Sodium-hydrogen exchanger 6 (NHE6) deficiency leads to hearing loss, via reduced endosomal signalling through the BDNF/Trk pathway

Acid-base homeostasis is critical for normal growth, development, and hearing function. The sodium–hydrogen exchanger 6 (NHE6), a protein mainly expressed in early and recycling endosomes, plays an important role in regulating organellar pH. Mutations in NHE6 cause complex, slowly progressive neurodegeneration. Little is known about NHE6 function in the mouse cochlea. Here, we found that all NHE isoforms were expressed in wild-type (WT) mouse cochlea. Nhe6 knockout (KO) mice showed significant hearing loss compared to WT littermates. Immunohistochemistry in WT mouse cochlea showed that Nhe6 was localized in the organ of Corti (OC), spiral ganglion (SG), stria vascularis (SV), and afferent nerve fibres. The middle and the inner ears of WT and Nhe6 KO mice were not different morphologically. Given the putative role of NHE6 in early endosomal function, we examined Rab GTPase expression in early and late endosomes. We found no change in Rab5, significantly lower Rab7, and higher Rab11 levels in the Nhe6 KO OC, compared to WT littermates. Because Rabs mediate TrkB endosomal signalling, we evaluated TrkB phosphorylation in the OCs of both strains. Nhe6 KO mice showed significant reductions in TrkB and Akt phosphorylation in the OC. In addition, we examined genes used as markers of SG type I (Slc17a7, Calb1, Pou4f1, Cal2) and type II neurons (Prph, Plk5, Cacna1g). We found that all marker gene expression levels were significantly elevated in the SG of Nhe6 KO mice, compared to WT littermates. Anti-neurofilament factor staining showed axon loss in the cochlear nerves of Nhe6 KO mice compared to WT mice. These findings indicated that BDNF/TrkB signalling was disrupted in the OC of Nhe6 KO mice, probably due to TrkB reduction, caused by over acidification in the absence of NHE6. Thus, our findings demonstrated that NHEs play important roles in normal hearing in the mammalian cochlea.


Results
NHE1 -9 genes are expressed in postnatal WT and Nhe6 KO mouse cochleae. The NHE proteins belong to a large family of transporters known as the solute carrier (SLC) gene superfamily 28,29 . We performed quantitative PCR (qPCR) to investigate NHE expression in cochlear samples from postnatal day 5 (P5) WT and Nhe6 KO mice. We found that all NHE isoforms were expressed in the inner ears of both strains (except NHE6 KO mice lacked Nhe6). However, Nhe3, Nhe5, and Nhe7 expression in the OC, and Nhe1 and Nhe8 expression in the SG were down-regulated in Nhe6 KO mice, compared to WT mice (Fig. 1). Considering these differences in gene expression, we have tested the protein expression levels of NHE3 and NHE6 and found no significant differences between WT and NHE6 KO ( Supplementary Fig. 2).
Adult Nhe6 KO and WT mice show similar cochlear microanatomy. Recently, physiological and pharmacological studies have indicated that NHEs participate in ion homeostasis in the inner ears of guinea pigs and gerbils 30,31 . We stained sections of the temporal bones of adult WT and Nhe6 KO mice with haematoxylin and eosin to study potential differences in cochlear microanatomy. We found no significant differences in inner ear microanatomy between WT (Fig. 2a-e) and Nhe6 KO mice (Fig. 2f-j). All displayed OCs with three rows of OHCs and one row of inner hair cells (IHCs). The basilar membrane, SG neurons, and the SV also showed similar morphology between mouse strains. Finally, the middle ear microanatomy was not significantly different between WT and Nhe6 KO mice ( Supplementary Fig. 1).
