Protection of cochlear from noise-induced excitotoxic trauma by blockade of Ca2+-permeable AMPA receptors

Ning Hua, Mark A. Rutherfordb, and Steven H. Greena,1

aDepartment of Biology, University of Iowa, Iowa City, IA 52242; and bDepartment of Otolaryngology, Washington University in St. Louis, St. Louis, MO 63110

Edited by David P. Corey, Harvard Medical School, Boston, MA, and accepted by Editorial Board Member David E. Clapham January 7, 2020 (received for review August 16, 2019) Exposure to loud sound damages the postsynaptic terminals the ribbon synapses between IHCs and SGNs (3, 4). This noise- of spiral ganglion neurons (SGNs) on cochlear inner hair cells induced cochlear synaptopathy (NICS) is detectable by counting (IHCs), resulting in loss of synapses, a process termed synaptopathy. synapses histologically or, noninvasively, by a persistent reduction Glutamatergic via α-amino-3-hydroxy-5- of auditory brainstem response (ABR) wave-I amplitude even methylisoxazole-4-propionic acid (AMPA)-type receptors is required after complete recovery of ABR thresholds. for synaptopathy, and here we identify a possible involvement of Available evidence strongly supports the presumption that NICS + GluA2-lacking Ca2 -permeable AMPA receptors (CP-AMPARs) using is excitotoxic damage caused by excessive release of glutamate IEM-1460, which has been shown to block GluA2-lacking AMPARs. from overstimulated IHCs. Direct application of glutamatergic In CBA/CaJ mice, a 2-h exposure to 100-dB sound pressure level agonists causes vacuolization of postsynaptic bouton terminals, octave band (8 to 16 kHz) noise results in no permanent threshold similar to the morphological changes seen after sound over- shift but does cause significant synaptopathy and a reduction in exposure (5, 6). Deficiency of the glutamate aspartate transporter auditory brainstem response (ABR) wave-I amplitude. Chronic intra- increases synaptic glutamate and exacerbates noise damage (7). cochlear perfusion of IEM-1460 in artificial perilymph (AP) into adult Conversely, blockade of glutamate receptors reduces noise dam- CBA/CaJ mice prevented the decrease in ABR wave-I amplitude and age to synapses (6), and genetic removal of vesicular glutamate the synaptopathy relative to intracochlear perfusion of AP alone. release from IHCs prevents loss after noise exposure (8). Interestingly, IEM-1460 itself did not affect the ABR threshold, pre- Although NICS requires activation of glutamate receptors, the sumably because GluA2-containing AMPARs can sustain sufficient specific receptor types, the downstream mechanism(s) of damage, synaptic transmission to evoke low-threshold responses during and the reasons for selective vulnerability of a subset of synapses blockade of GluA2-lacking AMPARs. On individual postsynaptic den- are unknown. sities, we observed GluA2-lacking nanodomains alongside regions SGNs express a variety of glutamate receptor types, in- with robust GluA2 expression, consistent with the idea that individ- cluding N-methyl-D-aspartate (NMDA), α-amino-3-hydroxy-5- + ual synapses have both CP-AMPARs and Ca2 -impermeable AMPARs. methylisoxazole-4-propionic acid (AMPA), kainate, and metabo- SGNs innervating the same IHC differ in their relative vulnerability to tropic receptors (9, 10). The AMPA type is responsible for the noise. We found local heterogeneity among synapses in the relative abundance of GluA2 subunits that may underlie such differences in Significance vulnerability. We propose a role for GluA2-lacking CP-AMPARs in noise-induced cochlear synaptopathy whereby differences among Noise can cause excitotoxic trauma to cochlear synapses by synapses account for differences in excitotoxic susceptibility. These triggering excessive release of the neurotransmitter glutamate data suggest a means of maintaining normal hearing thresholds from the auditory sensory hair cells. We report that a specific while protecting against noise-induced synaptopathy, via selective + class of glutamate receptors, Ca2 -permeable α-amino-3-hydroxy- blockade of CP-AMPARs. 5-methylisoxazole-4-propionic acid (AMPA) receptors (CP-AMPARs), is largely responsible for this trauma. Because cochlear syn- | cochlear synapse | spiral ganglion neuron | noise-induced + apses are heterogenous with respect to glutamate receptors, synaptopathy | Ca2 -permeable AMPA receptor the observation that a specific class is responsible might ex- plain the variability in susceptibility to noise among synap- piral ganglion neurons (SGNs) are bipolar neurons in the ses. Selective blockade of CP-AMPARs prevents excitotoxicity Scochlea that conduct auditory information from the sensory and noise-induced cochlear synaptopathy, while other glu- haircellstothebrain.Almostall(95%)SGNsaretypeIand tamate receptors continue to mediate neurotransmission and synapse on inner hair cells (IHCs); the remaining 5%, type II, allow hearing. synapse on outer hair cells. (Herein, “SGN” used without quali- fication refers to type I.) Each IHC in the rodent cochlea provides Author contributions: N.H., M.A.R., and S.H.G. designed research; N.H., M.A.R., and S.H.G. the sole afferent presynaptic input to ∼10 to 20 SGNs, depending performed research; N.H., M.A.R., and S.H.G. analyzed data; and N.H., M.A.R., and S.H.G. on species and tonotopic location. Each SGN has a myelinated wrote the paper. axon with an unbranched dendritic terminal making one synapse Competing interest statement: A provisional patent application titled “Targeting Calcium-Permeable AMPA Receptors for Inner Ear Therapy with IEM-1460 and Related with one IHC. Each of these synapses consists of a single post- Compounds” was filed on 23 December 2019. Some data presented in this paper was synaptic density (PSD) apposed to a presynaptic ribbon-type active cited in the application. zone in the IHC. Each PSD includes the scaffold protein PSD95 This article is a PNAS Direct Submission. D.P.C. is a guest editor invited by the and is estimated to contain a few thousand glutamate receptors Editorial Board. receiving excitation from the IHC ribbon synapse (1, 2). Published under the PNAS license. Exposure to sounds at high levels can destroy cochlear hair 1To whom correspondence may be addressed. Email: [email protected]. cells, resulting in permanent elevation of the hearing threshold. This article contains supporting information online at https://www.pnas.org/lookup/suppl/ Even at a sound pressure level (SPL) too low to permanently doi:10.1073/pnas.1914247117/-/DCSupplemental. impair hair cell function, exposure may impair hearing by destroying First published February 3, 2020.

