The technique used for the RNAscope^® analyses of HC sections was recently described (Liu and Rask-Andersen, 2022). In short, fixed-frozen human tissue sections underwent pretreatment with H₂O₂ (10 min, RT) and protease III (30 min, 40°C). After protease III incubation, the sections were exposed to a RNA-scope hybridization assay. Bio-Techne produced the probes. Hybridization was initiated by adding the RNA probe(s) (a mixture of probes termed C1, C2, and C3 channels) to the sections. Incubation was performed in a HybEZ™ Oven (Bio-Techne) for 2 h at 40°C. After hybridization, the slides were washed using 1x RNA-scope^® Wash Buffer. The sections were incubated with RNA-scope^® Multiplex FL v2 Amp 1, 2, and 3 (for 30 min/30 min/15 min, respectively) sequentially at 40°C to amplify the signal. For the RNA-scope^® Multiplex study, RNAscope^® Multiplex FL v2, HRP-C1, HRP-C2, and HRP-C3 were sequentially added to the sections (time 15 min). For signal detection, TSA-diluted Opal 520, 570, and 690 fluorophores were added to the sections after HRP-C1, C2, and C3, incubating the sections for 30 min each at 40°C. When the three Opal fluorophores were assigned to each channel, in present representation, two channels—C1 and C2—were assigned to GJB2 and GJB6, KCNJ10 and GJB2, and Na/K-ATP1A1 and Na/K-ATP1A2 (Table 1). After each fluorophore incubation and rinse with 1x RNA-scope^® Wash Buffer, the RNA-scope^® Multiplex FL v2 HRP blocker was added and incubated in an oven for 15 min at 40°C. Then, sections were counterstained withxs 4’,6-diamidino-2-phenylindole (DAPI), and the slides were cover-slipped with ProLong^® Glass Antifade Mountant (Thermo Fisher Scientific, Waltham, MA, United States). The RNA-scope ISH produces a puncta of signals that represent a single mRNA transcript (Grabinski et al., 2015). The RNAscope^® Multiplex Fluorescent v2 assay allows simultaneous detection of up to four RNA targets (Wang et al., 2015). The standard RNAscope probes have 20 zz pairs with each consisting of 35-50 nucleotides and resulting in 1000bp coverage for any given transcript. Both the probes target different regions and they do not cross-detect each other. The GJB6 is targeting exon 5¹ and GJB2 is targeting exon 2².

Four archival human specimens were sectioned semi-thin, stained, photographed, and re-analyzed. The specimens were fixed in 3% glutaraldehyde and decalcified in 10% Na-EDTA. The bone was post-fixed in 1% osmic acid (OsO4, 75632 Sigma-Aldrich), and dehydrated and embedded in Epon (Resolution Performance Products, Houston, TX, United States). The technique was described earlier (Tylstedt et al., 1997). Photo images of various turns were examined in parallel with the mRNAscope^®. Sections included different turns of the cochlea and were compared with the RNAscope findings at different regions (Figure 1).

Techniques used for fixation, decalcification, embedment, and cryo-sectioning for immunohistochemistry were described in earlier publications (Liu et al., 2016, 2017a,b). Here, we used data obtained from studies on archival sections and previously obtained material. This included immune labeling of Na/K-ATPase (NKA) α1, α2, β1, NKCC, occludin, lamininβ2, Claudin-11, Cx30, and Cx26 (Table 1).

