CHD7 and SOX2 act in a common gene regulatory network during mammalian semicircular canal and cochlear development

Significance

Little is known about the early molecular events that give rise to vestibular and auditory inner ear structures. We show that CHD7, a chromatin remodeler, and SOX2, a pioneer transcription factor, cooperate to help shape the inner ear. CHD7 regulates Sox2 expression, and both genes bind near transcription start sites of key otic development and patterning genes, including Pax2 and Otx2. We show that inner ear morphogenesis depends on proper dosage of both Chd7 and Sox2. Our findings uncover a gene regulatory network that helps define the early epigenetic landscape of otic progenitor cells. These findings are relevant for understanding and treating hearing and balance disorders and shed light on potential interactions between these genes in other organs.

Abstract

Inner ear morphogenesis requires tightly regulated epigenetic and transcriptional control of gene expression. CHD7, an ATP-dependent chromodomain helicase DNA-binding protein, and SOX2, an SRY-related HMG box pioneer transcription factor, are known to contribute to vestibular and auditory system development, but their genetic interactions in the ear have not been explored. Here, we analyzed inner ear development and the transcriptional regulatory landscapes in mice with variable dosages of Chd7 and/or Sox2. We show that combined haploinsufficiency for Chd7 and Sox2 results in reduced otic cell proliferation, severe malformations of semicircular canals, and shortened cochleae with ectopic hair cells. Examination of mice with conditional, inducible Chd7 loss by Sox2^CreER reveals a critical period (~E9.5) of susceptibility in the inner ear to combined Chd7 and Sox2 loss. Data from genome-wide RNA-sequencing and CUT&Tag studies in the otocyst show that CHD7 regulates Sox2 expression and acts early in a gene regulatory network to control expression of key otic patterning genes, including Pax2 and Otx2. CHD7 and SOX2 directly bind independently and cooperatively at transcription start sites and enhancers to regulate otic progenitor cell gene expression. Together, our findings reveal essential roles for Chd7 and Sox2 in early inner ear development and may be applicable for syndromic and other forms of hearing or balance disorders.

Genetic or environmental lesions that disrupt inner ear development cause auditory and/or vestibular dysfunction. Inner ear structures originate from the otic placode (op), a thickening of the surface ectoderm near the hindbrain, through a complex series of morphogenic events that convert the op to an otic cup, then a spherical otocyst, and ultimately into the mature inner ear. The dorsal inner ear vestibular organs include semicircular canals, utricle, and saccule, and their associated sensory structures are responsible for the perception of angular and linear movements. The organ of Corti, the cochlear sensory epithelium, is essential for sound perception. The organ of Corti is composed of one row of inner hair cells (IHCs) and three rows of outer HCs (OHCs) along with adjacent interdigitated supporting cells (SCs), including Deiters’ cells (DCs), whose nuclei reside basal to the OHC bodies (1). Proper development of the inner ear requires highly orchestrated action by Wnt, Shh, Bmp, Fgf, and other morphogens (2–4), transcription factors (5), and planar cell polarity (PCP) proteins (6).

CHD7, Chromodomain Helicase DNA-binding Protein 7, is an epigenetic regulator and chromatin remodeler. Chd7 is broadly expressed throughout the early developing otocyst and cochleovestibular ganglion (CVG) and is later localized to the vestibular and cochlear epithelia and associated spiral ganglion neurons (SGNs) (7). Haploinsufficiency of human CHD7 causes CHARGE syndrome, a multiple anomaly condition characterized by ocular coloboma, heart defects, atresia of the choanae, retarded growth and development, genital hypoplasia, and ear defects, including deafness and semicircular canal dysgenesis (8). Mice with heterozygous loss of Chd7 exhibit hypoplasia/dysplasia of posterior and lateral semicircular canals, abnormal innervation in the vestibular sensory epithelium (9), and reduced neuroblast proliferation (10). Chd7 heterozygous mutant mice exhibit mixed conductive and sensorineural hearing loss despite intact auditory sensory epithelia and associated innervation (11, 12). Conditional deletion of Chd7 using a pan-otic Cre leads to hypoplasia of semicircular canals and disorganized innervation of the organ of Corti (10, 13). Thus, Chd7 has pleiotropic roles in early development of auditory and vestibular structures.

SOX2 (HMG-box transcription factor sex-determining region Y-box 2) belongs to the group B family of SOX transcription factors and is one of the earliest genes to mark neurosensory progenitors in the otic epithelium. Sox2 is expressed in the lateral edge of the op as early as E8.5 in mice (14) and subsequently expands into the dorsal domain, marking presumptive anterior and lateral cristae and utricle, and into the ventral domain, which gives rise to the organ of Corti and saccule (15). Sox2 expression persists postnatally in differentiated SCs and SGNs (16). Mice with complete loss of Sox2 in the ear (Lcc/Lcc) fail to establish a prosensory domain, resulting in the absence of HCs and SCs (17). Sox2 also regulates the expansion of neurosensory progenitor cells and initiates differentiation of HCs by binding and activating Atoh1, a transcription factor essential for HC differentiation (18). Conditional deletion of Sox2 in the developing otocyst leads to neurogenesis defects and premature death of neurosensory progenitors (19). In addition, Sox2-expressing cells in the early otocyst give rise to large numbers of nonsensory structures throughout the inner ear (20), and early Sox2 loss impairs formation of nonsensory structures including semicircular canals and vestibular maculae (21).

CHD7 has been shown to bind to thousands of sites in the vertebrate genome and is enriched at enhancers and transcription start sites (TSSs) (22). In neural stem cells, SOX2 and CHD7 physically interact, bind to overlapping sites in the genome, and regulate a set of common target genes including Jag1, Gli3, and Mycn (23). SOX2 also collaborates with CHD7 to regulate the expression of regulator of cell cycle (Rgcc) and protein kinase C-θ (PKCθ) to activate oligodendrocyte precursor cells after spinal cord injury (24). Previous studies showed reduced Sox2 inner ear expression in Chd7 conditional knockout mice (25), suggesting that CHD7 and SOX2 act in a common genetic pathway and have overlapping or complementary functions during inner ear development.

Here, using mice with germline or conditional inducible deletions, we tested the hypothesis that a Chd7-Sox2 gene regulatory network functions to promote prosensory and sensory cell development in the ear. Our results show that CHD7 and SOX2 cooperate to regulate morphogenesis of the semicircular canals and the cochlea, via effects on cell proliferation and regulation of other developmentally critical genes. Genome-wide analysis shows that Chd7 acts upstream of Sox2 and regulates a broad set of target genes. CHD7 and SOX2 bind to unique and common regions of the genome, including near TSSs and/or enhancers of genes that are critical for inner ear formation and patterning. These studies provide a basis for understanding combined functions of Chd7 and Sox2 in the inner ear and help define the early otic transcriptional regulatory landscape.

Results

CHD7 and SOX2 Expression and Hierarchy during Inner Ear Development.

