DNA methylation in the mouse cochlea promotes maturation of supporting cells and contributes to the failure of hair cell regeneration

We hypothesized that systemic heterochromatinization contributes to the loss of transdifferentiation potential in supporting cells. To identify genomic regions undergoing heterochromatinization in maturing postnatal supporting cells, we examined DNA methylation, a feature of heterochromatin. We profiled DNA methylation using whole-genome bisulfite sequencing (WGBS) on cells from the organ of Corti at the following key developmental stages: embryonic day 13.5 prosensory progenitor cells (E13.5 PG), postnatal day 1 hair cells (P1 HC), postnatal day 1 supporting cells (P1 SC), postnatal day 6 supporting cells (P6 SC), and postnatal day 21 supporting cells (P21 SC). We used enzymatic digestion followed by fluorescence-activated cell sorting (FACS) mediated-purification to isolate E13.5 prosensory progenitors from p27Kip1-GFP mice, P1 hair cells from Atoh1-GFP mice, and P1 supporting cells from Lfng-EGFP mice. p27Kip1-GFP was used to isolate prosensory progenitor cells at E13.5 that will eventually give rise to both hair cells and supporting cells during development (9, 25). For P6 and P21 supporting cells, we used the NuTRAP homogenization method to isolate nuclei from Lfng-CreER; NuTRAP mice (SI Appendix, Fig. S1 and Materials and Methods). We refer to promoters of genes that sit within a CpG island as CGI promoters, and promoters of genes that do not sit within a CpG island as non-CGI promoters. CGIs are genomic regions of 500 to 1,500 bp long with a ratio of CpG dinucleotide sequences greater than 0.6, and they are often found at promoters and retrotransposable elements (26). As a positive control, we confirmed that DNA methylation is depleted at 13,564 CGI promoters and enriched at 8,586 non-CGI promoters (SI Appendix, Fig. S2 A and D). We observed distinct methylomes in the different cell types and time points. For example, we observed that the Atoh1 and Dll1 loci are more heavily methylated in P21 supporting cells compared to E13.5 prosensory progenitors, P1 hair cells, P1 supporting cells, and P6 supporting cells (Fig. 1A). Since the ATOH1 transcription factor is both necessary and sufficient for hair cell differentiation, we profiled ATOH1 chromatin-binding regions in E17.5 hair cells. We used CUT&RUN with anti-GFP antibodies to identify binding sites in hair cells obtained from Atoh1-EGFP mice. We found previously identified ATOH1-binding sites at the Atoh1 promoter and its 3′ autoregulatory enhancer overlapping with regions of DNA hypermethylation in P21 supporting cells (Fig. 1A). We saw a similar pattern at the Dll1 locus, which is specifically expressed in hair cells (Fig. 1A). DNA hypermethylation at these regulatory elements could serve as a mechanism to permanently silence the endogenous Atoh1 and Dll1 loci in mature supporting cells by preventing ATOH1 binding. Since ATOH1 and DLL1 play critical roles in hair cell and supporting cell fate specification, DNA hypermethylation of these gene loci suggests a larger genome-wide shutdown of gene programs responsible for the development of hair cells and supporting cells with increasing age.

We characterized the changes in DNA methylation on a genome-wide level in our samples using principal component analysis (PCA). We separated our DNA methylation data into different genomic features—CpG islands, promoters, and putative enhancers—prior to PCA analysis (Fig. 1B). Annotations of CpG islands and promoters were obtained from the UCSC Genome Annotation Database for GRCm38/mm10 (27). Putative enhancers were identified by calling peaks on H3K4me1 CUT&Tag signal (Fig. 1D). At CpG islands, E13.5 prosensory progenitors, P1 hair cells, and P1 supporting cells clustered together, while P6 supporting cells and P21 supporting cells clustered away from all other cell types and stages, and also from each other (Fig. 1 B, Left). This clustering pattern suggests that DNA methylation at CpG islands stays relatively constant among cochlear cell types at earlier stages of development, but diverges at the end of the first postnatal week. At gene promoters, all cell types and stages clustered separately, with E13.5 prosensory progenitors in the middle and hair cells and supporting cells diverging toward separate identities (Fig. 1 B, Middle). This suggests that there are distinguishing DNA methylation patterns at promoters between specific cell types and stages of development in the organ of Corti. At gene enhancers, P1 hair cells and P1 supporting cells cluster away from the prosensory progenitor state but are relatively similar to each other, suggesting they still share similar enhancer networks at this earlier stage of development, consistent with our previous data (Fig. 1 B, Right) (17). Overall, our data suggest that DNA methylation is positively correlated with the increasing maturation of supporting cells with age.