The RNA-scope sections were analyzed using an inverted fluorescence microscope (Nikon TE2000; Nikon, Tokyo, Japan) and a three-filter digital camera (emission spectra maxima at 358, 461, and 555 nm). NIS Element BR-3.2 (Nikon) imaging software was used. Laser confocal microscopy was performed with a three-channel laser emission system. The software used was Nikon EZ-C1 (ver. 3.80), SR-SIM [Elyra S.1 SIM system with a 63x/1.4 oil Plan-Apochromat objective (Zeiss), sCMOS camera (PCO Edge), and ZEN 2012 software]. Multichannel SR-SIM imaging was performed using a laser and filter setup: 405 nm laser of excitation coupled with BP 420–480 + LP 750 filter, 488 nm laser of excitation with BP 495–550 + LP750 filter, 561 nm laser of excitation with BP 570–620 + LP 750 filter, and 647 nm laser of excitation with LP 655 filter. The resolution was measured with sub-resolution fluorescent beads (40 nm, Zeiss) in the green channel (BP 495–550 + LP750). An average point spread function (PSF) value was obtained from multiple beads using the built-in experimental PSF algorithm of the ZEN software. The resolution of the SR-SIM gave a lateral precision of 80 nm (Gustafsson et al., 2008). The volume resolution of three-dimensional (3-D) SIM is almost eightfold that of conventional microscopy (Schermelleh et al., 2008). The 3-D reconstructions and rendering of Cx26 (red) and Cx30 (green) protein expression were performed in spot detection mode using Imaris 9.5.1 (Bitplane, Zurich, Switzerland). The grid size of the frame on the 3-D images was 5μm. The number of GJB2 gene puncta was counted using ImageJ particle analysis in the lateral wall at all three turns. Numbers were difficult to assess and had to be estimated in areas of transcription that burst in the cell nuclei. Both the total number of puncta and the number per cell (DAPI) were calculated and compared. Distortion of tissue may have influenced the results and focus remained primarily on obtaining comparative data of the three different turns.

An archival, semi-thin section of a decalcified HC fixed in 3% glutaraldehyde is demonstrated in Figure 1. This image shows the different anatomy of the OC and the lateral cochlear wall in the three turns. The spiral limbus and tectorial membrane are smaller and the spiral ligament is thicker against the basal high frequency region. Frequency mapping was estimated based on the number of cochlear turns. Confocal microscopy of Cx26 and Cx30 antibody co-labeling of the middle turn in an adult cochlea is shown in Figures 2A–C. Cx26 and Cx30 are both separately and co-expressed in the spiral limbus and lateral cochlear wall. Cx30 expression dominates in the OC (Figure 2B, not co-stained). SR-SIM using spot detection mode shows the distribution of Cx26 and Cx30 in the lateral wall (Figure 2D). There is a separate expression, with a band of Cx26 corresponding to the basal cell layer and the junction at the type I fibrocytes of the SV. Higher magnification confirms a close association between the two channel proteins (Figure 2D’,D”). The molecular arrangements of connexons (homomeric/heteromeric/heterotypic) cannot be assessed with certainty since the microscope resolution of 80 nm is far above the diameter of a Cx26 channel (Maeda et al., 2009). Cx30 is expressed solely in the intermediate cells, suggesting a homomeric molecular design. No GJ proteins were expressed in the marginal cells.

The RNAscope^® and ISH identified the single GJB2 and GJB6 mRNA oligonucleotide transcripts in the formaldehyde-fixed sections of the HC. Decalcification maintained the cell integrity, and gene transcripts could be localized in the cell nuclei and cytoplasm. Na/K-ATPase and Kir4.1 gene transcripts (ATPA1, ATP1A2, and KCNJ10) were used as additional controls. ATPA1 was only expressed in the marginal cells of the SV, while KCNJ10 was expressed in the intermediate cells and ATP1A2 in the basal cells (Figure 3).

The GJB2 and GJB6 gene transcripts were demonstrated in the cells of the spiral limbus, OC, and lateral wall in all three turns (Figures 4–8 and Table 2). The largest numbers of GJB2 and GJB6 puncta were detected in the outer sulcus epithelium, root cells, basal cells, and types II and I fibrocytes. In the SV, many GJB2 and some GJB6 gene transcript puncta were noted in the basal cells. Intermediate cells contained GJB6, but did not contain GJB2 puncta. At the suprastrial region and the junction between the Reissner’s membrane and the SV, type V fibrocytes displayed both GJB2 and GJB6 transcripts. Type I fibrocytes contained a larger number of GJB6 as well as GJB2 gene transcripts, especially the cells located near the basal cells. While the upper and middle parts of the spiral prominence epithelium contained few GJB2/GJB6 transcripts, the lower slope contained a large number of both GJB2 and GJB6 gene transcripts. The GJB2-containing basal cells merged with the spiral prominence epithelium and marginal cells in the apical turn (Figure 4D). A similar situation existed at the lateral insertion of the Reissner’s membrane (Figure 4C). A corresponding, but slightly different, anatomy was present in the middle and lower turns of the HC (Figures 5, 6).