NHE6 protein expression patterns in adult mouse cochlea. We first investigated the expression and distribution of NHE6 in the mouse cochlea. Previously, NHE activity was demonstrated in the guinea pig cochlea 32 . We stained formalin-fixed paraffin-embedded adult mouse cochlear sections with specific fluorescently-labelled antibodies against NHE6. Cochlear section of WT mice stained only with secondary antibodies served as negative control (Fig. 3a). As expected, we detected no NHE6 staining in KO mouse cochleae (Fig. 3b). In WT mouse cochleae, the NHE6 protein was found in the IHCs, OHCs, SV, and SG (Fig. 3c). Higher-magnification images revealed strong NHE6 staining in IHCs and OHCs (Fig. 3d). NHE6 staining was also detected in the SV and SG (Fig. 3e,f). Split single channel exposure to myosin VIIa, DAPI and NHE6 could be found in Supplementary Fig. 3.

Adult NHE6 KO mice show elevated auditory thresholds compared to WT littermates. To
extend these studies, we tested hearing function in vivo in WT and Nhe6 KO mice. Mice were tested with broadband click stimuli and pure tones to determine hearing thresholds. Nhe6 KO mice exhibited significantly elevated auditory brainstem response [33][34][35][36] thresholds for click stimuli and sound stimuli delivered at 4 kHz, 12 kHz, and 32 kHz frequencies, compared to WT littermates (Fig. 4a). Representative click-induced ABR recordings from WT and Nhe6 KO mice are shown in Fig. 4b. No differences were observed in the ABR waveforms, amplitudes (Fig. 4c), or inter-peak latencies (Fig. 4d) between WT and Nhe6 KO mice.  56 . In SV samples, no significant difference is observed between WT and KO mice. (Bottom) In SG samples, only in Nhe1 and Nhe8 was down-regulated in KO compared to WT mice. Results are the mean fold-change in transcript levels ± SD, compared to GAPDH, a housekeeping gene. qPCR was done in triplicate (n = 20 mice and 40 OC per strain was pooled) *p < 0.05 (Student's t-test).
are mainly regulated by Rab monomeric GTPases. Rabs regulate vesicular trafficking by controlling the transport, anchoring, and coupling of vesicles through effector binding 40 . Rab5, Rab7, and Rab11 are among the key GTPases known to be involved in BDNF/TrkB signalling 41 .
We detected no significant difference in BDNF protein contents between KO and WT mice (Fig. 5a). Quantification of Western blot signals show BDNF protein expression in OC explants of KO and WT mice (Fig. 5b). Western blot analyses indicated that phosphorylated TrkB (p-Trk) levels in the OC of Nhe6 KO mice neurons were significantly reduced compared to WT mice, while levels of total TrkB (t-TrkB) remain similar (Fig. 5c). These findings suggested that the absence of NHE6 promoted significant differences between WT and KO mice (Fig. 5d).
To investigate Akt activation, we performed Western blots with OC lysates from WT and KO animals. The phosphorylated Akt signal (p-Akt) was significantly lower in the OC of KO mice compared to WT mice (Fig. 6a), indicated by the ratio of p-Akt to total Akt (t-Akt). To characterize the NHE6 endosomal compartment further, we quantified the expression of Rab5 (early endosomes), Rab7 (late endosome), and Rab11 (recycling endosome). Rab5 expression was not significantly different between the strains. Furthermore, Rab7 protein levels were reduced and Rab11 levels were increased in Nhe6 KO mice compared to WT littermates (Fig. 6b,c).  www.nature.com/scientificreports www.nature.com/scientificreports/ cochleae. We performed qPCR in three biological replicates to investigate differences in gene expression. SG II neurons were specifically identified by the expression of peripherin (Prph), Plk5, and Cacna1g. SG Ia neurons were identified by the expression of Slc17a7, Calb1, and Pou4f1. SG Ib neurons were identified by the expression of Lypd1 and Pou4f1. SG Ic neurons were identified by the expression of Calb2. All these genes, except Cacna1g, were differentially expressed in Nhe6 KO mice, compared to WT mice (Fig. 7).