Chd7 and Sox2 are highly expressed in the developing mouse otocyst and play important roles in proneurosensory development (10, 17). Prior studies showed Chd7 expression at E8.5 in the neural tube (26) and in the E9.5 otocyst (27) and Sox2 expression in the proneural region of the E9.5 otocyst (15). To determine the precise timing of Chd7 and Sox2 coexpression in the developing ear, we performed double immunofluorescence on sectioned inner ears from E8.5-E11.5 wild-type (WT) embryos (Fig. 1). Chd7 and Sox2 were both expressed throughout the neural tube and op at E8.5 (Fig. 1 A–C). Over the next 2 d of embryonic development, CHD7 and SOX2 were enriched in the ventral otocyst and cochleovestibular and facial ganglia (E9.5-E10.5). Notably, Chd7 expression was broader than Sox2 in both otocyst and ganglia (Fig. 1 D–I). By E11.5, CHD7 and SOX2 were enriched in the ventral otocyst, and CHD7 was abundant in the CVG, whereas SOX2 was present in both the cochleovestibular and facial ganglion (Fig. 1 J–L). Given their complex expression patterns, we asked whether CHD7 and SOX2 participate in the same or distinct gene regulatory networks of the inner ear by assaying Chd7 expression in the absence of Sox2 and vice versa. To avoid the embryonic lethality of complete Sox2 deficiency (28), we induced Sox2 deletion in Sox2^CreER/flox embryos, in which one Sox2 allele contains a CreER fusion protein in place of the coding region, and the other allele has loxP sites flanking the coding region (29). Sox2 was deleted by maternal intraperitoneal injection of tamoxifen at E8.5, E9.5, and E10, followed by harvesting embryos at E10.5. SOX2 immunostaining of otocyst sections at E10.5 showed that Sox2 expression was severely diminished in Sox2^CreER/flox otocysts compared to Sox2^+/flox (SI Appendix, Fig. S1 A and B). CHD7 immunofluorescence in the otocyst was unchanged in Sox2^CreER/flox embryos compared to Sox2^+/floxlittermates (n = 5 ears; Fig. 1 M–O). By contrast, SOX2 immunofluorescence was reduced ~30% in Chd7^Gt/Gt otocysts compared to WT (n = 5 ears, P < 0.01; Fig. 1 P–R) (the Chd7^Gt allele expresses a gene trapped β-geo construct, resulting in a null allele) (27). These studies confirm that Chd7 is genetically upstream of Sox2 in otic development.

Fig. 1.

CHD7 colocalizes with and regulates Sox2 expression in the otocyst. (A–L) Transverse sections from E8.5-E11.5 WT embryos costained with anti-CHD7 (red) and anti-SOX2 (green). Dorsal is up and lateral is right for all sections. At E8.5, CHD7 and SOX2 broadly colocalize in the neural tube and op (A–C). At E9.5, CHD7 and SOX2 are both present in the ventral otocyst and cvg while CHD7 is also present in the facial ganglion (fg) (D–F). At E10.5, CHD7 and SOX2 are enriched in the ventrolateral otocyst; in the cvg and fg; CHD7 is present in a broader region than SOX2 (G–I). At E11.5, double positive (CHD7+/SOX2+) cells are abundant in the ventral otocyst while the cvg and the fg contain primarily singly labeled CHD7+ or SOX2+ cells, respectively (J–L). (M and N) E10.5 transverse sections from WT and Sox2^CreER/flox embryos treated with tamoxifen/progesterone at E8.5, E9.5, and E10 and stained with anti-CHD7. (O) Quantification of CHD7 immunofluorescence showed no difference between Sox2^+/flox control and Sox2^CreER/flox otocysts (n = 5 per genotype). (P and Q) E10.5 transverse sections from WT or Chd7^Gt/Gt embryos stained with anti-SOX2. (R) Quantification of SOX2 immunofluorescence showed a ~30% decrease in Chd7^Gt/Gt otocysts compared to WT (n = 5 per genotype). Dashed lines in (M, N, P, and Q) denote the otic vesicle. ***P < 0.01. (Scale bar, 50 μm.)

Chd7 and Sox2 Genetically Interact during a Critical Window of Inner Ear Development.

Chd7 loss results in highly penetrant semicircular canal defects that are detectable as early as E12.5 (10, 27). Sox2 is also required between E8.5 and E10.5 for semicircular canal formation (17, 21). Previous studies showed that CHD7 and SOX2 physically interact in brain-derived mouse neural stem cells (23), but whether these genes function cooperatively in the ear is unknown. To test this, we crossed Sox2^CreER and Chd7^flox alleles and induced Cre-mediated deletion at E8.5, E9.5, or E10.5 and assessed E14.5 inner ears by paint-fill. Since Sox2^CreER/+ mice have only one functional copy of Sox2, this cross provides insights into the timing of interactions between Sox2 and Chd7 as well as temporal requirements for Chd7 during inner ear development. After tamoxifen exposure at E8.5, E14.5 control (Chd7^flox/flox or Chd7^flox/+) and Sox2^CreER/+;Chd7^flox/+ ears exhibited normal morphology (Fig. 2A; n = 12 and Fig. 2B; n = 20, respectively), whereas Sox2^CreER/+;Chd7^flox/flox ears displayed malformed lateral semicircular canals (n = 2 of 6 ears) (Fig. 2C and SI Appendix, Fig. S2 H and L and Table S3). Interestingly, tamoxifen treatment of Sox2^CreER/+;Chd7^flox/+ears at E9.5 led to severely malformed lateral semicircular canals (n = 5 of 6 ears) (Fig. 2E and SI Appendix, Fig. S2 A–F and Table S3). Among Sox2^CreER/+;Chd7^flox/flox ears treated with tamoxifen at E9.5, all (n = 6 of 6 ears) lateral semicircular canals were aplastic or hypoplastic, and posterior semicircular canals were truncated (n = 2 of 6 ears) (Fig. 2F and SI Appendix, Fig. S2 M–R and Table S3). Tamoxifen treatment at E10.5 (Fig. 2 G–I and SI Appendix, Fig. S2 S–X and Table S3) resulted in low penetrance (n = 1 of 6 ears) lateral canal hypoplasia in E14.5 Sox2^CreER/+;Chd7^flox/flox embryos and normal inner ears in Chd7^flox/flox (n = 6 of 6 ears) and Sox2^CreER/+;Chd7^flox/+ (n = 6 of 6 ears) embryos. Together, these results indicate that the period around E9.5 is critical in mice for Chd7 function and genetic interactions between Sox2 and Chd7 that help shape the inner ear.

Fig. 2.

Chd7 and Sox2 genetically interact during a critical period of inner ear development. Paint-filled E14.5 inner ears from crosses between Sox2^CreER and Chd7^flox/flox mice after tamoxifen induction at E8.5 (A–C), E9.5 (D–F), or E10.5 (G–I). Conditional heterozygous embryos exhibited dysplastic lateral semicircular canals (*) after Cre induction at E9.5 (n = 5 of 6 ears; E) but not at E8.5 or E10.5 (B and H). In contrast, conditional knockout embryos exhibited dysplastic lateral semicircular canals after tamoxifen treatment at E8.5 (n = 2 of 6 ears), E9.5 (n = 6 of 6 ears), or E10.5 (n = 1 of 6 ears) and hypoplastic or truncated posterior semicircular canals (#) after tamoxifen treatment at E9.5 (n = 2 of 6 ears; F) but not at E8.5 or E10.5. Abbreviations: aa, anterior ampulla; asc, anterior semicircular canal; cd, cochlear duct; ed, endolymphatic duct; la, lateral ampulla; lsc, lateral semicircular canal; pa, posterior ampulla; psc, posterior semicircular canal; sa, saccule; ut, utricle. All ears are shown in lateral view. (Scale bar, 50 μm.)

