Vesicular Glutamatergic Transmission in Noise-Induced Loss and Repair of Cochlear Ribbon Synapses

Abstract

Noise-induced excitotoxicity is thought to depend on glutamate. However, the excitotoxic mechanisms are unknown, and the necessity of glutamate for synapse loss or regeneration is unclear. Despite absence of glutamatergic transmission from cochlear inner hair cells in mice lacking the vesicular glutamate transporter-3 (Vglut3^KO), at 9–11 weeks, approximately half the number of synapses found in Vglut3^WT were maintained as postsynaptic AMPA receptors juxtaposed with presynaptic ribbons and voltage-gated calcium channels (Ca_V1.3). Synapses were larger in Vglut3^KO than Vglut3^WT. In Vglut3^WT and Vglut3^+/− mice, 8–16 kHz octave-band noise exposure at 100 dB sound pressure level caused a threshold shift (∼40 dB) and a loss of synapses (>50%) at 24 h after exposure. Hearing threshold and synapse number partially recovered by 2 weeks after exposure as ribbons became larger, whereas recovery was significantly better in Vglut3^WT. Noise exposure at 94 dB sound pressure level caused auditory threshold shifts that fully recovered in 2 weeks, whereas suprathreshold hearing recovered faster in Vglut3^WT than Vglut3^+/−. These results, from mice of both sexes, suggest that spontaneous repair of synapses after noise depends on the level of Vglut3 protein or the level of glutamate release during the recovery period. Noise-induced loss of presynaptic ribbons or postsynaptic AMPA receptors was not observed in Vglut3^KO, demonstrating its dependence on vesicular glutamate release. In Vglut3^WT and Vglut3^+/−, noise exposure caused unpairing of presynaptic ribbons and presynaptic Ca_V1.3, but not in Vglut3^KO where Ca_V1.3 remained clustered with ribbons at presynaptic active zones. These results suggest that, without glutamate release, noise-induced presynaptic Ca²⁺ influx was insufficient to disassemble the active zone. However, synapse volume increased by 2 weeks after exposure in Vglut3^KO, suggesting glutamate-independent mechanisms.

SIGNIFICANCE STATEMENT Hearing depends on glutamatergic transmission mediated by Vglut3, but the role of glutamate in synapse loss and repair is unclear. Here, using mice of both sexes, we show that one copy of the Vglut3 gene is sufficient for noise-induced threshold shift and loss of ribbon synapses, but both copies are required for normal recovery of hearing function and ribbon synapse number. Impairment of the recovery process in mice having only one functional copy suggests that glutamate release may promote synapse regeneration. At least one copy of the Vglut3 gene is necessary for noise-induced synapse loss. Although the excitotoxic mechanism remains unknown, these findings are consistent with the presumption that glutamate is the key mediator of noise-induced synaptopathy.

Introduction

Glutamate is recognized as the afferent neurotransmitter between cochlear inner hair cells (IHCs) and auditory nerve fibers (ANFs) (Ruel et al., 2008; Seal et al., 2008; Akil et al., 2012). Excessive activation of AMPA-type glutamate receptors may be sufficient to initiate excitotoxic damage to ANF terminals, as demonstrated by application of AMPA to the cochlea. In pharmacological experiments and in response to sound overexposure, ANF postsynaptic terminals swell and burst, producing postsynaptic damage and vacuolization (Puel et al., 1994; Wang and Green, 2011). The excitotoxic mechanisms of glutamate in the organ of Corti are unknown and may involve Ca²⁺-permeable AMPA receptors (Puel et al., 1998; Eybalin et al., 2004; Sebe et al., 2017). The question of whether excess synaptic glutamate release drives noise-induced synapse loss has not been directly tested with a genetic approach. Moreover, the mechanisms of hair-cell synapse repair are not fully understood. Neurotrophin-3 may enhance repair after noise exposure (Wan et al., 2014), but the role of glutamate is unclear.

