The ATP-dependent chromatin remodeling enzyme CHD7 regulates pro-neural gene expression and neurogenesis in the inner ear

The vertebrate inner ear is a complex organ that is necessary for transducing hearing and balance signals. In mammals, the inner ear develops from the otic placode, an ectodermal thickening that is adjacent to the hindbrain that gives rise to the otocyst (Alsina et al., 2009; Bok et al., 2007). Axial specification of the otocyst, defined by regionalization of gene expression, is essential for generation of neuroblasts and normal development of inner ear structures (Alsina et al., 2009; Bok et al., 2007; Fekete and Wu, 2002). One of the earliest cell fate decisions during inner ear development is neural competence (Fekete and Wu, 2002). This event is concurrent with otic placode formation, and is defined by expression of the basic helix-loop-helix protein neurogenin 1 (Ngn1) (Alsina et al., 2009). Ngn1-null mice fail to express neural fate markers such as NeuroD (Neurod1), and lack all sensory neurons of the inner ear (Ma et al., 2000; Ma et al., 1998). Expression of Ngn1 and NeuroD, along with lunatic fringe (Lfng), Otx2 and the fibroblast growth factor Fgf10, defines the anterior region of the developing otocyst (Alsina et al., 2009; Bok et al., 2007; Fekete and Wu, 2002). Posteriorly expressed genes, such as Tbx1 (T-box transcription factor 1), stabilize the neurogenic region by inhibiting Ngn1, creating the otic anteroposterior axis (for a review, see Bok et al., 2007). The neurogenic region is also restricted ventrally by dorsal Wnt signaling from the neural tube and ventral Sonic hedgehog (Shh) signaling from the notochord and floor plate of the hindbrain (Riccomagno et al., 2002; Riccomagno et al., 2005).

After acquisition of neural competence and expression of pro-neural genes, cells within the anteroventral region of the otocyst delaminate and proliferate to form neurons of the statoacoustic (future vestibulo-cochlear) ganglion (Fritzsch et al., 2006; Fritzsch et al., 1999; Sanchez-Calderon et al., 2007). After delamination, ganglionic neuroblasts continue to proliferate and later differentiate into bipolar neurons that send projections to the brain and to inner ear sensory end organs (Fritzsch et al., 2005). As bipolar neurons develop, pro-neural fate markers are downregulated and genes that are essential for neuronal survival and extension, including TrkB and TrkC neurotrophin receptors, are expressed (Pirvola and Ylikoski, 2003; Sanchez-Calderon et al., 2007). Non-ganglionic progenitors in the neurogenic region of the otic epithelium are fated to develop into cells that occupy the sensory epithelium of the vestibular maculae (Raft et al., 2007).

Haploinsufficiency for CHD7, the chromatin remodeling enzyme mutated in CHARGE syndrome (Vissers et al., 2004), results in severe malformations of the lateral and posterior semicircular canals in both humans and mice (Adams et al., 2007; Bosman et al., 2005; Hurd et al., 2007). Adult Chd7^Gt/+ mutant mice also have defects in innervation to the adult posterior crista within the ear (Adams et al., 2007), and reduced neural stem cell proliferation in the olfactory epithelium (Layman et al., 2009). CHD7 belongs to a family of nine mammalian CHD proteins characterized by the presence of two chromodomains N-terminal to an SNF2 ATP-dependent helicase domain (Woodage et al., 1997). CHD proteins are proposed to repress or activate downstream target genes, via participation in chromatin remodeling complexes that have epigenetic functions in proliferation (Rodriguez-Paredes et al., 2009) and cell fate (Takada et al., 2007). Downstream targets have been identified for some CHD proteins in cultured cells (Nioi et al., 2005; Tai et al., 2003; Tong et al., 1998; Vertegaal et al., 2006), and recent research in Xenopus has shown that CHD7 associates with the PBAF (polybromo- and BRG1-associated factor-containing) complex to promote neural crest gene expression and cell migration (Bajpai et al., 2010). In vitro chromatin immunoprecipitation followed by microarray analysis (ChIP-chip) has also shown that CHD7 binds to sites of methylated histones (H3K4me) that are associated with sites of active transcription (Schnetz et al., 2009). CHD7 is thought to directly enhance transcription by chromodomain binding to H3K4me, which regulates access to DNA for other transcriptional activators via the ATP-dependent helicase domain (Schnetz et al., 2009). CHD7 is predicted to bind to thousands of sites throughout the mammalian genome, including transcription start sites and enhancer regions (Schnetz et al., 2009). Comparisons of CHD7 targets between different cell lines, along with the variable penetrance of CHARGE phenotypes, suggest that CHD7-binding specificity and target gene expression are likely to be tissue and developmental stage specific. In this study, we report the generation of the first Chd7^flox allele and its use in investigating the effect of otocyst-specific CHD7 deletion.