Relatively few gene transcripts were found in type IV fibrocytes, and none were detected in the type III fibrocytes. Type I and II fibrocytes, Hensen, Claudius, and outer sulcus cells contained many GJs on transmission electron microscopy (TEM). The spiral limbus contained many GJB2 and GJB6 gene transcripts, and many were contained in the cells lining the scala vestibuli floor (Figures 6A–C). A composite micrograph of the OC and lateral wall is shown in Figure 7. Claudius cell nuclei were crowded with GJB6 gene transcripts, but there were only a few in the cytoplasm (Figure 7A). None were found in Reissner’s membrane or hair cells. In Deiters’, Claudius, and outer sulcus epithelial cells, the number of transcripts were also concentrated to the cell nuclei.

Multiplex labeling of GJB2 and GJB6 transcripts in the HC shows that both transcripts are generally present in the epithelial and connective tissue cells (Figures 8, 9), except in the intermediate cells that only contained GJB6 transcripts. An exceptional large number of GJB2 and GJB2 gene transcripts were noted in the outer sulcus and root cells. Both GJB2 and GJB6 were present in these cell nuclei (Figures 8B,D,E). Generally, the GJB6 dominated over GJB2 except in the basal cells where GJB2 exceeded GJB6 (Figure 9). We tried to assess whether cells may contain either GJB2 or GJB6 transcripts separately or if both are transcribed in the same cell by analyzing serial stacks of images. These results showed that some cell nuclei seemed to contain only one of the two transcripts while others contained both transcripts (Figures 9B–D).

Results of ImageJ particle analysis of the number of GJB2 gene puncta in the lateral wall in the three turns are shown in Table 3. Results showed that there were fewer gene transcripts in the basal turn. This did not seem to result from a reduction of transcripts in each cell, but from a lower number of cells (DAPI) in the SV and spiral ligament at this level.

To better understand the relationship between the GJB2-containing basal cells and the spiral prominence epithelium, we analyzed the co-expression of the tight junction (TJ) protein Claudin-11 and Cx26 (Figure 10A). It showed that basal cells jointly expressed TJ and GJ proteins and that TJ-expressing cells formed a thin layer that separated the basal and marginal cells from each other (Figures 10B,C).

We found a few GJB2 and GJB6 transcript puncta present in the type I spiral ganglion cells as well as in the satellite glia cells (SGCs) (Figure 11). Larger stained areas were also seen in the peripheral cytoplasm of type I cells recognized as conceivably unspecific.

There is uncertainty of whether human GJ channels in the OC and lateral wall of the cochlea are organized as heteromeric or homomeric connexons. Recent studies submit that Cx26 and Cx30 can be separately expressed, forming closely associated GJ plaques with Cx26 subdomains that occasionally appear as “hybrid plaques” (Liu et al., 2016, 2017a). Different GJ channels could populate the same GJ plaque or closely interact with different plaques. This was determined by freeze-fracture immunogold labeling (Fujimoto, 1995) and astrocytes expressing Cx26, Cx30, and Cx43 in rats (Rash et al., 2001). The present study also supports this arrangement. A high degree of complexity of the transcriptional co-regulation mechanism may be present in these sophisticated human cellular networks (Kim et al., 2012). SR-SIM showed that both GJB2 and GJB6 gene transcripts were separately localized in different cells, but could also appear in the same cell. Cx26 and GJB2 gene transcripts were enriched in the basal cell layer, which forms a critical stratum for generating the endo-cochlear potential (EP) by passive movement of potassium ions across its apical membranes (Tasaki and Spyropoulos, 1959; Salt et al., 1987). Cx26 and Cx30 co-staining showed a distinct layer of Cx26-positive cells with less Cx30 (Figure 2). The marginal cells contained neither GJB2 nor GJB6 gene transcripts. The basal cells may comprise preferentially homomeric or homotypic Cx26 and Cx30 hemichannels, while a heterotypic molecular arrangement seems less likely considering the IHC and GJB2 and GJB6 transcript distribution. It may favor the presence of molecularly separate GJs and possibly hybrid junctions with plaques consisting of both Cx26 and Cx30 channels (Liu et al., 2017a,b). They may have different properties and act synergistically, including the generation of the EP, a prerequisite for human hearing. The rich content of the Cx26 protein is reflected by the large number of GJB2 gene transcripts and could clarify the major effects caused by GJB2 gene disruption on human hearing and EP that is not readily compensated for by Cx30. Cochlear cell degeneration of cCx26 null mice was more extensive and rapid than in Cx30 null mice (Sun et al., 2009). One would expect that genetic deletion of one of the genes would result in a similar pathology if it was heterotypically arranged. The dominating effect of Cx26 was also shown by the fact that hearing can be normal without Cx30 in cases where Cx26 levels are preserved (Ahmad et al., 2007; Boulay et al., 2013; Jagger and Forge, 2015).