In addition to glutamate release, sound overexposure may activate multiple damage pathways by releasing adenosine, aspartate, protons, and proinflammatory cytokines (Jäger et al., 1998; Peng et al., 2004; Wakabayashi et al., 2010; Kurabi et al., 2017). Upstream of glutamate release from the IHC, damage to the presynaptic active zone may involve Ca²⁺ influx through voltage-gated Ca²⁺ channels at the IHC ribbon synapses. Glutamate is concentrated into synaptic vesicles in IHCs by vesicular glutamate transporter type 3 (Vglut3), pathogenic variants of which cause nonsyndromic deafness DFNA25 in humans (Greene et al., 2001; Thirlwall et al., 2003; Ruel et al., 2008). To isolate the roles of Vglut3 in noise damage and repair in the inner ear, we studied mice having 0, 1, or 2 functional copies of the Vglut3 gene (SLC17A8).

Vglut3^WT and heterozygous (Vglut3^+/−) mice have similar hearing thresholds measured by auditory brainstem response (ABR), suggesting that one copy of Vglut3 is sufficient for hearing low-level sounds. Vglut3^KO mice are profoundly deaf but have no obvious functional vestibular deficit and otherwise develop relatively normally. In Vglut3^KO, the outer hair cells, responsible for cochlear amplification, still transduce sound and generate distortion product otoacoustic emissions (DPOAEs). The IHCs of Vglut3^KO mice form ribbon synapses with ANFs; they develop a mature complement of basolateral ion channels, including Ca_V1.3; and they exocytose the contents of synaptic vesicles upon depolarization, like Vglut3^WT, but in Vglut3^KO the synaptic vesicles lack glutamate (Ruel et al., 2008; Seal et al., 2008).

Here, to address the roles of vesicular glutamate in the mechanisms of synapse loss and regeneration after overexposure to sound, we assessed hearing function and synaptic anatomy in Vglut3 mice on C57BL/6J background at 0 h, 24 h, 14 d, and 28 d after noise exposure. Our findings show that glutamate is required, and that presynaptic Ca²⁺ influx alone is likely insufficient, for loss of synapse number. Interestingly, noise trauma to Vglut3^KO did induce changes in ribbon synapse morphology, suggesting glutamate-independent mechanisms.

After noise trauma in Vglut3^WT and Vglut3^+/− mice on C57BL/6J background, we found an immediate reduction of synapse number per IHC. This was followed by partial recovery of synapse number, contrasting with results in CBA/CaJ mice showing no recovery (Kujawa and Liberman, 2009; Liberman et al., 2015), but similar to findings in guinea pig showing partial recovery after noise trauma (Shi et al., 2013). Following noise trauma, recovery of hearing function and synapse number was more complete in Vglut3^WT than Vglut3^+/−, suggesting a role for Vglut3 and glutamate transmission in ribbon synapse repair and recovery of hearing function.

Materials and Methods

Animals.

Mice were used in accordance with protocols approved by the Animal Studies Committee of Washington University in St. Louis. Vglut3 mice (Slc17a8^tm1Edw) on C57BL/6J background were obtained from The Jackson Laboratory (RRID:IMSR_JAX:016931). We used mice of either sex in similar proportions aged 2–3 months (normative studies) or 2–9 months (to study age-related hearing loss on C57BL/6J background).

Acoustic overexposure.

We exposed awake unrestrained mice to 2 h of 8–16 kHz octave band noise at a sound pressure level (SPL) of 94 or 100 dB (Hirose et al., 2005), known to cause swelling in the peripheral terminals of ANFs in the basal (high-frequency) half of the cochlea. Cochlear responses to threshold-level tones define the tonotopic map (Müller et al., 2005). As SPL increases, the region of excitation broadens, and the focal region of excitation shifts toward mid-frequency regions corresponding to the center of the threshold audiogram (Lee et al., 2019). As 8–16 kHz is a relatively low-frequency range for mouse, with increasing SPL, the focal region broadens and shifts from more apical cochlear regions toward higher-frequency regions in the cochlear base (see Fig. 1). Mice were placed in individual cages on a suspended shelf in a custom-made acrylic chamber in which no two sides were parallel. The sound stimulus was produced by an RX6 processor (Tucker-Davis Technologies), filtered (Frequency Devices), amplified (Crown 75A power amplifier), and delivered to the acrylic chamber via a speaker horn (JBL). The SPL was measured through a 1/4-inch free-field microphone (ACO Pacific) calibrated to a 124-dB pistonphone (Bruel and Kjaer). Before the experimental noise exposure, four quadrants of the chamber were sampled with the 1/4-inch microphone, and sound pressure was confirmed to vary by no more than 0.5 dB across these measurement positions.