Double Haploinsufficiency of Chd7 and Sox2 Leads to Vestibular Defects and Shortened Cochleae with Decreased Cell Proliferation.

Previous studies have shown that Chd7 is critical for embryonic viability since homozygous null (Chd7^Gt/Gt) embryos do not survive beyond E10.5 (7, 27). In contrast, heterozygous loss of Chd7 (Chd7^Gt/+) results in mice that do survive after birth but exhibit circling behavior, cleft palate, growth delay, and feeding difficulties and variably reduced postnatal survival (27). Single-copy loss of Sox2 does not impact overall survival or inner ear development (29, 30). To further test for functional interactions, we generated mice with heterozygous deletions of both Chd7 (Chd7^Gt/+) and Sox2 (Sox2^CreER/+). Mice were observed on the day of birth (P0) and each subsequent day for 6 d (SI Appendix, Fig. S3 A and B and Table S2). Survival of Chd7^Gt/+ mice was lower (P < 0.05) than WT or Sox2^CreER/+ mice at P6, and survival of Chd7^Gt/+;Sox2^CreER/+ mice was significantly reduced compared to WT, Sox2^CreER/+ (P < 0.0001), or Chd7^Gt/+ (P < 0.01, log-rank test) mice. Eight of 14 Chd7^Gt/+;Sox2^CreER/+ mice died on the day of birth, another four died by P1, and only two mice survived the entire 6-d observation period. The cause of death is unknown, but Chd7 and Sox2 also exhibit overlapping expression in the developing brain, heart, and craniofacial structures, and functional cooperativity in these organs may be a contributing factor. These results demonstrate that Chd7 and Sox2 act together to promote critical aspects of mammalian development.

Chd7^Gt/+ mice exhibit highly penetrant inner ear defects of the lateral and posterior semicircular canals and associated cristae (9, 10, 25), whereas Sox2^CreER/+ mice have normal inner ear membranous labyrinths (21). Loss of two alleles of Chd7 or Sox2 (in conditional nulls) results in severely malformed vestibular and auditory structures (10, 21). In our study, paint-filled membranous labyrinths from E14.5 WT (Fig. 3A, n = 12) and Sox2^CreER/+ embryos (Fig. 3C, n = 4) showed normal inner ear morphology, whereas Chd7^Gt/+ mice exhibited the typical dysplasia of posterior and lateral semicircular canals (Fig. 3B, n = 12) (9, 10). Strikingly, Chd7^Gt/+;Sox2^CreER/+ inner ears displayed almost complete aplasia of the lateral and anterior semicircular canals, dysplasia of the posterior semicircular canal, malformed endolymphatic duct, absence of anterior, lateral, and posterior ampullae, absence of the utricle, and shortened saccule; in addition, the cochlea was undercoiled (Fig. 3D, n = 7). The inner ear abnormalities in double heterozygous Chd7 and Sox2 embryos bore a striking resemblance to those from Chd7 conditional knockout or Sox2 null mice.

Fig. 3.

Mice with heterozygous deletion of Chd7 and Sox2 exhibit vestibular defects and shortened cochleae with decreased cell proliferation. (A–D) Paint-filled membranous labyrinths of E14.5 WT (A, n = 12) and Sox2^CreER/+ (C, n = 4) mice show normal inner ear morphology, whereas Chd7^Gt/+ ears (B, n = 12) show dysplasia of posterior (#) and lateral (*) semicircular canals; Chd7^Gt/+;Sox2^CreER/+ears (D, n = 7) exhibit almost complete loss of lateral (*), anterior (**), and posterior (#) semicircular canals along with malformed endolymphatic duct, absent ampullae and utricle, shortened saccule, and hypoplastic cochlear duct (arrow). (E–H) Representative bright field images of whole-mounted cochleae from P0 WT (E), Chd7^{Gt/ +}(F), Sox2^CreER/+ (G), and Chd7^Gt/+;Sox2^CreER/+ (H) mice. (I) Bar graph showing reduced cochlear length in double heterozygous Chd7^Gt/+;Sox2^CreER/+ mice compared to WT. n.s., not significant; ****P < 0.0001, log-rank test. (J–K’) Representative bright field images of hematoxylin and eosin stained E18.5 transverse sections through the inner ear from WT (J and J’) and Chd7^Gt/+;Sox2^CreER/+(K and K’) mice. J’ and K’ depict magnified images of the Inset panels from (J and K). (L–O) Anti-phospho-Histone H3 and DAPI counterstaining of transverse sections through the otocyst from E10.5 (L and M) or E11.5 (N and O), Chd7^+/+;Sox2^+/+ (L and N) or Chd7^Gt/+;Sox2^CreER/+ (M and O) embryos. (P) Quantitation of phospho-Histone H3-positive cells in the otocyst at E10.5 and cochlear region at E11.5 from Chd7^Gt/+;Sox2^CreER/+ (red) embryos and Chd7^+/+;Sox2^+/+ (black) littermate embryos. n = 5 for each group. Abbreviations: pa, posterior ampulla; psd, posterior semicircular duct; rm, Reissner’s membrane; sg, spiral ganglion; sm, scala media; st, scala tympani; sv, scala vestibuli; sva: stria vascularis. *P < 0.05, Student’s t test. [Scale bar, 150 μm in (J and K); Scale bar, 50 μm in other images.]

To better understand the auditory defects in Chd7;Sox2 double heterozygous mutant mice, we examined the cochleae at P0, the latest time at which we could recover adequate tissues. Cochlear length at P0 was similar in WT (Fig. 3 E and I, 4,970±57 μm, n = 6 ears), Chd7^Gt/+ (Fig. 3 F and I, 4,442 ± 55 μm, n = 8 ears, P = 0.19), and Sox2^CreER/+ (Fig. 3 G and I, 4,740 ± 80 μm, n = 8 ears, P = 0.99, one-way ANOVA) mice. In contrast, cochlear length was 39.4% reduced in Chd7^Gt/+;Sox2^CreER/+ mice (Fig. 3 H and I, 2,901 ± 189 μm, P < 0.0001; n = 12 ears, one-way ANOVA) compared to WT littermates. We further examined the membranous labyrinth in E18.5 inner ear transverse sections stained with hematoxylin and eosin. Chd7^Gt/+;Sox2^CreER/+ embryos exhibited smaller inner ear capsules and fewer cochlear turns compared to WT littermates (Fig. 3 J and K); however, the cross-sectional appearance of the cochlear fluid spaces (scala vestibuli, scala media, and scala tympani) was normal, as were Reissner’s membrane and the stria vascularis (Fig. 3 J’ and K’).