A BAC clone containing Chd7 genomic sequences (RP23-121H12) was obtained from Children's Hospital Oakland Research Institute (Oakland, CA). Regions of homology were engineered as shown in Fig. 1. The linearized targeting vector was electroporated into W4 (129S6/SvEvTac) ES cells and selected for by neomycin resistance. Positive clones were identified by PCR and confirmed using Southern blot analysis. Targeted clones were injected into C57BL/6 blastocysts and founder males (Chd7^{floxfrtneokan/+}) bred with C57BL/6 females. Resulting F1 progeny were genotyped as shown in Fig. 1. Chd7^{floxfrtneokan/+} female mice were bred with FLPeR male mice (Jackson Laboratory #003496) to remove the neo/kan cassette to generate Chd7^+/flox mice.

Chd7^Gt/+ mice were generated as previously described (Hurd et al., 2007). Mice were maintained by backcrossing with 129S1/Sv1mJ (Jackson Laboratory) mice to generation N5-N7 or B6D₂F₁/J (Jackson Laboratory #100006) mice to N2-N3. Chd7^flox/+ mice were mated with Foxg1-Cre or EIIa-Cre mice. Foxg1-Cre mice were maintained on the Swiss Webster (Charles River Laboratories #F44281) background and genotyped for Cre as described (Hebert and McConnell, 2000).

Postnatal day 0 (P0) pups were euthanized and temporal bones dissected. The oval and round windows and apex of the cochlea were perforated, fixed in 2% paraformaldehyde for 1 hour at room temperature, and washed in PBS followed by X-gal wash buffer [sodium phosphate buffer (pH 7.4) with 2 mM MgCl₂ and 0.02% NP-40 (Sigma, St Louis, MO)]. Ears were stained in X-gal wash buffer containing 0.02 mg/ml X-gal (Roche, Indianapolis, IN), 5 mM potassium ferrocyanide (Fisher, Pittsburgh, PA), 5 mM potassium ferricyanide (Fisher), and 0.33% N-N-dimethylformamide (Sigma), post-fixed in 4% paraformaldehyde and cleared using a glycerol gradient.

Timed pregnancies were established and the morning of plug identification designated as E0.5. Embryos were harvested and amniotic sacs collected for PCR genotyping as described (Hurd et al., 2007). Embryos were fixed in 4% paraformaldehyde for 30 minutes (E9.5-E10.5) to 1.5 hours (E11.5-E12.5), then washed in PBS and dehydrated in ethanol. Embryos were embedded in paraffin and sectioned at 7 μm or incubated in 30% sucrose for frozen sectioning at 12 μm as described previously (Martin et al., 2004). For each antibody, multiple sections from at least four ears of each genotype were tested. The following antibodies were incubated at 4°C overnight: polyclonal rabbit anti-CHD7 antibody (1:1000, Abcam, Cambridge, MA); mouse anti-β-tubulin III (1:1500, TuJ1; Covance, Richmond, CA); rabbit anti-NeuroD (1:600, gift of Jacques Drouin); mouse anti-Islet1 (1:500, Developmental Studies Hybridoma Bank, Iowa City, IA); and rabbit anti-TBX1 (1:1000, gift of Bernice Morrow). Direct secondary antibodies conjugated to Alexa488 and Alexa555 were purchased from Invitrogen. Alternatively, secondary detection was performed using biotinylated antibodies (Vector Laboratories, Burlingame, CA) followed by HRP-streptavidin Tyramide signal amplification (TSA) (Invitrogen).

Timed pregnancies were established and embryos collected at E14.5, after cervical dislocation and hysterectomy. Embryos were washed briefly in PBS, fixed in Bodian's fixative and cleared in methyl salicylate. Heads were bisected, brains removed and ears were injected with 3% WhiteOut in methylsalicylate, as described previously (Hurd et al., 2007).

In situ hybridization was performed on paraffin embedded embryo sections using digoxigenin-labeled riboprobes as previously described (Martin et al., 2002) at annealing temperatures between 55°C and 65°C. For each probe, multiple sections from at least four ears of each genotype were tested. RNA probes were provided by Doris Wu (Otx2 and Lfng), Deneen Wellik (Eya1) and Brigid Hogan (Fgf10).

Wild-type and Chd7^Gt/+ E9.5-E10.5 embryos were processed as above for immunofluorescence. For cell proliferation assays, serial adjacent sections were labeled with rabbit anti-phosphohistone H3 (1:200, Millipore, Billerica, MA) followed by incubation with anti-rabbit AlexaFluor488-conjugated secondary antibody (Invitrogen) and co-stained with DAPI (1:50, Invitrogen). Cell death assays were performed on serial adjacent sections using Fluorescein-FragEL DNA fragmentation detection kit (Calbiochem, San Diego, CA) according to the manufacturer's instructions.

Cell counts were performed on at least four embryos (eight ears) of each genotype at each time point. One-way ANOVA with Tukey's post-hoc analysis was performed when comparing groups of three genotypes or more. Student's t-tests with Welch's correction were performed to determine statistical significance of H3-positive cells in the neurogenic region. Standard error of the mean (s.e.m.) was calculated for each value.