3828–3838 | PNAS | February 18, 2020 | vol. 117 | no. 7 www.pnas.org/cgi/doi/10.1073/pnas.1914247117 Downloaded by guest on September 25, 2021 large fast excitatory currents necessary for synaptic transmission nerve fibers spiking at sound onset. The ABR waveform is a series at cochlear afferent synapses (11–14) and for excitotoxicity (15). of peaks with latencies corresponding to electrical activity at AMPA-type glutamate receptors are tetrameric combinations of successive levels of auditory processing in the brain. The am- GluA1, GluA2, GluA3, and GluA4 subunits, of which primarily plitude of the first peak (i.e., wave-I amplitude) is proportional the latter three are expressed in the PSDs of SGNs (9, 16). Any to the number of synchronously spiking SGNs. Synapses lost as a combination of AMPA receptor (AMPAR) subunits makes a re- consequence of NICS reduce the number of SGNs responding, + + + ceptor permeable to Na ,K ,andCa2 ions, but combinations that resulting in a reduced wave-I amplitude but, if hair cells are + include GluA2 have greatly reduced Ca2 permeability (17, 18). intact, no change in auditory threshold (27). Because GluA2 appears to be expressed at every synapse, it might The experiment was designed to allow us to establish for each + be assumed that all synapses lack Ca2 -permeable AMPAR re- mouse in each experimental group the effects of noise and the ceptors (CP-AMPARs), but this does not appear to be the case. effects of IEM-1460 while correcting for any possible effects In synapses on frog ear and fish lateral line hair cells, CP- of surgery on the cochlea (Fig. 1A). ABR was measured five AMPARs in particular appear to play a major role in synaptic times in each mouse that underwent surgery and three times in transmission and excitotoxicity (16). This is potentially significant unoperated mice. The first measurement was done before + because studies of excitotoxicity in the brain have implicated Ca2 surgery and the second was done at 3 d postsurgery, to de- influx as a key mediator of excitotoxic trauma (19), and although termine any effects on the threshold or the wave-I growth curve + NMDA-type glutamate receptors play the major role in Ca2 in- (amplitude as a function of sound level). These two measure- flux, CP-AMPARs also have a role in glutamate toxicity in the ments provided a baseline for any effects of surgery or IEM- brain (20–22). Moreover, previous in vitro (23) and in vivo (24) 1460 on the ABR. The cannula connecting the minipump to the + studies have suggested that, unlike in the brain, Ca2 entry via round window was filled with artificial perilymph (AP) only, so NMDA-type glutamate receptors is not a major contributor to that the minipump contents reached the cochlea at 3.5 d after excitotoxicity in the cochlea. Rather, we suggest that cochlear the surgical implantation. Approximately 1.5 d after the entry excitotoxicity is mediated by GluA2-lacking CP-AMPARs. To of IEM-1460 into the cochlea, a third ABR measurement was directly test this hypothesis, we used IEM-1460, a selective performed to assessed the effect of IEM-1460 itself on hearing. blocker of GluA2-lacking AMPARs (25, 26), in noise-exposed This measurement also provided the prenoise baseline for mice, with the results implicating a role of GluA2-lacking/CP- comparison to and normalization of postnoise measurements AMPARs in NICS. for each mouse. At age 13 to 14 wk and with minipump contents already pre- Results sent in the cochlea for ∼2.5 d, two groups of mice were exposed NEUROSCIENCE Experimental Design and Timeline of ABR Measurements. To test the to noise on day 0: those with implanted minipumps containing role of CP-AMPARs in NICS, we infused a selective blocker of IEM-1460 in AP and those containing AP only. We used a CP-AMPARs, IEM-1460, directly into the cochlea and assessed previously described noise exposure (3) confirmed to cause the extent of NICS by measuring ABR wave-I amplitude and synaptopathy but not hair cell loss: 100-dB SPL for 2 h, in a 8- to counting synapses. The ABR is the field potential of auditory 16-kHz octave band. The effects of noise exposure on auditory

Fig. 1. In vivo assessment of IEM-1460. (A) Timeline for the noise exposure experiments showing times relative to noise exposure (day 0) for five ABR measures: (1) presurgery, to verify normal hearing; (2) postsurgery but before IEM-1460 reaching the cochlea, to determine any effect of surgery on the ABR; (3) prenoise/post–IEM-1460 reaching the cochlea, to determine any effect of IEM-1460 on the ABR and to establish a prenoise baseline for normalizing subsequent measures; (4) at 1 d postnoise, to measure temporary threshold shift; and (5) at 10 to 14 d postnoise, to ensure that there was no permanent threshold shift. (B–D) Measure of ABR thresholds (mean ± SEM) at 8, 16, and 32 kHz for each of the five timepoints shown in A (only three timepoints for unoperated controls) for ears perfused with IEM-1460 (n = 11; representative example in Fig. 2A), with control vehicle (AP; n = 11; representative example in Fig. 2B), or unoperated controls (contralateral ears and six other mice; n = 28; representative examples in Fig. 2 A and B). The threshold elevation at day −1is not significantly different between control mice receiving AP only and mice receiving AP with IEM-1460 (P = 0.3493, P > 0.9999, and P = 0.2997, respectively, for 8, 16, and 32 kHz; Student’s t test). These data show a mean threshold shift of ∼35 dB for 16-kHz tone bursts between 1 d prenoise and 1 d postnoise for operated ears and unoperated control ears.

Hu et al. PNAS | February 18, 2020 | vol. 117 | no. 7 | 3829 Downloaded by guest on September 25, 2021 threshold are referred to as temporary threshold shift (TTS) if the ABR wave-I amplitude relative to the third ABR measurement effects do not persist beyond several days, and as permanent performed at 1 d before the noise trauma. threshold shift (PTS) if the effect does not recover (28). The fourth ABR measurement was done on postnoise day 1 IEM-1460 Does not Significantly Affect Hearing Threshold. We first (PND1). Measuring TTS, the elevation in threshold due to the asked whether IEM-1460 itself affects ABR thresholds. The five ABR threshold measurements for the two experimental groups noise exposure, shows whether an appropriate noise exposure described above (noise/IEM-1460 or noise/AP-only) are shown has been accomplished. We had empirically determined that in Fig. 1 B‒D for 8-, 16-, and 32-kHz tone-burst stimuli. For synapse loss, but not hair cell loss, was achieved by a TTS of 30 to comparison with these measurements on operated ears (left 40 dB for 16-kHz tone bursts and used this as a criterion. The side),thesamemeasurementsweremadeforthesamestimuli fifth ABR measurement was done at 14 d postnoise (PND14), to delivered to unoperated contralateral ears (noise/control, right ensure a lack of PTS and quantify any permanent reduction in side) for three of the time points (prenoise, PND1, and PND14).

Fig. 2. Representative examples of ABR thresholds and waveforms for perfused and contralateral unoperated ears of two mice. Latency is in milliseconds. In all experiments, the operated ear of each mouse was the left ear, and the unoperated control ear was the right ear. Both ears were exposed identically to noise, but only the left ear was perfused. (A) IEM-1460 perfused mouse. (B) Control AP perfused mouse. (a) Auditory thresholds for 16-kHz tone bursts for the left and right ears from a mouse receiving IEM-1460 or AP control, recorded at the time points shown in Fig. 1A.(b–g) ABR waveforms from 85 to 45 dB SPL are shown for the unoperated (right) ears (b, c) and for the perfused (left) ears (d–g) for the indicated time points. Prenoise comparisons show an ∼10 dB threshold elevation at 3 d postsurgery relative to presurgery (a and d). The minipump contents enter the cochlea at ∼3.5 d postsurgery. Day −1 (1 d prenoise) is 5 d postsurgery, ∼1.5 d after IEM-1460 (or control AP) enters the cochlea. By day −1, the ABR threshold and wave-I amplitude have recovered to near presurgery values (a and e), showing recovery from surgery and demonstrating that IEM-1460 itself has no significant effect on the auditory nerve re- sponse (see text). Postnoise comparisons show that a TTS due to noise exposure is equally evident in both ears as ABR thresholds elevated by ∼35 dB (a) and ABR amplitudes reduced on PND1 (b and f). By PND14, ABR thresholds have recovered to the prenoise level in both ears (a). ABR wave-I amplitude has recovered to the prenoise level in the noise/IEM-1460 ear (A, g) but not in the noise/AP ear (B, g) or in the noise/control (right) ears (A, c and B, c), consistent with the conclusion that IEM-1460 prevented synaptopathy in this ear.

