Redox activation of excitatory pathways in auditory neurons as mechanism of age-related hearing loss

Age-related hearing (ARHL) loss affects a large part of the human population with a major impact on our aging societies. Yet, underlying mechanisms are not understood, and no validated therapy or prevention exists. NADPH oxidases (NOX), are important sources of reactive oxygen species (ROS) in the cochlea and might therefore be involved in the pathogenesis of ARHL. Here we investigate ARHL in a mouse model. Wild type mice showed early loss of hearing and cochlear integrity, while animals deficient in the NOX subunit p22phox remained unaffected up to six months. Genes of the excitatory pathway were down-regulated in p22phox-deficient auditory neurons. Our results demonstrate that NOX activity leads to upregulation of genes of the excitatory pathway, to excitotoxic cochlear damage, and ultimately to ARHL. In the absence of functional NOXs, aging mice conserve hearing and cochlear morphology. Our study offers new insights into pathomechanisms and future therapeutic targets of ARHL.


Introduction
Presbycusis or age-related hearing loss is a frequent, degenerative neurosensory disorder in ageing societies. Presbycusis affects more than a third of the elderly population at retirement age [1]. Hearing loss in this vulnerable population substantially contributes to cognitive decline, depression and social isolation [2,3] and represents an important socio-economic burden. Hearing aids and cochlear implants alleviate symptoms, but are not a causal treatment. Finding new treatments to reverse or slow down the progression of age-related hearing loss will have major consequences for the affected individuals and for society as a whole.
The etiology of presbycusis is multifactorial and involves a complex interaction of genes. It can be triggered by a variety of external and internal factors such as noise exposure, ototoxic molecules and medical conditions, among others [4]. The histopathological correlates of agerelated hearing loss in the inner ear are the loss of mechanosensitive hair cells, spiral ganglion neurons, supporting cells and cells in the stria vascularis with a gradient from base to apex [5]. Therefore, the typical age-related hearing loss is predominantly affecting the higher frequencies, which are coded at the cochlear base.
Oxidative stress, i.e. high levels of reactive oxygen species (ROS), has traditionally been thought to lead to tissue damage and eventually cell death through non-specific protein, lipid and DNA oxidation. However more recent concepts rather favor a role of physiological ROS levels in cellular signaling [6], and a dysregulation of signaling through excessive ROS levels [7]. Abundant evidence demonstrates changes in the cochlear redox environment with age [8]. For instance, several antioxidant genes, including SOD2, GST or UCP, have been associated to age-related hearing loss [9] and genetic mouse models deficient for antioxidant genes, including Nrf2 [10,11], Sod1 [12] or genes regulating the mTOR pathway [13], show accelerated age-related hearing loss. Furthermore, age-related hearing loss (ARHL) has been shown to be slowed by supplementation with antioxidants in laboratory animals, and a few studies have investigated the effect of antioxidants supplements against ARHL in humans [14]. While a role of excessive oxidant generation has been widely described as causative of hearing loss, the exact sources of oxidants are unclear. Age-related hearing loss in wild type A/J mice: functional and morphological alterations in auditory neurons precede the loss of sensory epithelium. A) Auditory brainstem response hearing thresholds of A/J mice were assessed from 4 to 26 weeks old animals, on a group of 11 WT A/J mice following clicks and pure tone frequencies stimulation. B) Elevation of hearing threshold is accompanied with a decrease of the ABR maximal amplitude (peak to peak amplitude, (p-p)), assessed at 75 dB SPL. C-F) Basal turn representative sensory epithelium immunostainings (Myo7a in green) of the WT A/J cochlea at 4 weeks old (C), 6 weeks old (D), 18 weeks old (E) and 26 weeks old (F). G-J) On the same samples, the number of synaptic ribbons between inner hair cells (nucleus in white) and spiral ganglion neurons was also determined using CtBP2 immunostaining (red dots), using 4 weeks old (G), 6 weeks old (H), 18 weeks old (I) and 26 weeks old (J) mice. K-N) Representative mid-modiolar hematoxilin-eosin staining of A/J mouse basal cochlear turns at 4 weeks old (K), 6 weeks old (L), 18 weeks old (M) and 26 weeks old (N). O) Bar graph showing the quantification of outer hair cells (green), synaptic ribbons (red), SGN density (pink), ABR wave I amplitude (light grey) and ABR threshold (black) in the 6 and 18 weeks old mice basal turn, relatively to the 4 weeks old group that serve as reference (100%). n = 11 animals. Scale bar (C-F) = 25 μm, (G-J) = 10 μm and (K-N) = 50 μm. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) NADPH oxidases (NOX) are a family of enzymes whose main biochemical function is the production of ROSnamely superoxide radical anion O 2 •and H 2 O 2 . In mammals, the NOX family consists of seven isoforms (NOX1-5, DUOX1, 2). NOX have several subunits and the p22 phox subunit is crucial for the function of several NOX isoforms, namely NOX1 to NOX4. In fact, p22 phox is essential for NOX stabilisation and activity: the CYBA CRISPR knockout is devoid of ROS generation in NOX1, NOX2, NOX3 and NOX4 expressing cells [15]. Therefore, p22 phox is a master regulator of ROS generation as it regulates most NOX-derived oxidants. The biological function of NOX-derived ROS is broad, from host defense, to cellular signaling and hormone biosynthesis. NOX-derived ROS and in particular H 2 O 2 , resulting from superoxide dismutation are important second messengers in cell signaling. Through a reversible reaction with H 2 O 2 , cysteine residues are oxidized, changing the function of the respective protein [16]. Thereby, ROS regulates the activity of protein tyrosine phosphatases [17], permeability of ion channels [18] or affinity of transcription factors for their target DNA sequence [19] and thus regulate important physiological function in the cell (i.e. proliferation, differentiation, survival metabolism or motility). While NOX have important physiological functions in virtually all organ systems, an over-activation of these enzyme systems can lead to oxidative stress through overproduction of ROS and ultimately to oxidative stress-driven disease (e.g. fibrosis, cardiovascular disease, neurodegeneration). We hypothesize that in the inner ear, NOX activation results in hearing loss. NOX3 is highly and exclusively expressed in the inner ear [20,21], and its function has been mainly attributed to otoconia formation in the developing vestibular system, whereas its function remains unknown in the cochlear tissues [20,22]. Nox3 mutant mice develop a vestibular deficiency, leading to a head-tilt phenotype. Mice with a loss of function mutation of the p22 phox subunit show a similar vestibular phenotype as NOX3 mutant mice with, in addition, a defect of innate immunity due to the absence of NOX2 [23]. Although growing evidence suggests a primary role of NOX and in particular NOX3 in different cochlear pathologies [24]; the contribution of NOX isoforms in agerelated hearing loss has never been investigated. The fact that NOX3 is exclusively expressed in the inner ear makes it a prime target for interventions aiming at slowing down ROS production and consecutive cell-damage in the inner ear. While NOX3 is only found in the inner ear, more broadly expressed NOX family members are also present in the inner ear and may contribute to hearing and balance disorders. Indeed, NOX2 is highly expressed in microglia cells, which are abundant in spiral ganglia [25], and inhibition of microglia activation was protective in a mouse model of neomycin-induced hearing loss [26]. NOX4 is strongly expressed in vascular endothelium and could play a role in the stria vascularis. And indeed, NOX4 overexpressing transgenic mice show an increased sensitivity to noise-induced hearing loss [27].
Given the central role of p22 phox in NOX activity, p22 phox deficient mice are a particularly useful tool for studying the role of NOX in inner ear [23]. Indeed, considering the absence of Nox5 in the mouse, p22 phox deficient mice can be defined as pan-Nox-deficient mouse model [28]. The nmf333 mouse model (A.B6 Tyr + nmf333jax; A/J genetic background) harbors a missense mutation in CYBA leading to a functional inactivation of NOX1-4 [29]. The A/J strain shows a very early age-related hearing loss and is therefore an interesting model to study oxidative-related pathologies of the inner ear and to evaluate the effects of potential otoprotective drugs [30].
In the present study, the A/J mouse model (nmf333) was used to investigate the impact of NOX in the progression of age-related hearing loss. Our data demonstrate that p22 phox is a key regulator of age-dependent hearing loss. Mechanistically, we show that p22 phox regulates the calcium release from intracellular stores in auditory neurons and that deletion of p22 phox protects from excitotoxic insult in auditory neurons. Since, p22 phox is a key regulator of redox pathways due to its main function as controlling NOX activity, our data suggest a role of oxidant-derived NOX in age-related auditory neuropathy and show promises for novel therapeutic approaches for future treatment of agerelated hearing pathologies.