We have previously reported the generation of a Chd7 genetrap (Chd7^Gt) allele in which Rosafary genetrap sequences are inserted in non-coding sequences upstream of the ATG-containing exon 2 (Hurd et al., 2007). Homozygous Chd7-null mice (Chd7^Gt/Gt) are lethal after E10.5, prohibiting examination of Chd7-null inner ears beyond the otocyst stage (Hurd et al., 2007). Thus, we generated a conditional Chd7^{floxfrtneokan} allele for use in tissue-specific deletions (Fig. 1A). We confirmed the generation of a conditional Chd7-null allele by crossing Chd7^+/flox mice with a ubiquitous Cre-expressing recombinase, EIIA-Cre, generating Chd7^+/− mice (Fig. 1). Chd7^+/− mice displayed circling and head-bobbing behaviors, and lateral semicircular canal abnormalities similar to those observed in Chd7^Gt/+ mice (data not shown). CHD7 immunofluorescence was present in E10.5 Chd7^+/− embryos and absent in Chd7^−/− embryos (Fig. 1D). Thus, Chd7^+/− mice have similar phenotypes to Chd7^Gt/+ mice, and the Chd7^flox allele can be used for conditional Chd7 mutagenesis.

To generate mice with inner ear-specific Chd7 deficiency, we used Foxg1-Cre mice that express Cre recombinase in the otic vesicle by E8.5 (Hebert and McConnell, 2000). We mated Chd7^Gt/+ mice, which express β-galactosidase, with Foxg1-Cre mice to generate Foxg1-Cre;Chd7^Gt/+ males. Foxg1-Cre;Chd7^Gt/+ males were then crossed with Chd7^flox/flox females for embryo collection. This breeding scheme required Cre recombination at only one Chd7 allele and enabled use of β-galactosidase as a reporter of Chd7-expressing cells (Fig. 2A). This mating also allowed us to analyze Foxg1-Cre;Chd7^+/flox (conditional heterozygous) phenotypes as important controls for possible Foxg1 heterozygous effects (Hebert and McConnell, 2000).

Foxg1-Cre;Chd7^Gt/flox (conditional null) mice displayed medially displaced eyes and shortened snouts compared with control littermates (see Fig. S1 in the supplementary material). Conditional null mice survived only until postnatal day 1, with no obvious milk in the stomach, suggesting early postnatal mortality due to inability to feed. We performed CHD7 immunofluorescence on E10.5 embryos (Fig. 2B-D) to confirm Cre recombination. CHD7 immunofluorescence was highest in the anterior ventral region of the E10.5 otocyst in Chd7^+/flox embryos, reduced in conditional heterozygotes, and absent in conditional null embryos (Fig. 2D).

Conditional heterozygous and Chd7^Gt/flox inner ears had defects similar to those previously described in Chd7^Gt/+ ears, including truncated or hypoplastic lateral semicircular canals with preserved anterior semicircular canals, utricles, sacculae and cochleae (Fig. 2E,F) (Hurd et al., 2007). Chd7^Gt/flox inner ears expressed β-galactosidase throughout all six sensory epithelia [maculae of the utricle and saccule; crista ampullaris of the lateral, anterior and posterior canal; and the cochlear epithelium (Fig. 2H)]. By contrast, conditional null inner ears displayed severe structural abnormalities in both vestibular and cochlear structures (Fig. 2G). The conditional null endolymphatic duct and anterior semicircular canal were hypoplastic, whereas the posterior and lateral canals and vestibular sensory organs (cristae, saccule and utricle) were missing (n=6) (Fig. 2G). The conditional null cochlea was also severely hypoplastic and undercoiled, with abnormal twisting at the apex. β-Galactosidase staining was observed in the conditional null cochlea and in a dorsal stripe in the vestibule behind the endolymphatic duct (Fig. 2J). The severity of the inner ear phenotype in conditional null inner ears compared with the mild phenotype observed in conditional heterozygous ears also suggests that these defects are not due to double heterozygous mutant effects.

To better understand the defects observed by loss of Chd7 function within the inner ear, we closely examined CHD7 protein between E9.5-E12.5. CHD7 was widespread in the mesenchyme, otic epithelium and developing vestibulo-cochlear ganglion at E9.5 (see Fig. S2A in the supplementary material). By E10.5, during axial specification, CHD7 was restricted to the anterior and ventral regions of the otocyst (corresponding to the neurogenic domain) and in the developing statoacoustic ganglion (Fig. 2B). At E11.5, CHD7 was highest in the ventral region of the otocyst, which gives rise to the saccule, utricle, cochlea and vestibulo-cochlear ganglion (see Fig. S3B-C in the supplementary material). By E12.5, CHD7 was observed in the vestibulo-cochlear ganglion and cochlea, and at lower levels in the lateral canal primordium (see Fig. S3D-F in the supplementary material). Low levels of CHD7 were also present in the periotic mesenchyme between E10.5 and E12.5 (Fig. 2B and see Fig. S3 in the supplementary material). CHD7 was absent from the endolymphatic duct, consistent with mild ductal hypoplasia in Chd7 conditional null mice (Fig. 2G). Thus, Chd7 expression was widespread at the otocyst stage with progressive restriction to the vestibulo-cochlear ganglion and pro-sensory tissues during later inner ear development.