3830 | www.pnas.org/cgi/doi/10.1073/pnas.1914247117 Hu et al. Downloaded by guest on September 25, 2021 Representative examples of the ABR measurements, comparing level ≥40 dB SPL. The largest reduction in wave-I amplitude mice receiving IEM-1460 and those receiving AP only, are shown was ∼40% at 90 dB, the highest stimulus level tested (Fig. 3A, in Fig. 2 A and B,respectively. 16 kHz). If IEM-1460 prevents NICS, then ABR wave-I amplitude At the third ABR measurement, done at 1 d before noise at PND14 should not be significantly reduced relative to baseline. exposure and after ∼1.5 d of IEM-1460 delivery, the threshold Indeed, ears infused intracochlearly with 0.5 mM IEM-1460 elevation relative to the presurgery baseline was generally 5 to showed no significant reduction in wave-I amplitudes at PND14 8 dB for mice included in the study. Mice with a >10 dB threshold relative to prenoise (Fig. 3B), further supporting the inference elevation at 8, 16, or 32 kHz were excluded. Significantly, this from Fig. 1 that IEM-1460 itself does not impair hearing. threshold elevation is attributable to the surgery, not to the IEM- A comparison of Fig. 3 A and B suggests that IEM-1460 1460, because there was no significant difference between the prevented the noise-induced decline in wave-I amplitudes. How- IEM-1460 and AP-only groups (Fig. 1 B‒D). An important im- ever, this comparison is between operated and unoperated ears plication of this result is that the GluA2-containing AMPARs, and does not address the concern that cannula placement surgery which are relatively insensitive to IEM-1460 (25, 26), maintain may have impaired hearing sufficiently to reduce noise-induced sufficient synaptic transmission in the presence of IEM-1460 to trauma, thus “protecting” the cochlea. Therefore, the appropriate drive spike activity in the auditory nerve in response to low-level comparison is between ears experiencing identical surgeries with sounds. We hypothesized that the GluA2-lacking class of AMPARs implantation of minipumps infusing AP with IEM-1460 vs. AP is necessary for noise-induced excitotoxicity in this context, and only. Indeed, the ears receiving AP only exhibited a significant thus asked whether IEM-1460 would protect against NICS. reduction in ABR wave-I amplitudes at PND14 relative to the prenoise baseline (Fig. 3C) in contrast to ears receiving IEM-1460 NICS Is Prevented by IEM-1460: Electrophysiological Measurements. (Fig. 3B), confirming a significant protective effect of IEM-1460. In the absence of PTS, a reduction of the ABR wave-I amplitude A comparison of prenoise baseline wave-I amplitudes across indicates loss of afferent synapses on IHCs. This is what was seen groups (Fig. 3A vs. Fig. 3 B and C) suggests a small reduction due in noise-exposed control ears in this study (Fig. 3A, noise/control): to the surgery, and a comparison of Fig. 3A and 3C shows that ABR wave-I amplitude, plotted as a function of stimulus level the effect of noise on wave-I amplitude is greater in unoperated (loudness), is depressed in the noise/control (i.e., unoperated) ears than in operated ears receiving AP only. These observations ears at PND14 relative to the same ears measured prenoise, are consistent with a small hearing impairment caused by the although ABR threshold recovers (Fig. 1 B and C). The de- surgery that reduced the impact of the noise exposure. None- crease in wave-I amplitude is significant for every stimulus theless, comparison of noise/IEM to noise/AP (Fig. 3B vs. 3C) NEUROSCIENCE

Fig. 3. IEM-1460 prevents noise-induced long-term reduction in ABR wave-I amplitude. Shown are ABR wave-I amplitude measurements (“growth curves”) made in the same mice, prenoise (blue) and at 14 d postnoise (PND14; red) for 8-kHz (a), 16-kHz (b), and 32-kHz (c) tone bursts at indicated sound levels. (A) Noise-exposed unoperated control (noise/control; n = 28). (B) Noise-exposed IEM-treated (noise/IEM; n = 11). (C) Noise-exposed vehicle-only control (noise/AP, n = 11). Data are mean ± SEM. The curves were constructed by fitting the data (by least squares) to a second-order polynomial. The significance of amplitude differences between prenoise and PND14 measures at each stimulus level was as shown: *P < 0.05, **P < 0.01, ***P < 0.001, repeated-measures two-way ANOVA over all stimulus levels and prenoise vs. PND14, Sidak’s multiple comparisons test. The overall difference between each pair of curves, prenoise vs. PND14, was derived from the repeated-measures ANOVA. Significant differences between prenoise and PND14 measures were found for Figures A, a–c and C, b and c; that is, in noise-exposed control mice and for noise-exposed mice treated only with AP, there was a significant decline in ABR wave-I amplitude at 14 d after noise exposure, but in noise-exposed, IEM-1460–treated mice, there was no difference in the ABR wave-I growth curve at 14 d postnoise.

Hu et al. PNAS | February 18, 2020 | vol. 117 | no. 7 | 3831 Downloaded by guest on September 25, 2021 shows that IEM-1460 is largely responsible for the prevention of stimulus levels for each mouse (Fig. 4B). The overall mean ABR noise-induced reduction of wave-I. wave-I amplitude change for each of the experimental groups The conclusions drawn from Fig. 3 are further supported by are compared with one another and with unexposed controls, normalizing the PND14 measurements to the baseline mea- demonstrating complete protection from noise exposure at all surements for each mouse (Fig. 4), as in the example shown in SI test frequencies in IEM-1460–treated ears (Fig. 4B). Although Appendix, Fig. S2B. Such normalization is possible because ABR wave-I amplitude was reduced in AP-only ears before the noise measurements obtained 2 to 3 wk apart are reproducible in mice exposure, which we attribute to the surgery, the difference between of this age, as shown in SI Appendix,Fig.S2A. The baseline- the IEM-1460 and AP-only groups after noise is significant. Sep- normalized PND14 wave-I amplitudes are shown in Fig. 4A using arating the overall data of Fig. 4B into two groups based on re- the third ABR for normalization—that is, the ABR measured 1 d sponses to higher-level or lower-level stimuli (≥35 dB or ≤30 dB before noise trauma, after minipump contents have reached the above the ABR threshold, respectively) shows that the mean wave- cochlea. A comparison of noise/control, noise/IEM-1460, and I amplitude reductions are similar for low and high stimulus levels noise/AP only shows the effectiveness of IEM-1460 in preventing (SLs; Fig. 4 C and D). the decline in wave-I amplitudes at 8, 16, and 32 kHz. Because some comparisons were made between operated and Two-way ANOVA of the results in Fig. 4A by stimulus intensity unoperated contralateral ears, we verified that there was no sig- and experimental condition shows an effect only of experimen- nificant difference in wave-I amplitude at PND14 in contralateral tal condition and not of stimulus intensity on any differences in noise/control ears between mice receiving AP-only and those re- the normalized amplitudes. Thus, we summarized the intergroup ceiving IEM-1460 (SI Appendix,Fig.S3), indicating that IEM-1460 comparisons of Fig. 4A with the overall amplitude percent re- infused intracochlearly via the round window did not affect the duction for each group, averaged across all suprathreshold contralateral ear.