Notably, intermediate cells only contained GJB6 gene transcripts and Cx30 protein. Transcripts encircled the blood vessels, which is consistent with murine studies showing that Cx30, but not Cx26, is present in the SV’s intermediate cells (Cohen-Salmon et al., 2002). They found that, in homozygous Cx30 mutations, a down-regulation of betaine homocysteine S-methyltransferase in the stria capillaries resulted in endothelial dysfunction and hearing loss. It is not known whether these channels represent GJs or hemichannels that form routes between the extracellular space and cytosol. Such channels have been implicated in paracrine signaling through plasma membrane release of various messengers, such as ATP and glutamate, nicotinamide adenine dinucleotide, and prostaglandins (Wang et al., 2013). They may also be involved in regulating inflammatory responses (Baroja-Mazo et al., 2013; Willebrords et al., 2016).

Supporting cells in the OC, including Hensen and Claudius cells, contain a moderate number of GJB2 and GJB6 gene transcripts, mostly located in the same cell nuclei. The precise number was difficult to assess since fluorescent puncta coalesced into larger assemblies, possibly representing transcriptional bursts. No transcripts were found in the hair cells or in the cells of the tympanic covering layer (TCL) located beneath the basilar membrane. The large number of transcripts in the outer sulcus epithelium, root cells, type I and II fibrocytes, and proteins may support the view that there is a lateral flow of K⁺ ions from the hair cell region to the lateral wall and back to the SV. Many cells contained a large number of transcripts, both in the cell nuclei and cytoplasm, while others contained a lower number. This suggests that cells show a variable transcription activity. SR-SIM and Cx30 immune staining indicated that open hemichannels are also present in the supporting cells in the human OC (Liu et al., 2017a). Inner and outer pillars and Deiters’ cells face cortilymph, and hemichannels could mediate autocrine/paracrine signaling by releasing ATP (Leybaert et al., 2003). They could also play an important role in maintaining low extracellular glutamate levels in order to avoid neuronal death from glutamate excitotoxicity and thereby maintain synaptic transmission (Ye et al., 2003). Cx30 greatly dominates Cx26 in the adult human OC.

GJB2 and GJB6 gene transcripts were also detected in the mesothelial lining of the floor of the scala vestibule and supra-limbus cells. These cells expressed Cx26 and Cx30 (Liu et al., 2016), suggesting that both proteins have the capacity to trap the sequestered K⁺ ions and bring them to the suprastrial fibrocytes in the lateral wall. This could be an additional route for recirculating K⁺ from the inner hair cell region, inner sulcus, and spiral limbus contributing to endolymph production from perilymph (Sterkers et al., 1982; Marcus et al., 2002; Konishi et al., 1978). The GJs were shown in both the supralimbal and mesenchymal cells lining the scala vestibuli in the rat (Kikuchi et al., 1995). More studies are needed in order to verify whether GJs are present between these cells in the HC.

