Connexin 26 Functions as a Direct Transcriptional Regulator During the Cochlea Development

ABSTRACT

Gap junction Beta 2 Protein (GJB2, Connexin26, Cx26), the primary genetic cause of hereditary hearing loss (25%–50% of cases), has been exclusively regarded as forming an intercellular channel that mediates rapid communication. Here, we redefine its biological role by discovering its nuclear localization and direct transcriptional regulatory function in cochlear structure development. We demonstrate that Cx26 could aggregate in the nucleus of cochlear support cell and cell lines. Cx26 can bind to the promoter transcription start point of genomic DNA and directly regulate gene transcription, thus controlling the structural development of the tunnel of Corti during cochlear development. Further, we provide strategies based on mechanisms to promote the TC development and hearing rescue in Cx26 deficient cochlea, which has important implications for the discovery and development of treatment strategies for hearing loss caused by Cx26 deficiency.

Connexin26 can not only form intercellular channels that mediate rapid communication on the cell membrane, but also enter the nucleus as a transcription factor to directly regulate the transcription of nuclear genes. In the developing cochlea, Cx26 can control the maturation of the molecular scissor ADAM10 by regulating the transcription of TspanC8, ultimately affecting the formation of the tunnel of Corti.

1. Introduction

Congenital hearing loss, the most prevalent sensory disorder worldwide [1], arises from genetic mutations in over 50% of cases [2]. Mutations in the Gap junction protein beta 2(GJB2) gene are the most common genetic cause of non‐syndromic hereditary deafness [3], accounting for more than 25%–50% of all cases of hereditary hearing loss [4, 5].

Connexin26 (Cx26, encoded by GJB2), was classically associated with intercellular gap junction channels (GJCs) on the cytomembrane between adjacent epithelial cells and supporting cells in the cochlea [6, 7], allowing intercellular communication and the transfer of ions, microRNA, secondary messengers, and metabolites ≤1.5 kDa [8]. Previous studies have attributed Cx26‐related deafness to disrupted cochlear gap junctions and following pathological changes, including K⁺ Recycling defect [9, 10], reduced endolymphatic potential (EP) [11, 12], disruption of miRNAs permeability [13, 14], redox imbalance and decrease in cochlear nutrient delivery [15, 16, 17], sensory hair cell degeneration [18, 19, 20]. However, emerging evidence challenges the classical membrane gap junction function impairment as the primary mechanism for congenital deafness [21]. While complete deletion of Cx26‐mediated GJC function after postnatal day 8 (P8) in mice causes progressive hearing loss rather than immediate deafness, critically, Cx26 deficiency during gestation or early postnatal period (before P4) leads to profound hearing loss coinciding with the onset of auditory function (P12–P18) [22]. The remarkable difference in both the severity and onset timing of hearing loss induced by Cx26 deficiency at different developmental stages suggests that the loss of Cx26 channel function alone may not fully explain the pathogenesis of congenital deafness, implicating potential non‐classical roles of Cx26 beyond intercellular channels.

Our and others previous studies have revealed a critical association between congenital deafness and cochlear structural developmental disorders caused by Cx26 defects [22, 23, 24]. During the cochlea development, the lateral membranes in the middle of inner pillar cells (IPCs) and outer pillar cells (OPCs) are separated, forming a triangular space filled with perilymph between adjacent pillar cells (PCs), which is the tunnel of Corti (TC) [25, 26]. We have found that Cx26 deficiency or mutation in the mouse cochlea before P4–5 results in failure of the tunnel of Corti (TC) formation and severe congenital deafness. In contrast, conditional knockout of Gjb2 (Cx26) after TC opening (post‐P8) does not cause already developed TC collapse and immediate severe hearing loss [22, 23, 27, 28, 29], suggesting that structural developmental disorders of TC caused by Cx26 defects, rather than Cx26 deficiency itself, are the main mechanism underlying profound hearing loss. This malformed organ of Corti (OC) without formed TC has also been observed in temporal bone sections of deafness donors with GJB2 mutants and cochleae sections of different GJB2 mutant(p.R75W, c.235del) transgenic mice models [18, 30, 31, 32, 33], which indicating that developmental disorders of OC are universal pathological changes in mammals with Cx26‐deficient hearing loss. We and others have reported that the cytoskeleton developmental abnormalities and the dysplasia of the non‐centrosomal microtubule‐organizing center occur in PCs in Cx26‐deficient mice, which may contribute to the deformity of OC structure and failure of TC formation [22, 23]. However, how Cx26 regulates cochlear development independently of its canonical channel function remains unexplained.

Increasingly evidence showed that Connexins (Cxs) could distribute in mitochondria, nucleus, nuclear envelope, cytoplasm, and other different cellular compartments [34], highlighting non‐canonical biological functions of connexins including the direct regulation of cell processes and fate [35, 36, 37, 38, 39]. Nonetheless, the view of subcellular localization of Cx26 in the cochlea still focuses on the intercellular cytomembrane.

In this study, we focus on the nuclear localization of Cx26 and its function in regulating gene transcription during cochlear structure and hearing development. We demonstrate that the inhibition of gap junction channels does not lead to hearing loss and TC deformity. Thus, the Cx26 deficiency‐related congenital deafness is not the result of loss of Cx26 intercellular channel function. We found that Cx26 localizes to the nucleus not only in cochlear supporting cells but also in various cell lines. The nuclear‐localized Cx26 can directly bind to the transcription start site of the promoter of genomic DNA, thereby directly controlling the transcriptional regulation of genes by regulating gene promoter activity. Cochlear Cx26 deficiency results in a decrease in the transcription level of the TSPANC8 gene, leading to TC structure disorders and hearing loss. These results redefine Cx26 as a molecule coordinating both junctional communication and gene transcriptional regulator during cochlear and hearing development. Further, we provide strategies based on mechanisms to promote the TC development and hearing rescue in Cx26 deficient cochlea, which has important implications for the discovery and development of treatment strategies for Cx26 deficiency diseases.

2. Results

2.1. Gap Junction Channel Dysfunction is Not the Primary Mechanism for TC Structural Disorders and Cx26 Deficiency‐Related Congenital Deafness

Consistent with our previous reports, the Cx26 conditional knockout (cKO) mouse model exhibited profound hearing loss at all tested frequencies (4, 8, 12, 16, 20, 24, 28, 32, and 40 kHz) at P18 (Figure 1A). Cochlear morphology revealed shows that congenital Cx26 deficiency leads to TC structural malformation, while clearly visible TC structures could be observed in the control group cochlea (Figure 1B). To clarify whether the impaired cochlear gap junction channels caused by Cx26 defects are related to hearing loss and TC structural disorders, classic gap junction inhibitor carbenoxolone (CBX) was injected intraperitoneally (20 µg/g body weight) into neonatal mice (Figure 1C). CBX treatment can effectively inhibit the permeability of Cx26 GJCs (Figure S1). Consistent with the control group, CBX treatment did not significantly alter Cx26 expression, and the separation of inner and outer pillar cells was observed (Figure 1D). TC structure developed normally after continuous intraperitoneal (IP) injection of CBX (20–30 mg/kg/day, P0‐P6) or single injection (2 µL, 5 µg/µL dissolved in 1× PBS) through the round window membrane (RWM) in P4. These results suggest that Cx26 gap junction channel dysfunction is not the primary mechanism for Cx26 deficiency‐related TC structural disorders (Figure 1E,F). Moreover, continuous IP of CBX (P0‐P17) treatment does not cause statistically significant hearing loss (Figure 1G).

FIGURE 1.

CBX‐mediated inhibition of cochlear Cx26 gap junction function does not lead to TC structural deformity and hearing loss in mice. Auditory brainstem response (ABR) test showed P18 Cx26 cKO mice have severe hearing loss (A, n = 6). TC structural deformity was observed in the resin slice specimen of Cx26 cKO mice cochlea (B). Schematic of the gap junction inhibitor CBX blocking Cx26 channels (C). Representative images of regions between IPCs and OPCs in CBX‐inhibited cochlea and control cochlea (D). Schematic diagram of continuous IP injection of CBX (E). Representative images of OC regions from cochlear longitudinal frozen section in CBX‐treated and control cochleae (F). Continuous CBX treatment do not affect the separation. Schematic of the continuous intraperitoneal (IP) CBX injection protocol (G). Representative images of OC regions from the cochlear frozen section of CBX‐inhibited cochlea and control cochlea (H). Continuous IP of CBX or a single injection of CBX through RWM do not affect the TC development. Schematic diagram of continuous IP injection of CBX until P18 (I). ABR test showed that continuous CBX IP treatment do not cause mice hearing loss (J, n = 6).

2.2. Cx26 Aggregates in the Cell Nucleus In Vivo and In Vitro

Several experiments were conducted to demonstrate that Cx26 can localize in the nucleus. We prepared subcellular (cytosolic and nuclear) fractions from mouse cochleae, and analyzed them for Cx26 by Western blotting. As shown in Figure 2A, immunoreactive bands of Cx26 corresponding to approximately 25 kDa were detected in nuclear fractions. To further examine the subcellular localization in cochlear Cx26 of support cells (SCs), we use antibodies to label Cx26 and use DAPI to label the nucleus. Immunofluorescence staining revealed two distinct patterns of Cx26 localization in mouse cochleae (Figure 2B–F). One pattern, the Cx26 staining localized to sites of intercellular contact. The other pattern was characterized by strong intranuclear Cx26 staining (Figure 2E; Yellow arrow). To examine the subcellular localization of Cx26 in other cell lines, we used BxPC‐3 cells with endogenous Cx26 expression, and HEK293T and HeLa cells stably expressing human Cx26 (HEK293T‐Cx26, HeLa‐Cx26). Unexpectedly, Cx26 was also observed localized to the nuclei in both BxCP‐3, 293T‐Cx26, and Hela‐Cx26 cell lines (Yellow arrow; Figure 2G–L). In addition, Cx26 could also be detected in nuclear fractions separated from BxPC‐3 cell, Hela‐Cx26 cell, and HEK293T‐Cx26 cell. GAPDH and H3‐histone were used as cytoplasmic marker and nuclear marker, respectively (Figure 2M,N). Sanger sequencing showed no mutation was detected in the exon region of the GJB2 gene in BxPC‐3 cells (Table 1). These experiments show independent evidence for nuclear localization of wild‐type Cx26 (WT‐Cx26) in the cochlea and other cell line from different tissue sources. The nuclear localization patterns of Cx26 and its truncated mutants during the mitotic cycle can be found in the supplementary materials (Figures S2–S5).

FIGURE 2.