Fig. 4. IEM-1460 prevents noise-induced long-term reduction in ABR wave-I amplitude: normalized amplitude growth curves. (A)(a–c)NormalizedABR wave-I amplitude growth curves for 8-kHz (a), 16-kHz (b), and 32-kHz (c)toneburstsoverstimuluslevels(mean± SEM). The PND14 amplitude measure at each stimulus level (from Fig. 3) for each individual mouse was normalized to the prenoise (day −1inFig.1A for operated ears) measure for that mouse to provide a within-subject comparison. A normalized value or ratio of 1 (dotted line ) indicates no change in wave-I amplitude at PND14 relative to the prenoise value. The figures compare normalized wave-I amplitudes among the noise-exposed unoperated control group (noise/Ctr), noise-exposed IEM- 1460–treated group (noise/IEM), and noise-exposed vehicle-only control group (noise/AP; n = 11). Two-way ANOVA was used to test the significant

differences of normalized amplitude among these groups and across stimulus levels. There is no significant difference across stimulus levels (F(5, 282) = 0.1948, P = 0.9644 at 8 kHz; F(6, 329) = 0.6437, P = 0.6952 at 16 kHz; F(5, 270) = 1.525, P = 0.1823 at 32 kHz). Differences are significant among the experimental groups (F(2, 282) = 96.53, P < 0.0001 at 8 kHz; F(2, 329) = 158.8, P < 0.0001 at 16 kHz; F(2, 270) = 196.6, P < 0.0001 at 32 kHz) with Sidak’s post hoc test for multiple comparisons. Significant differences at each stimulus intensity between noise/IEM and noise/AP are indicated in the figure by asterisks. *P < 0.05, **P < 0.01, ***P < 0.001, ****P = 0.0001. (B–D) The overall average amplitude decline at PND14 over the range of stimulus intensities was calculated by averaging the amplitude declines of all stimulus intensities exceeding threshold. These overall declines in the normalized PND14 amplitudes for all stimulus levels, expressed as percentage [= 100% × ((prenoise amplitude – PND14 amplitude)/prenoise amplitude)] are shown in B. The percentage amplitude declines for responses only to high-intensity stimuli, ≥35 dB SL (i.e., stimulus level relative to ABR threshold), are shown in C; the percentage amplitude declines for responses only to low- intensity stimuli, ≤30 dB SL, are shown in D. Results were similar for responses to high- and low-intensity stimuli. Significance of differences was determined by two-way ANOVA over all conditions with Tukey’s post hoc test for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001. Because the noise/AP ≥35 dB SL at 32 kHz results did not pass a D’Agostino–Pearson normality test, a Kruskal–Wallis test with Dunn’s multiple comparisons test was applied (P < 0.0001). Noise/ IEM is significantly different from noise/AP control at all three frequencies for all stimulus intensity groups, and noise/IEM is not significantly different from the no noise control at all three frequencies for all stimulus intensity groups, indicating that IEM-1460 effectively prevents noise-induced reduction in ABR wave-I amplitude.

3832 | www.pnas.org/cgi/doi/10.1073/pnas.1914247117 Hu et al. Downloaded by guest on September 25, 2021 NICS Is Prevented by IEM-1460: Immunohistological Measures. Pre- protection appears to be due to IEM-1460, because the number of vention of ABR wave-I reduction implies protection from synapses in the noise/AP-only ears was not significantly different synaptopathy. To directly test the hypothesis that IEM-1460 pre- from that in the noise/control ears at the 8- and 16-kHz locations. vents synapse loss, we labeled and counted synapses in cochlear At the 32-kHz location, AP-only perfused ears did show a small, whole-mount preparations from mice euthanized immediately yet significant, reduction in synapse loss relative to noise-exposed after the PND14 ABR measurements. Synapses were counted unoperated ears, but synapse number was nevertheless signif- in the cochlear regions corresponding to the three stimulus – SI icantly reduced relative to non noise-exposed ears or to noise- frequencies at which ABRs were elicited: 8, 16, and 32 kHz ( exposed IEM-1460–perfused ears. These results are consistent Appendix, Materials and Methods and Fig. S1). Fig. 5 A–D shows with those from ABR wave-I amplitudes, supporting the idea representative examples of images from all three cochlear lo- that IEM-1460 provides protection from NICS. cations from unexposed mice (Fig. 5A), from a noise-exposed Lower-magnification images (SI Appendix, Fig. S4) used to unoperated cochlea (Fig. 5B), from a noise-exposed cochlea perfused with IEM-1460 (Fig. 5C), and from a noise-exposed count inner and outer hair cells showed no hair cell loss at the cochlea perfused with AP only (Fig. 5D). three cochlear regions, consistent with the lack of PTS (Fig. 1). The results of the synapse counts (Fig. 5E) show that in Thus, the 100-dB SPL noise exposure was sufficient to cause unoperated ears, noise exposure significantly reduced synapse synapse loss unless protected by IEM-1460, but not to cause hair numbers by ∼20% at the 8-kHz location, by ∼25% at the 16-kHz cell death. However, not all synapses were lost, suggesting location, and by ∼40% at the 32-kHz location. In contrast, in ears heterogeneity among synapses in the vulnerability to NICS. perfused with IEM-1460 during noise exposure, there was no To investigate the possibility that differences in AMPAR significant change in synapse number relative to unexposed control subunits contribute to differences in synaptic vulnerability to ears. Following the same logic as for the ABR measurements, this excitotoxicity, we measured the spatial distribution and relative NEUROSCIENCE

Fig. 5. Number of synapses surviving at 14 d postnoise exposure. (A–D) Representative examples of organ of Corti whole-mount preparations from indicated regions (a,8kHz;b,16kHz;c, 32 kHz) of control or noise-exposed mouse cochleae. (A) No noise control. (B)Noise-exposed.(C) Noise-exposed/ IEM-1460–treated. (D) Noise-exposed/AP. Shown are 2D Z-projections, the actual synapse counts were done on the entire three-dimensional (3D) confocal image stacks. The preparations were labeled to detect PSDs (anti-PSD95; green), presynaptic ribbons (anti-CtBP2; red), and hair cells (anti-myosin VI/anti-myosin VIIa; blue), as described in SI Appendix, Materials and Methods. Magnification is the same for all panels. (Scale bar: 10 μm.) In the examples shown here, the numbers of synapses counted were as follows: A, a,15.1;A, b,19.5;A, c,17.9;B, a,12.3;B, b,13.5;B, c,12.8;C, a,14.6,C, b,20.0;C, c,17.0;D, a,10.6;D, b,16.8;D, c, 11.9. Synapse counts at the 8-, 16-, and 32-kHz locations for the indicated numbers of cochleae are summarized in E. Conditions under the same bracket are not significantly different from one another unless indicated. ***P < 0.001, two-way ANOVA over all conditions and frequencies, Tukey correction for multiple comparisons. The noise/control and noise/AP at 8 kHz groups failed a normality test, so a Kruskal–Wallis test was applied (Dunn’s multiple com- parisons test, P < 0.0001). Noise/IEM-1460 is significantly different from noise/AP control at all three locations and is not significantly different from the no noise control at all three locations, indicating that IEM-1460 effectively prevents noise-induced synapse loss.