Initial patterning of the otocyst dictates later morphogenetic events, including formation of the vestibulo-cochlear ganglion and semicircular canals (Bok et al., 2007). This morphogenetic patterning is established by dynamic gene expression changes that generate unique domains for prospective inner ear structures. We tested whether reduced CHD7 impacts expression of genes implicated in E9.5-E10.5 otocyst patterning. Otx2, which regulates cochlear patterning and ganglion development (Miyazaki et al., 2006; Morsli et al., 1999), was retained in E10.5 Chd7^Gt/+ and conditional heterozygotes, but was absent from conditional null and Chd7^Gt/Gt otocysts (Fig. 3A-E).

Lfng expression, which defines a ventral region within the otocyst that gives rise to the saccule and utricle and the organ of Corti, was present in all E9.5-E10.5 Chd7 mutants, and slightly reduced in the Chd7^Gt/Gt otocyst (Fig. 3F-J and see Fig. S2B-F in the supplementary material). We also detected normal expression of Eya1 in E10.5 Chd7 heterozygotes and conditional null otocysts (Fig. 3K,L,N,O); however, Eya1 expression was absent from Chd7^Gt/Gt otocysts (Fig. 3M). The fibroblast growth factor, Fgf10, which regulates Ngn1 expression in the otocyst (Alsina et al., 2004), was reduced in Chd7^Gt/+ and conditional heterozygous embryos and was completely absent from Chd7^Gt/Gt and conditional null otocysts (Fig. 3P-T). These data suggest that CHD7 is involved in controlling early neurogenic events in the developing ear.

To address whether CHD7 regulates initiation or maintenance of the neurogenic domain, we analyzed expression of the pro-neural marker Ngn1 in wild-type and Chd7 mutant tissues. We observed no difference in ventral expression of Ngn1 between Chd7 heterozygous mutant and wild-type embryos at E9.5 or E10.5 (Fig. 3Z,AA,EE,FF). However, Ngn1 expression was absent in E9.5-E10.5 Chd7^Gt/Gt embryos (Fig. 3BB,GG) and significantly reduced in E9.5-E10.5 Chd7 conditional null embryos (Fig. 3DD,JJ). Notably, Ngn1 expression was present in younger (E9.0) Chd7^Gt/Gt otocysts (Fig. 3HH). These observations suggest that CHD7 is required for maintenance, but not necessarily for initiation, of Ngn1 expression.

Because Ngn1 expression is inhibited by TBX1 (Raft et al., 2004), we also examined Chd7 mutants for TBX1 immunofluorescence. Within the dorsal otocyst, TBX1 was maintained in E10.5 Chd7^Gt/+ embryos but was expanded ventrally in Chd7^Gt/Gt and conditional null embryos (Fig. 3U-Y). These results suggest that Chd7 is important for early otic patterning by positively or negatively regulating expression of genes that promote (Ngn1, Otx2, Fgf10) or inhibit (Tbx1) development of the otic neurogenic domain, in a dose-dependent manner.

Vestibulo-cochlear ganglion neurons are formed in the mouse between E9.5 and E14.5, by delamination of ventral otic epithelial neural progenitors into the surrounding mesenchyme (Fritzsch et al., 1999). Delaminated ganglionic neuroblasts express Ngn1, Islet1, neuron-specific β-Tubulin III, and NeuroD, and continue to proliferate prior to aggregation and subsequent formation of separate vestibular and cochlear ganglia (Sanchez-Calderon et al., 2007). As Chd7 is highly expressed in the anterior ventral region of the otocyst and vestibulo-cochlear ganglion, we hypothesized that CHD7 may colocalize with markers of pro-neural genes and have a role in neurogenesis. We observed CHD7 in β-Tubulin III-positive neurons throughout the E10.5 vestibulo-cochlear ganglion (Fig. 4A-C). CHD7 also colocalized with NeuroD within the ganglion (Fig. 4D-F), although there a small number of CHD7-positive cells were NeuroD negative (arrow in Fig. 4F). By contrast, most Islet1-positive cells were CHD7 positive, both within the ganglion and also in the ventral otocyst (Fig. 4G-I). The majority of proliferating cells within the ganglion, otocyst and surrounding mesenchyme were also CHD7 positive, based on colocalization of CHD7 with anti-phosphohistone H3 (Fig. 4J-L).

Direct measurement of sectioned tissues showed a 1.5-fold reduction in the size of the E10.5-E12.5 vestibulo-cochlear ganglion between Chd7^Gt/+ and wild-type embryos (data not shown). β-Tubulin III and Islet1 immunofluorescence were also reduced in Chd7^Gt/+ and conditional null ganglia compared with wild type (Fig. 5). In Chd7^Gt/Gt ganglia, only a small number of cells were β-Tubulin III or Islet1 positive, consistent with severe ganglionic hypoplasia. Quantitation of Islet1-positive cells in the E10.5 vestibulo-cochlear ganglion revealed significantly fewer Islet1-positive cells in Chd7^Gt/+ (mean=343; s.e.m.=72; n=8) and conditional heterozygous (mean=426; s.e.m.=54; n=8) mice compared with wild type (mean=841; s.e.m.=64; n=8) (Fig. 5I and Table 1). The reduction in Islet1-positive cells was even more severe in Chd7^Gt/Gt (mean=14; s.e.m.=4; n=8) and conditional null (mean=130; s.e.m.=18; n=8) embryos (Fig. 5I and Table 1). Reduced Chd7 mutant ganglia and Islet1-positive neuroblasts together suggest that CHD7 is a positive regulator of inner ear neurogenesis, perhaps via effects on pro-neural gene transcription.