Since regulation of cellular apoptosis, proliferation, and cell cycle exit (CCE) are all essential for proper cochlear development, we sought to determine whether any of these biological processes were disrupted upon loss of Chd7 and Sox2. Anti-cleaved Caspase-3 was used to detect apoptotic cells. We observed similar anti-cleaved Caspase-3 immunostaining in the endolymphatic duct and ventral-most region of E10.5 WT and Chd7^Gt/+;Sox2^CreER/+ otocysts (SI Appendix, Fig. S3 C–H, arrows and arrowheads, respectively). There were no apoptotic cells detected in either Chd7^Gt/+;Sox2^CreER/+ or WT cochlear region at E11.5, consistent with a previous report on cell death in the WT ear (31). To determine whether the shortened cochlea was caused by premature CCE of prosensory cells, we injected pregnant females with the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) at E13.5 or E12.5, harvested cochleae at P0 and labeled HCs with anti-Myo7a antibody. EdU+ OHCs at the time of harvest were interpreted as having arisen from mitotic precursors at the time of injection. Reduced numbers of EdU+ HCs indicate premature CCE, whereas increased number of EdU+ HCs indicates delayed CCE. Chd7^Gt/+;Sox2^CreER/+ and WT cochleae exposed to EdU at E13.5 both showed the typical apical to basal gradient of EdU staining (CCE) (32) ( SI Appendix, Fig. S3I), whereas Chd7^Gt/+;Sox2^CreER/+ cochleae exposed to EdU at E12.5 showed minimal premature CCE in the middle and delayed CCE in the base of the cochleae (SI Appendix, Fig. S3J). To assay for changes in cell proliferation, we used anti-phospho-Histone H3 (pHH3), which marks cells in mitosis. In contrast to normal apoptosis and minimally changed CCE, anti-pHH3 immunofluorescence was reduced by 27% in Chd7^Gt/+;Sox2^CreER/+ E10.5 otocysts and by 32.1% at E11.5 compared to WT (Fig. 3 L–P). The lack of major changes in CCE at E12.5 and E13.5 in Chd7^Gt/+;Sox2^CreER/+ cochleae also suggests that cell proliferation is normal at these stages. From these experiments, we conclude that cell proliferation, but not CCE or apoptosis, is highly sensitive to combined loss of Chd7 and Sox2.

Chd7 and Sox2 Cooperate to Promote Cochlear Epithelial Organization.

The significantly shortened cochleae in Chd7^Gt/+;Sox2^CreER/+mice prompted us to examine cochlear epithelial and cellular structures. Previous studies showed that Chd7 is expressed in mature cochlear HCs (11) whereas Sox2 is expressed in early postnatal cochlear SCs (33). Given the extensive colocalization of these two proteins in early ear development, we asked when their expression in the sensory epithelium diverges. At E14.5, Chd7 was broadly expressed in the cochlear floor, including (but not restricted to) the SOX2 positive prosensory domain (Fig. 4 A–C). At E18.5, around the time that IHCs and OHCs are identifiable, Chd7 remained broadly expressed in the cochlear floor, including in both HCs and SCs, whereas Sox2 was still expressed in HCs but highly expressed in SCs (Fig. 4 D–F), which is consistent with published data (18).

Fig. 4.

Chd7 and Sox2 cooperate to establish proper hair cell numbers. WT cochleae from E14.5 (A–C) and E18.5 (D–F) embryos stained with anti-CHD7 (A and D) and anti-SOX2 (B and E). CHD7 is present in a broad cochlea region whereas SOX2 is restricted to the prosensory domain (brackets in A–C) at E14.5. At E18.5, CHD7 is abundant in IHCs and OHCs HCs (arrows) and SCs (arrowheads) while SOX2 is lower in HCs and enriched in SCs. (G1–J3’) Representative images of whole-mounted organ of Corti immunostained with anti-Myo7a (G1–J1, G2–J2, and G3–J3) and anti-NF (G1’–J1’, G2’–J2’, and G3’–J3’) in the basal (G1–J1’), middle (G2–J2’), and apical (G3–J3’) turns of WT (G1–G3’), Chd7^Gt/+ (H1–H3’), Sox2^CreER/+ (I1–I3’), and Chd7^Gt/+;Sox2^CreER/+ (J1–J3’) cochleae. Supernumerary OHCs are present in Chd7^Gt/+;Sox2^CreER/+ cochlea in both basal and middle turns (arrows in J1 and J2), whereas an additional row of OHCs is present in the apex (arrow in J3). Supernumerary IHCs in Sox2^CreER/+ and Chd7^Gt/+;Sox2^CreER/+ cochleae are indicated by arrowheads (I1, I2, I3, J2, and J3). NF staining shows normally organized neurite projections in all genotypes. Ectopic neurites project to supernumerary OHCs in the middle turn of the Chd7^Gt/+;Sox2^CreER/+ cochlea (arrows in J2’). (K) Quantification of supernumerary IHCs shows a significant increase in the number of IHCs solely in the apical turn for both Sox2^CreER/+ and Chd7^Gt/+;Sox2^CreER/+ cochleae compared to WTor Chd7^Gt/+;Sox2^+/+ cochleae. (L) Quantification of supernumerary OHCs shows significant increase in supernumerary OHCs in the middle and apical turns, but not in the basal turn in Chd7^Gt/+;Sox2^CreER/+ cochleae compared to all three other genotypes. In (K and L), *P < 0.05, ****P < 0.0001, one-way ANOVA. [Scale bar, 50 μm in (A–F) and 25 μm in (G1–J3).]

Prior studies have shown that cochlear anatomy is normal in Chd7^Gt/+ mice (11), whereas conditional deletion of Chd7 leads to abnormal innervation and organization of the cochlear epithelium (11, 13). Sox2 heterozygous loss results in supernumerary IHCs (30). We examined the HCs of the cochlea in all four genotypes (WT, Chd7^Gt/+, Sox2^CreER/+, and Chd7^Gt/+;Sox2^CreER/+) by staining P0 whole-mounted cochleae with the hair cell marker Myosin7a (Myo7a). In WT (n = 4 ears) and Chd7^Gt/+ (n = 4 ears) cochlear epithelia, anti-Myo7a labeled the typical three rows of OHCs and one row of IHCs with occasional supernumerary HCs primarily present in the apex and middle regions (Fig. 4 G1–G3 and H1–H3). As expected, Sox2^CreER/+ (n = 3 ears) cochleae exhibited occasional supernumerary IHCs located at the neural side of the single row of IHCs (Fig. 4 I1–I3) (30, 34). The number of supernumerary IHCs (1.6 ± 0.6/100 μm, n = 5 cochleae) in the apical turn was significantly higher in Sox2^CreER/+ cochleae compared to WT (P < 0.05, two-way ANOVA) or Chd7^Gt/+ cochleae (P < 0.05, Fig. 4K). In Chd7^Gt/+;Sox2^CreER/+ cochleae, the number of supernumerary IHCs in all three cochlear turns (base, middle, and apex) was comparable to those in the corresponding turns in Sox2^CreER/+ cochleae (Fig. 4K), suggesting that Sox2 regulates IHC number throughout the organ of Corti. No supernumerary OHCs were present in WT, Chd7^Gt/+ or Sox2^CreER/+ cochleae (Fig. 4 G1–I3), whereas abundant supernumerary OHCs were observed in the basal and middle turns of Chd7^Gt/+;Sox2^CreER/+cochleae (Fig. 4 J1–J3, arrows). Notably, one additional row of supernumerary OHCs was present in the apical turn of Sox2^CreER/+;Chd7^Gt/+ cochleae (Fig. 4J3, arrow). The number of supernumerary OHCs in Chd7^Gt/+;Sox2^CreER/+ cochleae was significantly higher in the middle turn (5.1 ± 1.5/100 μm, P < 0.05) and apical turn (8.8 ± 1.5/100 μm, P < 0.0001) in comparison to cochleae from all three other genotypes (Fig. 4L). These results indicate that supernumerary OHCs occur in Chd7^Gt/+;Sox2^CreER/+ cochleae in a graded manner with the most severe phenotypes in the apex and the least severe in the base.