Cx26 protein expression pattern and subcellular localization in vivo and vitro. Cx26 subcellular localization in cochlear cytoplasmic and nuclear fractions of WT mice was detected by immunoblotting (A). Images of 3D reconstruction of the OC structure based on Z‐axis scanning using a nikon confocal laser microscope (B). Immunofluorescence staining showed Cx26 protein expression and subcellular localization elected cross section in cochlear of WT mice (C,D). Cx26 staining localized to sites of intercellular contact between SCs and regions in the SC nucleus. Cx26 staining (E) and DAPI staining (F) in an elected cross‐section. Phalloidin is used to labeled the cytoskeleton (Red). DAPI is used to labeled the cell nucleus (Blue). Green: Cx26. Cx26 protein expression and subcellular localization in three cell lines (G: BxPC‐3; H: Hela‐Cx26; L: HEK293T‐Cx26 cell lines) were detected by immunofluorescence staining by capturing single confocal slice imgae. Cx26 protein expression and subcellular localization in three cell lines (G: BxPC‐3; H: Hela‐Cx26; L: HEK293T‐Cx26 cell lines) were detected by immunofluorescence staining. Cx26 protein could been observed to aggregate in the nucleus in both BxPC‐3, Hela, and HEK293T. Yellow arrow: Cx26 aggregating in the nucleus. Scale bar = 50 µm. Expression of Cx26 in cytoplasmic and nuclear fractions of three cell lines (J: BxPC‐3; K: Hela‐Cx26; L: HEK293T‐Cx26 cell lines) was detected by immunoblotting. Bands of Cx26 could been detected in nucleus fractions of both BxPC‐3, Hela, and HEK293T. M‐P:Cx26 subcellular localization in cytoplasmic and nuclear fractions form P5 mice cochleae (M), BxPC‐3 cell (N), Hela cell transfected with Cx26 (O), and HEK293T cell transfected with Cx26 (P) was detected by immunoblotting. GAPDH was used as a cytoplasmic marker. Lamin B was used as a nuclear marker.

TABLE 1.

Sanger sequencing results of GJB2 gene in BxPC‐3 cells.

Site on GJB2 gene(Homo sapiens)	Reference base	Dertected base
168	G	G/A
948	C	C/T
1247	A	G
1477	G	G/A
2129	G	G/C
2133	A	G
2138	G	G/A
2179	A	A/T
2228	G	G/A
2263	A	A/G
2356	T	T/C
2414	G	G/A
2766	A	A/T
2931	G	G/A
3004	C	A
3324	C	C/T
4123	T	T/C
5106	G	G/T
5191	G	A
5316	T	C

No SNP were detected in the coding DNA sequence(CDS) of GJB2 gene in BxPC‐3 cells.

2.3. Cx26 Directly Binding to Nuclear Proteins and Genomic DNA

To investigate the function of Cx26 in the nucleus, immunoprecipitation‐Mass spectrometry analysis (IP‐MS) and chromatin immunoprecipitation‐sequence (ChIP‐seq) were performed (Figure 3A). IP‐MS identified potential Cx26‐interacting proteins (Figure 3B–D; Table 2), and these results suggest that Cx26 is involved in cellular processes such as chromatin assembly, transcription and replication, RNA splicing and mRNA processing, hnRNPs, and primary miRNA processing (Figure 3D). Chromatin immunoprecipitation‐sequence (ChIP‐seq) experiments were performed to verify the direct interaction between Cx26 and genomic DNA (Figure 3E–G). Cx26 could be distributed in different genomic functional regions (Figure 3E). 9.28% of Cx26 peaks were located in promoter regions containing transcription start sites (TSS), while 1.52% were directly overlapping TSS regions. Figure 3F showed an obvious Cx26 peak near the TSS region. Potential binding DNA sequences were predicted using HOMER software, and the top five motifs were listed in Table 3. We used Gene Ontology (GO) enrichment analysis to reveal the specific biological functions and signaling pathways associated with the identified genes using ChIP‐seq (Figure 3G). We found that these signaling pathways were mainly associated with DNA replication, helicase activity, small GTPase‐mediated signal transduction, and cytoskeleton organization. These results suggested that Cx26 could interact with nuclear proteins and genomic DNA.

FIGURE 3.

Cx26 directly interacts with nuclear proteins and genomic DNA. IP‐MS (sample from mice cochleae) and ChIP‐seq (sample from HEK293T‐Cx26 cell) were performed to identify nuclear proteins and genomic DNA that directly interacts with Cx26 (A). Volcano plot (B), Venn Diagram (C), and analysis of protein potential functions (D) based on the results of proteins co‐immunoprecipitation with Cx26 using IP‐MS analysis. There are 12 proteins were identified to co‐immunoprecipitate with Cx26, and 5 proteins have been reported to be involved in nuclear function. The potential functions of the 5 proteins interact with Cx26, including RNA splicing and mRNA processing, chromatin transcription, and replication. Distribution of Cx26 peaks in genomic functional regions (E). TSS: transcription start sites. 9.28% and 1.52% of Cx26 peaks were identified in promoter‐TSS regions and TSS regions, respectively. Schematic visualization of Cx26 ChIP‐seq signal and Heatmap showed the distribution of Cx26 Peaks in upstream and downstream regions of TSS (F). A high peak of Cx26 could be observed near the TSS regions of the Schematic visualization of Cx26 ChIP‐seq signal. The GO analysis showed the specific biological functions and signaling pathways associated with the identified genes co‐immunoprecipitation with Cx26 (G).

TABLE 2.

IP‐MS detected the protein co‐precipitated with Cx26.

Identified proteins	Potential functions	Subcellular localization
WDHD1	Function as adaptor/regulatory modules in signal transduction, pre‐mRNA processing and cytoskeleton assembly	Nucleus, nucleoplasm
RA1L3	Predicted to be involved in mRNA splicing, via spliceosome.	Nucleus
HNRPQ	Regulate alternative splicing, polyadenylation, and other aspects of mRNA metabolism and transport	Nucleus; Cytosol; Endoplasmic reticulum
ROA2	Complex with heterogeneous nuclear RNA (hnRNA)	Nucleus; Extracellular
HNRPK	Complex with heterogeneous nuclear RNA (hnRNA)	Cytoplasm; Nucleus; Extracellular
HNRPU	Function in the formation of ribonucleoprotein complexes	Nucleus; Cytosol
TMPS9	Function as a membrane‐bound type II serine polyprotease	Cell membrane
MIA40	Function as a component of human mitochondria	Mitochondrion intermembrane space
NEO1	Function as multi‐functional cell surface receptor regulating cell adhesion	Cell membrane
GCC1	Play a role in the organization of trans‐Golgi network subcompartment involved with membrane transport	Cytoplasm, Golgi apparatus membrane, Peripheral membrane protein
RBM14	Functions as a general nuclear coactivator, and an RNA splicing modulator	Nucleus
SRRT	Involved in primary miRNA processing	Nucleus, nucleoplasm

A total of 8 nuclear localization proteins co‐precipitated with Cx26 were identified, and the potential functions of these proteins were analyzed and listed.

TABLE 3.

Top 5 predicted Cx26 binding motifs with the most significant P‐values.

Rank	Motif	p‐value
1		1e‐11
2		1e‐10
3		1e‐10
4		1e‐10
5		1e‐8

2.4. Cx26 Funtion as a Direct Transcriptional Regulator of Tetraspanin 5, 15, and 17

To identify potential genes transcripitional regulated by Cx26 during cochlea development, differential gene expression in the cochleae between postnatal day (P)5 control group mice and P5 Cx26 cKO mice was screened using Spatial Transcriptomics (ST; Figure 4A–E). We constructed two spatial RNA‐seq libraries of a P5 control group mouse cochlea and a P5 Cx26 cKO mouse cochlea. Figure 4A,B showed the spatial characteristics of number counts of transcription (nCount_RNA) and genes(nFeature_RNA) in the control cochlea and in Cx26 cKO cochlea, respectively. We found that the hotspots of gene and transcript expression are concentrated in the regions of stria vascularis (SV), basement membrane, and Rosenthal's canal in both control cochlea and Cx26 cKO cochlea, and the number of genes and gene transcript levels in both groups of cochleae are on the same order of magnitude. A total of 13 cell clusters in the control cochlea (Figure 4C) and 12 cell clusters (Figure 4D) in the Cx26 cKO cochlea were identified after dimensional reduction and using visualization of Uniform Manifold Approximation and Projection (UMAP). The spatial arrangement of cell groups was displayed in control cochlea and Cx26 cKO cocolea slices to provide intuitive visualization. Cells at OC and SV regions (Cx26 highly expressed) in the control cochlea correspond to cluster 6, and the Top3 high expression gene of cluster 6 are Ibsp, Ptn, and Smpd3 (Figure 4E). In Cx26 cKO cochlea, cells at OC and SV regions were correspond to cluster 1, and the Top3 high expression gene of cluster 1 are Acp5, Ctsk, and Mmp9 (Figure 4F). The top 3 high expression gene of each cluster were shown in the representative heatmap (Figure 4E,F). These results of ST demonstrated the effect of Cx26 knockout on differential gene expression in the cochlea at the critical time of cochlear TC development. While the throughput and resolution constraints of the ST platform limited our ability to detect genes with subtle expression changes, specifically the finite capacity for analysis units (bins) and probes, the ST data provided critical spatial context by revealing that the most significant transcriptional alterations occurred specifically in the sensory epithelial region where Cx26 is expressed and the tunnel of Corti forms.

FIGURE 4.

Characterization of DEGs and cell clusters in Cx26 cKO cochlea and WT cochlea using Spatial transcriptomics. ncounts plot (gene expression statistics; (A) and nfeature plot (gene number statistics; (B) and their spatial distribution of Cx26 cKO cochlea and WT cochlea (n=1 for each group). UMAP classifying 13 clusters in WT cochlea (C) and 12 clusters in Cx26 cKO cochlea (D), and their spatial distribution based on the transcriptomes of overall gene expression relationships. Different cell clusters are color‐coded. Heatmap showing the top3 highly expressed genes in each cluster identified (E, F).

To detailed identify potential genes transcripitional regulated by Cx26 during cochlea development, we screened the differential gene expression in cochleae of Cx26 cKO mice and cochleae of WT mice using RNA‐sequence (Figure 5A–J). RNA‐seq results reveals genes with statistically significant expression changes (TOP100) and the relevant pathways (anlyzed by GO barplot analysis) in the cochlea of Cx26 cKO mice. In the RNA‐seq results, the transcripts level of TSPAN14 and TSPAN5 genes showed statistically significant changes in the cochlea of Cx26 cKO knockout mice when compared with the control group. TSPAN14 and TSPAN5 are members of the tetraspanin superfamily. Together with TSPAN10, TSPAN15, TSPAN17, and TSPAN33, they form the TSPANC8 subfamily, which regulates ADAM10 exit from the endoplasmic reticulum and its substrate selectivity [40]. Therefore, qPCR was used to detect the relative TspanC8 gene mRNA level in cochleae of Cx26 cKO mice and in cochleae of WT mice. Compared with the control group, the relative Cx26 mRNA level reduced 72.66 ± 33.85% (P<0.0001) in Cx26 cKO cochleae, suggesting high KO efficiency of Cx26 in Cx26 cKO cochleae. And the relative mRNA expression level of TSPAN5 (Figure 5E; 31.31 ± 27.83%, P = 0.0018), TSPAN15 (Figure 5H; 24.40 ± 32.70%, P = 0.0317) and TSPAN17 (Figure 5I; 30.56 ± 34.55%, P = 0.0103) was significantly decreased in Cx26 cKO mice cochleae at P5 compared with WT mice, respectively. However, the relative mRNA expression level of other members of the TSPANC8 family, such as TSPAN10 (Figure 5F), TSPAN14 (Figure 5G), and TSPAN33 (Figure 5J), in the Cx26 cKO mice group showed no statistical significance differences (P > 0.05) when compared to the control group. These results suggest that cochlea Cx26 deficiency reduces transcription of TSPANC8 mRNA (especially TSPAN5, TSPAN15, and TSPAN17) during cochlea development. Moreover, it was found that TSPAN5, TSPAN14, TSPAN15, and TSPAN17 are the main components of TSPANC8 expressed in the P5 cochleae (Figure 5K) by comparing the relative mRNA levels of TspanC8 in P5 WT mice cochlea. This suggests that Cx26 may be involved in cochlear development by regulating the transcription levels of the main component of TSPANC8 in the cochlea. Besides, ChIP‐seq analysis reveals the direct binding of Cx26 to genomic DNA at the TSPAN15 locus, while ChIP‐qPCR demonstrated the direct binding of Cx26 to the promoter of TSPAN5, TSPAN15, and TSPAN17, respectively (Figure S6). These results suggest the potential direct regulatory role of Cx26 in the transcription of the TSPANC8 gene in the mice cochlea.

