we assayed for FM1-43 uptake in explants. FM1-43 is a fluorescent styryl dye that can be taken up by hair cells through their mechanotransduction channels, which develop postnatally in the mouse cochlea, in a gradient from base (~P1) to apex (~P3) (Lelli et al., 2009). Rac1 mutant explants exhibited similar levels of FM1-43 uptake compared to control explants, consistent with the presence of transduction channels in Rac1 mutant hair cells (Supplementary Figure 7A–B). Together, these results argue that, during hair bundle maturation, Rac1 is specifically required for hair bundle cohesion, but is dispensable for differential elongation of the stereocilia and the acquisition of the mechanotransduction channels.
To investigate the mechanisms by which Rac1 regulates the cohesion of the developing hair bundle, we first examined if pPAK localization is altered in Rac1 mutants. At E17.5, while a robust pPAK crescent was observed in control OC (Figure 8A), in Rac1 mutants, a significant fraction of hair cells showed abnormal pPAK staining. In some hair cells, the membrane localization of pPAK was diffused or diminished (Figure 8B, arrow). In others, the crescent of pPAK staining lost its lateral orientation (Figure 8B, open arrowheads). By E18.5, in controls, in addition to the membrane localization (Figure 8C), pPAK was enriched in a region surrounding the lateral side of the hair bundle base and the kinocilium insertion site, and also on some stereocilia (Figure 8E and G). In this region, which was largely devoid of α-spectrin, a component of the cuticular plate, pPAK staining partially overlapped with microtubules (Supplementary Figure 8). In E18.5 Rac1 mutants, similar to E17.5, membrane staining of pPAK was diffused in some hair cells (Figure 8D). While some cells had normal pPAK staining around the hair bundle base, in others, particularly those with deformed bundles, pPAK staining around the bundle base became diffused and disorganized (Figure 8F, H and I). The staining pattern of pPAK supports the hypothesis that PAK acts downstream of Rac to regulate hair bundle morphogenesis.
To test this hypothesis, we applied a specific chemical inhibitor of PAK, IPA-3 (Deacon et al., 2008), to mouse embryonic cochlear explants at different developmental stages and examined the effect of blocking PAK signaling on hair bundle morphology. Strikingly, IPA-3 treatment at the equivalent of E18.5 resulted in hair bundle fragmentation and kinocilium mispositioning in a dosage-dependent manner (Figure 8J–L, Supplementary Figure 7D), phenocopying the bundle defects of Rac1 mutant explants (Figure 7, Supplementary Figure 7C). Specifically, treatment with 10µM IPA-3 resulted in 89% dysmorphic bundles, the majority of which were split (Supplementary Figure 7D). To determine the critical period of PAK signaling for hair bundle cohesion, we also treated explants at the equivalent of E16.5 or E17.5 with IPA-3. Treatment at E16.5 and E17.5 caused very slight and mild bundle deformation, respectively (data not shown). Therefore, we conclude that Rac-PAK signaling is required for the cohesion of the developing hair bundle during the late phase of bundle morphogenesis.
Many of the deafness genes implicated in hair bundle morphogenesis were originally identified through genetic analysis of hearing-impaired individuals or mouse mutants (Friedman et al., 2007; Leibovici et al., 2008). These genetic studies have provided great insights into the mechanisms underlying normal hair cell development and function, as well as the pathophysiology of deafness and balance disorders. However, the function of essential genes in hair cell development, such as ubiquitous or global regulators of the actin cytoskeleton, is still poorly understood due to the early lethality associated with loss-of-function mutations in these genes. Our results in this study demonstrate a requirement of Rac1, a central regulator of actin dynamics, in cochlear morphogenesis and hair bundle formation, and shed new light on the exquisite genetic control of hair bundle morphogenesis.
The pleiotropic effects of Rac1 deletion indicate that Rac1 participates in multiple developmental pathways to coordinate cochlear epithelial morphogenesis. Rac1 mutants have a severely shortened cochlea with a reduced number of hair cells and support cells, consistent with an early function of Rac1 in specification of the OC precursor cells, which awaits further studies. Rac1 deletion also had a deleterious effect on cell-cell interactions in the OC, with abnormalities in cell shape, spacing and arrangement. Although we did not detect gross changes in the distribution of adherens junction components, subtle defects in cell-cell adhesion and/or cell contractility may account for these phenotypes.
Rac1 is also necessary for normal PCP signaling. Rac1 mutants exhibited misoriented hair bundles similar to mouse PCP mutants. We found that the activity of PAK proteins, downstream effectors of Rac, is asymmetrically localized according to the PCP axis and is likely spatially regulated by PCP signaling during bundle orientation. However, we were unable to reliably examine the role of PAK in bundle orientation in vitro, as wild type cochlear explants grown on artificial substrates often have slight orientation defects (data not shown). We also found that asymmetric localization of Fz3 was disrupted in Rac1 mutants, which is likely a secondary effect of epithelial/cytoskeletal disorganization. There are precedents for a role of the cytoskeleton in PCP regulation (Blair et al., 2006; Shimada et al., 2006; Yan et al., 2009). Together, these results support both a direct role of Rac1 as a downstream effector of the PCP pathway and perhaps an indirect role in regulating Fz3 localization via the cytoskeleton. The multiple functions of Rac1 in regulating PCP in the inner ear parallels those of Rho1 in Drosophila, where Rho1 acts both downsteam and upstream of the core PCP genes (Yan et al., 2009). Our results provide further evidence that Rac GTPases are required for PCP signaling in vertebrates (Habas et al., 2003). This stands in contrast to Drosophila, where Rac genes appear to have very minor, if any, function in PCP regulation (Hakeda-Suzuki et al., 2002; Munoz-Descalzo et al., 2007). We speculate that, like other vertebrate-specific PCP regulators, the Rac genes are co-opted by the core PCP pathway to effect complex cytoskeletal remodeling during vertebrate tissue morphogenesis.