Researchers resembled the basal cell layer with the placenta barrier, where both GJs and TJs regulate the exchange of fluids and small metabolites between blood supplies (Metz et al., 1978; Kikuchi et al., 1995). A breach of this barrier could be detrimental, and a reciprocal regulation between GJ and TJ proteins was suggested (Giepmans and Moolenaar, 1998; Kojima et al., 2002). Gap junctions were shown to influence TJ permeability, such as Claudin and Occludin which also co-immuno-precipitate (Kojima et al., 2001; Dbouk et al., 2009). A similar interaction between GJ and TJ may occur in the cellular networks of the SV. Claudin-11 was shown to be indispensable for hearing and for maintaining the EP, but not of a high K⁺ concentration in the endolymph (Kitajiri et al., 2004a,b). Human studies show that Cx26 partly co-assembles with the TJ Claudin-11 along the basal cell layer. Unexpectedly, GJB2-containing basal cells reached the upper slope of the spiral prominence epithelium and seemed to face endolymph. Staining showed that Claudin-11 insulates basal cells from spiral prominence and endolymph (Liu et al., 2017b) (Figures 10A–C). The TJ protein forms laminar insulations between the GJB2-positive basal cells and spiral prominence epithelium (Figure 10). A similar arrangement exists at the insertion of Reissner’s membrane. This compartmentalization could be pivotal to separating the spiral prominence from the SV in order to polarize K⁺ flow into it. The spiral prominence may be critical as a circuit for ions entering and leaving this region.

There were few GJB2 and GJB6 gene transcripts puncta that were detectable in the spiral ganglion (Figure 11). This is surprising since IHC showed positive staining of both proteins (Liu et al., 2009, 2019, 2021b). Cx30 was localized in discrete deposits in type I spiral ganglion cell bodies while Cx26 was more diffusely stained. Differences may depend on the antibodies that were used. In an earlier study, larger areas were found with assumed GJB6 expression (Liu et al., 2021b). A similar pattern was noted in the present study, but with few transcript puncta present in type I spiral ganglion cells and satellite cells (Figure 11). Further studies using additional molecular techniques are necessary in order to establish the degree to which GJB2 and GJB6 transcripts are present in the human spiral ganglion cells. There was no positive Cx30 staining after double-labeling with TUJ1 or in parvalbumin-positive nerve endings beneath the hair cells. This suggests that Cx30 may not be involved in nerve excitation in the human OC. Cx36 is the principal neuronal connexin in the mammalian CNS, but the Cx36 gene transcript, GJD1, was not detected in an earlier investigation (Liu et al., 2021b).

A noteworthy number of GJB2 and GJB6 gene transcripts in outer sulcus and spiral ligament cells may suggest a high transcription rate. The GJ connexins have been shown to have a remarkably short half-life of just a few hours (Laird, 2006). Connexin proteins are continually synthesized and degraded depending on physiological demands, and genes may be upregulated with increased protein levels and altered localization (Risek et al., 1990). Gap junctions can remodel, such as in cardiac disease where Cx43 serves to electrically convey impulses between myocytes (Dhein and Salameh, 2021). Connexin channels may also change between an active and inactive state, regulated by phosphorylation processes (Bukauskas et al., 2000). A high turnover rate indicates that there is a rapid synthesis and replenishment essential to upholding function. Cell-cell coupling was previously considered stationary, but it is now realized that gene expression and transcription can be highly irregular, with spontaneous variability occurring in some cells due to environmental conditions. Such “burst” transcription has been described in several mammalian tissues (Bahar Halpern et al., 2015). Insults may remodel GJ coupling with rapid degradation in the plasma membrane, removal, and turnover, which are critical for function. Noise- and age-dependent hearing loss have been associated with Cx26 dysregulation (Wang et al., 2014; Wu et al., 2014; Zhou et al., 2016; Fetoni et al., 2018). We found a reduction in the number of expressed genes in the basal turn (Table 3), supporting an earlier finding of diminished Cx26 and Cx30 activity at the high-frequency level (Zhao and Yu, 2006). Further elucidation is also need on whether this relates to age-related hearing loss. Inflammatory changes in spiral ligament cells could also interrupt GJ K⁺ recycling and cause sensorineural hearing loss. Methods to pharmacologically target and modify GJ trafficking have been suggested as a way to restore altered coupling and transcellular electrical activity in other organs. Whether such intriguing strategies could also be applied to the inner ear to treat hearing impairment remains to be determined.