Next, we sought to identify the genomic regions driving the variance between the cell types that could contribute to distinct gene silencing patterns. We used dmrseq to identify differentially methylated regions (DMRs). We identified DMRs between E13.5 prosensory progenitors and P1 hair cells, P1 hair cells and P1 supporting cells, and P1 supporting cells and P21 supporting cells to obtain sets of regions that were hypermethylated or hypomethylated in one group versus the other. We further intersected the DMR sets to obtain four mutually exclusive sets of regions, which we named according to the dynamics of their DNA methylation signature (Fig. 1C). All four sets of DMRs showed evolutionary conservation, suggesting that they possess some regulatory function and giving confidence to the peak calling method (SI Appendix, Fig. S2C). We identified 2,850 preestablished DMRs, which are characterized by being hypomethylated in E13.5 prosensory progenitors, P1 hair cells and P1 supporting cells, becoming progressively hypermethylated in P6 and P21 supporting cells. We identified 6,746 de novo hair cell–specific DMRs, which are uniquely hypomethylated in P1 hair cells. We identified 3,058 de novo supporting cell–specific DMRs, which become most heavily hypomethylated in P21 supporting cells. Finally, we identified 9,683 de novo common DMRs, which are hypomethylated in P1 hair cells, P1 supporting cells, P6 supporting cells, and P21 supporting cells (Fig. 1C). Both the preestablished and de novo hair cell–specific DMRs showed enrichment of the ATOH1 E-Box and POU4F3 homeobox motifs, suggesting that these DMRs are part of the hair cell gene regulatory network (SI Appendix, Fig. S2B). Strikingly, the DMRs appear to be highly specific to organ of Corti cells, as the DMRs are completely hypermethylated in intestinal tissues, even though both tissues express ATOH1 and regulate cell fate by Notch signaling (SI Appendix, Fig. S2E) (28–30). The de novo distinction we used is important, because the prosensory progenitors, hair cells, and supporting cells are postmitotic at all developmental stages examined in this study. Thus, the mechanism of DNA hypomethylation is most likely actively mediated by TET enzyme activity, rather than by passive DNA demethylation during cell division (31, 32). Taken together, our analysis shows that DNA methylation patterns coincide with distinct developmental stages and cell types.

To determine if the DMRs identified in our analysis were part of the hair cell gene regulatory network, we overlaid the DMR sets with binding sites for two key hair cell transcription factors, ATOH1 and POU4F3. We used published ATOH1 and POU4F3 CUT&RUN data from FACS-purified E17 hair cells (17, 33) (Fig. 1D). POU4F3 is a downstream target of ATOH1 and has pioneer transcription factor activity required for accessing closed ATOH1 target enhancers and allowing hair cell differentiation to proceed (33). The de novo hair cell–specific DMRs and the de novo common DMRs contain both ATOH1- and POU4F3-binding sites, suggesting that they might be downstream targets of ATOH1 and POU4F3, and are therefore required for the differentiation of hair cells. On the other hand, the preestablished DMRs and the de novo supporting cell–specific DMRs are only bound by ATOH1, and not by POU4F3 (Fig. 1D). Since the preestablished DMRs are hypomethylated in E13.5 prosensory progenitors, we suggest these regions may serve as initial targets of ATOH1, which are up-regulated within the prosensory domain at E14.5 to initiate hair cell differentiation (10, 33). To further categorize these DMR sets, we performed CUT&TAG on the promoter-specific mark H3K4me3 and the promoter-enhancer mark H3K4me1 in P1 hair cells and P1 supporting cells (34). We found the DMR sets are depleted of the promoter mark H3K4me3 but are enriched for the enhancer mark H3K4me1 in both hair cells and supporting cells (Fig. 1D), suggesting that the DMRs are putative hair cell–specific and supporting cell–specific enhancers.