To address whether reduced expression of pro-neural genes in Chd7 mutant inner ears reflects a reduction in delaminating neuroblasts, we quantitated the number of NeuroD-positive cells in the E9.5-E11.5 ganglion and otic epithelium (Fig. 6 and Table 2). NeuroD-positive cells in the Chd7^Gt/+ ventral otic epithelium were significantly reduced at E10.5 (mean=65; s.e.m.=6; n=8) compared with wild-type littermates (mean=91; s.e.m.=7; n=8), but were unchanged at E9.5 and E11.5 (Fig. 6J and Table 2). There were also significantly fewer NeuroD-positive cells in the Chd7^Gt/+ ganglion at E9.5 (mean=87; s.e.m.=11; n=8) and E10.5 (mean=509; s.e.m.=72; n=8) compared with wild type (E9.5: mean=245; s.e.m.=48; n=8; E10.5: mean=1010; s.e.m.=82; n=8), with no differences observed at E11.5 (see Table 2). Interestingly, NeuroD-positive cells did not accumulate in the Chd7^Gt/+ epithelium between E9.5-E11.5, suggesting normal delamination from the otocyst. These data suggest a defect in primary neurogenesis in the E10.5 Chd7^Gt/+ otic epithelium and ganglion, and are consistent with reduced Islet1-positive and TUJ1-positive neuroblasts (Fig. 5). Together, these observations indicate a defect in ganglionic neuronal production or specification. By E11.5, Chd7^Gt/+ embryos have normal numbers of neuroblasts in the epithelium and ganglion, suggesting that the requirement for CHD7 is temporally restricted to earlier timepoints.

In contrast to Chd7^Gt/+, analysis of Chd7^Gt/Gt embryos showed severe reductions in NeuroD-positive neuroblasts, in both the epithelium and the ganglion at E9.5 and E10.5 (see Fig. 6 and data in Table 2). At E11.5, conditional null embryos also had severe reductions in NeuroD-positive cells in both the epithelium (mean=21; s.e.m.=4; n=8 versus wild type: mean=62; s.e.m.=10; n=8) and ganglion (mean=346; s.e.m.=35; n=8 versus wild type: mean=1082; s.e.m.=66; n=8) (Fig. 6, Table 2). Like Chd7^Gt/+, conditional heterozygotes also had unchanged NeuroD-positive cells in both the epithelium and ganglion at E11.5 (see Fig. 6, Table 2). Thus, homozygous loss of Chd7 function, whether in the germline or by Foxg1-Cre, results in a more severe defect in inner ear neurogenesis than in Chd7 heterozygotes and may be related to disruption of the neurogenic domain (Fig. 3BB,GG,DD,JJ).

Prior studies showed that Chd7 haploinsufficiency is associated with reduced proliferation of olfactory neural stem cells (Layman et al., 2009). As CHD7 is in proliferating cells of the developing otocyst and stato-acoustic ganglion (Fig. 4J-L), we examined cellular proliferation in E9.5-E10.5 Chd7 mutant ears using anti-phosphohistone-H3 immunofluorescence (Fig. 7 and Table 3). We observed no change in the number of proliferating cells in the epithelium of Chd7^Gt/+ embryos at either E9.5 or E10.5 compared with wild type (Fig. 7 and Table 3). In the Chd7^Gt/+ ganglion, cellular proliferation was reduced at E9.5 (mean=7; s.e.m.=2; n=8 versus wild type: mean=16; s.e.m.=3; n=8) but was unchanged at E10.5 (mean=21; s.e.m.=3; n=8 versus wild type: mean=24; s.e.m.=3; n=8) (Fig. 7 and Table 3). Likewise, proliferating cells in the E10.5 conditional heterozygous otic epithelium were reduced (mean=53; s.e.m.=4; n=8) compared with Chd7^+/flox controls (mean=66; s.e.m.=4; n=8), whereas proliferating cells in the ganglion were unchanged (see Fig. 7 and Table 3). We also quantitated mitotic cells within the neurogenic domain, after staining adjacent sections with Ngn1 and anti-phosphohistone H3 (Fig. 7C-G and see Table S1 in the supplementary material). There were significantly fewer proliferating cells within the neurogenic domain of Chd7^Gt/+ ears at E9.5 (mean=23; s.e.m.=3; n=8 versus wild type: mean=38; s.e.m.=4; n=8) and E10.5 (mean=33; s.e.m.=3; n=8 versus wild type: mean=59; s.e.m.=5; n=8). Thus, despite normal expression of Ngn1 in the Chd7^Gt/+ neurogenic domain (Fig. 3AA,FF), it contains fewer dividing neuroblasts. This is consistent with fewer migrating neuroblasts (Fig. 5B,F and Fig. 6B,E,H) and a smaller vestibulo-cochlear ganglion.