Figure 1.

Better recovery of ABR thresholds in Vglut3^WT than Vglut3^+/− after 100 dB exposure. A, ABR thresholds (mean ± SD) for unexposed mice aged 9–11 weeks. Vglut3^KO are profoundly deaf. Red arrowheads indicate thresholds > 100 dB SPL. Unexposed Vglut3^WT and Vglut3^+/− had similar thresholds. B, Comparison of ABR thresholds before (unexposed, filled circles) and 1 d after exposure (exposed, open circles) to 2 h of 100 dB SPL, 8–16 kHz octave-band noise. Yellow rectangle indicates frequency band of noise. Threshold shifts exceeded 40 dB in the mid-frequencies for Vglut3^WT and Vglut3^+/− mice. Threshold shifts occur for test frequencies within the noise band and also for higher frequencies that probe more basal cochlear locations. C, Two weeks after exposure (open squares), ABR thresholds had recovered better in Vglut3^WT than Vglut3^+/−. N = 6–19 animals per group. *Frequencies for which thresholds of Vglut3^WT are significantly lower than Vglut3^+/− (p < 0.05).

ABR and distortion product otoacoustic emissions.

Hearing was assessed 3 d before acoustic overexposure and reassessed at 24 h after exposure (1 d group), 14 d after exposure (2 week group), and 28 d after exposure (4 week group). The ABR is a sound-evoked potential reflecting activation of the ANFs by the IHCs (wave I), and subsequent activation of the ascending auditory pathways (waves II-V). Responses were recorded from subcutaneous electrodes located at the vertex (active electrode) and behind the right pinna (reference electrode), with ground electrode placed on the back, under ketamine (100 mg/kg)/xylazine (20 mg/kg) anesthesia, using a sampling frequency of 25 kHz. ABR thresholds in response to tone pips at frequencies presented in half-octave intervals (5.6, 8, 11.2, 16, 22.6, 32, 45.2, and 64 kHz) were determined using 5 ms tone pips (including 0.5 ms cosine² rise/fall) at a repetition rate of 40 s⁻¹. The responses were amplified (×10,000), filtered (100 Hz to 3 kHz), and averaged using custom computer software (System 3; Tucker-Davis Technologies). Stimuli were presented in 5 dB steps from 15 dB to 100 dB SPL in ascending order to the right ear. At each level, 1024 responses were averaged, with stimulus polarity alternated. Response waveforms were rejected if the peak-to-peak voltage exceeded 15 μV. Threshold was determined by a single observer, who noted the lowest sound level at which a recognizable waveform was present. Waveforms were confirmed as auditory-evoked responses by their increasing amplitude and decreasing latency with increasing SPL. If hearing threshold was not detected at 100 dB SPL, the threshold value was assigned as >100 dB. For wave I measurements, ABRs were evoked with tone pips presented at a rate of 21 s⁻¹ and measured using Tucker-Davis Technologies hardware in conjunction with BiosigRZ software (Tucker-Davis Technologies). Stimuli were presented in 5 dB steps in descending order to the right ear. Responses to 500 stimulus presentations at each level were used to construct average ABR waveforms. ABR wave I amplitudes were quantified offline at 8, 22.6, and 45.2 kHz as the difference between the prestimulus baseline and the first positive peak. To assess the status of the cochlear outer hair cells, we measured DPOAEs with a microphone in the ear canal. DPOAEs were recorded from the right ear using Emav software (S. Neely, Boys Town National Research Hospital) in conjunction with Tucker-Davis Technologies and custom hardware, using a sampling rate of 192 kHz. We measured cochlear emissions at 2f₁-f₂, with f₂ frequencies from 5 to 40 kHz (total of 25 tone pairs, 8 per octave). The f₁/f₂ ratio was 1.22. Tone levels were set to 75 dB SPL (L₁) and 65 dB SPL (L₂). The mean level of the DPOAE across frequencies per mouse was used to calculate group means as a function of time after exposure in Figure 9C–E. The frequency-specific level of the DPOAE was used to calculate group means as a function of f₂ in Figure 9A,B.