Hu et al. PNAS | February 18, 2020 | vol. 117 | no. 7 | 3833 Downloaded by guest on September 25, 2021 abundance of AMPAR subunits within and among synapses in and GluA4 is ∼0.15 μm for synapses from all three cochlear unexposed mice. regions, ranging from 0.022 to 0.34 μm at individual synapses (Fig. 6 D and E). The coefficient of variation (SD/mean) of AMPAR Subunit Distribution Is Spatially Nonuniform within Synapses. GluA2 to GluA4 intercentroid distances ranged from 0.44 to The inhibition of NICS by partial blockade of AMPARs with 0.53 for six images analyzed from the three regions. For compar- IEM-1460 implies that cochlear afferent synapses have glutamate ison, intercentroid distances from presynaptic ribbons to GluA receptors that lack GluA2. Because GluA2 seems to be present at subunits (Ribbon-GluA2 or Ribbon-GluA4) are similarly hetero- all synapses, we hypothesize that at each PSD there are GluA2- geneous among synapses, ranging over an order of magnitude, deficient nanodomains, consistent with what we have previously with medians from 0.18 to 0.24 μm. Synapses in the basal cochlea observed in rats (16). It would be at these nanodomains that sig- tend to have closer Ribbon-GluA intercentroid distances than + nificant Ca2 could enter the postsynaptic bouton during synaptic apical or midcochlear synapses (Fig. 6E). transmission, unless the GluA2-lacking CP-AMPARs were blocked To independently confirm that GluA2 and GluA4 distribu- by IEM-1460. Therefore, we asked whether such nanodomains tions differ within a PSD, in separate experiments (SI Appendix, are detectable anatomically in the mouse. Fig. S5), we used a different GluA4 antibody combined with an We imaged apical, midcochlear, and basal regions of adult antibody that recognizes either the GluA2 or the GluA3 subunit CBA/CaJ mice with confocal microscopy after immunolabeling (anti-GluA2,3). Considering the ribbon as the center of the syn- of the presynaptic ribbon (anti-CtBP2) and postsynaptic AMPARs apse, the AMPARs encompass a relatively large volume of space on acutely dissected samples. We labeled GluA2 and GluA4 extending beyond the ribbon. At most synapses, the spatial dis- with subunit-specific antibodies (Figs. 6 and 7). If the different tributions of fluorescence intensities for the two GluA subunit AMPAR subunits occupy distinct spatial distributions within antibodies appear different. Line profiles on two-dimensional (2D) the PSD, then we would expect separation between the centroids maximum-intensity projections reveal that for some synapses, the of the immunofluorescent puncta (Fig. 6C). If synapses display two peaks in GluA fluorescence reside on opposite sides of the heterogeneity of intercentroid distances, then such differences may ribbon (SI Appendix, Fig. S5A). For clarity, the ribbon signal is influence the vulnerability to synaptopathy. Measured in three omitted in SI Appendix,Fig.S5B, to emphasize the GluA distri- dimensions, the median intercentroid distance between GluA2 butions in another subset of synapses from the same image. These

Fig. 6. Distance between ribbons and AMPAR subunits within synapses. (A) Afferent synapses between IHCs and auditory nerve fibers from the mid- cochlea of a P30 mouse, immunolabeled to detect presynaptic ribbon (anti-CtBP2; red), postsynaptic glutamate receptor subunit GluA2 (anti-GluA2; green), and GluA4 (anti-GluA4; blue); maximum intensity Z-projection of a confocal stack. (B) White box in A enlarged. (C) Synapse puncta in B replaced with markers showing centroids identified in three dimensions. (D) Cumulative histograms of 3D intercentroid distances between synaptic puncta in each synapse, for two images in the apical cochlea (8-kHz region; n = 285 synapses), midcochlea (23-kHz region; n = 423 synapses), and basal cochlea (47-kHz region; n = 381 synapses). GluA2-GluA4, black; Ribbon-GluA2, green; Ribbon-GluA4, blue. (E) For comparison of cochlear tonotopic regions, the overall distributions from apical, midcochlear, and basal regions (thick, dashed, and thin lines, respectively). Same colors as in D. (Scale bars: 5 μminA;0.3μm in B and C.)

3834 | www.pnas.org/cgi/doi/10.1073/pnas.1914247117 Hu et al. Downloaded by guest on September 25, 2021 NEUROSCIENCE

Fig. 7. Heterogeneous abundance of AMPAR subunits among and within synapses. (A) Afferent synapses between IHCs and auditory nerve fibers from the midcochlea of a P30 mouse, labeled with antibodies to the presynaptic ribbon (anti-CtBP2; red), the postsynaptic glutamate receptor subunit GluA2 (anti- GluA2; green), and GluA4 (anti-GluA4; blue). Individual channels are shown in gray scale for the region of interest in the white box. (Scale bar: 3 μm.) Maximum intensity projection. (B) Fluorescence intensity (in arbitrary units, a.u.) of GluA4 (y-axis) vs. GluA2 (x-axis) showing that the relative abundance of AMPAR subunits per synapse was positively correlated over a >10-fold range. (C) Fluorescence intensity of GluA4 (blue) or GluA2 (green) vs. CtBP2 intensity. (D) Distributions of intensity per synapse for GluA4 (blue) and GluA2 (green, top axis). The GluA4/GluA2 ratio per synapse ranged from <0.5 to >1.5 (black, bottom axis). (E) GluA4/GluA2 ratio as a function of CtBP2 intensity (magenta, top axis) or overall synapse volume (black, bottom axis). Data are pooled from three images centered at 23-, 30-, and 45-kHz cochlear locations; n = 745 synapses. (F) For the image presented in A, comparison of GluA4/GluA2 ratios among synapses (solid orange columns; n = 138 synapses) and within synapses (open black columns; n = 1,242 nanodomains).

+ data further support a conclusion that the spatial distributions of relative Ca2 permeability among synapses. We used fluores- GluA subunit types within synapses is nonuniform; the GluA4/ cence ratios to measure heterogeneity in the relative abundance GluA2,3 fluorescence ratio tends to peak near the perimeter of of GluA subunits among synapses in a given image (Fig. 7A), with the postsynaptic labeling, relatively far from the ribbon center an antibody specific to GluA2, as in Fig. 6. Synaptic GluA2 (SI Appendix,Fig.S5). This is consistent with a hypothesis that immunofluorescence spanned a >10-fold range of intensity, GluA2-deficient nanodomains, presumably enriched in CP- and GluA4 fluorescence per synapse was positively correlated AMPARs, exist in the PSD and may contribute to synaptic with GluA2 fluorescence (Fig. 7B; R2 = 0.78). On average, susceptibility to excitotoxicity. synapses with greater ribbon fluorescence (anti-CtBP2) have greater GluA2 and GluA4 fluorescence than synapses with less AMPAR Subunit Relative Abundance Is Heterogeneous Among and ribbon fluorescence (Fig. 7C; R2 = 0.35 and 0.28 for CtBP2 vs. within Synapses. Differential vulnerability to NICS among syn- GluA2 and GluA4, respectively); that is, glutamate receptor con- apses implies differential susceptibility to excitotoxicity. We hy- tent is loosely proportional to the overall intensity of the rib- pothesize that this is due to differing ratios of GluA2 to GluA3 bon, with GluA2 somewhat more correlated than GluA4. Despite and GluA4 among synapses, as this ratio would determine the this correlation, the ratio of GluA4/GluA2 immunofluorescence