Results
A/J mouse as model for age-related hearing loss. In order to characterize the early onset hearing loss in A/J mice, we investigated hearing thresholds starting over an age range from 4 to 26 weeks ( Fig. 1A and B). Click-evoked auditory brain stem response (ABR) hearing thresholds were close to 45 dB SPL in young adult mice (4 weeks old) and progressed up to 75 dB SPL in old mice (Fig. 1A). The pure tone frequencies measurement (2-32 kHz) was more informative: while young mice audiogram was comprised between 2 and 32 kHz (with a threshold of 30 dB SPL at 11.3 kHz) ( Fig. 1A; empty squares) we observed a progressive loss of hearing at high frequencies narrowing the hearing spectrum (from 32 kHz at 4 weeks to 5.7 kHz at 26 weeks old). Accordingly, maximal amplitude of the ABR decreased with age (Fig. 1B). The most important hearing loss was observed at around 11.3-16 kHz (Fig. 1A). Consistent with the high frequency hearing loss phenotype, we observed a progressive degeneration of the sensory epithelium from the base to the apical turn ( Supplementary Fig. 1). While the apical turn was relatively well conserved in old mice, there was an important cellular degeneration in the basal region ( Fig. 1C-F). This was accompanied by a dramatic decrease in the number of synaptic ribbons per inner hair cell (IHC) (Fig. 1G-J) and a marked decrease in the neuronal density in the spiral ganglion with age ( Fig. 1K-N). Interestingly, the kinetics of degeneration, as observed in the basal turn of the cochlea, showed synaptic and post synaptic morphological alteration (decrease in number of ribbons per IHC, auditory neuron density and in the ABR wave I) occurring prior to the sensory epithelium degeneration ( Fig. 1O; the dotted line represent control values obtained with mice at 4 weeks of age). Therefore, we conclude that A/J mice have an early onset age-related hearing loss which is mainly caused by neural degeneration and subsequent loss of the sensory epithelium.
Pattern of NOX expression in the cochlea. While NOX3 is known to be highly enriched in the inner ear, the presence of other NOX isoforms was not studied [24,31]. Therefore, we addressed the expression of NOX isoforms in the cochlea of the human fetus ( Fig. 2A) and the adult A/J mouse ( Fig. 2B-K). By qPCR, we found relatively high levels of NOX2, NOX3 and NOX4 mRNA (CT values below 30) in both human ( Fig. 2A) and mouse cochlea (Fig. 2B). In both species, NOX1 and DUOX1/2 levels in the cochlea were below detection threshold. NOX5 levels in the human cochlea were low (Ct values [33][34][35]. Note that NOX5 gene is absent in mice and rats [28]. The fetal cochlear tissues used for real time qPCR is heterogenous and contained spiral ganglion neurons, as well as parts of the stria vascularis. We therefore aimed to obtain more precise information about the localization of the NOX mRNA. There are major issues with antibodies for the different NOX isoforms [32]. We therefore rather performed in situ hybridization (using the RNAscope technology) in order to localize the mRNA of the different isoforms ( Fig. 2C-K). We compared expression of Nox2, Nox3 and Nox4 in three different subparts of the cochlea, namely the organ of Corti, the stria vascularis and the spiral ganglia. Signals with the Nox2 and Nox4 probes ( Supplementary Fig. 2) were detected throughout the cochlea without a strong signal in specific parts. In contrast, the NOX3 probe gave a very strong signal in the spiral ganglia (Fig. 2H), comparable or even stronger than the control gene (Ppib, Fig. 2F; Supplementary Fig. 3). The Nox3 signal also appeared above background in the stria vascularis (Fig. 2K), but not in hair cells (Fig. 2E).
p22 phox deficient A/J mice are protected from early onset hearing loss. As several NOX isoforms are present in the inner ear, we next investigated hearing of A/J mice, harboring a loss of function mutation on the NOX1-4 subunit p22 phox (nmf333) (Fig. 3). For comparison, the results of wild type mice, as already described in Fig. 1, are added. Hearing thresholds of mice were investigated monthly from 4 to 22 weeks old ( Fig. 3A-F). Four weeks old p22 phox mutant mice, exhibited an audiogram profile similar to wild type littermates, indicating that there is no basal difference in hearing between p22 phox mutant and WT mice (Fig. 3A). As already described in Fig. 1, from an age of 6 weeks, high frequencies hearing loss was observed in WT mice ( Fig. 3B-F, Fig. 3K). In contrast, only minor hearing impairment was observed in mutant mice. In particular the high frequency hearing loss was almost completely prevented in p22 phox deficient animals up to 18 weeks old (Fig. 3K). In fact, the pattern of hearing loss was distinct in p22 phox deficient mice. At 22 weeks old, p22 phox mutant showed only a minor (statistically not significant) hearing loss. This minor hearing loss appeared to be equally distributed along the tonotopic axis (Fig. 2F). This is in contrast to the WT animal where the amplitude of the ABR recorded at 75 dB SPL consistently showed age-dependent decrease at high frequencies ( Fig. 3H; Supplementary Fig. 4) and following click stimuli (Fig. 3G). In addition to the hearing threshold elevation, WT mice exhibited significant delay in the ABR wave I with age ( Fig. 3I and J). Together, these experiments clearly demonstrate the role of p22 phoxdependent NOX in age-related hearing loss in A/J mice.
Conserved architecture of the sensory epithelium in p22 phox deficient mice. Mice were sacrificed at different ages (6, 18 and 26 weeks old) for evaluation of cochlear histomorphology (Fig. 4, Fig. 5 and Supplementary Fig. 5). In young animals (6 weeks old), approximately at the onset of hearing loss, the sensory epithelium was overall well conserved and comparable between WT and p22 phox animals ( Fig. 4A-E). However, in contrast to p22 phox deficient mice, we    Remarkably, architecture of the sensory epithelium of p22 phox deficient mice was fully preserved. In the apical turn, few missing OHC were observed in both WT and p22 phox deficient mice but without significant differences. Accordingly in the basal turn of the cochlea, synaptic ribbons were virtually absent in wild type mice, while they were well conserved in p22 phox deficient mice (Supplementary Figs. 5F-J and Fig. 5E-J). No significant differences were observed in the apex. Overall, we observed that cochlear morphological and functional changes observed in A/J mice appeared first at the synaptic/postsynaptic level, anteceding the alteration of the sensory epithelium. These damages were prevented by p22 phox (Cyba) loss of function.