We next performed gene ontology (GO) enrichment analysis on the DMR sets using GREAT (35). The preestablished DMRs include known organ of Corti developmental genes such as Sox2, Atoh1, Gfi1, Notch1, and Dll1. The preestablished DMRs are enriched for GO terms relating to “regulation of epithelial cell differentiation,” “stem cell population maintenance,” and “regulation of auditory receptor cell differentiation” (Fig. 1E). The de novo hair cell–specific DMRs are enriched for terms relating to the maturation of hair cells, such as “neuromuscular process controlling balance,” “mitochondrial membrane organization”, and “auditory receptor cell stereocilium organization”. Genes within the de novo hair cell–specific DMRs include early hair cell developmental genes like Atoh1, Pou4f3, Gfi1, Myo7a, Myo6, and Jag2, as well as genes involved in the maturation of hair cells such as Pcdh15, Cdh23, and Clic5. The de novo supporting cell–specific DMRs have GO terms relating to epithelium formation with terms such as “cell–cell junction organization,” and “positive regulation of cytoskeleton organization.” Key genes include Sox2, Sox4, Sox9, Gjb2, Gjb6, Coch, KCN (potassium channel) family genes, and SLC (solute carrier) family genes. Finally, the de novo common DMRs have GO terms relating to more general epithelial maturation, such as “cytoskeleton-dependent intracellular transport” and “adherens junction organization” and broadly include KCN, SLC, and CDH (cadherin) family genes. The GO analyses demonstrate that each organ of Corti cell type has a distinct DNA methylation signature at each age that correlates with its known gene expression and function. Taken together, we suggest that dynamic DNA methylation patterns are used to reinforce gene program switches as cells transition through different states and identities during their lifetime. For instance, as supporting cells lose the ability to transdifferentiate postnatally, we observed a switch from a hair cell differentiation gene program to a supporting cell maturation gene program, where the former becomes hypermethylated and the latter becomes hypomethylated.

The gradual DNA hypermethylation of preestablished DMRs in postnatal supporting cells during the first postnatal week correlates with the period during which supporting cells lose the potential to transdifferentiate in response to DAPT (N-[N-(3, 5-difluorophenacetyl)-l-alanyl]-s-phenylglycinet-butyl ester) (13, 36). Initially, the percentage of methylated CpGs at preestablished DMRs decreased from 45% in E13.5 prosensory progenitors to 30% in P1 hair cells and P1 supporting cells, which coincided with the start of differentiation of cochlear hair cells and supporting cells (Fig. 2A). The 30% mCpG, although an average of all preestablished DMRs, corresponds to the lowest point of the DNA methylation valleys observed at the Atoh1, Pou4f3, and Gfi1 loci (Fig. 2D). The CpG percent methylation then increased gradually to 38% and 50% by P6 and P21 in supporting cells, coinciding with loss of transdifferentiation potential (Fig. 2A). To better understand the role of DNA hypermethylation in supporting cells, we annotated the preestablished DMRs to genomic features (37), and found that they are enriched around CpG islands and their corresponding shores and shelves, which are defined as iterative 2 kb regions flanking the CpG island (Fig. 2 B and C). In agreement with this, the preestablished DMRs are also enriched around genic regions, spanning from 5 kb upstream of the promoter, through the gene body, and all the way to the 3′ UTR (Fig. 2C). The preestablished DMRs are not enriched at intergenic regions, suggesting that they are unlikely retrotransposons or distal enhancers. Rather, the majority of preestablished DMRs appear to identify proximal and genic enhancers that occur within the neighborhood of CpG island–containing promoters, but not necessarily occurring at the promoter itself since they do not possess H3K4me3 peaks. Therefore, we focused our attention on preestablished DMRs overlapping CGI promoters in our subsequent analyses.

CGI promoter genes serve important roles as transcriptional hubs and tend to be developmentally regulated. They sit within DNA methylation valleys, also referred to as DNA methylation canyons, which are largely hypomethylated (38). These CGI promoter genes are frequently silenced by the repressive histone modification H3K27me3. The recruitment of PRC2, a H3K27me3-specific methyltransferase, to CG rich sequences at promoters also repels DNA methyltransferases (DNMTs) (39). Thus, CGI promoters, used for lineage specification during the process of cell differentiation, are maintained in a DNA hypomethylated state in differentiated cells. CGI promoter repression is understood to be mediated by H3K27me3 for both the prelineage commitment bivalent state, as well as the subsequent repression of nonlineage genes (40). DNA hypermethylation has been shown to occur at promoters that are CG-poor and lack CGIs, which tend to be somatic tissue–specific genes expressed in more mature cell types (38). Building on this model, we found that cochlear supporting cells undergo a switch between the two modes of repression—from H3K27me3 methylation to DNA methylation—at developmentally regulated hair cell–specific CGI promoter genes (Fig. 2D). For example, at three key hair cell–specific transcription factor gene loci, Atoh1, Pou4f3, and Gfi1, we find evidence of both DNA hypermethylation and heterochromatinization. Additional examples of DNA hypermethylation were observed at Tead4, Eya1, Notch1, and Fgfr3 (SI Appendix, Fig. S3A). At the genic level, the DNA hypermethylation of preestablished DMRs manifests as DNA methylation encroachment of the DNA methylation valleys in which the respective CGI promoter gene sits (Fig. 2D, WBGS).