Nhe6 KO mice exhibit a reduction in afferent cochlear neurons. Neurotrophin signalling controls cochlear innervation. A lack of neurotrophin signalling in the cochlea has been well documented in early postnatal animals. This condition results in the loss of cochlear sensory neurons and a region-specific reduction in target innervation along the tonotopic axis 42 . Disruptions in BDNF/Trkb signalling lead to impaired neural growth. Here, we sought to evaluate neural growth by staining cochlear sagittal sections of WT and Nhe6 KO adult mice with an anti-neurofilament (anti-NF200) antibody to visualize afferent innervations (Fig. 8a). Signal intensity quantifications revealed a significant reduction in NF200-positive fibres in the cochlear nerves of Nhe6 KO mice, compared to observations in WT mice (Fig. 8b). www.nature.com/scientificreports www.nature.com/scientificreports/ Discussion NHE6 participates in regulating cytosolic and organellar pH and cell volume. NHE6 also contributes to whole body volume and acid-base homeostasis. Mutations in the NHE6 genes cause neurological disease and contribute to the pathophysiology of multiple human diseases 3 . Little is currently known about NHEs in the inner ear. In this study, we investigated the expression and function of NHE6 in mouse cochlea with the Nhe6 KO mouse. We found that NHE6 mRNA was expressed in all compartments of the mouse cochlea (OC, SG, SV). These results were consistent with those from studies performed in guinea pigs and gerbils 30,31 . Despite the fact that we did not find any morphological differences in the middle or inner ears of WT and Nhe6 KO mice, the Nhe6 KO mice displayed significant hearing loss compared to WT littermates. Nhe6 KO mice had elevated hearing thresholds in click-induced and frequency-specific ABR measurements, compared to WT littermates. The ABR measurements were, on average, about 15 dB different between Nhe6 KO and WT mice, in the acoustic spectrum of 4 kHz to www.nature.com/scientificreports www.nature.com/scientificreports/ 32 kHz. Interestingly, our data did not show any differences in the latencies or amplitudes of the evoked waves between strains. This finding suggested that the auditory stimuli could travel normally along the successive nuclei in the central auditory pathway, once sound levels surpassed the elevated threshold. These findings indicated that hearing loss in Nhe6 KO animals was due to cochlear damage.
As noted, NHE6 loss can lead to a too-steep decrease in endosomal pH and reduced signalling 24,43 which in turn can reshape neuronal arborization through changes in endosomal neurotrophin signalling [44][45][46] . Embryonic cochlear innervation relies on neutrophins and their respective Trks 42,47 , and in model organisms, cochlear sensory neurons are lost without appropriate signalling. Loss of signalling also is associated with region-specific reductions in innervation, decreases that track with the axis of frequency detection 42 . Here, mice lacking Nhe6 had reduced levels of phosphorylated Trk and Akt proteins in the OC compared to WT animals, yet knockout animals had no differences in BDNF protein levels versus WT mice. BDNF, like other neutrophins, is secreted from both the soma and distal neuronal regions, so that signalling to the nucleus can be either local or long distance. One group proposed that the OC does not itself produce BDNF but rather acquires it via the systemic circulation, explaining how levels could still be the same despite strain differences 47 . The PI3K/AKT signalling pathway controls neuronal growth through its effects on axonal and dendritic protein production and cytoskeleton dynamics 38 , and NHE6-associated endosomes have been identified in growing axons and dendrites 24 .
To characterize the NHE6 endosomal compartment in mouse cochlea further, we quantified the levels of Rab5 (early endosomes), Rab7 (late endosomes), and Rab11 (recycling endosomes) in the OCs of both strains. Rab5, 7, and 11 were previously shown to be important for normal neuronal migration and maturation in the cortex, through the regulation of N-cadherin trafficking 48 . That finding revealed that the endocytic pathway played a physiological role in development. We found that Rab5 expression was similar in both strains, but the level of Rab7 protein was reduced in KO, compared to WT animals. These two endosomal proteins mediate endocytosis and promote neuronal growth through BDNF/Trk signalling. The reduction in Rab7 signalling was a consequence of interrupted signalling. In contrast, Rab11 expression was increased in KO samples, compared to WT samples. This result might be explained by findings from Rink et al., who showed that Rab11 was overexpressed in injured neurons, which led to neuronal autophagy. That study showed that neuronal exosomes enriched with miR-21-5p could inhibit neuronal autophagy by targeting Rab11a, which suppressed trauma-induced, autophagy-mediated nerve injury in vitro 13 .