Hu et al. PNAS | February 18, 2020 | vol. 117 | no. 7 | 3835 Downloaded by guest on September 25, 2021 per synapse ranged over a factor of three, from ∼0.5 to ∼1.5 (Fig. synapses can potentially explain differences in susceptibility within 7D), consistent with the possibility that differences in relative both low- and high-threshold synapse populations. abundance of AMPAR subunits may underlie differences in A second hypothesis is that the ability of IEM-1460 to protect vulnerability to NICS. The GluA4/GluA2 ratio appears to be synapses from NICS while not significantly affecting auditory independent of overall synapse volume and of ribbon size, as threshold or wave-I amplitude after chronic application for days assessed by CtBP2 immunofluorescence intensity (Fig. 7E). is due to the presence of GluA2-lacking and GluA2-containing We noticed that regions within synapses appeared to be equally AMPAR tetramers within each synapse, as implied by the or more heterogeneous in relative abundance compared with the nonuniform spatial distribution of GluA immunofluorescence synapses as a whole. To investigate GluA4/GluA2 ratios within nanodomains within synapses (Figs. 6 and 7F and SI Appendix, synaptic nanodomains, we separated each synapse into nine Fig. S5). GluA2-deficient domains presumably are responsible square-shaped nanodomains of 325 × 325 nm. The range of for NICS, and when GluA2-lacking AMPARs are blocked by nanodomain ratios was 0.25 to 2.9, and the range of synaptic ratios IEM-1460, the AMPARs in the GluA2-abundant domains was 0.54 to 1.3. Thus, our analyses strongly suggest that synaptic mediate synaptic transmission sufficient to drive spiking in the nanodomains are significantly more heterogenous than synaptic auditory nerve fibers, even at low sound levels (Figs. 2A and 3B). microdomains (coefficient of variation, 0.30 vs. 0.17). An additional consideration is that chronic infusion of IEM-1460 may cause synaptic adaptation to the blockade of CP-AMPARs Discussion that could contribute to the resilience of ABR thresholds to IEM- While environmental noise can cause both hair cell loss and 1460. More work is needed to understand the acute and chronic synaptopathy, this study is focused on the latter. Consistent with effects of CP-AMPAR blockade. previous studies (3, 4), we show that if the only consequence of A third hypothesis, implied by IEM-1460 blocking excitotoxicity + noise is synaptopathy, hearing thresholds appear unaffected, al- through GluA2-lacking Ca2 -permeable AMPARs, is that an + though ABR wave-I amplitude is reduced. We calibrated our noise excitotoxic mechanism involving Ca2 influx is the direct cause of + exposure to be sufficient to cause some synapse loss but avoid hair the synaptopathy. Ca2 influx has been implicated in excitotoxicity cell death. The noise exposure we used had been previously shown in the brain (19). More significant to excitotoxicity than the cyto- + + to cause no loss of hair cells and no elevation of ABR threshold solic Ca2 load is the route of Ca2 influx; for example, in cultured + (3). In the long term, synaptopathy is associated with a gradual loss cortical neurons, Ca2 entry via certain glutamate receptors is toxic + of SGNs even with an intact hair cell population (29). but influx via voltage-gated Ca2 channels is not toxic (36). While + This phenomenon, first described in animal models, appears the focus has been on Ca2 influx via NMDA receptors in stud- to be relevant to humans. A reduction of ABR wave-I amplitude ies of excitotoxicity in the brain (19, 37, 38), GluA2-lacking CP- correlated with a history of noise exposure has been observed in AMPARs have also been definitively implicated (20, 21), and humans (30), and loss of IHC synapses has been found in human our studies here using the adamantane derivative IEM-1460 postmortem cochlear specimens (31). Primary loss of SGNs (25, 26) to block GluA2-lacking glutamate receptors impli- without corresponding loss of hair cells has also been observed in cate CP-AMPARs as the glutamate receptor subtype primarily human postmortem studies (32). Therefore, cochlear synaptopathy responsible for NICS. The presence of CP-AMPARs on SGNs is a potential cause of human hearing impairment. While the + is consistent with cobalt ion (Co2 ) entry into guinea pig SGNs issue is not resolved in the current literature, some studies sug- following excitotoxic exposure to kainate (although not to AMPA) gest that synaptopathy may cause tinnitus and impair speech-in- in vivo (12). Nevertheless, further investigation is needed to de- noise performance (27). Previous studies have implicated AMPA-type glutamate re- termine the exact mechanism. GluA2-lacking AMPARs and CP- AMPARs appear to be essentially equivalent (17, 18, 39), but this ceptors in synaptopathy (23, 24). Here we implicate a particular 2+ subset of AMPARs, GluA2-lacking AMPARs, by showing that does not necessarily imply that Ca entry is the direct cause of synaptopathy. GluA2-lacking AMPARs differ from GluA2- IEM-1460 significantly reduces NICS. This allows formulation of 2+ at least three novel hypotheses about NICS. One of these hy- containing AMPARs not only in being Ca -permeable, but also in having significantly higher conductances and, conse- potheses involves the differential susceptibility among afferent + synapses on IHCs. For example, with the noise exposure used quently, significantly higher inward Na currents (17, 18). In- here, 70 to 75% of the synapses remain intact. One observation vestigation of GluA2 knockout mice (40) has shown increased susceptibility of cortical neurons to excitotoxicity correlated of differential susceptibility among synapses is that synapses + 2+ + made by SGNs with high thresholds are more susceptible to with increased Na rather than Ca influx. Increased Na NICS than those with low thresholds, as shown by studies in influx could account for the swelling of SGN terminals due to guinea pigs (33). However, even within the high- and low- excitotoxic trauma in vivo (5). threshold groups, synapses show differential susceptibility, sug- Although IEM-1460 by i.p. injection has been used as an gesting that additional factors contribute to NICS. Data presented anticonvulsant in immature rats (41), we chose chronic intra- here suggest that differences in GluA2 expression, evident as a cochlear perfusion rather than systemic injection to ensure that greater than threefold difference between the highest and lowest the IEM-1460 could gain access to the cochlea and to minimize GluA4/GluA2 ratios (Fig. 7D), may be one molecular determi- off-target effects. However, intracochlear perfusion does in- nant of differential vulnerability, presumably with GluA2-deficient troduce a source of potential artifact. Inserting a cannula into synapses being more susceptible and GluA2-abundant synapses the round window can cause a conductive hearing loss that less susceptible. Apropos, GluA2 relative abundance on SGN would diminish the intensity of the sound reaching the cochlea terminals appears to be independent of SGN response threshold. and so reduce noise damage to the cochlea regardless of the Studies in cat (34) and mouse (35) have shown that high-threshold contents of the cannula. Reduced synapse loss could be accounted synapses, which are located on the modiolar (medial) side of hair for by this noise suppression rather than by an effect of IEM-1460. cells, tend to have larger ribbons than low-threshold synapses on In fact, we did detect some reduction in NICS attributable to the the pillar (lateral) side. Here we show that variation in the GluA4/ surgery. For this reason, we base our conclusions on comparisons GluA2 ratio among synapses is independent of ribbon size (Fig. of mice receiving either IEM-1460 or vehicle but implanted with 7E) and, by implication, is also independent of threshold. Thus, identical cannulae. This comparison, which compensates for any differences in GluA2 expression level do not appear to account for effect of surgery, shows a significant protective effect of IEM-1460 the difference in susceptibility between high- and low-threshold against NICS, evidenced by preservation of ABR wave-I amplitude synapses. Nevertheless, differences in GluA2 expression among and synapse number.