FIGURE 5.

Cx26 mediates TSPANC8 mRNA expression in mice cochleae. Cx26^flox/flox; Rosa26CreER mice were injected subcutaneously with TMX at P0 and P1, and were sacrificed at P5 to prepare RNA from cochleae (A, n = 4 for each group). Volcano plot based on genes differentially expressed in Cx26 cKO and WT mice using RNA‐Seq analysis (B). GO barplot analysis of the relevant pathways enriched in Cx26 cKO and WT mice using RNA‐seq (C). RNA‐seq experiment includes 3 biological replicates. The relative mRNA expression of Cx26 (C), TSPAN5 (D), TSPAN10 (E), TSPAN14 (F), TSPAN15 (G), TSPAN17 (H), and TSPAN33 (I) in cochlea of Cx26 cKO mice were detected by qPCR (n ≥ 6 for each group). Compared to control group, the relative mRNA expression of Cx26 (C), TSPAN5 (D), TSPAN15 (G), TSPAN17 (H) in Cx26 cKO group reduced 72.66 ± 33.85% (P < 0.0001), 31.31 ± 27.83%(P = 0.0018), 24.40 ± 32.70% (P = 0.0317) and 30.56 ± 34.55% (P = 0.0103), respectively. There are no statistical significance differences (P > 0.05) of the relative mRNA expression of TSPAN10 (E), TSPAN14 (G), TSPAN33 (J) between the Cx26 cKO group and the control group. The relative mRNA expression of TSPANC8 in the cochlea of P5 WT mice (J). n.s: P > 0.05. ^*: P < 0.05. ^**: P < 0.01.****: P < 0.00001. Statistical significance difference evaluated by Student's t‐test.

Further experiments demonstrated the control effect of Cx26 on transcriptional regulation of the main component of TSPANC8 (TSPAN5, TSPAN15, and TSPAN17) in cochlea. Different siRNA sequences were generated and transfected into BxPC‐3 cell to knock down the Cx26 expression. The siRNA sequence with the highest knockdown efficiency (reduce 84.56 ± 45.10%, P < 0.0001) in BxPC‐3 cells was selected for subsequent experiments (Figure 6A). Cx26 knockdown significantly reduced mRNA levels of TSPAN5 (Figure 6B, 48.73 ± 45.29%, P = 0.0013), TSPAN15 (Figure 6C, 50.59 ± 5.06%, P < 0.0001) and TSPAN17 (Figure 6D, 57.18 ± 7.79%, P < 0.0001), indicating that Cx26 deficiency downregulates TspanC8 transcription in vitro. We next examined the regulatory effect of increased Cx26 expression on TSPANC8 mRNA transcription. A GJB2 recombinant vector was transfected into HEK293T and HeLa cells to overexpress Cx26. Cx26 level increased 539.72 ± 34.21% (P < 0.0001) in HEK293T‐Cx26 cell and 865.05 ± 22.06% (P < 0.0001) s in Hela‐Cx26 cells compared to the control group, respectively. In Hela‐Cx26 cell, the mRNA expression level of TSPAN5, TSPAN15, and TSPAN17 increased 361.46 ± 67.21% (Figure 6F; P < 0.0001), 300.27 ± 45.06% (Figure 6G; P < 0.0001), and 425.73 ± 39.74% (Figure 6H; P < 0.0001), respectively. In HEK293T‐Cx26 cell cell, the mRNA expression level of TSPAN5, TSPAN15, and TSPAN17 increased 368.74 ± 65.55% (Figure 6J; P < 0.0001), 567.37 ± 78.00% (Figure 6K; P < 0.0001), and 28.52 ± 33.44% (Figure 6L; P = 0.1239), respectively. These results demonstrate that Cx26 regulates TspanC8 transcription in vitro, consistent with our in vivo findings.

FIGURE 6.

Cx26 directly regulates TSPAN5, Tpsan15, and TSPAN17 gene transcription. Efficiency of siRNA used to silence the expression of Cx26 in BxCP‐3 cell lines was detected by qPCR (A). The relative mRNA expression of Cx26 reduce 84.56 ± 45.10% (P < 0.001) after Cx26 siRNA treatment in BxPC‐3 cell. mRNA expression change of TSPAN5, Tpsan15 and TSPAN17 after Cx26 knockown (KD) using Cx26 siRNA in BxCP‐3 cell lines were detected by qPCR (B–D). Compared to the control group, the relative mRNA expression of TSPAN5 (B), TSPAN15 (C), TSPAN17 (D) in Cx26 KD group was reduced 48.73 ± 45.29% (P = 0.0013), 50.59 ± 5.06% (P < 0.0001), and 57.18 ± 7.79% (P < 0.0001), respectively (n ≥ 3 for each group). mRNA expression change of Cx26, TSPAN5, Tpsan15 and TSPAN17 after Cx26 overexpression in Hela cell lines were detected by qPCR(F‐H, n ≥ 8 for each group). Compared to control group, the relative mRNA expression of Cx26 (F), TSPAN5 (F), TSPAN15 (G) and TSPAN17 (H) in Cx26 overexpression group increased 539.72 ± 34.21% (P < 0.0001), 361.46 ± 67.21% (P < 0.0001), 300.27 ± 45.06% (P < 0.0001) and 425.73 ± 39.74% (P < 0.0001), respectively. mRNA expression change of Cx26, TSPAN5, Tpsan15 and TSPAN17 after Cx26 overexpression in HEK293T cell lines were detected by qPCR (I–L, n ≥ 8 for each group). Compared to control group, the relative mRNA expression of Cx26 (F), TSPAN5 (F), Tpsan15 (G) and TSPAN17 (H) in Cx26 overexpression group increased 865.05 ± 22.06% (P < 0.0001), 368.74 ± 65.55% (P < 0.0001), 567.37 ± 78.00% (P < 0.0001) and 28.52 ± 33.44% (P = 0.1239), respectively. Dual‐luciferase reporter assay was used to measure the effects of Cx26 on promoters of TSPAN5, Tpsan15, and TSPAN17 (M‐O, n = 4 for each group). The relative luciferase activities of TSPAN5 promoter reporter vector (M), Tpsan15 promoter reporter vector (O), and TSPAN17 promoter reporter vector (N) increased 401.47 ± 36.57% (P = 0.0002), 177.54 ± 54.77% (P = 0.0262), and 64.24 ± 29.63%(P = 0.0349) after Cx26 overexpression in HEK293T cell, respectively. mRNA expression change of TSPAN5, Tpsan15, and TSPAN17 after Cx26 p.leu79CysfsTer3 overexpression in Hela cell lines were detected by qPCR (P–R, n ≥ 8 for each group). There are no statistical significance differences (P > 0.05) of the relative mRNA expression of TSPAN10 (P), TSPAN14 (Q), TSPAN33 (R) between Cx26 p.leu79CysfsTer3 overexpression group and control group. mRNA expression change of TSPAN5, Tpsan15 and TSPAN17 after Cx26 p.leu79CysfsTer3 overexpression in HEK293T cell lines were detected by qPCR (S–U, n ≥ 8 for each group). There are no statistical significance differences (P > 0.05) of the relative mRNA expression of TSPAN10 (S), TSPAN14 (T), TSPAN33 (U) between Cx26 p.leu79CysfsTer3 overexpression group and the control group. n ≥ 3.^*: P < 0.05. ^**: P < 0.01.^****: P < 0.0001. Statistical significance difference evaluated by Student's t‐test.

Moreover, to identify the control mechanism of Cx26 on transcriptional regulation of the TSPANC8, we used a dual‐luciferase reporter gene assay to detect whether Cx26 could control the promoter activities of TSPANC8. TSPAN5, TSPAN15 and TSPAN17. Promoter reporter vectors for TSPAN5, TSPAN15, and TSPAN17 were individually co‐transfected with either a Cx26 overexpression vector or an empty vector (negative control) into HEK293T cells (Figure 6M–O). The assays demonstrated that Cx26 overexpression significantly enhanced the promoter activities of TSPAN5 (Figure 6M; 401.47 ± 36.57%, P = 0.0002), TSPAN15 (Figure 6N; 177.54 ± 54.77%,P = 0.0262) and TSPAN17 (Figure 6O; 64.24 ± 29.63%, P = 0.0349) compared to negative control group. Collectively, these results support a role in which Cx26 functions as a direct transcriptional regulator of TspanC8 genes.

2.5. Cx26 Regulates the Structural Development of the Tunnel of Corti by Transcriptionally Controlling the TSPANC8/ADAM10 Complex

TSPANC8 tetraspanins are regulators of a disintegrin and metalloproteinase 10 (ADAM10) intracellular trafficking and enzymatic maturation [40]. Many studies have highlighted the role of TSPANs as extrinsic factors in regulating maturation, transport, cleavage activity, and an essential subunit of an ADAM10 scissor complex [41]. Here, to identify the results of Cx26 transcriptional regulation on TSPANC8 during cochlea TC structural development, we further investigated the association between cochlea Cx26 expression and ADAM10 maturation. In P5 mouse cochleae, TSPAN5, TSPAN15, and TSPAN17 each co‐immunoprecipitated with ADAM10, respectively (Figure 7A–C). This is consistent with the results observed by other researchers in the kidneys, brain, blood system, or other tissues [40, 41, 42]. We also found that relative Cx26 expression level reduce 34.89 ± 13.12% (P = 0.0007; Figure 7D,F) in the cochleae of P5 Cx26 cKO mice, leading to a 47.01 ± 37.16% (P = 0.006) descrease of mature ADAM10 (Figure 7E,G) compared to the control group. The precursor‐ADAM10 expression level increased 34.60 ± 13.79% (P = 0.0046) in the cochleae of P5 Cx26 cKO mice, while there are no statistical significance differences (P > 0.05) of the relative all‐form ADAM10 protein expression between the Cx26 cKO group and the control group. These results suggests that cochlea Cx26 deficiency could lead to ADAM10 maturity disorder via down‐regulating the transcription of TSPANC8. Immunofluorescence staining of ADAM10 in the OC region of Cx26 cKO mice also showed a 71.71% ± 44.98 (P < 0.0001) decrease compared to the control group (Figure 7J–L). In addition, there was no statistically significant difference (P > 0.05; Figure 7M) in ADAM10 mRNA levels in the cochlea of Cx26 cKO mice. These results suggest that Cx26 may control the maturation and functional processes of ADAM10 by regulating the transcription of TSPAN C8 during cochlea development, and Cx26 deficiency in the mouse cochlea leads to a decrease of mature ADAM10 in cochlea.