Auditory nerve sound signal processing was hypothesized to originate from the diverse biophysical properties of class I and II SG neuronal fibres. Therefore, we examined the expression of specific markers for types I and II SG neurons. Petipré at al identified four types of SG neurons (i.e., three type I subclass neurons and one type II class), along with numerous new marker genes. They also described a comprehensive genetic framework that www.nature.com/scientificreports www.nature.com/scientificreports/ could shape synaptic communication patterns. In addition, they characterized the differential projection patterns of distinct of type I subclasses to hair cells 27 . In our study, genes specific to both types I and II neurons were up-regulated in the SGs of Nhe6 KO mice. In contrast, immunohistochemistry showed a reduced number of neuronal fibres in the cochlear nerve of Nhe6 KO mice. Additionally, mouse brains with disrupted NHE6 gene displayed reduced axonal and dendritic branching, reduced synapse numbers, and reduced circuit strength 24,49 . We assumed that these findings might be explained by the fact that the SG samples were derived from postnatal mice, and there is evidence that perinatal neuronal diversification occurs before the onset of hearing 27,50 . We showed that BDNF/Trk signalling was impaired in the cochlea of Nhe6 KO mice, which resulted in reductions in neuronal production and development. We found that the other genes described by Petipré et al. were up-regulated in the present study, which might reflect a compensatory effect in the SG of Nhe6 KO mice.
In conclusion, our study showed that Nhe6 KO mice displayed significant, moderate hearing loss. NHE6 was expressed in all WT cochlear tissues. Our findings indicated that the depletion of NHE6 had an effect on Trk protein turnover and endosomal signalling. These results suggested that the failure of dendritic and axonal growth Western blotting. We have previously described preparation of protein samples from the cochlea of P5 pups 53 . Briefly, samples were placed in cell lysis buffer with a protease inhibitor cocktail (Sigma C3228, P8340), followed by homogenization on ice for 1 min. We used the NanoDrop 1000 (Thermo Scientific) to measure proteins, with mouse brain lysate serving as control. Lysates then were mixed with Laemmli sample buffer (Sigma) in equal amounts and heated for 5 min at 95 °C. After sample resolution on SDS-PAGE gels (10 μg protein per lane), gels were blotted onto polyvinylidene fluoride membranes. Following blockage of nonspecific sites with 5% nonfat dry milk in PBS for 1 h at room temperature (RT), we incubated membranes with primary antibodies in PBS-Tween. Primary antibodies were as follows: mouse monoclonal anti-p-TrkB (1:1000, Santa Cruz Biotechnology, sc-8058), mouse monoclonal anti-t-TrkB (1:1000, Invitrogen, MA5-14903), rabbit polyclonal anti-p-Akt 1:1000 (Cell Signalling, #9275), rabbit polyclonal anti-t-Akt 1:1000 (Cell Signalling, #9272), rabbit polyclonal anti-BDNF (1:1000, Sigma AG, AB1779SP), and rabbit monoclonal anti-Rab5, 7, 11 (1:1000, Cell Signalling, #3547; #9367; #5589). Secondary antibodies were anti-mouse (1:3000, Cell Signalling, #7076P2) and horseradish peroxidase-linked anti-rabbit (1:2000, Cell Signalling, #7074P2). Membranes were incubated overnight with primary antibodies at 4 °C, washed with PBS-Tween (3 × 10 min), and incubated with an appropriate secondary antibody at RT for 1 h. After washings, Super Signal West Dura Extended Duration Substrate (Thermo Scientific, Switzerland) was used to visualize bands, with rabbit polyclonal β-actin (1:1000, Cell Signalling, #4967) to control for protein loading.