Severe cochlear hypoplasia and supernumerary OHCs in Chd7^Gt/+;Sox2^CreER/+ mice prompted us to ask whether cochlear innervation might also depend on proper dosage of Chd7 and Sox2. Prior studies showed that single-copy Chd7 loss does not impact cochlear innervation, while two-copy loss due to deletion with Foxg1^Cre/+, Pax2Cre, or Ngn1Cre results in severely misguided cochlear neurites (13). Sox2 regulates neurogenesis in the inner ear in a dosage-dependent manner (19) and is critical for hair cell development (17, 21); mice with heterozygous loss of Sox2 show normal hearing (30), and no innervation defects have been reported. To assess cochlear innervation, we applied the neurite marker anti-neurofilament (NF) to P0 whole-mounted cochleae to detect neurites projected from SGNs that synapse on IHCs or pass through the IHC region to synapse on OHCs. In the region of OHCs, neurite projections typically arc toward the basal region of the cochlea, with multiple neurites gathering beneath each row of OHCs to form outer spiral bundles (Fig. 4 G1’–J2’). These outer spiral bundles are still immature in the apical turn of the cochleae at early postnatal stages (Fig. 4 G3’–J3’) since the development of SGNs and neurites in the cochleae begins in the basal turn and progresses toward the apical turn (35). Cochlear innervation in P0 Chd7^Gt/+ (Fig. 4 H1’–H3’) and Sox2^CreER/+ (Fig. 4 I1’–I3’) mice appeared similar to WT (Fig. 4 G1’–G3’). In addition, cochlear innervation in Chd7^Gt/+;Sox2^CreER/+ mice (Fig. 4 J1’–J3’) appeared normal, except that ectopic outer spiral bundles were present in association with supernumerary OHCs (Fig. 4J2’). These results indicate that single-copy loss of Chd7 and/or Sox2 is not sufficient to disrupt cochlear innervation and that supernumerary OHCs of the double heterozygous mutant receive neurite innervation from their associated SGNs.

Since HCs and SCs arise from common progenitor cells, defects that give rise to changes in the number or integrity of HCs might also disrupt the number of SCs. Loss or reduction of the Notch signaling genes Dll1, Jagged2, or Notch1 in mice leads to supernumerary HCs at the expense of SCs (36). To determine whether the supernumerary OHCs observed in Chd7^Gt/+;Sox2^CreER/+ cochleae were associated with changes in the number of SCs, we examined the cellular structure of the organ of Corti. In WT mice, phalloidin labeled the stereociliary bundle located on one side of HCs (SI Appendix, Fig. S4A1), whereas Prox1 antibody labeled inner pillar cells, outer pillar cells, and three rows of DCs at early postnatal stages (SI Appendix, Fig. S4A1’) (37). Supernumerary Prox1-positive SCs were observed in the middle turn of Chd7^Gt/+;Sox2^CreER/+ cochleae, in the same region of the cochlea where supernumerary OHCs were present (SI Appendix, Fig. S4 D2’ and D2, respectively, dots). In the apical turn, where one additional row of OHCs was identified (SI Appendix, Fig. S4D3, arrow), an additional row of Prox1-labeled SCs was also present (SI Appendix, Fig. S4D3’, arrow). These results demonstrate that in Chd7^Gt/+;Sox2^CreER/+ cochleae, supernumerary SCs are present in areas of the cochleae where supernumerary OHCs are also found, suggesting that supernumerary OHCs production does not occur at the expense of nearby SCs.

Previous studies have demonstrated the importance of the PCP signaling pathway in cochlear development. Mice with mutations in various members of the PCP pathway display shortened cochleae, and some also exhibit disrupted stereociliary bundle orientation (6). To determine whether Chd7 and Sox2 functionally interact to regulate PCP signaling, we analyzed bundle orientation in cochleae from mice of all four genotypes (WT, Chd7^Gt/+, Sox2^CreER/+, and Chd7^Gt/+;Sox2^CreER/+). The cartoon in SI Appendix, Fig. S4E (adapted from ref. 38) illustrates the method used to measure bundle orientation. Specifically, the orientation of individual stereociliary bundles was scored as 0° when the line drawn from the vertex through the symmetric bundle ran parallel to the neural-abneural axis. When the bundle was rotated clockwise (or counterclockwise), bundle orientation was scored as positive (or negative) degrees, respectively. Bundle orientation of most IHCs and OHCs was −90° to +90°, with −30° to +30° regarded as normal (SI Appendix, Fig. S4F) (38). Stereociliary bundle orientation distribution histograms were generated from these measurements. The distribution of stereociliary bundle orientation was similar across all four genotypes (one-way ANOVA, SI Appendix, Fig. S4G). We also plotted and analyzed exact angle measurements for each individual OHC and IHC (SI Appendix, Fig. S4H). The distribution of stereociliary bundle orientation was similar across all four genotypes using the nonequal concentration parameter approach ANOVA test (39). These data suggest that neither Chd7 nor Sox2 are required for regulation of stereociliary bundle orientation. Double heterozygous loss of Chd7 and Sox2 leads to supernumerary IHCs and OHCs that appear to be innervated, associated with SCs, and exhibit normally formed and oriented stereociliary bundles.

RNA Sequencing and CUT&Tag Identify Putative CHD7 and SOX2 Target Genes and Genomic Binding Sites.

Dynamic expression of Chd7 and Sox2 during inner ear development, combined with Chd7 regulation of Sox2 and robust inner ear phenotypes in double heterozygous mutant ears, led us to hypothesize that CHD7 and SOX2 may regulate downstream gene expression in potentially synergistic ways. To address this, we performed bulk RNA sequencing on E10.5 otocysts from all four genotypes (WT, Chd7^Gt/+, Sox2^CreER/+, and Chd7^Gt/+;Sox2^CreER/+) (Fig. 5A). Chd7 expression was down-regulated in Chd7^Gt/+ and Chd7^Gt/+;Sox2^CreER/+ otocysts compared to WT (SI Appendix, Fig. S5A), and Sox2 was down-regulated in Chd7^Gt/+;Sox2^CreER/+ otocysts relative to WT, Chd7^Gt/+, or Sox2^CreER/+ otocysts (SI Appendix, Fig. S5B) (40). In contrast, Sox2 was not down-regulated in Chd7^Gt/+ or Sox2^CreER/+ otocysts compared to WT (SI Appendix, Fig. S5B), suggesting that SOX2 autoregulates its own transcription even in the context of haploinsufficiency, as previously shown in other tissues and cell types. This result is also consistent with previous single-cell qPCR studies in Chd7^Gt/+ vs. WT E10.5 otocysts (41).

Fig. 5.