To test for dose effects, we also examined cellular proliferation in Chd7^Gt/Gt otocysts. Chd7^Gt/Gt embryos had severe reductions in cellular proliferation at E9.5 and E10.5 in both the epithelium and the ganglion (see Fig. 7A,B and Table 3). Analysis of E10.5 conditional null embryos also showed reduced proliferation in both the epithelium (conditional null: mean=34; s.e.m.=2; n=8 versus Chd7^+/flox control: mean=66; s.e.m.=4; n=8) and ganglion (conditional null: mean=16; s.e.m.=2; n=8 versus Chd7^+/flox control: mean=36; s.e.m.=2; n=8) (Fig. 7A,B and Table 3). Reduced proliferation in the otic epithelium and ganglion of Chd7^Gt/Gt and Chd7 conditional null embryos is consistent with significantly reduced or absent expression of neurogenic markers (Fig. 3C,E,R,T,BB,DD), suggesting that Chd7 has multiple roles that include otic patterning and regulation of proliferation and/or differentiation.

Changes in cell proliferation are often accompanied by alterations in cell death. To test this, we quantified numbers of apoptotic cells in E9.5-E10.5 Chd7 mutant embryos. We observed no changes in the number of apoptotic cells in E9.5-E10.5 ganglion or otic epithelium of Chd7^Gt/+ embryos compared wth wild type (see Fig. S4 and Table S2 in the supplementary material). By contrast, there were significantly fewer apoptotic cells within the Chd7^Gt/Gt ganglion compared with wild type at E9.5 (Chd7^Gt/Gt: mean=8; s.e.m.=3, n=8 versus wild type: mean=57; s.e.m.=10, n=8) and E10.5 (Chd7^Gt/Gt: mean=11; s.e.m.=4, n=8 versus wild type: mean=41; s.e.m.=8, n=8) (see Fig. S4 and Table S2 in the supplementary material). Thus, CHD7 appears to promote early neuroblast proliferation without concomitant increases in cell death.

Here, we describe the first Chd7^flox allele and its use in inner ear specific loss-of-function studies. We show that Chd7 conditional null embryos have severe inner ear malformations that affect both vestibular and auditory structures. Complete absence of CHD7 results in decreased expression of patterning and pro-neural (Otx2, Fgf10, Ngn1, NeuroD and Islet1) genes within the otic epithelium and ganglion (Fig. 8), and reduced proliferation of developing neuroblasts. Our observations support the hypothesis that CHD7 function is essential for proper inner ear morphogenesis and neural development.

A number of genes involved in inner ear neurogenesis, including Fgf10, Ngn1, NeuroD and Isl1 are downregulated by Chd7 deficiency, suggesting that CHD7 is required to maintain transcription of these genes. NGN1 is crucial for inner ear neurogenesis, as Ngn1^−/− mice do not display any inner ear sensory neurons (Ma et al., 2000; Ma et al., 1998). NGN1 activates the basic helix-loop-helix protein NeuroD, which is important for neuronal differentiation and survival (Kim et al., 2001). The Ngn1-expressing region of the otocyst gives rise to the maculae of the utricle and saccule (Raft et al., 2007). Ngn1 expression is absent from E9.5-E10.5 Chd7^Gt/Gt otocysts but is present at low levels in the E9.0 Chd7^Gt/Gt otocyst and is significantly reduced in E9.5-E10.5 conditional null embryos. This suggests that CHD7 acts upstream of NGN1 within the neurogenesis signaling pathway to maintain Ngn1 expression, even though it is not required for Ngn1 induction. Preserved Ngn1 expression in Chd7 heterozygous mutant mice (Fig. 3AA) is also consistent with the normal ultrastructure of sensory maculae of the utricle and saccule in these mice (Adams et al., 2007), but does not explain the semicircular canal and other morphogenetic defects.

Fgf10 expression, which is absent from Chd7-null mutant embryos, is required for otic Ngn1 expression (Alsina et al., 2004). It is conceivable that absence of Fgf10 expression from Chd7 conditional null and Chd7^Gt/Gt otocysts could lead to decreased expression of Ngn1. Interestingly, Fgf10-null mice have similar inner ear phenotypes to those in Chd7 heterozygotes, including defects in innervation to the posterior crista (Adams et al., 2007; Pauley et al., 2003). However, Chd7-conditional null ear phenotypes are much more severe, supporting the idea that CHD7 functions upstream of Fgf10. Fgf10 has been shown to enhance the Foxg1-null inner ear phenotype (Pauley et al., 2006); however, it is unlikely that the Chd7 conditional null inner ear defects reported here can be attributed to the presence of Foxg1-Cre. Foxg1^+/− mice are reported to have no inner ear defects, and neurosensory domain specification is unaltered in Foxg1^−/− mice (Hwang et al., 2009; Pauley et al., 2006).