Next, we identified the epigenetic changes occurring across 103 gene promoters that overlapped with the previously identified set of 2,850 preestablished DMRs (Fig. 2 E–G). We observed a switch from H3K27me3-mediated repression to a DNA methylation–mediated silencing in supporting cells between P6-P8 and P21 (Figs. 2E and 3F). This finding agrees with previous work describing the antagonistic relationship between DNA methylation and PRC2-mediated H3K27me3 deposition (41, 42). The switch from H3K27me3-mediated repression to DNA methylation–mediated silencing was also observed at both HoxA and HoxB loci (SI Appendix, Fig. S4B). In addition to DNA hypermethylation, DNA accessibility measured by Cleavage Under Targeted Accessible Chromatin (CUTAC) (43) drops precipitously between P1 and P21 in supporting cells (Fig. 2G). While the increase of H3K27me3 between P1 and P8 explains the decreased accessibility within that same time window, DNA hypermethylation explains the decrease in accessibility between P8 and P21, suggesting that DNA hypermethylation is required in the final stages of maturation of supporting cells to complete heterochromatinization of their hair cell–specific CGI promoters. Besides CGI promoters, we also see similar changes in DNA methylation, H3K27me3, and accessibility at pre-established DMRs that overlap with non-CGI promoters, gene bodies, and intergenic regions, albeit less prominently (SI Appendix, Fig. S3B). As a positive control, we confirmed that H3K27me3 signal at promoters is comparable across samples (SI Appendix, Fig. S4A). Altogether, we propose switching between H3K27me3 and DNA methylation is a mechanism for long-term silencing of hair cell–specific CGI promoter genes and the complete shutting down of the hair cell gene regulatory network.

Neonatal supporting cells still retain the potential to transdifferentiate into hair cells after hair cell loss, or by blocking Notch signaling with the gamma-secretase inhibitor DAPT (8, 13, 36, 44, 45). However, from our DNA methylation data, we observe that the de novo hair cell–specific DMRs are hypermethylated in P1 supporting cells (Fig. 1C). Based on the previously described GO terms, the hair cell–specific DMRs are essential for hair cell differentiation (Fig. 1E). Since DNA methylation is a repressive epigenetic feature, we hypothesized that in cases where P1 supporting cells differentiate into hair cells, hypermethylated hair cell–specific DMRs in P1 supporting cells must first undergo demethylation and derepression to successfully up-regulate hair cell–specific genes and transdifferentiate. To test this, we induced P0 supporting cell transdifferentiation with DAPT and performed WGBS to measure mCpG percentages in DAPT-responsive supporting cells compared to DAPT-unresponsive cells. Briefly, we used a transgenic mouse line with Lfng-CreERT2 (46), and a ROSA tdTomato (TDT) Cre reporter (47) to lineage trace supporting cells, as well as Atoh1-GFP (48) to label hair cells and supporting cells undergoing transdifferentiation into new hair cells (17). Lineage-traced TDT+ supporting cells that transdifferentiate into hair cells upon DAPT treatment up-regulate hair cell–specific transcription factors ATOH1-GFP and POU4F3 (Fig. 3A and SI Appendix, Fig. S5A). Thus, supporting cells undergoing transdifferentiation exhibit TDT+/ GFP+ double labeling, whereas supporting cells not responding to DAPT are only labeled with the TDT reporter (Fig. 3A). Cochlear sensory epithelium was explanted and cultured with or without DAPT for 48 h, followed by FACS purification of single-labeled nonresponsive TDT+ supporting cells and double-labeled TDT+/GFP+ transdifferentiated supporting cells (Fig. 3 B and C). We processed these purified cells for bisulfite treatment for WGBS analysis. We also performed this experiment in supporting cells at P6 when the supporting cells no longer respond to DAPT.