RNA sequencing and CUT&Tag identify putative CHD7 and SOX2 target genes and genomic binding sites. (A) Multiomic approach using bulk RNA sequencing and CUT&Tag on microdissected E10.5 otocysts. (B) Volcano plot of DEGs in Chd7^Gt/+;Sox2^CreER/+ vs. WT ears. Unaffected genes are shown in black, down-regulated genes in blue and up-regulated genes in red. Sox2 was down-regulated in Chd7^Gt/+;Sox2^CreER/+ ears. (C) Table showing Top enriched GO terms among up-regulated (red) and down-regulated (blue) genes in Chd7^Gt/+;Sox2^CreER/+ vs. WT otocysts. (D) Identification of CHD7 and SOX2 target genomic regions by CUT&Tag in WT otocysts. Bedtools was used to define three categories, CHD7+SOX2+, CHD7+, and SOX2+, based on CHD7 and SOX2 occupancy. Heatmaps of CHD7 and SOX2 CUT&Tag signal centered at 21,403 CHD7+SOX2+, 11,449 CHD7+ and 132,639 SOX2+ genomic regions. (E) CHD7+SOX2+ regions associated with TSS or ENCODE-defined enhancers. Heatmaps of CHD7 and SOX2 CUT&Tag signals at 4,968 CHD7+SOX2+ TSS or 5,822 CHD7+SOX2+ enhancer regions. Heatmap signals centered ±5 kb around the summit of SEACR-defined peaks and represented as reads per kilobase per million (RPKM). (F) Venn diagrams display the intersection between CHD7+SOX2+ peaks at TSS or enhancer regions and show 4,968 CHD7+SOX2+ TSS and 5,822 CHD7+SOX2+ enhancers. (G) Differential motif discovery using HOMER revealed known SOX2 (P = 1 × 10⁻⁷) and de novo SOX2 (P = 1 × 10⁻⁴⁹) motifs using a nonspecific control CUT&Tag as background.

RNA sequencing revealed 63 differentially expressed genes (DEGs) in Chd7^Gt/+ otocysts compared to WT, including Chd7 (SI Appendix, Fig. S5C). These DEGs included 45 down-regulated and 18 up-regulated genes, with relevant GO terms including sensory cell morphogenesis, neurogenesis, and epithelial cell development (SI Appendix, Fig. S5E). In contrast to Chd7, heterozygous loss of Sox2 had minimal impact on the transcriptional landscape of the E10.5 otocyst, with only five genes differentially expressed between Sox2^CreER/+ and WT samples (SI Appendix, Fig. S5D). We identified 130 DEGs (55 up-regulated and 75 down-regulated) in Chd7^Gt/+;Sox2^CreER/+ otocysts compared to WT, including Chd7 and Sox2 (Fig. 5B). This represents a twofold higher number of DEGs in Chd7^Gt/+;Sox2^CreER/+compared to Chd7^Gt/+ otocysts (SI Appendix, Fig. S5C). GO term analysis of these 130 DEGs indicated biological processes including regulation of Smoothened and Notch signaling, sensory organ and nervous system morphogenesis, and cell fate commitment (Fig. 5C).

RNA sequencing showed several DEGs in E10.5 Chd7^Gt/+; Sox2^CreER/+ otocysts that are known to specify the dorsal region and contribute to vestibular development including Bmper (41), Wnt7b (42), Fgf3 (43), and Fgf10 (44). We also found that Aldh1a3 is up-regulated in Chd7^Gt/+;Sox2^CreER/+ ears, which is consistent with a previous study showing that Aldh1a3 is up-regulated in Chd7^Gt/+ and Chd7^Gt/Gtotocysts (45). Ventral specifying genes, including Otx2 (46) and Etv5 (47), were also differentially expressed in E10.5 Chd7^Gt/+;Sox2^CreER/+ otocysts. Additional DEGs, including Pax2 (48) and Fgfr2 (49), are also known to be involved in both vestibular and cochlear morphogenesis, suggesting that changes in the expression of critical patterning genes contribute to the phenotypes observed with loss of Chd7 and Sox2.

Since CHD7 and SOX2 physically interact in neural stem cells, we asked whether CHD7 and SOX2 bind to similar genomic regions during otic development. To identify CHD7 and SOX2 binding sites, we performed Cleavage Under Targets & Tagmentation (CUT&Tag) on WT otocysts microdissected from E10.5 embryos (Fig. 5A). Our CUT&Tag dataset, together with bulk RNA sequencing data, have been deposited into the NCBI Gene Expression Omnibus (GEO: GSE239363) database. Sparse Enrichment Analysis for CUT&RUN (SEACR) was employed to identify regions (peaks) that displayed statistically significant enrichment of CHD7 or SOX2. SEACR identified 33,035 CHD7 peaks and 155,548 SOX2 peaks from CUT&Tag libraries. Since we were interested in how CHD7 and SOX2 function together or individually, we identified three distinct types CUT&Tag reads based on occupancy by both CHD7 and SOX2 (21,403), CHD7 only (11,449) or SOX2 only (132,639) (Fig. 5D). Profile plots showed a similar distribution (SI Appendix, Fig. S6A). CHD7+SOX2+ sites were further examined to identify a common genetic network. Since CHD7 and SOX2 are known to bind cis-regulatory elements, we determined whether CHD7+SOX2+ signals were present in regions near TSSs or enhancers. As expected, heatmaps of CHD7+SOX2+ regions near TSSs and enhancers showed both CUT&Tag CHD7 and SOX2 signals (Fig. 5E). Profile plots also confirmed the distribution at these sites (SI Appendix, Fig. S6B). We found 4,968 CHD7+SOX2+ sites associated with TSSs and 5,822 CHD7+SOX2+ sites associated with enhancers (Fig. 5F). Thus, the majority of the CHD7/SOX2 bound sites may correspond to cis-regulatory regions away from TSSs or enhancers that are yet to be defined. We then used HOMER to identify enriched motifs at CHD7+SOX2+ genomic regions (50). The known SOX2 motif (P < 1 × 10⁻⁷) (Fig. 5G) in 2.81% of the CHD7+SOX2+ sites contained a consensus sequence bound by the high mobility group binding domain of SOX2 (ATTGTT) (51), which was previously confirmed by large-scale transcription factor binding analysis (52). De novo motif analysis also identified a SOX2 binding site in 0.32% of the CHD7+SOX2+ binding sites (P < 1 × 10⁻⁴⁹) which contained a sequence (ACAATG) identified using in vitro binding assays (Fig. 5G) (53). CHD7 contains a DNA binding domain (54) but no known consensus DNA binding site, suggesting that its binding is context dependent. Motif analysis confirmed that some CHD7+SOX2+ genomic regions harbor SOX2 motifs. These known and de novo motifs represent a small fraction of identified CHD7+SOX2+ sequences with SOX2 consensus binding sites and likely reflect the influence of CHD7 activity or the requirement for SOX2 to cooperatively partner with other DNA binding factors (55). Together, these results help to describe a CHD7 and SOX2 gene regulatory network in E10.5 otic cells.

Integrative Analysis Identifies Common CHD7 and SOX2 Regulators of Inner Ear Development.