Gene expression profiling in the cochlea from wild type and p22 phox deficient mice. To further explore molecular pathways underlying the protective effect of p22 phox deletion in age-related hearing loss, we performed cochlear transcriptome analysis of p22 phox deficient mice and WT littermates (Fig. 6). We chose young animals (6 weeks) to identify early transcriptome alterations, anteceding the onset of hearing loss. Our results revealed genes associated with several molecular functions that were statistically significantly downregulated in p22 phox mutants (Fig. 6A). The most redundant molecular function which was lower in CYBA mutant mice, is related to Ca 2+ homeostasis, and in particular ryanodine dependent Ca 2+ release (FDR < 0.05). Relevant genes include all 3 isoforms of the ryanodine receptors (Ryr1, 2, 3), involved in Ca 2+ release from intracellular organelles, Otoferlin, Vamp1 and Snap25, involved in Ca 2+ dependent pre-synaptic vesicle fusion and Slc17a6, involved in glutamate transport (Fig. 6B). We also identified other genes involved in the auditory synapse function that showed a trend towards lower expression albeit not significant in the RNAseq analysis: the inner hair cells glutamate transporter (Slc17a8), glutamate ionotropic receptor AMPA type subunit 2 (Gria2) and Adenosine A1 receptor (Adora1) (data not shown). qPCR results confirmed the significantly lower expression of ryanodine receptor 1, 2, 3, otoferlin, Slc17a6, Slc17a8 and Gria2 (Fig. 6C). Except for genes associated with ribbons (presynaptic: Vamp1, Snap25 and Otof), all genes were significantly enriched in the SGN (postsynaptic Ryr1-3 Gria2, Adora1) ( Supplementary Fig. 6). Note that Ryr3 was poorly expressed in both pre and post-synapse (Ct value of 32.6). Interestingly, downregulated genes were all centered on Ca 2+ and glutamate signaling and might have a major role in the excitatory pathway of auditory neurons, a wellknown molecular pathway of hearing loss [8].
Analysis of the impact of adenosine and ryanodine on glutamate-induced excitation of auditory neurons. To address the functional relevance of these genes in the excitatory response, we measured Ca 2+ mobilization in the cytosol of auditory neurons in response to glutamate ( Fig. 6D and E). Auditory neurons were obtained from spiral ganglia sphere forming stem cells as previously described [33]. While the genesis of spiral ganglia from this system has been described previously, so far no studies addressing the glutamatergic function have been performed on such a model. Upon glutamate stimulation, robust and immediate increase in cytosolic Ca 2+ concentration was observed in the auditory neurons (Fig. 6D). This excitatory response was dose dependent and showed an EC 50 in the micromolar range (Fig. 6E), similar to glutamate dose response curves reported in other systems [34,35]. In cortical neurons, adenosine, through adenosine receptor type I (Adora1), was shown to antagonize the glutamatergic pathway [36]. Interestingly, Adora1 tended to be upregulated in mutant mice cochlea (FDR = 0.1443). To investigate the functional relevance of ryanodine receptors, we pre-incubated auditory neurons for 30 min with ryanodine, which under these conditions inhibits glutamate-induced Ca 2+ release from the ryanodine-sensitive Ca 2+ stores. We observed a dose-dependent inhibition of Ca 2+ release by ryanodine with a maximal inhibition of 60%, suggesting that a substantial portion of glutamate-induced Ca 2+ elevations in auditory neurons originates from ryanodine-sensitive Ca 2+ stores (Fig. 6F, blue line and symbols). We next investigated the functional impact of activation of adenosine receptors. We observed a dose-dependent inhibition of cytosolic Ca 2+ elevation by adenosine with a maximal inhibition of 40%. Thus, similar as seen in other types of neurons, adenosine receptors are inhibitors of glutamatergic cell activation stores (Fig. 6F, magenta line and symbols).
Decreased Ca 2+ induced Ca 2+ release (CICR) in p22 phox deficient auditory neurons confers a protection against glutamatemediated excitotoxicity. To further characterize the excitatory function of p22 phox −/− auditory neurons, we compared the expression of key genes involved in glutamate response, namely ryanodine receptors and glutamate ionotropic receptor Gria2 (Fig. 7A). The data demonstrate significant downregulation of ryanodine receptor 1 and 2, as well as Gria2 in p22 phox -deficient auditory neurons (Fig. 7A). To understand the physiological relevance of this differential expression, we compared glutamate-induced Ca 2+ release in sphere-derived auditory neurons from WT and p22 phox −/− mice (Fig. 7B). Upon glutamate stimulation, we observed a dose-dependent increase of Ca 2+ cytosolic concentration with a comparable EC 50 in both WT (5.635 μM) and p22 phox −/− (5.183 μM) auditory neurons. However, the overall amplitude of the signal was significantly lower in p22 phox −/− auditory neurons (Fig. 7B; Supplementary Fig. 7). Furthermore, in absence of p22 phox , the ryanodine-sensitive Ca 2+ release was significantly decreased ( Fig. 7C and D), consistent with the lower expression of Ryr1 and 2 (Fig. 7A). Thus, we identified a functional role for p22 phox in auditory neurons, namely the control of Ca 2+ release from ryanodinesensitive stores. Based on the results of the gene expression analysis, as well as the functional characterization of the p22 phox −/− and WT spiral ganglia cells, we hypothesized that the lower amplitude of the CICR in p22 phox deficient auditory neurons is protective against glutamate induced excitotoxicity. We therefore compared the impact of glutamate on level of cellular integrity (ATP levels) and morphology of WT and p22 phox −/− spiral ganglia cells (Fig. 7E-G). Upon glutamate exposure, WT auditory neurons showed dramatic morphological changes, with, in particular loss of neurites (Fig. 7E). Following glutamate exposure, neurite length in the WT cells was decreased by 50% (Fig. 7F). Remarkably, p22 phox deficient auditory neurons were almost fully protected from glutamate induced morphological changes with only minor decrease in average neurite length (not statistically significant). The data also showed significant decrease in ATP concentration in WT SGN in a glutamate concentration dependent manner (Fig. 7G). By contrast, the ATP content did not significantly decrease after glutamate exposure in p22 phox -deficient cells. Together, the data demonstrate that downregulation of the excitatory pathway confers a protection against glutamate mediated toxicity in p22 phox deficient auditory neurons.