For histological evaluations, 10-µm sections were made using a Leitz microtome and mounted on Superfrost-Plus slides (Menzel, Braunschweig, Germany). Immunohistochemistry has been described before 54 , but briefly, sections were deparaffinized, rehydrated, and washed in PBS for 5 min, followed by incubation in blocking solution (TBS: 50 mM Tris, 0.9% NaCl; 0.5% Triton X-100, pH 7.35; 3% normal goat serum [NGS]) for 1 h at RT. For antibody binding, we incubated the sections with a mouse monoclonal antibody against myosin VIIa (1:350, Abcam, ab150386), primary anti-NF200 (1:500, Sigma AG, N4142), and rabbit anti-NHE6 (1:1000). To generate a rabbit polyclonal antibody targeting the NHE6 carboxyl terminus, we used cysteine bonding to couple maleimide-activated KLH with a synthetic peptide containing the last 13 amino acids of mouse NHE6. Rabbit immunization took place at UT Southwestern, following IACUC-approved protocols 55  www.nature.com/scientificreports www.nature.com/scientificreports/ of antibodies in TBS with 1% NGS at 4 °C and then three TBS washes, sections were incubated with the appropriate Alexa-conjugated secondary antibodies (1:250; Molecular Probes, Lubio Science, Switzerland). Antibodies were diluted in TBS with 1% NGS, and incubations were done at RT for 2 h. Following another TBS wash and DAPI counterstain, sections were mounted on glass slides using Mowiol. Negative controls were samples of WT tissues. To visualize binding, we used an Olympus AX-70 microscope equipped with a spot digital camera and adjusted recorded images for brightness and contrast using Image-Pro Plus and Photoshop.
Haematoxylin and eosin staining also was performed on some paraffin-embedded sections. As we previously described 52 , haematoxylin staining (Sigma Aldrich, Switzerland) was performed on sections for 5 min, followed by a 15-min rinse in tap water and immersion in eosin 1% aqueous (Eosin Y disodium salt, in deionized water; Sigma Aldrich, Switzerland) for 1-2 min. Before use, the solutions were filtered (Baxter grade 363, qualitative), and sections were rinsed in tap water until the water was clear. After dehydration in ascending alcohol concentrations (50%, 70%, 80%, 2× 95%, and 2× 100%), xylene clearance (2 times), and mounting with Eukitt (Fluka, Sigma Aldrich, Switzerland), sections were imaged using an Olympus IX50 microscope with a spot digital camera. As above, images were adjusted for brightness and contrast using Image-Pro plus and Photoshop.
Hearing function tests. We administered the ABR test to all animals before treatment, as described previously 53 (Tucker-Davis Technologies; TDT -RZ6-A-P1 hardware and software; Alachua, FL, USA), followed by their anesthetization with intraperitoneal ketamine (80 mg/kg body weight [BW]; Graeub AG, Switzerland), 12 mg/kg BW xylazine (Graeub AG, Switzerland), and 2 mg/kg BW acepromazine (Arovet AG, Switzerland). Testing was performed in sound-attenuating acoustic chambers, with infrared warming pads (Kent Scientific Corporation, USA) to keep animal body temperature at 38 °C. BioSig software was used to synthesize the tone burst acoustic stimuli (duration, 10 ms; rise/fall time, 0.5 ms; Blackman window), with an SA1 audio amplifier and a Multi Field Speaker-Stereo (MF1-S) as transducer. For testing, we placed an inverting electrode over the mastoid of the target ear, a non-inverting needle electrode at the midline vertex, and a ground electrode on the upper hindlimb. The speaker (MF1-S) was placed in the ear canal. Electrodes collected the signals, which then were amplified 20× with band-pass filters (100 Hz to 5 kHz) and input into a real-time processor (RA4PA) for software processing (BioSig RP software; TDT). We evaluated thresholds at 4, 12, or 32 kHz. Signals were preamplified with a gain of 20, and for each trial, we averaged 1000 sweeps. The starting level for stimuli presentation was 90 dB, which was incrementally decreased by 5 dB to threshold (the lowest intensity at which a visible, repeatable ABR wave could be observed in two averaged runs) and the ABR wave disappeared for each frequency. We calculated the amplitude and latency growth input-output function slopes (i.e., amplitude and latency as a function of stimulus intensity), as described before 53-55 . Statistical analysis. Statistical analyses were performed with the Student's t-test for unpaired samples.
When more than two groups were compared, we performed a one-way analysis of variance (ANOVA), followed by the Bonferroni post-hoc test. A p value <0.05 was considered statistically significant. All statistical tests were two-sided. Data were analysed with GraphPad Prism 7 software (La Jolla, CA, USA).