FIGURE 7.

Cochlea Cx26 deficiency leads to a decrease in mature ADAM10 and TC structural developmental disorders. Co‐immunoprecipitation showing biochemical interaction in cochlear between TSPAN C8(TSPAN5 (A), Tpsan15 (B), and TSPAN17 (B) and ADAM10, respectively. Expression of Cx26 in cochlear of Cx26 cKO mice and WT mice at P5 were detected by western blot (D). Expression of ADAM10 in cochlear of Cx26 cKO mice and WT mice at P5 were detected by western blot (E; n = 4). Statistical analysis chart of protein expression change of Cx26 (F), mature‐ADAM10(G), precursor‐ADAM10 (H) and all forms of ADAM10 (I) in cochlear of Cx26 cKO mice and WT mice (F–I). Compared to control group, the relative protein expression of Cx26 (D, n = 4) and mature‐ADAM10 (E, n = 4) in the Cx26 cKO group reduced 34.89 ± 13.12% (P = 0.0007) and 47.01 ± 37.16% (P = 0.006), respectively (n = 4 for each group). The relative protein expression of precursor‐ADAM10(H) in Cx26 cKO group increased 34.60 ± 13.79% (P = 0.0046). There is no statistical significance differences (P > 0.05) of the relative all‐form ADAM10 protein expression between Cx26 cKO group and the control group. Representative images of ADAM10 around the OC region in the basal turn of the cochlear in WT mice (J) and Cx26 cKO mice (K) at P5. Statistical analysis chart of relative ADAM10 expression around the OC region in the basal turn of the cochlear in Cx26 cKO mice and WT mice at P5 (L). Immunofluorescence staining shows that ADAM10 expression in OC regions reduced 71.71% ± 44.98(P < 0.0001). M: mRNA expression change of ADAM10 in cochleae in Cx26 cKO mice at P5. There is no statistical significance differences (P > 0.05) of the relative mRNA expression of ADAM10 between Cx26 cKO group and the control group. Inhibition of ADAM10 in cochleae leads to delayed destruction of P‐cadherin and the opening process of TC (N‐O). The targeted inhibitor GI254023X of ADAM10 or saline(control group) were injected into the left cochlea via the round window membrane at P4 (N). Inhibition of ADAM10 in cochleae leads delayed opening process of TC (O, n = 4 for each group). Scale = 10 µm. The yellow curve was used to mark TC. ^*: P < 0.05. ^**: P < 0.01.^****: P < 0.0001. ns:P > 0.05. Statistical significance difference evaluated by Student's t‐test.

Further experiments were conducted to investigate the effects of ADAM10 maturation disorders caused by Cx26 deficiency on cochlear TC structural development. ADAM10 inhibitor GI254023X(2ul volume, dissolved in DMSO, 1 µm) was injected into the perilymph of WT mice through the unilateral (left) round window at P4 (Figure 7N,O), to obstruct the function of ADAM10 in the cochlea. In the frozen sections of right cochlea (without GI254023X treatment), there is obvious space between IPCs and OPCs at the basal turn of the membranous cochlea duct (Figure 7O), while no formed TC could be observed in all turns of cochlea after GI254023X treated. Similar results can also be observed in flattened cochlear preparations. We found that the lateral membranes of inner pillar cells (IPCs) and outer pillar cells (OPCs) were still adjacent, and no formed TC could be observed in all turn of cochlea after ADAM10 was inhibited (Figure 8A,B). In the contralateral untreated cochleae of the same mouse, a local IPC and OPC detachment and an obvious space between adjacent IPCs and OPCs could be observed (Figure 8A,B). These results indicate that the separation of IPCs and OPCs could be control by the TSPANC8/ADAM10 complex. It should be noted that a single dose of GI254023X treatment merely delays rather than permanently prevents the separation of PCs, and TC finally fromed in subsequent days later. These results indicate that the separation of IPCs and OPCs depends on ADAM10. Consistent with our previous report, failure of separation of IPCs and OPCs could also be seen in Cx26 cKO mice (Figure 8C,D), suggesting that the disorder of pillar cells separation in the Cx26 deficiency cochlea results from the decrease of the TSPANC8/ADAM10 transcriptional regulated by Cx26.

FIGURE 8.

Delayed destruction of adhesive junction and separation of IPCs and OPCs in ADAM10‐inhibited cochlea and Cx26 cKO cochlea. The targeted inhibitor GI254023X of ADAM10 or saline(control group) were injected into the left cochlea via the round window membrane and observed at P5(A). Representative images of P‐cadherin between IPCs and OPCs in ADAM10‐inhibited cochlea and control cochlea(B). In the basal turn in mice cochlear with ADAM10 inhibitor(GI254023X) treatment, the lateral membranes at the middle regions of IPCs and OPCs were still adjacent, and P‐cadherin remains at the connected apical and middle regions between IPCs and OPCs. But in control cochleae, IPC and OPC have separated already, and an obvious space between adjacent IPCs and OPCs could be observed, and P‐cadherin on the lateral membranes in the middle region of OPCs have been disrupted (n = 4 for each group). Scale=10 µm. Cx26flox/flox; Rosa26CreER mice were injected subcutaneously with tamoxifen(TMX) at P0 and P1 and sacrificed at P5(C). Representative images of P‐cadherin between IPCs and OPCs in the Cx26 cKO cochlea and control cochlea(D). Consistent with ADAM10‐inhibited cochlea, the lateral membranes at the middle regions of IPCs and OPCs were still adjacent in Cx26 cKO cochlea, and P‐cadherin remains at the connected apical and middle regions between IPCs and OPCs (n = 4 for each group). Scale=10 µm. Expression of P‐cadherin in cochleae of Cx26 cKO mice and WT mice at P5 were detected by western blot(E, n = 4 for each group). Statistical analysis chart of protein expression change of P‐cadherin in cochleae of Cx26 cKO mice and WT mice(F). There is no statistical significance differences(P > 0.05) of the relative P‐cadherin protein expression between Cx26 cKO group and the control group. Co‐immunoprecipitation showing biochemical interaction between P‐cad and ADAM10 at P5(G). ns:P > 0.05. Statistical significance difference evaluated by Student's t‐test. No significant association was observed between E‐cadherin and N‐cadherin and TC opening in the cochlea of Cx26 cKO mice and WT mice(See also Figures S5 and S6).

Adhesion junctions, which were reported to be substrates of TSPANC8/ADAM10 complex, are highly dynamic intercellular interactions mediating cell–cell adhesion and separation to maintain the 3D structure of tissues [42]. We and Defourny et al. have observed that the separation of adjacent pillar cells and the opening of TC companied with the destruction of adhesion junctions, consisted of cadherin [26, 43]. In this study, several experiments were conducted to verify the association between the separation of PCs, ADAM10, and adhesive junctions during cochlea development. We examined the expression patterns of three classical cadherin, E‐cadherin, N‐cadherin, and P‐cadherin, between IPCs and OPCs at P5 in Cx26 deficiency group and the control group, respectively. We found that in the control group, PCs have separated and P‐cadherin have been disrupted on the lateral membranes in the middle region of OPCs, while in Cx26 deficiency group, adhesive junctions consisted of P‐cadherin still existed between the unseparated PCs (Figure 8D). These results suggest a correlation between the disruption of P‐cadherin and the separation of IPCs and OPCs. The failure of separation of PCs and formation of TC in the Cx26 deficiency cochlea may result from delayed disruption of P‐cadherin. We also found that cochlea ADAM10 inhibited by GI254023X results in delay destruction of P‐cadherin (Figure 8B). Adhesive junctions consisted of P‐cadherin still remian on the adjacent lateral cytomembrane in the middle region of PCs cochlea ADAM10 was inhibited, while in control cochleae, while few P‐cadherin could be observed on the cytomembrane of OPCs which were already separated with IPCs. However, there was no statistically significant difference(P > 0.05) in the protein expression level of P‐cadherin in the cochlea of Cx26 knockout mice compared to the control group, which may be due to the fact that the destruction of P‐cadherin only occurred locally in the OC region (Figure 8E,F). We further validated the interactions between ADAM10 and P‐cadherin during morphological development of TC, and we found that ADAM10 and P‐cadherin co‐immunoprecipitation in P5‐P6 mouse cochlear extract (Figure 8G). These findings demonstrate that ADAM10‐mediated cleavage of P‐cadherin between IPCs and OPCs drives the separation of PCs and formation of TC. Cx26 regulates TC structural development by regulating the destruction of the TSPANC8/ADAM10 complex to adhesive junctions between pillar cell. Cochlea Cx26 deficiency induced TC structural development disorders via down‐regualting the transcription of TSPANC8 and maturation of ADAM10.

In addition, we found that E‐cadherin remains on the lateral membranes in the middle region of OPCs in the Cx26 deficiency group and control group, regardless of whether pillar cells are separated or not (Figure S9). N‐cadherin is mainly expressed in the greater epithelial ridge region and in inner hair cells (IHCs), and few N‐cadherin was observed in PCs (Figure S10). These results suggest that E‐cadherin and N‐cadherin do not involve in Cx26 deficiency related TC structural development disorder.

2.6. Regulating ADAM10 Expression to Rescue Cx26 Deficiency Induced TC Structural Developmental Disorders and Hearing Loss

There are currently no available targeted activators for ADAM10. However, ADAM10 promoter contains two putative retinoic acid binding sites, ligands of the potential retinoid X receptors (RXR) dimerization partner such as thyroid receptor (TR) could regulate ADAM10 expression by stimulating the ADAM10 promoter [44, 45]. We investigated whether Triiodothyronine (T3), the active form of thyroid hormone, can promote TC opening in Cx26 deficient mice by enhancing ADAM10 expression [46]. Compared to control group, the levels of FT3 in the serum of T3 treated mice significantly increased (>500 pmol/L vs. 46.38 ± 0.84 pmol/L; P < 0.0001; Figure 9A,B) after 24 h T3 treatment, and the levels of FT4 in the serum of T3 treated mice also increased(176.37 ± 40.36 pmol/L vs. 14.37 ± 4.55 pmol/L; Figure 9C), while the levels of TSH in the serum no significantly difference (0.106 ± 0.017mIU/L vs. 0.099 ± 0.011mIU/L; P > 0.05; Figure 9D). T3 treatment also increased 19.93 ± 13.30% (P = 0.0459) ADAM10 mRNA level in cochleae of T3 treated mice (Figure 9E) compared to the control group, suggesting that T3 treatment could increase the transcription of ADAM10. To clarify whether T3 can restore mature ADAM10 levels in Cx26 deficient cochlea, T3 was subcutaneously injected into Cx26 cKO mice at P5. The relative protein expression levels of mature ADAM10 and all forms of ADAM10 in the cochleae of Cx26 cKO mice were restored after 96 h T3 treatment to the similar levels as the control group (P > 0.05; Figure 9F–J). These results indicate the potential of T3 [47], as a drug to increase ADAM10 expression, in rescuing Cx26 deficiency induced TC structural development disorders and hearing loss.

FIGURE 9.