Discussion
Our results provide the new evidence that the absence of functional p22 phox in the inner ear robustly protects from age-related hearing loss and related morphological changes in mice. Mechanistically, we found that the protection is mediated by decrease of the excitatory calcium/ glutamate signaling pathway in auditory neurons.
Our data provide an in-depth description of the early and progressive hearing loss in A/J mice over the first 6 months of life, beyond and in line with earlier reports [37,38]. In this background, onset of hearing loss occurs at 6 weeks of age together with predominant signs of synaptic and post-synaptic damages, including loss of auditory neurons and synaptic ribbons, as well as a decreased amplitude of ABR wave 1. Consistent with recent report in human [39], the auditory neuropathy precedes loss of hair cells, which occurs at a later stage in our model. This model mimics the progression of human presbycusis as the hearing loss in A/J mice starts in the high frequencies while lower frequencies are only affected later in life. Three mutations known to affect hearing are present in A/J mice, thereby providing structural and metabolic explanations for the severe and rapid age-related hearing loss. These include: i) a mutation in the age-related hearing loss-1 (ahl1) locus, which codes for cadherin 23, a structural protein of the stereocilia; ii) a mutation in ahl4, coding for citrate synthase, a mitochondrial enzyme of the Krebs cycle, important for ATP and NADPH synthesis and iii) mitochondrially encoded tRNA arginine, leading to a decreased efficiency in amino acid incorporation. Interestingly, a recent in vitro study demonstrated that decreased citrate synthase expression leads to increased production of ROS [37]. This increase of ROS was proposed to be due to a decrease in the cellular antioxidant system. However, the fact that p22 phox mutation can almost fully prevent the age-related hearing loss in spite of the presence of the above-mentioned mutations suggests that p22 phox and/or its downstream pathways are major driver of age-related hearing loss.
The A/J mouse model is a powerful tool for the study of age-related hearing loss and otoprotection. As the absence of p22 phox abolishes NOX1-4 catalytic activity, our data strongly support a role of NOXderived ROS in presbycusis. However, the NOX isoform involved in hearing loss remains to be identified. Nevertheless as of today, the differential distribution of NOX isoforms in the inner ear has not been clearly elucidated. Given the problems of specificity of commercially available NOX antibodies [32], we focused our efforts on detection of NOX mRNA by qPCR and in situ hybridization providing a striking similarity of NOX isoform expression between human and mouse. The qPCR experiments confirmed high expression levels of NOX3 and p22 phox , but also provided evidence of high levels of NOX2 and NOX4 in the cochlea. The in-situ hybridization experiments using the high-resolution technology RNAscope® yielded relevant new information. Nox2 and Nox4 mRNA appeared equally distributed throughout the cochlea. We had rather expected some strong Nox2-positive microglia cells in the region of the spiral ganglion. Note however that quiescent microglia express only low levels of Nox2, which most likely explains our observations. The most striking result was the high level of Nox3 mRNA in the spiral ganglion cells and its virtual absence in hair cells. This finding supports a role of Nox3 in the morphological changes observed in the cochleae of 6 weeks old animals, where at the first onset of hearing loss, ribbons and SGN density are affected but not hair cells. At this point, it is not clear whether the p22 phox -dependent degeneration of hair cells is secondary to degeneration of SGN or whether it is an independent process that occurs later in time. Concerning the first option, it is unlikely that degeneration of SGN by itself leads to loss of outer hair cells, as at least in the case of peripherin knock-out mice, SGN, but not OHCs are lost [40]. However we cannot exclude some NOX3-dependent efferent signaling mechanism that leads to a secondary OHC death. Yet, the second option, namely a delayed SGN-independent OHC death, certainly appears possible. While under physiological conditions in 6 week old mice, expression levels of NOX3 was low, it might be upregulated with aging. Indeed, there are indications that upon stress, Nox3 might be upregulated in hair cells [31]. Alternatively other p22 phox -dependent NOX enzymes might be upregulated in the aging mice. Cell type-specific NOX knock-out will be necessary to definitively clarify this issue.
The magnitude of the protective effect of p22 phox deficiency is remarkable. Indeed, up to 6 month of age, there was only a minor hearing loss in p22 phox mutant mice, while wild type A/J mice were virtually deaf at this age. This raises the question how the lack of p22 phox can have such a profound effect. Indeed, NADPH oxidase are not the only source of ROS in the cochlea (e.g. mitochondria are probably important), and mechanisms other than ROS are also likely to play a role in cochlear damage [8]. Thus, p22 phox may represent "the straw that breaks the camel's back", i.e. the cumulative effect of many insults leads to hearing loss and by removing one of them (i.e. p22 phox ), the function of the cochlea is maintained.