Figure 9.

DPOAE levels were reduced in Vglut3^WT, Vglut3^+/−, and Vglut3^KO mice after 94 or 100 dB exposure. A, B, DPOAE levels (mean ± SD) as a function of input frequency (f₂) in unexposed mice (left), 1 d after (middle), or 28 d after (right) 94 dB exposure (A) or 100 dB exposure (B) to 8–16 kHz octave-band noise for 2 h. C–E, DPOAE levels (mean ± SD) over time in unexposed mice (C), and in mice exposed to 94 dB (D) or 100 dB SPL noise (E). For each group (genotype × exposure condition), the mean DPOAE level across all frequencies per mouse was used to calculate the group means. N = 4–13 mice per group, exposed at 9–11 weeks of age. Bottom markers, Statistical comparison between three genotypes for each exposure condition (dB SPL group × time after exposure). Top markers, Statistical comparison of genotype over time within each dB SPL group. The color of the asterisks at each time indicates the genotypes having significantly smaller mean DPOAE levels relative to unexposed.

Immunohistofluorescence, confocal microscopy, synapse counts, volume, and intensity measurements.

Synapse loss and recovery after 100 dB SPL noise exposure were quantified on left cochleae immediately after the terminal ABR/DP assessment on right cochleae. Briefly, after cardiac perfusion with 4% PFA, for whole-mount preparations, organ of Corti and spiral ganglion were isolated from the cochlea in three pieces containing the apical, middle, and basal thirds. Antibodies: CtBP2 mouse (BD Biosciences; RRID:AB_399431), Ca_V1.3 rabbit (Alomone Labs; RRID:AB_2039775), GluA3 goat (Santa Cruz Biotechnology; RRID:AB_2113895), and Myosin7a rabbit (Proteus Biosciences; RRID:AB_2314838) primary antibodies were used with species-appropriate secondary antibodies conjugated to AlexaFluor (Invitrogen) fluorophores excited by 488, 555, or 647 nm light in triple-labeled samples as previously described (Jing et al., 2013; Sebe et al., 2017). Samples were batch processed using the same reagent solutions in six cohorts, each including exposed and unexposed Vglut3^WT, Vglut3^+/−, and Vglut3^KO mice. For synapse counts, intensity, and volume measurements, confocal stacks were collected without saturation of pixel intensity and sampled with a z step of 0.38 μm and with pixel size of 100 nm in x and y, on an LSM 700 with a 63 × 1.4 NA oil objective lens (Carl Zeiss). From the left cochlea of 34 mice, we collected 2 or 3 images at each of 3 cochlear regions centered near tonotopic characteristic frequencies of 9, 22.5, and 44.5 kHz (Müller et al., 2005). The numbers of IHCs and synapses were counted manually using the maximum-intensity projection (ImageJ software, National Institutes of Health) and the individual optical sections (Zen software, Carl Zeiss) to calculate the mean number of synapses per IHC in each image. For each group, grand means were calculated across images. Synapses were identified as juxtaposed pairs of presynaptic ribbons (labeled with anti-CtBP2) and postsynaptic AMPA-type glutamate receptor puncta (labeled with anti-GluA3), which appear to partly overlap at confocal resolution (Rutherford, 2015). CtBP2 puncta were identified as unpaired ribbons when they appeared to be in the IHC cytoplasm, completely overlapping with Myosin7a but without juxtaposed GluA3. For unpaired ribbons, we did not attempt to distinguish membrane-anchored from unanchored. GluA3 puncta were identified as unpaired “ribbonless” synapses when they lacked juxtaposed CtBP2 but appeared to be on the surface of the Myosin7a-labeled IHC. For unpaired GluA3, we distinguished ribbonless synapses on the cell surface from GluA3 puncta that may have been internalized in the IHC cytoplasm by excluding those puncta that appeared to completely overlap with Myosin7a in three dimensions. Synapses were quantified from mice killed 2–3 d before exposure, 1 d after exposure, or 2 weeks after exposure. The volumes of paired presynaptic ribbons and postsynaptic GluA3 were measured in Imaris software (Bitplane) by segmenting the synaptic elements from the 3D image as “surface” objects, using identical settings for each image stack, including the “local contrast background subtraction” algorithm for calculating threshold. This automatic thresholding algorithm compensates for differences in overall luminance between stacks and avoids potential subjective bias of setting a user-defined arbitrary threshold value. In some experiments, anti-Myosin7a was replaced with anti-Ca_V1.3 to assess presynaptic Ca²⁺ channels in unexposed ears or at 0 h after noise exposure. Intensities of synaptic ribbons, voltage-gated Ca²⁺ channels, and AMPA receptors were measured in Imaris software as the sum of pixel intensities within surface objects (Sebe et al., 2017). For display only, some images were acquired on an LSM 880 (Carl Zeiss) with Airyscan processing using a 100× 1.5 NA oil objective lens, a z step of 0.15 μm, and a pixel size of 40 nm in x and y.