To gain further insight into genes directly regulated by CHD7 and SOX2, we filtered candidate DEGs based on CHD7 and SOX2 CUT&Tag signals within ±5 kb of TSSs or within enhancers. Of the 75 genes down-regulated in Chd7^Gt/+;Sox2^CreER/+ otocysts, 49 were cobound by CHD7 and SOX2, and 35 of these were cobound by CHD7 and SOX2 near their TSSs and/or in enhancers. Eleven of these genes are known to be expressed in the otocyst, and three (Otx2, Sox2, and Dusp6) are associated with human hearing loss (Fig. 6A). Of the 55 genes up-regulated in Chd7^Gt/+;Sox2^CreER/+, 27 were cobound by CHD7 and SOX2, and among these, 23 were cobound by CHD7 and SOX2 near their TSSs and/or in enhancers. Thirteen of these genes have reported expression in the otocyst, and three (Pax2, Fgfr2, and Sall1) are associated with human hearing loss (Fig. 6A). Gene ontology (GO) analysis using Database for Annotation, Visualization, and Integrated Discovery (DAVID) was performed on the 49 down-regulated genes and 27 up-regulated genes, revealing biological processes such as ear development and morphogenesis (SI Appendix, Fig. S6C). To further investigate binding of CHD7 and SOX2 at these genes, genomic tracks of target genes were visualized. No CHD7 and SOX2 protein enrichment were found within ±5 kb of the Chd7 TSS. However, enrichment was observed in the first intron of Chd7 (Fig. 6B), a region consistent for the location of a distal enhancer. Enrichment of CHD7 and SOX2 proteins around Sox2 TSS is consistent with autoregulation of Sox2 gene expression in otic cell types (Fig. 6C). CHD7 and SOX2 enrichment near the TSSs of Pax2 and Otx2 also suggest direct regulation (Fig. 6 D and E). We confirmed Pax2 and Otx2 expression dysregulation by comparing immunofluorescence in double heterozygous otocysts relative to WT (Fig. 6F). The mean immunofluorescence intensity of PAX2 staining in doubly heterozygous ears (n = 6 ears) was 20% higher than in WT ears (n = 4 ears, P < 0.05, unpaired t test). The percentage of OTX2+ positive region in double heterozygous otocysts (n = 6 ears) was reduced 83.3% compared to WT ears (n = 6 ears, P < 0.01, unpaired t test). CHD7 and SOX2 proteins were also enriched near the TSSs of other genes involved in human deafness (Dusp6, Fgfr2, and Sall1) and known to be expressed in the E10.5 otocyst (Etv1, Irx3, and Shisa2) (SI Appendix, Fig. S7A). CHD7 and SOX2 proteins are also enriched at the enhancers of some genes including Fgf10 and Aldh1a3 (SI Appendix, Fig. S7B). Taken together, these results suggest that CHD7 and SOX2 establish a regulatory network during early otic development that includes Pax2, Otx2, and other critical genes.

Fig. 6.

Integrative analysis identifies common CHD7 and SOX2 regulators of inner ear development. (A) Integration of RNA sequencing and CUT&Tag results based on analyzing 130 DEGs with a series of criteria listed on the Left. The number of DEGs are shown after each filtering step. Down-regulated genes Otx2, Sox2, and Dusp6 and up-regulated genes Pax2, Fgfr2, and Sall1 meet all filtering criteria. (B–E) Genomic tracks of CHD7 and SOX2 CUT&Tag signals at the specified genomic coordinates of Chd7, Sox2, Pax2, and Otx2 genes. Highlighted regions mark overlapping CHD7 and SOX2 peaks. Closed red arrowheads indicate the canonical TSS of the specified gene, and open red arrowheads indicate the alternative TSS. (F) Immunofluorescence of PAX2 and OTX2 in E10.5 otocysts from WT and Chd7^Gt/+;Sox2^CreER/+ double heterozygous embryos with quantification of PAX2 and OTX2 staining. (G) Cartoon showing inner ear and cochlear sensory epithelial defects associated with various dosages of Chd7 and Sox2 alleles. (H) Model showing how Chd7 and Sox2 cooperate to regulate gene expression during inner ear development. CHD7 and SOX2 form a complex, bind near TSSs and/or enhancers, and promote expression of Sox2, Otx2, Dusp6, and other genes. CHD7 and SOX2 can also repress expression of Fgfr2, Pax2, Sall1, and other genes in the developing otocyst. In (G), * indicates lateral semicircular canal defects; ** indicates anterior semicircular canal; # indicates posterior semicircular canal defects; arrowheads show shortened cochleae. Red dots represent normal OHCs and IHCs. Light and dark blue dots represent supernumerary OHCs and supernumerary IHCs, respectively. Dashed line circles with crosses in panel Gg’ represent loss of all HCs. N.D. = not determined. [Scale bar in (F), 50 μm.]

Discussion

In this study, we found that Chd7 and Sox2 are expressed in the developing mammalian inner ear in a highly dynamic manner, with extensive colocalization in the E10.5 otocyst and subsequent segregation of the two proteins in both HCs and SCs (CHD7) vs. SCs (SOX2). We also uncovered unique phenotypes related to Chd7 and/or Sox2 deficiency in the developing vestibular and auditory portions of the inner ear. Double Chd7;Sox2 heterozygotes showed severe hypoplasia of all three semicircular canals and cochlea, likely related to reduced cell proliferation in the developing otocyst. Double Chd7;Sox2 heterozygotes also exhibited supernumerary cochlear HCs, normal stereociliary bundle orientation, and innervation. A critical developmental period (~E9.5) of increased sensitivity to combined Chd7 and Sox2 loss was determined using conditional and inducible loss of Chd7 by Sox2^CreER. Genome-wide expression and DNA binding assays identified several candidate target genes for Chd7 and Sox2 in the otocyst, including the early otic patterning genes Otx2 and Pax2. A model (Fig. 6H) of our findings depicts CHD7 and SOX2 binding near the TSSs or enhancers to promote expression of genes including Sox2, Otx2, and Dusp6 and inhibit expression of Pax2, Fgfr2, and Sall1, among others.

These studies shed light on roles for Chd7 and Sox2 in vestibular and cochlear morphogenesis, summarized in Fig. 6G. Normal semicircular canal formation relies on correct dosage of both Chd7 and Sox2. Loss of one Chd7 allele in Chd7^Gt/+ mice results in abnormal lateral and posterior semicircular canals (25, 27), while loss of both Chd7 alleles in Chd7 conditional nulls (using Pax2Cre or Foxg1^Cre/+) leads to severe hypoplasia of anterior, posterior, and lateral semicircular canals (10, 13). Loss of a single Sox2 allele is associated with normal vestibular development, while loss of both Sox2 alleles in Lcc/Lcc mutants leads to severely hypoplastic semicircular canals and cochleae (17). In the current study, we found that loss of one copy each of Chd7 and Sox2 results in inner ear abnormalities that resemble those observed in mice with loss of both alleles of Chd7 or Sox2. Previous reports also showed hypoplastic cochleae in Chd7 conditional knock-out mice using either Foxg1^Cre/+ (10) or Pax2Cre (13), and in mice with inner ear-specific deletion of Sox2 prior to E10.5 (21) or loss of Sox2 in Lcc/Lcc mutants (17). Neither Chd7^Gt/+ nor Sox2^CreER/+ mice exhibited shortened cochleae or cochlear cellular abnormalities beyond the supernumerary IHCs in Sox2^CreER/+ mice. However, Chd7^Gt/+;Sox2^CreER/+ embryos exhibited shortened cochlea and supernumerary OHCs and IHCs. Taken together, these results show that combined loss of Chd7 and Sox2 genes is detrimental to both cochlear and vestibular morphogenesis.