Our results demonstrate that p22 phox loss of function is strikingly important in sensory neurons as it leads to a lower expression of genes of the neuronal excitatory pathway in the cochlea. This supports a key role of NOX-derived ROS in this mechanism. Downregulated genes harbor functions in Ca 2+ signaling (ryanodine receptors 1,2,3), in synaptic transmission (Otoferlin, Snap25 and Vamp1), as well as in the glutamatergic pathway (Slc17a6, Slc17a8 and Gria2). The cytosolic free Ca 2+ concentration plays a key role in pre-and post-synaptic auditory signal transmission. On the presynaptic side, Ca 2+ influx through Ca 2+ channels activates synaptic vesicle fusion and subsequent glutamate release [41][42][43]. On the post-synaptic side, glutamatergic transmission involve the glutamate ionotropic AMPA type subunit 2 (Gria2) leading to Ca 2+ increase in auditory neurons [44]. The lower expression of neuronal excitatory genes may be explained by a delay in development as previously described [45] or to a downregulation of these genes in p22 phox mutant mice. Interestingly, lower expression levels of ryanodine receptors in p22 phox mutant mice are sufficient for normal hearing. However, our data also suggest that enhanced expression of ryanodine receptors leads to glutamate-induced toxicity and subsequent hearing loss, which was mitigated in p22 phox loss of function mice. In our in vitro auditory neuron model, we provide direct evidence that ryanodine receptors are involved in the post-synaptic Ca 2+ signal: glutamate-induced Ca 2+ elevations are inhibited by ryanodine in a dose-dependent manner. From the relative inhibition of the Ca 2+ signal, we can conclude that ryanodine receptors are important mediators of the F. Rousset, et al.
Redox Biology 30 (2020) 101434 excitatory signal in the cochlea at the post synaptic level. This observation is consistent with previous findings that AMPA-type glutamate receptors and intracellular Ca 2+ stores are coupled via RyR-dependent CICR in primary auditory neurons [46]. Furthermore, intracochlear perfusion of ryanodine in guinea pigs in vivo has been shown to decrease auditory nerve compound action potentials, indicating the importance of RyRs in cochlear transduction [47]. Direct activation of Ryr2 by ROS is well-documented [48,49], however, in the present study, transcriptomic data suggest that NOX-derived ROS rather act at the transcriptional level. Further characterization of the promoters and the transcriptional machinery will be required to understand whether specific redox-sensitive transcription factors are involved.
To explore key finding of our study, namely a protection of p22 phoxdeficient animals from age-related hearing loss, other findings need to be discussed. The A1AR adenosine receptor has been shown to be linked to hearing loss and NOX signaling: mice deficient for the A1AR exhibit greater susceptibility to noise exposure [50] and A1AR activation confers protection against cisplatin ototoxicity through downregulation of NOX3 [51]. Our results show only a trend towards increased levels of A1AR in p22 phox -deficient mice (p = 0.0064; FDR = 0.1443). Interestingly, A1AR is an inhibitor of the excitatory pathway [52,53], and in our hands, adenosine significantly decreased the excitatory signal in auditory neurons (Fig. 6). This might provide an additional element for the understanding why p22 phox -deficient mice are protected.
To the best of our knowledge, this is the first study investigating age-related hearing loss in p22 phox deficient mice. However, the impact of Nox3 deletion has respectively been addressed in the context of cisplatin [31] and noise [38] induced hearing loss. The first study reports a protective impact in cisplatin-induced hearing loss following injection of Nox3 siRNA in the middle ear while the second publication by Lavinsky and collaborators using Nox3 mutant mouse concluded that Nox3 might be protective against noise-induced inner ear damage. In terms of pattern of hearing loss, the Lavinsky study did not show an impact of Nox3 deficiency on noise-induced hearing loss at high frequencies, but a small protective effect of Nox3 at low frequencies, statistically significant only at 8 kHz. Our results studying age-related hearing loss showed a protective effect of p22 phox deficiency at frequencies above 8 kHz. The reason for these apparently conflicting data may be due to i) differences in the type of hearing loss (noise-induced, ototoxicity and age-related), ii) mice in a different genetic background were studied (black6 vs A/J), and iii) importantly, a different type of NOX-deficient mouse was investigated (NOX3-deficiency vs. p22 phox deficiency, corresponding to a functional NOX1-4 knock-out). Future studies addressing each isoform regulated by p22 phox may provide a better understanding of such differences.
Our study provides first conclusive evidence that p22 phox is key regulator of age-related hearing loss, most likely through regulation of the activity of ROS-generating NADPH oxidases. The extent of the protection of p22 phox -deficient animals is astonishingly strong. Our data suggest that p22 phoxmaster regulator of the NOX pathwayis not only regulating the levels of non-specific ROS-induced damage in the cochlea. In fact, we demonstrate that level of expression of genes of the excitatory pathway is strikingly diminished in p22 phox mutant mice. Thus, our data suggest that NOX-enhanced excitotoxicity contributes to age-related hearing loss. We believe that targeting NOX catalytic activity is a promising strategy to prevent age-related hearing loss. Small molecules NOX inhibitors would be interesting tools, but might be difficult to target to the inner ear. Local delivery of molecular therapies (i.e. RNA knock-down of p22 phox ) may provide long term protection against presbycusis and other related inner ear damage, but will equally have to be precisely targeted [54].