3836 | www.pnas.org/cgi/doi/10.1073/pnas.1914247117 Hu et al. Downloaded by guest on September 25, 2021 These data support a role for blockade of CP-AMPARs in IEM-1460 in AP was perfused into the left cochlea via the round window by protection against noise trauma. Although synapses lost because a minipump/cannula system. The cannula length was calculated so that the of excitotoxic trauma or acoustic overexposure can regenerate minipump contents reached the cochlea ∼3.5 d after surgical implantation. after treatment with NT-3 (23, 42–44), full functional regeneration The adult male CBA/CaJ mice are then exposed to 2 h of 100-dB SPL 8- to may be challenging. Transtympanic delivery of NT-3 was not ef- 16-kHz octave band noise at 6 d after the surgery. fective in all subjects (43), and high levels of NT-3 in the cochlea ABRs were measured before surgery, at 3 d and 5 d after surgery, on PND1, may damage otherwise healthy elements (45). In guinea pigs, and on PND14. These timepoints correspond to measurements of the baseline synapses repaired after NICS appear to be functionally deficient in ABR, postsurgery ABR, IEM-1460–exposed/prenoise ABR, the temporary temporal processing (46). We raise here the possibility of an al- threshold shift, and the permanent threshold shift for the within-subject ternative to regeneration: prevention of degeneration by selectively comparison. Collectively, these allowed measurement of the effect on blocking the neurotransmitter receptor class largely responsible for ABR threshold and amplitude of IEM-1460 itself, of the noise exposure, the underlying excitotoxic trauma, CP-AMPARs. and of any mitigation of acoustic trauma by IEM-1460. An important advantage of CP-AMPAR blockers over less Synapse number and other histological features were assessed in cochlear selective glutamate receptor blockers is that the CP-AMPAR whole-mount preparations. Presynaptic ribbons, PSDs, and hair cells were blocker that we used does not appear to significantly impair labeled with antibodies to CtBP2, PSD95, and myosin VI/VIIA, respectively. hearing thresholds or wave-I amplitudes at the concentration used. Glutamate receptor subunits GluA2 and GluA4 were labeled with the re- Nevertheless, therapeutic use of IEM-1460 or similar protective spective antibodies. The labeled preparations were fluorescently labeled with appropriate secondary antibodies and viewed by confocal microscopy. agents will require a less invasive means of delivery that also avoids – deleterious off-target effects, such as agents passing the blood Data Availability Statement. Data supporting the findings of this paper are brain barrier and interfering with normal glutamatergic trans- contained within the paper and SI Appendix. Detailed descriptions of the missionintheCNS.Forthisreason,andtoavoidsubtleeffects surgical procedures, ABR measures and analysis, noise exposure, cochlear on cochlear synaptic signaling not detected by ABR, thera- whole-mount preparation, immunofluorescent labeling, confocal imaging and peutic use would be most valuable acutely, immediately before quantitation, and statistics are also provided in SI Appendix. expected noise exposure. ACKNOWLEDGMENTS. We thank Dr. Amy Lee and members of the S.H.G. Materials and Methods lab for valuable comments on the manuscript. Support for this study was Protocols for all experiments involving animals were approved by the Uni- provided by grants from the NIH (DC02961, to S.H.G., and DC014712, to versity of Iowa’s Institutional Animal Care and Use Committee or the Animal M.A.R.), the Department of Defense (W81XWH-14-1-0494, to S.H.G.), and

Studies Committee of Washington University in St. Louis. the American Hearing Research Foundation (to N.H.). NEUROSCIENCE