To establish a baseline, we compared the CpG methylation percentage at de novo hair cell–specific DMRs to endogenous P1 supporting cells (Fig. 3D). After DAPT treatment, we found that only successfully transdifferentiated supporting cells at P0 showed a decrease in mCpG percentage at de novo hair cell–specific DMRs (Fig. 3D). No decrease in mCpG percentage was observed at the preestablished, de novo supporting cell–specific, or de novo common DMRs, suggesting that the demethylation is highly specific (SI Appendix, Fig. S5B). Nonresponsive TDT+ supporting cells in both the P0 and P6 cochlear explants had comparable percentage of mCpG to baseline in vivo P1 supporting cells. Quantifying the change, we found a 10% decrease in mCpG percentage from 72.37% in nonresponsive TDT+ supporting cells to 62.33% in transdifferentiated TDT+/GFP+ supporting cells (P-value: 2.97e-61; Fig. 3E). The 10% decrease in DNA methylation is notable because the demethylation process occurred within a relatively brief window of 48 h after DAPT exposure. In addition, earlier studies demonstrate that 5 to 10% DNA methylation differences can be biologically significant, for example in the case of the Oct4 locus between human embryonic stem cells, their derived cell populations, normal somatic tissues, and disease conditions (20, 49). We used de novo common regions as a control and found no significant change in DNA methylation levels (P-value: 1; Fig. 3E). The preestablished DMRs and the de novo supporting cell–specific DMRs also show no signs of demethylation (SI Appendix, Fig. S5B), indicating that the DNA demethylation is specific to the de novo hair cell–specific DMRs. Since the transdifferentiating TDT+/GFP+ supporting cells are postmitotic, the only mode for de novo DNA demethylation is through TET enzyme activity, and not through the lack of maintenance by DNMT1 during cell division (50). Together, this suggests that the activation of the hair cell GRN in transdifferentiating supporting cells is partly mediated by DNA demethylation of hair cell–specific DMRs, which is most likely driven by TET enzyme activity.

TET enzymes are responsible for the stepwise conversion of 5-methylcytosine to 5-hydroxymethylcytosine to 5-formylcytosine to 5-carboxylcytosine (51). The action of thymine-DNA-glycosylase (TDG) and base excision repair (BER) ultimately restores the unmethylated cytosine (51). The WGBS method we use detects both 5-methylcytosine and 5-hydroxymethylcytosine. Since we detected DNA demethylation after supporting cell transdifferentiation, we concluded that the 5-methylcytosine is being converted to at least 5-formylcytosine by TET enzyme activity, which may be sufficient to derepress hair cell–specific regulatory elements. We tested whether TET-mediated DNA demethylation is required for supporting cells to transdifferentiate into hair cells. We treated P0 cochlear explants with 10 µM of DAPT and a dose series of Bobcat339 (52), a TET1 and TET2 enzyme inhibitor for 48 h. Cochleas were then dissociated after 48 h for FACS-mediated quantification. We hypothesized that blocking TET enzyme activity with Bobcat339 would prevent supporting cells from successfully up-regulating ATOH1 and transdifferentiating in response to DAPT. This would lead to a decrease in the proportion of double-labeled GFP+/TDT+ cells (which we quantified as % permissive supporting cells) compared to total TDT+ cells. In response to 10 µM DAPT treatment, 56.97% of supporting cells were permissive to transdifferentiation. However, supporting cell transdifferentiation rate decreased in a dose-dependent manner with increasing concentrations of Bobcat339 cotreatment (Fig. 3F). We found 45.00% permissive supporting cells with 90 µM Bobcat339 cotreatment (n = 3, P-value = 0.085, n.s.), 30.09% permissive supporting cells with 180 µM Bobcat339 cotreatment (n = 3, P-value = 0.023, *), and 14.94% permissive supporting cells with 270 µM Bobcat339 cotreatment (n = 3, P-value = 0.030). This demonstrates that supporting cells require TET enzyme activity to demethylate and derepress hair cell–specific DMRs before they can be activated during transdifferentiation.