Animal procedures
A.B6 Tyr + -Cyba nmf333 /J mice were purchased from Jackson lab (strain 005445) and colonies were maintained at the animal facility of the University of Geneva through heterozygous x heterozygous mating, ensuring equal proportion of WT and p22 phox deficient littermates. Both male and female, were subjected to hearing threshold determination using the auditory brainstem response at different time points from 4 weeks old to 26 weeks old. At the end of the experiments, animals were sacrificed for cochlea histology or mRNA extraction. For the auditory tests, or before sacrifice, mice were anesthetized with a mix of Ketamine (100 mg/kg) and Xylazine (10 mg/kg).

Auditory brainstem response
Animals were anesthetized by intraperitoneal injection Ketamine 100 mg/kg and of Xylazine 5 mg/kg, and placed upon a heating pad to maintain body temperature. Depth of anesthesia was tested every 30 min by testing the pedal withdrawal reflex. If necessary, additional injections of about 50% of the initial dose were given. ABR recordings were performed in a sound proof chamber (IAC Acoustics, Illinois IL, USA). For stimulus generation and recording of responses, a multifunction IO-Card (National Instruments, Austin TX, USA) was used, housed in an IBM compatible computer. An integrated software package for stimulus generation and recording (Audiology_lab; Otoconsult, Frankfurt, Germany) was used. Sound pressure level was controlled with an attenuator and amplifier (Otoconsult, Frankfurt, Germany). Stimuli were delivered to the ear in a calibrated open system by a loudspeaker (AS04004PR-R, PUI Audio, Inc., Dayton, USA) placed 3 cm lateral to the animals' pinna. Sound pressure was calibrated online prior to each measurement with a microphone probe system (Bruel &Kjaer 4191) placed near the animals' ear. Recorded signals were amplified and bandpass filtered (80 dB; 0.2-3.0 kHz) using a filter/ amplifier unit (Otoconsult, Frankfurt, Germany). For the recordings, silver electrodes were placed subcutaneously on the mouse forehead (+), on the mastoid of the recorded ear (−) and a reference electrode on the back. ABR were recorded, following stimulation with 100μs clicks or 3 ms tone pipes (2.0-45.2 kHz at a resolution of 2 steps per octave). For all frequencies, ABRs were recorded from 0 to 90 dB SPL in 3 dB steps. Electrical signals were averaged over 256 repetitions of stimulus pairs with alternating phase. Hearing thresholds were defined as the sound pressure level where a stimulus-correlated response was clearly identified by visual inspection of the averaged signal.