T3 treatment promoted P‐cad destruction and separation of IPCs and OPCs in the cochlea of Cx26 cKO mice by increasing ADAM10 expression. T3 or saline(control group) were subcutaneously injected into P4 mice pups, and the free T3(FT3), free T4(FT4) and TSH level in serum of this mice were detected 24 h after T3 subcutaneous injection(A). Compared to the control group, the level of FT3 in serum of T3 treated mice were significantly increased (>500 pmol/L vs. 46.38 ± 0.84 pmol/L; P < 0.0001; B). Compared to control group, the level of FT4 in serum of T3 treated mice was significantly increased(176.37 ± 40.36 pmol/L vs. 14.37 ± 4.55 pmol/L; C). D: There is no statistical significance differences(P > 0.05) of the relative TSH level in serum between the T3 group and the control group (0.106 ± 0.017mIU/L vs. 0.099 ± 0.011mIU/L; D). Expression of ADAM10 mRNA in the cochlear of mice with T3 or saline treatment detected by qPCR(E). Compared to the control group, the relative protein expression of ADAM10(E) in the T3 group increased 19.93±13.30% (P = 0.0459). Cx26flox/flox; Rosa26CreER mice were injected subcutaneously with TMX at P0 and P1 and injected subcutaneously with T3 at P5, and finally were sacrificed at P9(F). Expression of ADAM10 protein in the cochlear of Cx26 cKO mice with T3 treatment(Cx26 cKO+T3 group) and WT mice(control group) detected by Western bloting(G). Statistical analysis chart of protein expression change of mature‐ADAM10(H), precursor‐ADAM10(I), and all forms of ADAM10 in the cochlear of Cx26 cKO+T3 group and control group. There are no statistical significance differences (P > 0.05) in the relative protein expression of mature‐ADAM10(H), precursor‐ADAM10(I), and all forms of ADAM10(J) between Cx26 cKO+T3 group and the control group. Representative images of P‐cadherin destruction between IPCs and OPCs in cochlea flattened preparations of Cx26 cKO+T3 mice(E). Scale=10 µm. Immunofluorescence staining shows that there are obvious spaces between IPCs and OPCs at the basal turn in cochleae of the P9 control group and Cx26 cKO+T3 group, while IPCs and OPCs remain adjacent in P9 Cx26 cKO mice treated with saline (Cx26 cKO group). P‐cadherin on the lateral membranes in the middle region of OPCs in the Cx26 cKO+T3 group has been disrupted. Representative images of TC open in the cochlear frozen section of Cx26 cKO+T3 mice(F). Immunofluorescence staining shows that the formed TC could be clearly observed in the control group and of Cx26 cKO+T3 group, while no TC structure can be observed in the Cx26 cKO. Scale=10 µm. RWM injection of T3 into the left mouse cochleae via the round window membrane can also promote TC opening in Cx26 cKO mice(See also Figure S7). The yellow curve was used to mark TC. ns:P > 0.05. ^*:P < 0.05. ^**:P < 0.01. ^****:P < 0.0001. Statistical significance difference evaluated by Student's t‐test.

To verify whether recovery express of ADAM10 induced by T3 can rescue TC structural development disorders in Cx26 deficiency cochleae, Cx26 cKO mice pups were given a single dose of T3 via subcutaneous injection at P5 (Figure 9F). Flattened cochlear preparations showed that there are obvious spaces between IPCs and OPCs at the basal turn in P9 WT mice (Control group) and in P9 Cx26 cKO mice treated with T3 (Cx26 cKO+T3 group; Figure 9K), while IPCs and OPCs remain adjacent in P9 Cx26 cKO mice treated with saline (Cx26 cKO group; Figure 9K). Frozen section of cochleae showed that TC could be clearly observed in control group and of Cx26 cKO+T3 group, while no open TC structure can be observed in the Cx26 cKO (Figure 9L). These results suggest that T3 could promote TC structural development in Cx26 deficiency cochlea. In addition, we found destruction of P‐cadherin on the cytomembrane of OPCs in Cx26 cKO+T3 group (Figure 9K,L). These results demonstrate suggesting the destruction of P‐cadherin is the core event of the separation of PCs, and T3 can rescue Cx26 deficiency induced TC structural development disorders via increasing ADAM10 expression and promoting P‐cadherin destruction.

To investigate whether different injection routes of T3 would affect the treatment outcome, we also have injected T3 through the round window membrane into the unilateral(left) cochlea of Cx26 cKO mice. We found that while local injection of T3 can promote TC opening in all turn of the cochleae treated with T3. In contrast, formed TC could be observed in the middle and basal turns of the contralateral cochlea, while the TC in the apical turn remains unopened. This result suggests that the T3 signal in the cochlea needs to be amplified to a certain extent or sustained for a long enough time in order to achieve complete opening of TC from the basal turn to the apical turn (Figure S11).

To verify whether T3 can rescue Cx26 deficiency induced congenital hearing loss. Auditory brainstem response(ABR) tests were performed at P18 (Figure 10A). ABR thresholds in Cx26 cKO mice evoked by Tone Burst stimulations at frequencies of 4, 8, 12, 16, 28, 32, and 40 kHz were both above 90 dB SPL, with an average ABR threshold of 89.17 ± 1.86 dB SPL and 88.33 ± 2.36 dB SPL at frequencies of 20 and 24 kHz(Figure 10B), respectively. This result suggests Cx26 cKO mice had severe hearing loss at the tested frequencies, which is consistent with previous reports by us and other [48]. In contrast, the average ABR thresholds in Cx26 cKO+T3 group were 86.67 ± 3.73, 60.83 ± 13.36, 45.83 ± 12.72, 33.33 ± 11.06, 39.17 ± 16.18, 47.5 ± 17.02, 45 ± 18.48, 52.5 ± 14.07, 82.5 ± 8.04 dB SPL, respectively (Figure 10B). Distortion product otoacoustic emission (DPOAE) tests were performed at P18 (Figure 10C–E). The amplitude of DPOAE of Cx26 cKO mice decreased significantly compared with the control group at 16 kHz and was almost unrecorded(Figure 10D), while the amplitude of DPOAE of T3 treated Cx26 cKO mice is close to the control group (Figure 10C,E). These results indicate that T3 can rescue Cx26 deficiency induced hearing loss and active cochlear amplification at the tested frequencies.

FIGURE 10.

T3 rescues hearing loss and HC loss in Cx26 cKO mice by rescuing cochlea structure. Cx26flox/flox; Rosa26CreER mice were injected subcutaneously with TMX at P0 and P1 and injected subcutaneously with T3 at P5, and finally accepted ABR test and sacrificed at P18(A). ABR threshold were measured in WT mice (control group), WT mice with T3 treatment (WT+T3 group), Cx26 cKO mice with saline treatment (Cx26 cKO group), and Cx26 cKO mice with T3 treatment (Cx26 cKO+T3 group) at P18, respectively (B). The ABR threshold of Cx26 cKO+T3 mice significant decrease compared to Cx26 cKO mice(n = 6 for each group). DPOAE evoked by two‐tone stimulation (f0 = 16 kHz, I1/I2 = 60/55 dB SPL) were measured in control group (D), Cx26 cKO group(E) and Cx26 cKO+T3 group(F) at P18(n=3). No DPOAE peak visible in Cx26 cKO mice (D). The peak of DPOAE (2f1‐f2) in Cx26 cKO+T3 mice was close to those in WT mice(C,E). Representative images of HCs of different turns in cochleae of the control group, Cx26 cKO group, and Cx26 cKO+T3 group at P18 (F). T3 rescued the loss of IHCs and OHC in the middle turn and basal turn in Cx26 cKO+T3 group. Phalloidin was used to label the HCs. Scale = 100µm(G) Scale = 10µm(H). The yellow curve was used to mark the region where IHCs and OHC significantly loss in Cx26 cKO group. Representative images of Cx26 expression in the OC region of WT mice, Cx26 cKO mice, and Cx26 cKO+T3 mice at P18(G). Significantly reduction of Cx26 expression was observed in PCs of Cx26 cKO mice and Cx26 cKO+T3 mice. OHCs count in WT mice, Cx26 cKO mice, and Cx26 cKO+T3 mice at P18 (H). The percent of survival OHCs in apical turn, middle turn and basal turn incresed 8.85±2.71%(P < 0.0001), 76.12±36.60%(P < 0.0001), and 11.29± 5.28%(P < 0.0001) compared to those in Cx26 cKO group, respectively. There is no statistical significance differences(P > 0.05) of the percent of survival OHCs between the control group and the Cx26 cKO+T3 group. Counts of Cx26‐null PCs of WT mice, Cx26 cKO mice, and Cx26 cKO+T3 mice (I). There is no statistical significance differences (P > 0.05) of percent of Cx26‐null cell between the Cx26 cKO group and the Cx26 cKO+T3 group. Relative distance between IPCs and OPCs in WT mice, Cx26 cKO mice, and Cx26 cKO+T3 mice (J). There is no statistical significance differences (P > 0.05) of the relative distance between PCs between the control group and the Cx26 cKO+T3 group. ns:P > 0.05. ^*:P < 0.05. Statistical significance difference evaluated by Student's t‐test.

To observe whether the morphological development of the cochlea in Cx26 deficient mice can be restored after T3 treatment, cochlear pathology were observed at P18 (Figure 10F–J; Figure S12A–G). Consistent with our previous report [47], a significantly loss of HCs could be observed in all cochleae of Cx26 cKO mice, especially in the middle turn (Yellow box, Figure 10F,H). Hardly no loss of HCs was observed in the Cx26 cKO+T3 group (Figure 10F,H). The percent of survival OHCs in apical turn, middle turn, and basal turn of Cx26 cKO+T3 mice cochlleae incresed 8.85 ± 2.71% (P < 0.0001), 76.12 ± 36.60% (P < 0.0001), and 11.29±5.28% (P < 0.0001) compared to those in the Cx26 cKO group, respectively. And there is no statistical significance differences(P > 0.05) of the percent of survival OHCs between the control group and Cx26 cKO+T3 group. These results suggest that T3 treatment rescue the loss of cochlear IHCs and OHCs in Cx26 cKO mice, especially the loss of OHCs in the middle turn (Figure 10H). We have reported that reduced postnatal expression of cochlear Cx26 could lead to hearing loss and loss of hair cells in a dose‐dependent manner. To examine the residual Cx26 in the cochlea of Cx26 cKO mice treated with T3, the percent of Cx26‐null pillar cells was counted. More than 90% of SCs had lost Cx26 expression in the cochlear Cx26 cKO mice, and there was no significant difference(P > 0.05) in the proportion of PCs losing Cx26 expression between Cx26 cKO mice with or without T3 treatment (Figure 10G–I). More cochlear pathological changes were observed to provide evidence for T3 rescue of cochlear structure in Cx26 cKO mice. The relative distance between the nuclei of IPCs and OPCs in T3 treated Cx26 cKO mice showed no no significant difference (P > 0.05) at P18 in comparison with controls (Figure 10J), which suggested that T3 treatment promoted the separation of PCs. And the IPCs and OPCs were still apposed in the Cx26 cKO mice cochlear without T3 treatment (Figure 10G). Frozen section of P18 cochlear was prepared to observed the structure of the OC and Spiral Ganglion Neurons (SGNs). Noticeable TC could be found in cochlear of Cx26 cKO mice with T3 treatment (Figure S12A–C). The relative distance between the feet of the IPCs and OPCs was also increased in the T3 treatment group compared to Cx26 cKO mice (Figure S12C,E). In accordance with previous reports [25], actin filaments labeled by phalloidin in the feet of PCs decrease in P18 Cx26‐null mice (Figure S8C,F). Meanwhile, actin filaments in the feet of PCs showed no significant difference (P > 0.05) compared with the control, which suggested that T3 may accelerates the development of the cytoskeleton in PCs. In addition, no substantial spiral ganglion neuron loss was observed in Cx26 cKO mice with or without T3 treatment (Figure S12D,G). These results suggested that T3 could rescue hearing and hair cell of Cx26 cKO mice by promote TC re‐open, and the realization of the rescue effect is related to TC opening and restoration of normal cochlear structure.