The results from our study may help distinguish among four distinct mechanisms regulating development of cochlear structures: cellular growth, cell proliferation, convergent extension, and radial intercalation. Beginning around E11 in the mouse, most cell types that line the cochlear duct continue to proliferate throughout at least early postnatal stages such that addition of new cells contributes to cochlear outgrowth (1, 6). Beyond E16, physical growth of HCs also contributes to continued cochlear extension (1, 6). We observed no obvious differences in HC size between WT and Chd7^Gt/+;Sox2^CreER/+ cochleae. In contrast, we observed decreases in cell proliferation in the Chd7^Gt/+;Sox2^CreER/+ E10.5 otocyst and E11.5 cochlear primordium that likely contribute to the shortened cochlea, especially since there was no concomitant increase in cell death or major alteration in the rate of progenitor CCE. In addition to cellular growth and proliferation, convergent extension and radial intercalation are key mechanisms that help transform the cochlea from a dense, pseudostratified epithelium to a bilayered structure, in which HCs locate luminally and SCs span the basal-luminal axis. Multiple signaling pathways, including Shh, Wnt, and core PCP are involved in these aspects of cochlear outgrowth (1, 6). Our RNA sequencing in E10.5 Chd7^Gt/+;Sox2^CreER/+ otocysts uncovered several DEGs which are components of Shh signaling (Gli1) or Wnt signaling (Wnt7b and Fzd4). These DEGs might account for the abnormal cochlear outgrowth in Chd7^Gt/+;Sox2^CreER/+mice, but additional studies at later time points during cochlear development are necessary to clarify their precise spatiotemporal requirements. Mutations in some PCP pathway genes, for example, Vangl2, lead to shortened cochleae and abnormal stereociliary bundle orientation (38), and it will be interesting to determine whether PCP is also dependent upon Chd7 and/or Sox2. Early insights into this question suggest that PCP pathways may be resistant to Chd7/Sox2 loss since stereociliary bundle orientation was unchanged in the early postnatal Chd7^Gt/+;Sox2^CreER/+ cochlea. Taken together, our results suggest that Chd7 and Sox2 control neurosensory progenitor proliferation and regulate expression of signaling pathway components involved in convergent extension and radial intercalation, but may be less important for cellular growth.

Our RNA sequencing and CUT&Tag results revealed several DEGs in the Chd7^Gt/+;Sox2^CreER/+ E10.5 otocyst that were also cobound by CHD7 and SOX2 near their TSSs and/or enhancers. Several of these genes are critical for cochlear outgrowth and sensory progenitor differentiation, including Pax2 (48), Otx2 (46), Etv5 (47), Sall1 (56), and Kremen1 (57). These gene expression changes, among others, may help explain the shortened cochlea and supernumerary HCs in Chd7^Gt/+;Sox2^CreER/+ cochleae, which exhibit increased supernumerary HCs and SCs. The Notch signaling pathway seems less likely to be involved in Chd7^Gt/+;Sox2^CreER/+ supernumerary OHC formation, since Notch mutants exhibit supernumerary HCs with concomitant reductions in SCs (36, 58). Our RNA sequencing also revealed dysregulation of some extracellular matrix (ECM) genes, such as Adamts15, which is expressed in the adult rat cochlea (59). Future studies will help determine whether Adamts15 or other ECM genes are involved in generating these cochlear phenotypes.

Our studies provide evidence that Chd7 acts upstream of and cooperatively with Sox2 in a dosage-sensitive manner during a critical period of inner ear development. Sox2^CreER mediated loss of both Chd7 alleles at E8.5, E9.5, or E10.5 resulted in malformed E14.5 lateral and posterior semicircular canals that were most penetrant after tamoxifen treatment at E9.5. In contrast, Sox2^CreER mediated loss of only one Chd7 allele resulted in canal defects after tamoxifen administration at E9.5, but not at E8.5 or E10.5. These differences in conditional heterozygous inner ear phenotypes that occur after deleting one copy of Chd7 on E8.5 vs. E9.5 may reflect dynamic changes in the expression of Sox2 compared to Chd7, or differences in the timing of Chd7/Sox2 genetic interactions. Sox2 expression in the E8.5 lateral otocyst starts to decline at E9.5 and shifts to medial and ventral expression at E10.5 (20, 21). Since Chd7 otic expression is broader than Sox2 throughout E8.5-E10.5, (Fig. 1), there may be sufficient Chd7 WT otic cells remaining in E8.5 tamoxifen-treated Sox2Cre^ER;Chd7^flox/+ mice to allow for normal inner ear development. Alternatively, the inner ear may be most sensitive to combined Chd7 and Sox2 loss around E9.5. Future experiments tracking the fates of Chd7 lineage cells will help distinguish among these possibilities.

The chromatin remodeling activity of CHD7, combined with evidence that SOX2 acts as a pioneer transcription factor (60), raises the possibility that these two proteins mediate changes in chromatin architecture by influencing the levels and timing of target gene expression. Together, these results support a gene regulatory network involving an important chromatin remodeling factor and a key transcription factor during inner ear morphogenesis. Genetic interactions between Chd7 and Sox2 may also have implications for human genetic disorders. Autosomal dominant pathogenic variants in human CHD7 are the major cause of CHARGE syndrome, and pathogenic variants in human SOX2 lead to microphthalmia. The pattern of malformations and clinical features varies greatly among individuals with CHARGE syndrome and with SOX2-related microphthalmia, even among members of the same family with the same sequence variant (61, 62). This phenotypic variability could also be related to the consequences of modifier genes that alter the penetrance and expressivity of certain clinical features. As an example of modifier gene activity, nonsyndromic autosomal recessive deafness-12 (DFNB12) is caused by homozygous or compound heterozygous pathogenic variants in the Cadherin-23 gene, and a variant in the ATP2B2 gene modifies the severity of sensorineural hearing loss (63). Further investigations using emerging genomic data on individuals with syndromic vestibular and/or hearing losses will help clarify whether SOX2 and CHD7 can also act as modifier genes to promote early mammalian vestibular and auditory development.

Materials and Methods

The following mouse strains were used in this study (with their Jackson Laboratory identification numbers): Chd7^Gt/+ (030659), Chd7^flox/flox (030660), Sox2^CreER/+ (017593), and Sox2^flox/flox (013093). Chd7^Gt/+and Sox2^CreER/+ mice were maintained by backcrossing with B6129SF1/J mice (101043). Chd7^flox/flox mice were maintained on a mixed C57BL/6;129S1/SvlmJ background. All procedures were approved by The University of Michigan Institutional Animal Care & Use Committee.

Details on Cre induction, immunofluorescence, paint-fills, EdU injection, RNA sequencing, and CUT&Tag are provided in SI Appendix, SI Materials and Methods.

Sequencing datasets for RNA sequencing and CUT&Tag were deposited into the NCBI GEO (GSE239363) database.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The data reported in this paper have been deposited in the GEO database, https://www.ncbi.nlm.nih.gov/geo (accession no. GSE239363) (64).