DNA methylation is considered to be the final layer of epigenetic silencing to permanently shut down DNA and make it refractory to transcription factors and transcriptional machinery (53). We have demonstrated that preestablished CpG island promoters undergo increasing de novo DNA methylation and decreasing accessibility in postnatal maturing supporting cells (Fig. 2D). Further, we showed that the transdifferentiation of P1 supporting cells both induces de novo DNA demethylation (Fig. 3D) and is attenuated by inhibiting TET enzyme activity (Fig. 3F), suggesting that DNA methylation plays a key role in silencing the hair cell gene regulatory network through heterochromatin formation. Next, we wanted to understand the implications of this heterochromatinization in the pathological context of hearing loss. Specifically, we wanted to understand whether there are further changes to the supporting cell chromatin landscape in the adult stage, as well as in a long-term deafened state, which is clinically relevant to human hearing loss. Since supporting cell subtypes are functionally distinct at adult stages, we used scMultiome to simultaneously profile gene expression and chromatin accessibility at the single-cell level. Comparing accessibility between organ of Corti supporting cell subtypes allows us to identify cell type–specific enhancers, as well as supporting cells that are more receptive to conversion into new hair cells. We collected scMultiome datasets for cochlear supporting cells in wildtype P1, P8 (54), and P70 mice, as well as in P70 mice that were deafened at P21 (Fig. 4A). For the P70 dataset, Lfng-CreERT2;NuTRAP;Pou4f3-DTR mice were used to label and sort supporting cells. Single-cell data were clustered using “uniform manifold approximation and projection” (UMAP) and “weighted-nearest neighbor” (WNN) using the Seurat R package, which integrates both the gene expression and the accessibility information to define a “joint” cellular state (55). WNN analysis is an improvement over previous methods of clustering that were based on RNA or ATAC data alone by improving our ability to resolve cell states (SI Appendix, Fig. S6A). scMultiomic profiling is especially well-suited for resolving the diversity of postnatal organ of Corti supporting cells, which no longer express distinguishing developmental genes and instead share common morphological and molecular features (56).

We annotated the clusters using gene expression data cross-referenced with previously published literature. Since gene programs change as the cochlea matures between P1, P8, and P70, we used different sets of genes to identify clusters at each developmental time point. Genes used to identify clusters from the P1, P8, and P70 cochlea 10× multiome datasets can be found in SI Appendix, Figs. S7–S9. After using the single-cell gene expression data to identify and annotate clusters, we compared the single-cell ATAC data between clusters and time points. We found that cochlear developmental genes such as Atoh1 were more accessible at P1 in both hair cells and supporting cells, whereas genes important for the physiological function of the cochlea such as Gjb2 and Gjb6 were more accessible at P70 in supporting cells (Fig. 4B). Altogether, this suggests that maturing supporting cells undergo a gene program switch, where the hair cell gene program is silenced in favor of a differentiated and functional supporting cell gene program. Additional examples of hair cell–specific genes Pou4f3, Jag2, Dll1, and Myo7a, and supporting cell–specific genes Sox2, Notch1, and Jag1 are shown in SI Appendix, Figs. S9B and S10A, respectively.

Next, we associated the single-cell ATAC data back to our DNA methylation data. We separated chromatin accessibility (ATAC) signal by cell types, and quantified it over the preestablished, hair cell–specific, supporting cell–specific, and common DMRs (Fig. 5A). We found supporting cell subtypes showed progressively decreasing accessibility at preestablished and hair cell–specific DMRs between P1, P8, and P70 time points, but not at the SC-specific or common DMRs (Fig. 5A). We noted that accessibility increased in the P70 long-term deafened supporting cells compared to the P70 wild-type supporting cells at hair cell–specific DMRs in BC, PC, and DC2 clusters (Fig. 5A). The most prominent example of this can be seen at the locus of the hair cell gene Myo7a, where increased accessibility in the deafened BC cluster also correlated with low Myo7a mRNA expression (Fig. 5B). To validate the Myo7a mRNA detected in the deafened BC cluster, we performed RNAscope RNA FISH. We identified Epyc as a putative BC marker from our P70 10× multiome dataset and confirmed the expression of Epyc in BCs in intact tissue using RNAscope (Fig. 5 C and D). We were able to detect Myo7a transcripts coming from the surviving Epyc+ BC after hair cell killing (Fig. 5D). This suggests that the loss of hair cells may cause supporting cells to modify their chromatin structure following damage, which may allow for easier reactivation of the hair cell GRN in the deafened state.

J.D.N., J.L., J.G.C., and N.S. designed research; J.D.N., J.L., and T.S. performed research; J.D.N. and J.L. contributed new reagents/analytic tools; J.D.N., T.S., A.K.G., and N.S. analyzed data; and J.D.N., A.K.G., and N.S. wrote the paper.

The authors declare no competing interest.