Cochlea histology (mid modiolar cut, cytocochleograms and ribbons staining)
At the end of the last ABR measurement, anaesthetized mice were sacrificed by cervical dislocation followed by decapitation. Temporal bones were isolated from the skull in order to extract the inner ears, and after proper dissection cochlea were placed in 4% paraformaldehyde overnight at room temperature. Cochleae were decalcified using USE-DECALC solution (Medite commercial solution) under sonication for 48 h (following ultrasonic technique of USE 33 decalcification machine from Medite, Cat. No. 03-3300-00). Finally, decalcified cochlea were microdissected to perform cytocochleograms and immunohistochemistry, or mid modiolar cuts (embedded in paraffin) with Hematoxilin-Eosin staining. As general rule, one ear from each animal was used for cytocochleograms and the other ear for mid-modiolar staining.

Immunohistochemistry and confocal microscopy (cytocochleograms)
Decalcified cochleae were microdissected with thin forceps under binocular microscope. The bony shell was removed to expose the Organ of Corti (OC), followed by the removal of the stria vascularis and separation of the sensory epithelium from the spiral ganglion. The basal and middle turns of the OC were cut in two pieces each one and the apical turn was preserved as one piece, resulting in 5 pieces transferred into 400 μL PBS solution in a 48 well plate. Samples were then permeabilized (3% Triton-X 100 in PBS 1X) for 30 min at room temperature and immersed in a blocking buffer, containing 2% bovine serum albumin (BSA) and 0.01% Triton-X 100 in PBS, for 1 h at room temperature. Explants were incubated with anti- MyoVIIa

Mid modiolar hematoxylin eosin staining
Decalcified cochlea were dehydrated and embedded in paraffin. Mid-modiolar cuts of 5 μm were processed and loaded onto gelatincoated slides. Harris' Hemalun/eosin staining was performed on the slides. The samples were qualitatively analyzed for density of spiral ganglia using image J software.

Quantitative analysis of cochlea morphology
The SGN density was determined by nuclear counting, employing the hematoxylin-eosin stained slides, of the Rosenthal's canal at each different cochlear turn (apical, medial and basal). Five non-consecutive sections were evaluated. To normalize the data, the total neuronal amount was divided between each Rosenthal's area (in mm 2 ), and the measured densities were averaged. For the synaptic ribbons count, 10-15 μm Z stack (0.7 μm steps) were performed with the 63X objective and merged in a single picture, to ensure a proper counting of the ribbons distributed along the Z axis. For each section of the OC, ribbons belonging to two different segments of 10-15 inner hair cells were counted to obtain an average of each cochlear turn. Additionally, 20X pictures were employed to evaluate hair cells (HC) survival (cytocochleogram). Inner hair cells (IHC) and outer hair cells (OHC) population were recorded in two representative areas of 100 μm from each cochlear section, in order to average the HC survival rate of each turn (apical, medial and basal). The open source image J program was used for the image analysis.

Mouse cochlea RNA extraction
Cochlea were quickly dissected in cold PBS, carefully removing remaining blood and surrounding tissues, immediately frozen in liquid nitrogen and kept at −80°C for further RNA extraction. Cochlea were physically homogenized with clean steel balls and tissueLyser (Qiagen) for 30 s at 30 rpm, and the DNA and RNA precipitated with 750 μl Trizol. After adding 150 μl chloroform, (shaking, resting and centrifuging step) we collected the aqueous phase and performed column purification with Qiagen RNeasy Micro kit, ignoring the lysis and homogenization steps described in the kit protocol (adding directly 70% ethanol) (adapted from Ref. [55]).

Human foetal cochlea RNA extraction
The inner ear was isolated from aborted human fetuses ranging from W8 to W12 post conception. Procurement and procedures were performed with full approval by the Ethics Committee of the State Geneva, Switzerland and following signed informed consent of the donors. The postmenstrual date was used for the calculation of the fetal stage. Tissue dissection was performed in ice-cold Hanks' balanced salt solution (HBSS). The rest of the procedure was similar as for mouse cochlea RNA extraction (see paragraph above).