Interestingly, OTX2 is implicated in mouse cochlear morphogenesis (Morsli et al., 1999). Otx2 expression is absent from Chd7^Gt/Gt and conditional null otocysts, which may contribute to the cochlear phenotype observed in Chd7 conditional null embryos. Furthermore, Otx2 is an important early otic patterning gene and is required for establishment of the neurogenic domain by repressing Fgf10 expression (Miyazaki et al., 2006). As both Otx2 and Fgf10 expression are absent from Chd7 null otocysts, CHD7 may regulate Fgf10 in an OTX2-dependent manner.

EYA1 is required for formation of the cochlea and vestibulo-cochlear ganglion, and for development of the sensory epithelium (Xu et al., 1999; Zou et al., 2008). Eya1 expression is absent in Chd7^Gt/Gt embryos, but maintained in conditional null inner ears. Hypomorphic Eya1 (Eya1^bor/bor) and Chd7^Gt/+ embryos show similarly reduced vestibulo-cochlear ganglion size and reductions of NeuroD and Islet1 in delaminating neuroblasts (Friedman et al., 2005). Analysis of Eya1-null and hypomorphic alleles showed that EYA1 promotes expression of Fgf10 and Otx2 in a dose-dependent manner (Zou et al., 2008). Otx2 expression is only maintained in the ventral otocyst if Eya1 expression is greater than 25% of normal values (Zou et al., 2008). In heterozygous Chd7 mutant ears, we observed no changes in expression of Otx2 or Fgf10, whereas Otx2 and Fgf10 expression was absent in Chd7 homozygous null and conditional null ears. Eya1^−/− and Chd7^Gt/Gt mutants also have expansion of Tbx1 expression into the ventral region of the otocyst (Fig. 3W) (Friedman et al., 2005). As TBX1 is a suppressor of neural cell fate within the dorsal otocyst (Raft et al., 2004), expansion of Tbx1 expression into the neurogenic domain could account for the reduced Ngn1 expression observed in both Eya1 and Chd7 mutant mice. It is therefore tempting to attribute the decreased expression of neural-fate markers observed in Chd7^Gt/Gt otocysts to absent Eya1 expression. However, in Chd7 conditional null otocysts, Eya1 expression is retained, yet Otx2 and Fgf10 expression are absent and Tbx1 expression is expanded. This suggests that absence of Eya1 expression within Chd7^Gt/Gt embryos is either influenced by tissues other than the otocyst, or is a consequence of the hypoplastic nature of the Chd7^Gt/Gt embryo. It also implies that changes in expression of Otx2, Fgf10 and Tbx1 in Chd7 mutant embryos are not regulated through EYA1, but via a parallel molecular genetic pathway (Fig. 8). In addition, Chd7^+/−;Tbx1^+/− double mutants display a more severe inner ear phenotype than do either single heterozygous mutation (Randall et al., 2009), suggesting that CHD7 and TBX1 could regulate the expression of one another. Interestingly, mutations in human TBX1 contribute to DiGeorge syndrome, a multiple anomaly condition that has phenotypic overlap with CHARGE Syndrome (de Lonlay-Debeney et al., 1997).

Lfng expression also correlates with neural fate markers, and defines a ventral region within the otocyst that ultimately gives rise to the saccule and utricle (Morsli et al., 1998). Lfng expression is absent from Eya1^−/− otocysts and is expanded in Tbx1^−/− nulls, consistent with changes in neural fate markers in the otocysts of these mutant mice (Raft et al., 2004; Zou et al., 2008). By contrast, Lfng expression is preserved in Chd7-null mutants, whereas pro-neural genes are decreased, suggesting that absence of CHD7 specifically affects neuroblast production in the neurogenic domain without completing disrupting patterning of the ventral otocyst. Our data support the hypothesis that CHD7 acts upstream of FGF10 and OTX2, and in parallel with EYA1, to promote inner ear neurogenesis in a dose-dependent manner via inhibition of Tbx1 and/or activation of the Ngn1/NeuroD/Isl1 transcriptional regulatory cascade (Fig. 8).

The observed effects of CHD7 deficiency on otocyst gene expression and neuroblast development appear to be highly sensitive to Chd7 gene dose. Loss of Chd7 in the germline results in extreme hypoplasia of the embryo and the developing vestibulo-cochlear ganglion, whereas heterozygous loss leads to milder phenotypes. Conditional Chd7 mutants have intermediate phenotypes, with reduced neuroblast proliferation and pro-neural gene expression. The reasons for these intermediate phenotypes could be related to the timing of onset or tissue specificity of Cre expression. There is Cre activity in the developing Foxg1-Cre hindbrain (Hebert and McConnell, 2000), and we observed reduced CHD7 in the hindbrain of Chd7 conditional mutant embryos (Fig. 2C,D). Thus, aberrant signaling from CHD7-dependent pathways in the rhombencephalon could also contribute to the intermediate phenotypes seen in Foxg1-Cre;Chd7 conditional mutant embryos.