1. S. Safieddine, A. El-Amraoui, C. Petit, The auditory hair cell ribbon synapse: From 18. T. A. Verdoorn, N. Burnashev, H. Monyer, P. H. Seeburg, B. Sakmann, Structural deter- assembly to function. Annu. Rev. Neurosci. 35, 509–528 (2012). minants of ion flow through recombinant glutamate receptor channels. Science 252, 2. M. A. Rutherford, T. Moser, “The ribbon synapse between type I spiral ganglion 1715–1718 (1991). neurons and inner hair cells” in The Primary Auditory Neurons of the Mammalian 19. K. Szydlowska, M. Tymianski, Calcium, ischemia and excitotoxicity. Cell Calcium 47, Cochlea, A. Dabdoub, B. Fritzsch, A. N. Popper, R. R. Fay, Eds. (Springer, New York, 122–129 (2010). 2016), pp. 117–156. 20. M. V. Bennett et al., The GluR2 hypothesis: Ca(++)-permeable AMPA receptors in 3. S. G. Kujawa, M. C. Liberman, Adding insult to injury: Cochlear nerve degeneration delayed . Cold Spring Harb. Symp. Quant. Biol. 61,373–384 after “temporary” noise-induced hearing loss. J. Neurosci. 29, 14077–14085 (2009). (1996). 4. H. W. Lin, A. C. Furman, S. G. Kujawa, M. C. Liberman, Primary neural degeneration in 21. D. E. Pellegrini-Giampietro, J. A. Gorter, M. V. Bennett, R. S. Zukin, The GluR2 (GluR-B) the guinea pig cochlea after reversible noise-induced threshold shift. J. Assoc. Res. hypothesis: Ca(2+)-permeable AMPA receptors in neurological disorders. Trends – Otolaryngol. 12, 605 616 (2011). Neurosci. 20, 464–470 (1997). 5. R. Pujol, M. Lenoir, D. Robertson, M. Eybalin, B. M. Johnstone, Kainic acid selectively 22. H. Tanaka, S. Y. Grooms, M. V. Bennett, R. S. Zukin, The AMPAR subunit GluR2: Still – alters auditory dendrites connected with cochlear inner hair cells. Hear. Res. 18, 145 151 front and center-stage. Brain Res. 886, 190–207 (2000). (1985). 23. Q. Wang, S. H. Green, Functional role of neurotrophin-3 in synapse regeneration by ’ 6. J. L. Puel, J. Ruel, C. Gervais d Aldin, R. Pujol, Excitotoxicity and repair of cochlear spiral ganglion neurons on inner hair cells after excitotoxic trauma in vitro. J. Neurosci. – synapses after noise-trauma-induced hearing loss. Neuroreport 9, 2109 2114 31,7938–7949 (2011). (1998). 24. J. L. Puel, R. Pujol, F. Tribillac, S. Ladrech, M. Eybalin, Excitatory amino acid antagonists 7. N. Hakuba, K. Koga, K. Gyo, S. I. Usami, K. Tanaka, Exacerbation of noise-induced protect cochlear auditory neurons from excitotoxicity. J. Comp. Neurol. 341, 241–256 hearing loss in mice lacking the glutamate transporter GLAST. J. Neurosci. 20, 8750–8753 (1994). (2000). 25. D. B. Tikhonov, M. V. Samoilova, S. L. Buldakova, V. E. Gmiro, L. G. Magazanik, 8. K. X. Kim et al., Vesicular glutamatergic transmission in noise-induced loss and repair Voltage-dependent block of native AMPA receptor channels by dicationic com- of cochlear ribbon synapses. J. Neurosci. 39, 4434–4447 (2019). pounds. Br. J. Pharmacol. 129, 265–274 (2000). 9. A. S. Niedzielski, R. J. Wenthold, Expression of AMPA, kainate, and NMDA receptor 26. L. G. Magazanik et al., Block of open channels of recombinant AMPA receptors and subunits in cochlear and vestibular ganglia. J. Neurosci. 15, 2338–2353 (1995). native AMPA/kainate receptors by adamantane derivatives. J. Physiol. 505, 655–663 10. L. Klotz et al., Localization of group II and III metabotropic glutamate receptors at (1997). pre- and postsynaptic sites of inner hair cell ribbon synapses. FASEB J. 33, 13734–13746 27. M. C. Liberman, S. G. Kujawa, Cochlear synaptopathy in acquired sensorineural (2019). – 11. J. Ruel, C. Chen, R. Pujol, R. P. Bobbin, J. L. Puel, AMPA-preferring glutamate receptors hearing loss: Manifestations and mechanisms. Hear. Res. 349, 138 147 (2017). 28. A. F. Ryan, S. G. Kujawa, T. Hammill, C. Le Prell, J. Kil, Temporary and permanent in cochlear physiology of adult guinea pig. J. Physiol. 518, 667–680 (1999). 12. J. Ruel, R. P. Bobbin, D. Vidal, R. Pujol, J. L. Puel, The selective AMPA receptor an- noise-induced threshold shifts: A review of basic and clinical observations. Otol. – tagonist GYKI 53784 blocks action potential generation and excitotoxicity in the Neurotol. 37, e271 e275 (2016). guinea pig cochlea. Neuropharmacology 39, 1959–1973 (2000). 29. S. G. Kujawa, M. C. Liberman, Acceleration of age-related hearing loss by early noise – 13. J. L. Puel, J. Ruel, M. Guitton, J. Wang, R. Pujol, The inner hair cell synaptic complex: exposure: Evidence of a misspent youth. J. Neurosci. 26, 2115 2123 (2006). Physiology, pharmacology and new therapeutic strategies. Audiol. Neurotol. 7,49–54 30. G. C. Stamper, T. A. Johnson, Auditory function in normal-hearing, noise-exposed – (2002). human ears. Ear Hear. 36, 172 184 (2015). 14. E. Glowatzki, P. A. Fuchs, Transmitter release at the hair cell ribbon synapse. Nat. 31. L. M. Viana et al., Cochlear neuropathy in human presbycusis: Confocal analysis of – Neurosci. 5, 147–154 (2002). hidden hearing loss in post-mortem tissue. Hear. Res. 327,78 88 (2015). 15. Z. Chen, M. Peppi, S. G. Kujawa, W. F. Sewell, Regulated expression of surface AMPA 32. C. A. Makary, J. Shin, S. G. Kujawa, M. C. Liberman, S. N. Merchant, Age-related receptors reduces excitotoxicity in auditory neurons. J. Neurophysiol. 102, 1152–1159 primary cochlear neuronal degeneration in human temporal bones. J. Assoc. Res. (2009). Otolaryngol. 12, 711–717 (2011). + 16. J. Y. Sebe et al., Ca2 -permeable AMPARs mediate glutamatergic transmission and 33. A. C. Furman, S. G. Kujawa, M. C. Liberman, Noise-induced cochlear neuropathy is excitotoxic damage at the hair cell ribbon synapse. J. Neurosci. 37, 6162–6175 selective for fibers with low spontaneous rates. J. Neurophysiol. 110,577–586 (2017). (2013). + 17. M. Hollmann, M. Hartley, S. Heinemann, Ca2 permeability of KA-AMPA–gated glu- 34. A. Merchan-Perez, M. C. Liberman, Ultrastructural differences among afferent syn- tamate receptor channels depends on subunit composition. Science 252, 851–853 apses on cochlear hair cells: Correlations with spontaneous discharge rate. J. Comp. (1991). Neurol. 371, 208–221 (1996).

Hu et al. PNAS | February 18, 2020 | vol. 117 | no. 7 | 3837 Downloaded by guest on September 25, 2021 35. L. D. Liberman, H. Wang, M. C. Liberman, Opposing gradients of ribbon size and 41. E. Szczurowska, P. Mareš, An antagonist of calcium permeable AMPA receptors, AMPA receptor expression underlie sensitivity differences among cochlear-nerve/hair- IEM1460: Anticonvulsant action in immature rats? Epilepsy Res. 109, 106–113 (2015). cell synapses. J. Neurosci. 31, 801–808 (2011). 42. G. Wan, M. E. Gómez-Casati, A. R. Gigliello, M. C. Liberman, G. Corfas, Neurotrophin-3 36. R. Sattler, M. P. Charlton, M. Hafner, M. Tymianski, Distinct influx pathways, not calcium regulates ribbon synapse density in the cochlea and induces synapse regeneration load, determine neuronal vulnerability to calcium neurotoxicity. J. Neurochem. 71, after acoustic trauma. eLife 3, e03564 (2014). 43. J. Suzuki, G. Corfas, M. C. Liberman, Round-window delivery of neurotrophin 3 2349–2364 (1998). regenerates cochlear synapses after acoustic overexposure. Sci. Rep. 6,24907 37. M. Tymianski, M. P. Charlton, P. L. Carlen, C. H. Tator, Source specificity of early calcium (2016). neurotoxicity in cultured embryonic spinal neurons. J. Neurosci. 13, 2085–2104 (1993). 44. D. J. Sly et al., Applying neurotrophins to the round window rescues auditory function 38. R. Sattler, M. Tymianski, Molecular mechanisms of glutamate receptor-mediated ex- and reduces inner hair cell synaptopathy after noise-induced hearing loss. Otol. – citotoxic neuronal cell death. Mol. Neurobiol. 24, 107 129 (2001). Neurotol. 37, 1223–1230 (2016). 39. N. Burnashev, H. Monyer, P. H. Seeburg, B. Sakmann, Divalent ion permeability of 45. M. Y. Lee et al., Viral-mediated Ntf3 overexpression disrupts innervation and AMPA receptor channels is dominated by the edited form of a single subunit. Neuron hearing in nondeafened guinea pig cochleae. Mol. Ther. Methods Clin. Dev. 3, 8, 189–198 (1992). 16052 (2016). 40. K. Iihara et al., The influence of glutamate receptor 2 expression on excitotoxicity in 46. L. Shi et al., Ribbon synapse plasticity in the cochleae of guinea pigs after noise-induced Glur2 null mutant mice. J. Neurosci. 21, 2224–2239 (2001). silent damage. PLoS One 8, e81566 (2013).

3838 | www.pnas.org/cgi/doi/10.1073/pnas.1914247117 Hu et al. Downloaded by guest on September 25, 2021