3. Discussion

In this study, we show that Cx26 could be distributed and enriched in the nucleus of pillar cells during cochlea development. Further studies in vitro showed that Cx26 is distributed and enriched in the nucleolus during Telophase to interphase during the cell mitotic cycle. IP‐MS revealed nuclear proteins that interact with Cx26, and ChIP‐seq showed that a considerably proportion of Cx26 binds to the TSS region of the genome promoter. These results support that Cx26 could be a directly transcriptional regulator during cochlea development. This is the first report revealing the revolutionizing concept of non‐canonical functions of Cx26 in the direct regulation of gene transcription and cell fate. The high conservation of Cxs across different vertebrates and the highly similar amino acid sequence of Cxs may reflect a common mechanism of gene regulation during the development process of tissues and organs [38].

Here, we show that Cx26 could directly regulate transcription of TSPAN5, TSPAN15, and TSPAN17 gene. Cx26 deficiency causes a reduction of TSPAN5, TSPAN15, and TSPAN17 mRNA level in the cochlea and in BxPC‐3 cell, while restore Cx26 leading to a rescue of TSPAN5, TSPAN15, and TSPAN17 mRNA level in vitro cell lines. Such regulation was mainly attributed to Cx26 function in the nucleus and enhanced TSPAN promoter activity. TSPAN5, TSPAN15, and TSPAN17 all belong to the TSPANC8 subgroup of tetraspanins, which were reported to be the essential regulators of ADAM10 maturation and trafficking to the cell surface [49, 50]. In our study, co‐immunoprecipitation identified specific ADAM10 interactions with TSPAN5, TSPAN15, and TSPAN17 in cochlea at P5, while P‐cadherin, a potential substrate of TSPANC8/ADAM10 complexes, was also identified interactions with ADAM10 in the cochlea. Moreover, the destruction of P‐cadherin between IPCs and OPCs is linked to the separation of PCs and opening of TC. These results indicated that Cx26 regulates TSPANC8 gene transcription and ADAM10 maturation to control the disruption of adhesion junctions, achieving PC separation and TC opening during neonatal mice cochlear development. Cochlea Cx26 deficiency results in a decrease of mature ADAM10 in the OC region at P5, leading to the delay of P‐cadherin destruction between IPCs and OPCs, which induces the failure of TC to open.

While studies have highlighted the novel transcriptional roles of Cx43, an intriguing question arises regarding whether other connexins, particularly Cx30, which is co‐expressed with Cx26 in cochlear supporting cells, share this functionality. Canonically, connexin have strong evolutionary conservation in vertebrates. Thus, there may intuitively exist a common phenomenon of gene regulation by connexin family members. Interestingly, although the amino acid sequences of Cx26 and Cx30 are highly similar, the Cx26 mutation mouse model and Cx26 deletion mouse model exhibit complete hearing loss, while the mouse model with Cx30 deficiency but Cx26 expression preserved showed progressive hearing loss. Thus, in the specific Cx30 knock‐out mouse model with Cx26 expression reduction, defective Cx26 expression is considered as the likely cause of deafness. These studies suggested that Cx30 may not be significantly involved in regulating gene transcription during the cochlear development stage of newborn mice, or Cx30 may have transcriptional regulatory activity, but the loss of this transcriptional regulatory activity does not have a significant impact on the development of the cochlea in newborn mice.

Localization pattern of Connexin‐43(Cx43) in the nucleus and cellular compartment has attracted a lot of attention [40, 51, 52]. Although Cx26 and Cx43 both belong to the Cxs family and have similar structures, the nuclear localization pattern of Cx26 and Cx43 is distinct, which may suggest the differences in regulatory mechanisms and specific function among Cxs. Nuclear pore complexes allow passive diffusion of proteins < 40 kDa [53], and the molecular weight of Cx26 is small enough (26 kDa). However, Cx26 accumulated in the nucleolus against a concentration gradient could be the result of active transport instead of free diffusion given. In this study, we demonstrated the dynamic changes of Cx26 accumulation in the nucleolus during the cell cycle in vitro, also suggesting that nucleolus localization of is results of strict regulation. Nuclear localization signal (NSL) is needed to binds with appropriate nuclear transport receptors when macromolecules enter the nucleus [53, 54, 55]. Although the NLS of the Cx26 is still unknown, it may contain N‐terminal residues of Cx26 instead of the C‐terminal since Cx26 p.Leu79CysfsTer3, Cx26 p.Glu178_Val226del, WT‐Cx26 both was found in the nucleus in vitro (Figures S3–S6). The difference in the distribution in the nucleus of Cx26 p.Leu79CysfsTer3 and Cx26 p.Glu178_Val226del suggests the signal guiding nucleolar localization and function is contained in No.79–No.178 residues (Figures S3–S6). It has also been reported that the integrity of Transmembrane Domain 4 is crucial for the cytomembrane localization of Cx26 [56, 57, 58]. These findings have preliminarily explored the structural domains that guide the subcellular localization of Cx26.

Dynamic changes in Cx26 nuclear localization were observed in Bxpc‐3 cells, Hela cells transfected with Cx26, and 293T cells transfected with Cx26, accompanied by the disappearance and reconstruction of the cell nucleus. However, no dynamic changes in nuclear localization of Cx26 were observed in vivo. Cochlear SCs are post‐mitotic and terminally differentiated in neonatal mice [57], and these supporting cells no longer undergo the process of disintegration and reconstruction. However, Cx26 aggregation has only been observed in the nucleus of partly (not all) support cells, and these nuclear localization Cx26 is transported against the concentration gradient, which suggests that the Cx26 in the nucleus actually undergoes dynamic nuclear transport under strict regulation, rather than exist statically in the nucleus.

Posukh et al. have summarized data of pathogenic or likely pathogenic (PLP) variants of GJB2 [58], and one of their interesting discoveries is that the distribution of PLP variants across Cx26 protein domains is uneven. PLP variants seem to be more concentrated in the four Cx26 domains near the N‐terminus (N‐terminal, TM1, EL1, and TM2), and the syndromic dominant variants are also enriched in the N‐terminus and EL1. As we have mentioned, the nuclear localization and function of Cx26 also depend on these domains or residues. Therefore, the heterogeneity of phenotypes associated with GJB2 variants may result from differences in nuclear localization patterns and nuclear functions of different Cx26 truncated mutants [59, 60, 61]. Disappearing TC and malformed OC were observed in GJB2 R75W transgenic mouse model [26, 62] and a tetraploid transgenic mouse model with a shorter Cx26 mutant, Cx26 p.Gly12ValfsTer2 (GJB2 c.35delG) [30]. Both two variants lead to amino acid changes or deletion in domains that may be related to nuclear function or nuclear localization.

ADAM10 interacts with TspanC8 subgroup comprising six tetraspanins, including Tspan5, Tspan10, Tspan14, Tspan15, Tspan17, and Tspan33 [63]. Different TspanC8/ADAM10 complexes functions as a “molecular scissor”, causing ectodomain shedding by cleaving the extracellular regions from its different membrane protein substrates [64]. The substrates of TspanC8/ADAM10 complex including Notch, amyloid precursor protein, cadherins, and growth factors. The main role of tetraspanins is to bind to their specific partner proteins and regulate their intracellular trafficking, lateral mobility in the membrane, and clustering into nanodomains [50]. Studies have reported that different TspanC8/ADAM10 complexes have distinct substrates and TspanC8 determines the substrate specificity of ADAM10. Tspan15/ADAM10 preferentially cleaves neural(N)‐cadherin [65, 66], whereas Tspan5 and Tspan14 preferentially cleave and activate Notch [67, 68]. ADAM10 interacting with Tetraspanins Tspan5 and Tspan17 could cleave vascular endothelial(VE)‐Cadherin [69]. Expression Each cell type expresses its own complement of tetraspanins. Tspan10 is relatively weakly expressed at the mRNA level or absent from most cell types [40], and low Tspan10 mRNA expression was detected in mice cochlea in this study. The mRNA expression of Tspan15 is relatively high in the cochlea of P5 mice. Tspan17 mRNA is expressed by most cell types including cochlea SCs, but at relatively low levels. N‐cadherin and P‐cadherin are classical cadherin sharing a similar protein structure [70]. Therefore, Tspan15/ADAM10 complex may be the key to the destruction of P‐cadherin during the process of pillar cell separation.

In this study, T3 treatment at P5 could partly rescue hearing of Cx26 cKO mice at mid and high frequencies by promoting open of TC, provides direct evidence for cochlear developmental disorders as the primary deafness mechanism of Cx26 deficiency. In developed cochlea, Cx26 expressed and located in coordination with Cx30, and Cx30 GJCs between supporting cells still exist after Cx26 cKO, which may partly compensate for intercellular communication and transfer [71]. However, there is a difference in the permeability of Cx26 GJCs and Cx30 GJCs to ions and miRNA [14]. The remaining Cx30 cannot fully make up for the communication and transfer dysfunction after Cx26 cKO, which may be the reason why T3 treatment cannot fully restore hearing in Cx26 cKO mice after TC open.

Here, we reported the nuclear localization and function in transcriptional regulation of Cx26 in cochlea. Our results demonstrate that Cx26 could function as directly transcriptional regulator during cochlea development, instead of just forming junctional communication. Cochlea Cx26 deficiency results in a decrease of TSPANC8 genes transcription and ADAM10 maturation in cochlear, and T3 can promote TC opening and hearing rescue in Cx26 cKO mice by increasing the expression of ADAM10. Our study provides a new treatment strategy for GJB2 mutation hearing loss with cochlear structural developmental defects.

4. Experimental Section/Methods

4.1. Mouse Model

Wild‐type CBA/CaJ mice were obtained from SPF Biotechnology Co., Ltd.(Beijing, China). Cx26loxP/loxP mice and Rosa26CreER mice were provided by Prof. Xi Lin at Emory University. The generation method of the Cx26 cKO mouse model have been mentioned in detail in our previous article. In brief, tamoxifen‐inducible Cx26 cKO mice were generated by crossbreeding of the Cx26loxP/loxP mice and Rosa26CreER mice. Mouse genotyping of was performed by PCR amplification of tail genomic DNA, and the genotyping primers were as follows:

Cx26 (F): 5′‐ACAGAAATGTGTTGGTGATGG‐3′,

Cx26 (R): 5′‐CTTTCCAATGCTGGTGGAGTG‐3′,

Rosa26CreER (F): 5′‐AGCTAAACATGCTTCATCGTCG GTC‐3′,

Rosa26CreER (R): 5′‐TATCCAGGTTACGGATATAGTTC ATG‐3′.