Recent observations showed that CHD7 promotes olfactory neural stem cell proliferation (Layman et al., 2009). Here, we demonstrate that heterozygous loss of Chd7 results in reduced proliferation within the neurogenic domain of the developing otocyst and fewer ganglionic neuroblasts available to proliferate and occupy the developing vestibulo-cochlear ganglion. Conditional Chd7-null embryos also have defects in ganglionic and epithelial neuroblast proliferation, in association with a more severe inner ear phenotype. Although CHD7 does not seem to directly regulate apoptosis, the number of apoptotic cells was increased (relative to wild type) in E10.5 mutant ears when normalized to account for the reduced number of NeuroD+ neuroblasts in each region (see Table 2; see Fig. S4 and Table S2 in the supplementary material). Thus, primary reduction of H3+ cells in the neurogenic domain results in fewer neuroblasts available to occupy a smaller mutant ganglion, and this smaller mutant ganglion exhibits relatively increased cell death.

Other members of the CHD protein family have been implicated in proliferation and/or apoptosis (Bagchi et al., 2007; Gaspar-Maia et al., 2009; Nishiyama et al., 2009; Rodriguez-Paredes et al., 2009). In vitro inhibition of Chd1 creates abnormally high levels of neural differentiation, and is required for maintenance of pluripotency (Gaspar-Maia et al., 2009). CHD8, which has close homology to CHD7, positively controls genes expressed in the G1/S transition phase, and is required for normal proliferation (Rodriguez-Paredes et al., 2009). In addition, both CHD5 and CHD8 have been implicated as tumor suppressor genes (Bagchi et al., 2007; Nishiyama et al., 2009). Thus, the CHD family of proteins may function as tissue-specific regulators of cell cycle progression. Interestingly, EYA1 is involved in cell cycle maintenance during inner ear neurogenesis. Eya1-null embryos display increased apoptosis within the vestibulo-cochlear ganglion and epithelium of the developing ear (Zou et al., 2004; Zou et al., 2006), supporting the hypothesis that CHD7 and EYA1 could act in parallel pathways by complementary regulation of cell cycle progression.

In Drosophila, the CHD7 ortholog kismet regulates transcriptional elongation via occupancy at sites of active transcription and RNA polymerase II binding (Srinivasan et al., 2005; Srinivasan et al., 2008). Kismet is required for transcription of the pro-neural factor atonal (Atoh1) and regulation of retinal photoreceptor cell development (Melicharek et al., 2008). CHD family members also play important roles in proliferation and/or apoptosis, raising the possibility that chromatin regulation of mitosis and programmed cell death may be widespread during development. The global hypoplasia observed in Chd7^Gt/Gt embryos, combined with in vitro evidence that CHD7 binds a large number of target genes (Schnetz et al., 2009), also suggests that CHD7 may positively regulate proliferation in many tissues. Thus, germline loss of Chd7 may not accurately reflect the tissue-specific roles of CHD7 in gene regulation, highlighting the importance of examining conditional loss of Chd7 in vivo.

Using our newly generated Chd7^flox allele, we showed that CHD7 deficiency activates or represses expression of a number of genes essential for inner ear neurogenesis in a dose-dependent manner. It is not known whether CHD7 acts directly or indirectly, alone or in combination with other co-factors to alter transcription of inner ear neural-fate genes. ChIP-chip data in pluripotent cell lines suggest that CHD7 deficiency leads to many changes in gene expression of small effect sizes (Schnetz et al., 2009). By contrast, our results show that CHD7 can exert large effects on transcription of a small number of crucial genes during development. The tissue-specific phenotypes that result from CHD7 deficiency could be related to changes in these crucial developmental regulator genes. Future studies will help determine whether CHD7 functions differently in multipotent versus differentiated cells, and whether its ability to promote proliferation and neurogenesis involves distinct molecular mechanisms or co-factors.

In summary, our data provide a basis for understanding how chromatin regulatory proteins influence the proliferative potential of inner ear progenitors during development. Still, unresolved issues remain. It is not known whether overexpression of Ngn1 or other pro-neural genes rescues Chd7 mutant phenotypes, or whether CHD7 exerts its effects via changes in transcriptional initiation or elongation. The composition of CHD7-containing complexes that mediate CHD7 effects in neural tissues is also not known, and it will be interesting to see whether this complex changes as progenitors switch to the postmitotic state. There is evidence for chromatin protein subunit switching during neurogenesis (Lessard et al., 2007; Yoo and Crabtree, 2009; Yoo et al., 2009), and further studies will determine whether similar or related mechanisms occur in developing inner ear neuroblasts.

We thank the University of Michigan Transgenic Animal Model Core for help with generation of the Chd7^flox allele. Doris Wu, Andy Groves and Dan Bochar provided critical comments on the manuscript. Elyse Reamer provided technical assistance. This work was supported by the A. Alfred Taubman Medical Research Institute (Y.R.), the Berte and Alan Hirschfield Foundation (Y.R.), NOHR (D.M.M.), and NIH/NIDCD grants P30 DC05188 (Y.R.) and DC009410 (D.M.M.). Deposited in PMC for release after 12 months.