Experimental design and statistical analysis.

Data were tested for normality with the Shapiro–Wilk test or assumed to be not normally distributed due to sample size. Parametric or nonparametric tests were applied as appropriate. All statistical values are presented as mean ± SD; mixed-model ANOVA: analysis with autoregressive covariance structure and individual subjects as a random factor was implemented using PROC MIXED procedure (Littell et al., 1998). The mixed-model approach is particularly useful for repeated measures on subjects over time (longitudinal studies) because it makes maximum use of the data available at any assessment time (deals with missing data). Here, it was used to investigate differences between animal groups and changes within groups over subsequent assessment times after noise exposure; exact p values are given. Dunnett's test was used for pairwise comparisons of multiple experimental groups to one control group; the family-wise error rate was kept below α. Planned bivariate analyses were used for comparisons between groups at specific probe frequency or level, or cochlear location. Wilcoxon rank-sum test was used for comparisons between two groups within assessment times at specific probe frequencies or levels; exact p values are given. To compare two samples of synapse volumes, we used the Kolmogorov–Smirnov test; exact p values are given; effect size was estimated as the absolute value of the difference between medians divided by the pooled SD. N is the number of animals or cochleae, as indicated in each figure legend or table. All statistical tests were two-sided and evaluated at the α level of 0.05 in SAS 9.4 (SAS; RRID:SCR_008567), Excel (Microsoft; RRID:SCR_016137), or IGOR Pro 7 (Wavemetrics; RRID:SCR_000325). For multiple comparisons with Student's t test, the α level was adjusted with the Bonferroni correction and the exact p value or range is reported. Graphs were plotted in IGOR.

Results

Cochlear function in Vglut3 mice before and after 100 dB exposure

The ABR is a measure of neuronal activity of the ascending auditory pathway. It depends on synchronous activation of ANFs by synaptic release of glutamate from IHCs at the onset of sound. The minimum sound level required to evoke an ABR at a given frequency relates closely to the threshold of perception defined by the behavioral audiogram in mice and human (Elberling and Don, 1987; Longenecker et al., 2016). ABR thresholds in unexposed Vglut3^WT and Vglut3^+/− mice at 9–11 weeks of age were not significantly different from each other at any of the 8 test frequencies (Fig. 1A). In contrast, Vglut3^KO mice typically showed no ABR (Fig. 1A), although some mice exhibited very small responses to high-SPL tone bursts (Fig. 1A; see 7C).

In Vglut3^WT and Vglut3^+/− mice, 2 h of 8–16 kHz noise at 100 dB SPL caused a threshold shift at 1 d that recovered incompletely by 2 weeks after exposure. ABR thresholds measured 1 d after exposure were significantly elevated relative to unexposed mice at test frequencies between 11.2 and 45.2 kHz, reaching a maximum shift of ∼50 dB at 16 kHz (Fig. 1B). Although ABR thresholds of Vglut3^WT and Vglut3^+/− mice were not significantly different from each other for any test frequency at 1 d after exposure, by 2 weeks after exposure, recovery of ABR threshold was significantly better in Vglut3^WT than Vglut3^+/− for three mid-frequency probe tones: 11.2, 16, and 32 kHz (Fig. 1C; Table 1). Next, we looked at the ribbon synapses between IHCs and ANFs, where the electrical signal of the ABR is initiated.