Tamoxifen (T5648‐1G, Sigma–Aldrich, St. Louis, MO, United States) was subcutaneously injected at a dose of 1.5 mg/10 g body weight(distributed in corn oil) at postnatal day(P) 0 and P1 to induce Cx26 condition knock out.

All mice were raised in the specific‐pathogen‐free Experimental Animal Center of Huazhong University of Science and Technology. All experimental procedures were conducted in accordance with the policies of the Committee on Animal Research of Tongji Medical College, Huazhong University of Science and Technology(IACUC Number: 4148).

4.2. Auditory Brainstem Response(ABR) and DPOAE Recordings

ABR hearing thresholds and DPOAE were recorded at P18 by use of a Tucker–Davis Technologies System (RZ6, Tucker–Davis Tech., Alachua, FL, USA). The details of the ABR and DPOAE test were as described in our previous study. Mice were anesthetized by intraperitoneal injection with 1.25% tribromoethanol (0.2 mL/10 g body weight), and body temperature was maintained at 37–38°C using a heating pad. ABR responses were evoked by tone burst stimuli at frequencies of 8, 16, 24, 32, and 40 kHz in decreasing 10 dB steps with an MF‐1 speaker(Tucker–Davis Tech., Alachua, FL, USA). ABR hearing thresholds were recorded as the lowest sound intensity that could be recognized. For DPOAE recording, the frequency was presented by a geometric mean of f1 and f2 [f0 = (f1 × f2)1/2] with a ratio of f1:f2 = 1:1.2. The intensity of f1 was set at 5 dB SPL higher than that of f2. The distortion product was averaged by 200 times. The 2f1‐f2 component was measured and recorded.

4.3. Cochlear Tissue Preparation and Immunofluorescent Staining

Mice were deeply anesthetized and sacrificed. The cochleae were dissected and fixed in 4% paraformaldehyde in 0.01 m phosphate‐buffered saline (PBS) at 4°C for 12 h. To prepare frozen sections, cochleae were decalcified with EDTA‐Na2 for 48 h and then dehydrated with 10%, 20% and 30% sucrose solution for 1.5 h each. After that, the prepared cochlea was embedded in Optimal Cutting Temperature Compound at 4°C overnight and cut into sections at −20°C with a thickness of 10 µm. To prepare flattened cochlear preparations, cochleae was carefully dissected in 0.01 m PBS after decalcification. The frozen sections or flattened cochleae preparations were incubated in a blocking solution (10% donkey serum with 0.1% Triton X‐100) for 1 h at room temperature and then incubated with primary antibodies (1:100–200 dilution) at 4°C overnight. Primary antibodies used in this experiment including Rabbit anti‐Cx26 polyclonal antibody (710500, Invitrogen), Rabbit anti‐Lamin B1 Polyclonal antibody (12987‐1‐AP, Proteintech), Rabbit anti‐ADAM10 Polyclonal antibody (A10438, ABclonal), Goat anti‐P‐cadherin Polyclonal antibody (AF761, R&D Systems), Rabbit anti‐HA‐tag Polyclonal antibody (51064‐2‐AP, Proteintech), Mouse anti‐E‐cadherin Monoclonal antibody (MAB7481, R&D Systems), Rabbit anti‐N‐cadherin Monoclonal antibody (GTX127345, GeneTex), each was diluted in 0.01 m PBS with 0.1% Triton X‐100. After washing with PBS, the samples were stained with secondary antibodies (1:200 dilution), including Alexa Fluor 647 donkey anti‐goat IgG (ANT033, AntGene), Alexa Fluor 488 donkey anti‐rabbit IgG (ANT024s, AntGene), Alexa Fluor 647 Donkey anti Mouse IgG (ANT034, AntGene). DAPI (C1005, Beyotime Biotechnology) and phalloidin (P5282, Sigma–Aldrich) were used for nuclear and Actin filament staining. Images were obtained with a laser scanning confocal microscope (Nikon, Tokyo, Japan). Quantitative analysis of ADAM10 expression in PCs was achieved through ADAM10 immunofluorescence staining. Images of the OC region were randomly obtained in the basal turn of the cochleae(two to three sections per mouse, four mice per group). Relative fluorescence was quantified by normalizing the ratio of the average fluorescence of target cells in the Cx26 cKO group to that in the control group. The distance between IPCs and OPCs was measured using ImageJ software. Relative distance was quantified by the length between IPCs and OPCs between IPCs and OPCs in the basal turn in the Cx26 cKO group to that in the control group. To assess the patterns of damage to hair cells, we quantified the percentage of surviving OHCs in the apical, middle, and basal turns of flattened cochlear preparations. Images of OHCs were captured at 60x magnification with the laser scanning confocal microscope in different turns of the flattened preparations. A total number of approximately 40 OHCs from each turn were taken for counting (n = 6 mice in each group).

4.4. Injection Through the Round Window

The mice pups were anesthetized by gradually lowering the ambient temperature. After treatment with povidone iodine and 75% alcohol, the skin behind the left ear area was cut, and a 1 cm incision was made. The fat and muscle were gently removed to expose the otocyst and round window. The drugs (volume≤1.5 mL) were injected through the round window membrane. The skin wound was sealed using Tissue Adhesive (1469C, 3 m Vetbond). The mice pups were placed on a 37°C heating pad to be revived.

4.5. T3 Treatments

T3 (T2877, Sigma) was dissolved in saline (adjust the pH to 7.4 before T3 dissolved) at a storage concentration of 5 mg/ml. The acquired solution was further diluted with saline to final concentrations of 1 mg/ml. We and Douglas F et al. have reported that subcutaneous injection of T3 into mice before P3 can cause severe hearing loss in WT mice, which may be related to disordered stereocilia of the OHCs. As a result, T3 (5 µl/1 g body weight), or the equivalent volume of saline, was subcutaneously injected into Cx26 cKO mice pups or WT mice pups at P5. The dosage of T3 injection was determined based on previous reports of Forrest D et al. and us.

4.6. Detection of Serum T3, T4, TSH Levels

After the mice were deep anesthesia and sacrificed, capillary glass tubes were used to collect blood. Free T3, free T4, and TSH levels in serum were measured in 10 µL serum samples from each mouse. Elecsys FT3 III Reagent Kit (09005811190, Roche), Elecsys FT4 IV Reagent Kit (09043284190, Roche), and Elecsys TSH Reagent Kit (08443432190, Roche) were used to measured the FT3, FT4, and TSH levels in serum according to the manufacturer's instructions. At least 3 samples from different mice should be tested in each group or at each time point.

4.7. Cell Culture and Transfection

BxPC‐3 cell line(CL‐0042, RRID: CVCL_0186), HEK293T cell line(CL‐0005, RRID: CVCL_0030), and Hela cell line(CL‐0101, RRID: CVCL_0063) were obtained from Pricella Biotechnology Co., Ltd. (Wuhan, China). Cell Line STR authentication was conducted to ensure that the cell line was contamination‐free. Cells were cultured in a 37°C incubator with 5% CO2 in the standard culture medium consisted of Dulbecco's Modified Eagle Medium (DMEM, 11965092, Gibco) supplemented with 10% Fetal Bovine Serum(A5670701, Gibco). Subcultures of cells were performed at 80%–90% confluence using 0.25% trypsin‐EDTA (25200072, Gibco). To achieve Cx26 expression in 293T cells and Hela cells, the GJB2 recombinant vector (GOSL0263268) was produced by Genechem Co.,Ltd.(Shanghai, China). GM easyTM Lentiviral Packaging Kit(Gmeasy‐10, Genomeditech) was used to package the vector with lentiviral capsids. After 293T or Hela cells were cultivated to 50%–60% confluence, the DMEM was replaced with culture medium. The recombinant lentivirus vector were incubated with cells for 24 h, and then the DMEM was replaced with standard culture medium as described above. Purimycin(1 µg/mL) was used to screen stable transfected Cx26 cell lines. To achieve Cx26 knockdown in BxPC‐3 cells, siRNA‐GJB2 was designed and produced by Tsingke Biotechnology Co., Ltd. (Beijing, China), to knock down the expression of Cx26 in BxPC‐3 cells, and an siRNA encoding a nonsense sequence was designed as the negative control. After BxPC‐3 cells were cultivated to 50%–60% confluence, the culture medium was shifted to Opti‐MEM (31985070, Gibco). Lipofectamine 3000 (L3000075, Gibco) was used following the manufacturer's instructions to provide superior transfection efficiency. About transfection for 8 h, the Opti‐MEM was replaced with standard culture medium as described above. Transfected BxPC‐3 cells were collected after 72‐96 h to detect Cx26 knockdown efficiency. The siRNA‐Cx26 sequence were as follows:

siRNA‐1 sense:5′‐GCAAGAACGUGUGCUACGA‐3’

siRNA‐1 antisense:5′‐UCGUAGCACACGUUCUUGC‐3’

siRNA‐2 sense:5′‐CCUUCAUGUACGUCUUCUA‐3’

siRNA‐2 antisense:5′‐UAGAAGACGUACAUGAAGG‐3’

siRNA‐3 sense:5′‐GCAAGAACGUGUGCUACGATT‐3’

siRNA‐3 antisense:5′‐UCGUAGCACACGUUCUUGCTT‐3’.

4.8. RNA Preparation and Real‐Time Quantitative Polymerase Chain Reaction (qPCR)

For cochleae tissues, after mice were deeply anesthesia and sacrificed, the membranous labyrinths were dissected from cochleae in cold 0.01 m PBS. Then the membranous labyrinths from 4 cochleae (From 2 mice) was used to generate one sample. For cells, 1 x 10^6 cells were collected as one sample. Total RNA was extracted with TRIzol Reagent(15596026CN, Invitrogen) and was reverse transcribed using a PrimeScript RT Reagent Kit with gDNA eraser (RR047A, Takara). Real‐time PCR was performed in a Roche LightCycler 480 instrument with TB Green Premix Ex Taq(RR420A, Takara). The mRNA expression was normalized to the mRNA expression of GAPDH. Relative gene expression data were analyzed using the 2–ΔΔCP method. The following were primers used in this experiment:

For cell Lines

GJB2(F):5′‐TCGCATTATGATCCTCGTTGTG‐3’

GJB2(R):5′‐GGGGAAGTAGTGATCGTAGCAC‐3’

GAPDH(F):5′‐GGAGCGAGATCCCTCCAAAAT‐3’

GAPDH(R):5′‐GGCTGTTGTCATACTTCTCATGG‐3’

TSPAN5(F):5′‐CGGGAAGCACTACAAGGGTC‐3’

TSPAN5(R):5′‐CCACCACAAGGAAGAGCCAA‐3’

TSPAN15(F):5′‐TCCCTCCGTGACAACCTGTA‐3’

TSPAN15(R):5′‐CCGCCACAGCACTTGAACT‐3’

TSPAN17(F):5′‐CGGGAGTCTTCATGGGCATC‐3’

TSPAN17(R):5′‐TTCCATTTGCTCGAGGTTCTGGG‐3’

For mice cochleae

Gjb2(F):5′‐CTCGGGGGTGTCAACAAACA‐3’