Rac1 function is essential for hair bundle morphogenesis. The split bundle phenotype, where stereocilia form separate clumps on the hair cell apex, resembles the phenotypes of mice mutant for Usher syndrome type I (USH1) genes, which encode myosin VIIa, harmonin, sans, cadherin 23 and protocadherin 15 (Lefevre et al., 2008). USH1 is characterized by congenital deafness, vestibular dysfunction and blindness caused by retinitis pigmentosa. Interestingly, Dock4, a guanine nucleotide exchange factor (GEF) for Rac1 (Lu et al., 2005), has recently been shown to localize to hair cell stereocilia, where it binds to harmonin (Yan et al., 2006). We observed no changes in myosin VIIa and harmonin localization in Rac1 mutants (Supplementary Figure 5 and data not shown), suggesting that Rac1 is not required for the localization of USH1 proteins. Although the function of Dock4 in hair bundle formation is not known, it is tempting to speculate that USH1 proteins act upstream to regulate Rac1 activity via Dock4 during bundle morphogenesis.
Flat bundles were also frequently observed in Rac1 mutants at E18.5, where the kinocilium had an abnormal off-center position. This phenotype suggests that Rac1 regulates interactions between the kinocilium and stereocilia in the developing hair bundle. There is increasing evidence for the role of the kinocilium and its associated basal bodies as the ‘hair bundle organizing center’ to instruct the intrinsic structural polarity of the hair bundle. Indeed, it has been reported that mouse mutants for genes implicated in human ciliopathy, such as the Bardet-Biedl syndrome (BBS), have ‘flattened’ bundles with mispositioned kinocilium (Ross et al., 2005). Furthermore, genetic ablation of the kinocilium in the mouse leads to formation of hair bundles with a flattened morphology (Jones et al., 2008; C. Sipe and X. Lu, unpublished results). Because Rac1 mutant explants had a predominant split bundle phenotype at the equivalent of P3, we infer that flat bundles proceed to become fragmented. Thus, our results highlight the requirement of kinocilium-stereocilium interaction for maintenance of hair bundle integrity.
In spite of bundle fragmentation, in Rac1 mutant explants we still observed formation of the bundle staircase and relatively normal FM1-43 uptake. These results argue that developmental pathways that regulate distinct aspects of hair bundle morphogenesis, such as hair bundle cohesion, differential elongation of the stereocilia and acquisition of the mechanotransduction apparatus, operate independently of one another. Consistent with this idea, FM1-43 uptake occurs normally in the Snell’s waltzer myosin VI mutant mouse, which displays severe bundle structural defects (Self et al., 1999). Likewise, mechanotransduction is normal in myosin XVa mutants with abnormally short stereocilia (Stepanyan et al., 2006). Nevertheless, disruption of any of the aforementioned aspects of bundle morphogenesis will disable hearing.
We identified PAK kinases as important downstream effectors of Rac1 during auditory hair bundle morphogenesis. Rac1 deletion appeared to reduce, rather than abolish pPAK staining. Together with the fact that some but not all Rac1 mutant hair cells displayed hair bundle defects, it suggests that other small GTPases may function redundantly with Rac1 to activate PAK signaling during bundle morphogenesis. Indeed, we found by RT-PCR analysis that, in addition to Rac1, Rac3 and the closely related Cdc42 were also expressed in the developing cochlea (Supplementary Figure 1A and data not shown). Furthermore, to determine if Rac1 has a similar function in vestibular hair cells, we examined Rac1 mutant utricles and observed essentially normal bundle morphology, FM1-43 uptake and mechanotransduction currents (data not shown). We suspect that functional redundancy is also in play in utricular hair cells.
How may PAK activity control bundle morphogenesis? We showed that pPAK staining was enriched on the lateral membranes, and in a region adjacent to the base of the bundle and the kinocilium insertion site, where it partially overlapped with microtubules. In explant assays, when PAK activity was inhibited by low-dose IPA-3, we observed flat bundles with off-center kinocilium, whereas higher doses of IPA-3 resulted in fragmentation of the bundle. Based on the staining pattern and the inhibitor experiments, we favor a model where PAK regulates the kinocilium-stereocilia interaction via microtubule-based mechanisms thereby maintaining hair bundle cohesion. Future identification of molecular targets of PAK will reveal the underlying mechanisms by which Rac-PAK signaling regulates hair bundle morphogenesis.
We thank Paul Adler, Jeff Holt, Bettina Winkler and Jason Kinchen for helpful comments on the manuscript, Jan Reddick for assistance with electron microscopy, Lixia Liu for technical support, Thomas Parsons for plasmids, and James Bartles, Thomas Friedman, Jeremy Nathans, Yanshu Wang and Bette Dzamba for antibodies. This study was supported by NIH grants RO1 DC009238 (to X.L.) and DC008853 (to G.S.G.G.). C.G–M. is supported by a fellowship from the Hartwell foundation. C.S. is supported by NIH training grant 5T32 GM008136-24 to the University of Virginia.
The authors declare no conflicts of interest.