Real time quantitative polymerase chain reaction
Following the cochlear extraction, RNA concentration was determined using a Nanodrop. 500 ng of RNA was used for cDNA synthesis using the Takara PrimeScript RT reagent Kit, following manufacturer's instruction. Real-time PCR was performed using SYBR green assay on a 7900HT SDS system from ABI. The efficiency of each primer was verified with serial dilutions of cDNA. Relative expression levels were calculated by normalization to the geometric mean of the three house-keeping genes (Eef1a, Tubb and Actb for mouse and EEF1a, B2M and GAPDH for human samples). The highest normalized relative quantity was arbitrarily designated as a value of 1.0. Fold changes were calculated from the quotient of means of these RNA normalized quantities and reported as ± SEM. Sequences of the primers used are provided in Supplementary Table 1

RNA seq
TruSeq ribodepleted stranded mRNA from 6 weeks old A/J mice (3 WT vs. 3 p22 phox deficient) was applied to eliminate ribosomic RNA and sequenced using Illumina TruSeq protocol. The sequencing quality control was done with FastQC v.0.11.5. The sequences were mapped with the TopHat v.2.0.11 software to the UCSC mm10 mouse reference. The biological quality control and summarization were performed using the PicardTools v.1.141. The table of counts with the number of reads mapping each gene feature of UCSC mm10 was prepared with HTSeq v0.6p1. The differential expression analysis was performed with the statistical analysis R/Bioconductor package edgeR v. 3.14.0. Briefly, the counts were normalized according to the library size and filtered. The genes having a count above 1 count per million reads (cpm) in at least 4 samples were kept for the analysis. The initial number of genes in the set was 23′420 and after the poorly or non-expressed genes were filtered out, 15′126 genes were left. The p values of differentially expressed gene analysis were corrected for multiple testing problems using the Benjamini-Hochberg (BH) procedure. All RNA-sequencing data files were submitted to the arrayexpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under the accession number E-MTAB-8668.

In vitro culture of mouse SGN
Mouse spiral ganglion neurons were isolated from p2-p5 A/J pups, cultured and passaged as previously described [33]. Briefly, neurospheres from spiral ganglia were maintained in DMEM:F12 with 1x N2 and B27 supplement in presence of bFGF, IGF1, Heparan sulfate and EGF. For passages or differentiation, cells were dissociated with acutase and mechanically disintegrated. Following filtration with a 70 μm filter, retaining non dissociated spheres, SGN were counted and plated in a 384 or 96 well plate coated with matrigel (hESC qualified, Corning, New York, USA) with medium without growth factors, in order to induce differentiation. After 5 days of differentiation, cells were employed for cytosolic Ca2+ measurement and excitotoxicity assay.

Measurement of Ca 2+ cytosolic release
16000 cells/well were differentiated to SGN on a 96 well plate coated with matrigel, as described above (in vitro culture of spiral ganglion neurons). Differentiated cells were loaded with FLUO-8 (Interchim, Montluçon, France) according to the manufacturer protocol. After 45min incubation at 37°C, the glutamate-induced cytosolic calcium release was assessed in a FDSS/μCELL Functional Drug Screening System (Hamamatsu, Yokohama, Japan). The neuronal kinetics of calcium release was followed over 10min following glutamate addition, with 1 measure every 0.5 s. The impact of ryanodine or adenosine (0-100 μM), respectively added 30min or 24 h prior glutamate, was also investigated.

Impact of high glutamate concentration on auditory neuron ATP content
16000 cells/well were differentiated to SGN on a 96 well plate coated with matrigel, as described above (in vitro culture of spiral ganglions). Differentiated cells were then treated with increasing concentrations of glutamate for 6 h and the SGN viability was assessed using the ATPLite kit (PerkinElmer, Wellesley, USA), following manufacturer's instructions using spectra L reader.

Impact of high glutamate concentration on auditory neuron morphology
Auditory neuron progenitors were seeded on matrigel coated coverslips and differentiated into auditory neurons for 5 days. Cells were then exposed to 1 mM glutamate for 6 h and fixed for 10 min with PFA 4%. Following the fixation, neurons were permeabilized (0.2% Triton-X100 in PBS 1X) for 30 min at room temperature and immersed in a blocking buffer containing 2% BSA and 0.01% Triton-X 100 for 1 h at room temperature. SGNs were incubated with the primary antibody anti-BIII tubulin (1:1000, mouse; Biolegend, USA) in blocking buffer overnight at 4°C. On the following day, tissues were rinsed three times with PBS and incubated with the secondary antibody anti-mouse Alexa Fluor 555 (1:500; Invitrogen, USA) in blocking buffer for 2 h at room temperature. Coverslips were washed 3 times with PBS and mounted on a glass slide with Fluoroshield containing DAPI (Sigmaaldrich, USA). The labelled cells were visualized with a confocal laser-scanning microscope (Zeiss LSM700) equipped with a CCD camera (Leica Microsystems) with Plan-Neofluar 10X/0.30 NA objective. Neurite length evaluation was performed using image J software.

Statistics
All data were analyzed using Two-way ANOVA followed by Bonferroni multiple comparison test using GraphPad Prism software unless where stated otherwise in the Figure legend. Values with p < 0.05 was considered as statistically significant. *; p < 0.05, **p < 0.01, ***; p < 0.005, ****p < 0.0005.

Study approval
All procedures were approved by the local veterinary office and the Commission for Animal experimentation of the Canton of Geneva, Switzerland, authorization number GE/106/18.