Gjb2(R):5′‐CACGAGGATCATGATGCGGA‐3’

Gapdh(F):5′‐GAAGGTCGGTGTGAACGGAT‐3’

Gapdh(R):5′‐CTCGCTCCTGGAAGATGGTG‐3’

TSPAN5(F):5′‐ACTTCGGGAAAACACCTTTCTT‐3’

TSPAN5(R):5′‐GGTTCCAATCATCAGCTCCAAAA‐3’

TSPAN10(F):5′‐CCTACAGACCCTATGCTGATGC‐3’

TSPAN10(R):5′‐ACAGGCAGCTATTCTCACAGA‐3’

TSPAN14(F):5′‐TGGCTGGAGTTGTCTTCCTTG‐3’

TSPAN14(R):5′‐GGGTCGATTCCATGCAACC‐3’

TSPAN15(F):5′‐TGTCAGTGGGGATCTACGCA‐3’

TSPAN15(R):5′‐CCACCAATAAGCTCCATGACC‐3’

TSPAN17(F):5′‐GGGGAGTCATGTCAGTGTTGG‐3’

TSPAN17(R):5′‐AAGGTCGAGATCATCCCGGTA‐3’

TSPAN33(F):5′‐GGGGACGAGTTCTCCTTCG‐3’

TSPAN33(R):5′‐TGCTTCTGCGTGCTTCATTAG‐3’

ADAM10(F):5′‐ATGGTGTTGCCGACAGTGTTA‐3’

ADAM10(R):5′‐GTTTGGCACGCTGGTGTTTTT‐3’

4.9. Scratch Loaded Dye Transfer Experiment

This experiment was conducted through the transfer of lucifer yellow dye utilizing a scrape‐loading assay to evaluate of intercellular coupling mediated by Cx26 gap junctions after CBX treatment. The method has been described in detail in our previous studies. For CBX treatment, cells were initially incubated with calcium‐free medium containing 1 mmol CBX for 30 min.

4.10. Protein Extraction and Western Blotting

For total proteins extraction, after the mice were deep anesthesia and sacrificed, the membranous labyrinths were dissected from cochleae in cold 0.01 m PBS to extract total protein. Membranous labyrinths from 2 cochleae(from the same mouse) were used to generate one sample. For cells, 1 x 10^6 cells were collected as one sample. Cells and cochleae samples were lysed with cold RIPA Lysis Buffer (P0013B, Beyotime) with 1% PMSF (ST506, Beyotime). The mixture was centrifuged at 12 000 g for 15 min at 4°C, and the supernatant was collected. For nuclear proteins extraction, a nuclear and Cytoplasmic Protein Extraction Kit(P0027, Beyotime) was use to extract nuclear protein of cochleae. The product instructions were used to guide experimental operations steps. Fasttest blocking buffer (HYC00811, HYCEZMBIO) were used to incubate and block the PVDF membrane. BCA Protein Assay Kit (P0009, Beyotime) was used to measure the protein concentration. The proteins were separated by 10%–12% sodium dodecyl sulfate polyacrylamide gel electrophoresis and then transferred to 0.45 µm PVDF membranes. Primary antibodies used in this experiment including Rabbit anti‐Cx26 polyclonal antibody (710500, Invitrogen), Rabbit anti‐ADAM10 Monoclonal antibody (T56805, Abmart), Rabbit anti‐GAPDH Polyclonal antibody (A19056, ABclonal). GAPDH as the reference protein. Horseradish peroxidase(HRP)‐conjugated donkey anti‐rabbit IgG (72‐8099, Antgene), HRP‐conjugated Donkey anti‐Mouse IgG(72‐8097, Antgene), and HRP‐conjugated Donkey anti‐Goat IgG(72‐8082, Antgene) was used as the secondary antibody. An enhanced Chemiluminescent kit (P10300, NCM Biotech) was used to detect the immune complexes according to the manufacturer's instructions. The bands were visualized and measure the intensities by using ImageJ software.

4.11. Immunoprecipitation(IP)

For Co‐IP, membranous labyrinths from 14 cochleae(from the same mouse) were used to generate one sample. After carefully dissected in cold 0.01 m PBS, the membranous labyrinths were lysed with 1 mL cold cell lysis buffer for Western and IP (P0013J, Beyotime) with 1% PMSF (ST506, Beyotime). Grind the lysate at 4°C temperature to fully break the membrane labyrinths tissues. The mixture was centrifuged at 12 000 g for 15 min at 4°C, and the supernatant was collected. Then the produtions were incubated with 5 µg Rabbit anti‐ADAM10 Monoclonal antibody (T56805, Abmart) or IgG antibody for 24 h at 4°C. Protein A/G Magnetic Beads (B23201, Selleck) were used to immunoprecipitate the ADAM10 immune complex according to the manufacturer's instructions and were finally precipitated with the immune complex using a magnetic stand. The magnetic beads were boiled at 95°C for 10 min with 1x SDS loading buffer to denature proteins. After co‐immunoprecipitated proteins were eluted, western blotting was performed following the steps described to get the images of bands. Primary antibodies used in this experiment including Goat anti‐P‐cadherin Polyclonal antibody (AF761, R&D Systems), Rabbit anti‐TSPAN5 Polyclonal antibody (12122‐1‐AP, Proteintech), Rabbit anti‐TSPAN15 Polyclonal antibody (A14395, ABclonal), Rabbit anti‐TSPAN17 Polyclonal antibody (A15423, ABclonal). Secondary antibodies used in this experiment including Goat Anti‐Rabbit IgG HRP(M21006, Abmart), Goat Anti‐Rabbit IgG HRP(M21007, Abmart), HRP‐conjugated Donkey anti‐Goat IgG(72‐8082, Antgene), and Goat Anti‐Mouse IgG HRP(M21005, Abmart).

For IP‐MS, 293T cells stably transfected with Cx26 were lysed with 1 mL cold cell lysis buffer for Western and IP (P0013J, Beyotime) with 1% PMSF (ST506, Beyotime). The mixture was centrifuged at 12 000 g for 15 min at 4°C, and the supernatant was collected. The productions were incubated with 5 µg Rabbit anti‐Cx26 Polyclonal antibody (710500, Invitrogen) or IgG antibody for 24 h at 4°C. Protein A/G Magnetic Beads (B23201, Selleck) were used to immunoprecipitate the ADAM10 immune complex according to the manufacturer's instructions and were finally precipitated with the immune complex using a magnetic stand. The magnetic beads were boiled at 95°C for 10 min with 1x SDS loading buffer to denature proteins. The solution was used for Mass spectrometry analysis. The Mass spectrometry analysis experiment was conducted by Seqhealth Technology Co., Ltd., Wuhan, China. MS raw data were analyzed with MaxQuant (version 1.6.6.0) using the Andromeda database search algorithm. Proteins that could not be distinguished based on unique peptides were merged by MaxQuant into a single protein group. Search results were filtered with 1% FDR at both peptide and protein levels.

For ChIP‐seq, ChIP experiments were performed by a Sonication ChIP Kit (RK20258, ABclonal) according to the manufacturer's instructions. In brief, HEK293T stable transfected Cx26 cell lines were fixed with 1% paraformaldehyde at 25°C for 10 min to cross‐link proteins and DNA, and a glycine solution was used to terminate the crosslinking reaction. Cells were collected after washing with PBS, and then treated with cell lysis buffer and ultrasonic. The production was centrifuged at 12 000 g for 10 min, and the supernatant was collected. This supernatant contains cross‐linked chromatin. The sample was divided into two tubes. 10% of the sample was stored and marked as the input group. 80% of the sample was immunoprecipitation with Rabbit anti‐Cx26 polyclonal antibody (710500, Invitrogen) and marked as the IP group. 10% was incubated with rabbit IgG and marked as the IgG group. High through‐put sequencing and data analysis were conducted by Seqhealth Technology Co., LTD (Wuhan, China) on DNBSEQ‐T7 sequencer (MGI Tech Co., Ltd. China) with the PE150 model. Raw sequencing data were first filtered by Trimmomatic (version 0.36), low‐quality reads were discarded, and the reads contaminated with adaptor sequences were trimmed. The clean reads were used for protein binding site analysis. They were mapped to the reference genome of the house mouse from GRCm39. Gene ontology (GO) analysis and Kyoto encyclopedia of genes and genomes (KEGG) enrichment analysis for annotated genes were both implemented by KOBAS software (version: 2.1.1), and P < 0.05 was considered to be statistically significant enrichment.

For ChIP‒qPCR, immunoprecipitation was performed at 4°C overnight with anti‐Cx26(710500) or normal rabbit IgG antibody. Then, qRT‒PCR was utilized to quantify the immunoprecipitated DNA, and the data were normalized to the input. The following were primers used for ChIP‒qPCR.

Tspan5‐F 5’‐TGTGTGTGTGTTGGGAGGAG‐3’

Tspan5‐R 5’‐CCGGAGGGACCCTGAAAAAG‐3’

Tspan15‐F 5’‐CCTGAGTCAGTCCAGGGAGA‐3’

Tspan15‐R 5’‐GATCCGGCCACCGCTC‐3’

Tspan17‐F 5’‐GTGGTAGTCCGAGCGCC‐3’

Tspan17‐R 5’‐CTTCATGGCCAGAGCCACA‐3’

TSPAN5‐F 5’‐GTCTCCACCAGCATCTGCC‐3’

TSPAN5‐R 5’‐CTAGCCCCGAACACAAAGCG‐3’

TSPAN15‐F 5’‐GTCAGTCCCGGAGAGAACG‐3’

TSPAN15‐R 5’‐TCTCCGGCTCGGTCCC‐3’

TSPAN17‐F 5’‐CTCCCTAAATAATCCCGAGGC‐3’

TSPAN17‐R 5’‐AACCTGCACCTAGCACGC‐3’

4.12. Luciferase Reporter Assay

TSPAN5, TSPAN15, and TSPAN17 promoter luciferase reporter vectors were produced by Genechem Co.,Ltd.(Shanghai, China). For the luciferase reporter assay, GJB2 recombinant, GJB2 c.235delC vector or empty vectors were co‐transfected with different luciferase reporter vectors into 293T cells, respectively. Transfection efficiency was normalized by co‐transfection with a vector containing a Renilla luciferase gene. A Dual Luciferase Reporter Gene Assay Kit (G1701, Servicebio) was used to quantified the Firefly luciferase and Renilla luciferase activities.

4.13. Statistical Analysis

All data were expressed as means ± SD. Statistical analyses and related figures was performed on GraphPad Prism (Version 10.0.2). Parametric and nonparametric data comparisons were performed using one‐way ANOVA or Student's T‐tests after assessment of normality and variance, and p < 0.05 was considered to be statistically significant. ^*: P < 0.05. ^**: P < 0.01. ^***: P < 0.001. ^****: P < 0.0001.

Funding

National Natural Science Foundation of China (No. 82430035), the National Key Research and Development Program of China (Nos. 2021YFF0702301, 2024YFC2511101, 2023YFE0203200), the Foundation for Innovative Research Groups of Hubei Province (No. 2023AFA038), the Fundamental Research Funds for the Central Universities (No.2024BRA019), the National Natural Science Foundation of China (No.82501422).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: advs73560‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.