Table 1.

ABR thresholds (dB SPL, mean ± SD) for test frequencies showing significant differences between Vglut3^WT and Vglut3^+/− 2 weeks after 100 dB exposure^a

	Tone frequency (kHz)
	11.2	16	32
Vglut3WT	33.3 ± 4.1	50.8 ± 7.4	72.5 ± 10.4
Vglut3+/−	48.3 ± 14.4	70.9 ± 10.7	90.0 ± 12.3
p (Wilcoxon rank sum test)	0.048	0.012	0.039

^aSee also Figure 1.

Ribbon synapse number in Vglut3 mice before and after 100 dB exposure

Presynaptic ribbons and postsynaptic AMPA receptors are juxtaposed on either side of the IHC-ANF synaptic cleft. Figure 2 compares ribbon synapses from a mid-basal cochlear location in 10-week-old unexposed Vglut3^WT and unexposed Vglut3^KO IHCs. Vglut3^+/− (data not shown) and Vglut3^WT cochleae (Fig. 2A) had ∼15 ribbon synapses per IHC, populating the basolateral membrane extending ∼10 μm below the IHC nucleus. In contrast, in the basal half of the cochlear spiral of Vglut3^KO mice, we noted truncation of IHC basolateral membranes (Fig. 2B), which we also observed in plastic mid-modiolar sections (data not shown). Unexposed Vglut3^KO had fewer ribbon synapses per IHC, and they appeared to be larger in size than unexposed Vglut3^+/− or Vglut3^WT, as quantified below.

Figure 2.

Synapses are fewer in number in unexposed Vglut3^KO mice. A, B, Confocal images of 10-week-old unexposed Vglut3^WT (A) and Vglut3^KO (B) IHCs labeled with anti-Myosin7a (blue), anti-CtBP2 (red), and anti-GluA3 (green). Imaged with LSM 880 Airyscan. Scale bars, 5 μm. Boxes are 3 × 3 μm.

Ribbons without paired AMPA receptors, or vice versa, were rare in unexposed mice (Fig. 3A). Based primarily on studies in CBA/CaJ^WT mice, 2 h of 100 dB SPL 8–16 kHz noise exposure is expected to cause swelling of the ANF postsynaptic terminals and disintegration of the corresponding presynaptic and postsynaptic elements on IHCs in the (high-frequency) basal half of the cochlea (Kujawa and Liberman, 2009). One day after the 100 dB SPL noise exposure in Vglut3^WT and Vglut3^+/− cochleae, there were many unpaired synaptic puncta, and few paired synapses that persisted (Fig. 3B). By 2 weeks after exposure, many synaptic puncta had apparently repaired (Fig. 3C). Unpaired synaptic puncta were rarely observed 2 weeks after exposure, but they remained in greater proportion than in unexposed animals. While Figure 2 illustrates the phenomenon in Vglut3^WT mice, similar observations were made in Vglut3^+/−.

Figure 3.

Synapses are lost and repaired following 100 dB noise exposure in Vglut3^WT. Confocal images of 10-week-old Vglut3^WT. Left, IHCs are labeled with anti-Myosin7a, presynaptic ribbons with anti-CtBP2, and postsynaptic AMPA receptors with anti-GluA3. Right, White rectangular regions are enlarged. Midcochlear location, ∼22.5 kHz. A, Unexposed cochleae. Horizontal arrow indicates a lone ribbon. Vertical arrowheads point to ribbonless synapses. B, One day after exposure. Diagonal arrows indicate synapses that remain paired. C, Two weeks after exposure. Horizontal arrows indicate lone ribbons. Vertical arrowheads point to ribbonless synapses. Similar observations were made in Vglut3^+/− (data not shown). Imaged with LSM 700. Scale bars, 5 μm.

We quantified hair-cell synapse loss and repair by calculating the mean number of paired synapses per IHC in three cochlear regions. Unexposed Vglut3^WT and Vglut3^+/− mice had more synapses per IHC than Vglut3^KO in all three regions (Fig. 4A–C). The ABR threshold shift for middle and high frequencies at 1 d after exposure (Fig. 1B) was accompanied by ≥50% loss of synapse number in Vglut3^WT and Vglut3^+/− cochleae in the midcochlear and basal regions corresponding, respectively, to characteristic frequencies centered at ∼22.5 and 44.5 kHz (Fig. 4B,C). Relative to unexposed controls, on average, there were 9.5 fewer synapses per IHC in the Vglut3^WT midcochlea and 9.6 fewer in the Vglut3^+/− midcochlea at 1 d after exposure (Table 2). In the basal cochlea, noise-induced synapse loss per IHC was more extreme: 11.1 in Vglut3^WT and 12.0 in Vglut3^+/−, on average.

Figure 4.

No loss of synapse number in Vglut3^KO, and greater recovery of synapse number in Vglut3^WT than Vglut3^+/− after 100 dB exposure. A–C, Paired synapses (mean ± SD) per IHC in unexposed mice (open circles) and as a function of time after exposure (filled circles) for each genotype at the three cochlear locations: A, apical; B, mid-cochlea; C, basal. For 9 groups (3 genotypes × 3 assessment times): total 34 mice, N = 3–5 per group; 10–12 complete IHCs per image; 205 images (73 apical, 71 mid-cochlear, 61 basal); 5–11 cochleae per group.

Table 2.

Paired synapses (mean ± SD) per IHC in unexposed, 1 d, or 2 weeks after 100 dB SPL exposure in Vglut3^WT, Vglut3^+/−, and Vglut3^KO in apical, middle, and basal cochlea^a

		Unexposed	1 d	2 weeks
Tonotopic region (kHz)
Apical (9)	Vglut3WT	15.2 ± 0.4	15.4 ± 0.7	14.8 ± 0.6
	Vglut3+/−	13.2 ± 0.7	14.3 ± 0.4	13.3 ± 0.4
	Vglut3KO	9.5 ± 0.2	9.4 ± 0.3	9.7 ± 0.8
Middle (22.5)	Vglut3WT	18.3 ± 0.2	8.8 ± 1.2	14.3 ± 1.3
	Vglut3+/−	18.0 ± 0.6	8.4 ± 0.8	11.5 ± 1.1
	Vglut3KO	9.1 ± 0.2	9.0 ± 0.2	8.2 ± 0.8
Basal (44.5)	Vglut3WT	15.4 ± 1.2	4.3 ± 0.3	10.6 ± 3.0
	Vglut3+/−	16.4 ± 0.4	4.4 ± 0.5	6.9 ± 0.3
	Vglut3KO	7.8 ± 0.2	5.9 ± 0.9	6.2 ± 0.7

^aSee also Figure 4.

At 2 weeks after exposure, we observed incomplete recovery of synapse numbers in Vglut3^WT and Vglut3^+/− IHCs at the 22.5 and 44.5 kHz locations (Fig. 4B,C). Although synapse loss was similar at 1 d after exposure, recovery of synapse number was greater in Vglut3^WT than Vglut3^+/− by 2 weeks after exposure. On average, Vglut3^WT recovered 5.5 synapses per IHC at the 22.5 kHz location and 6.3 at the 44.5 kHz location, whereas Vglut3^+/− cochleae recovered only 3.1 and 2.5 synapses per IHC, respectively (Table 2). Results of the mixed-model ANOVA revealed a significant difference between Vglut3^WT and Vglut3^+/− in number of synapses per IHC at 2 weeks after exposure (p = 0.0089). Synapses per IHC in the apical region (∼9 kHz) of Vglut3^WT, Vglut3^+/−, or Vglut3^KO cochleae appeared unchanged in response to the 100 dB SPL exposure (Fig. 4A). In mid-frequency and high-frequency regions in Vglut3^KO, unlike in Vglut3^WT or Vglut3^+/−, noise exposure did not result in loss of synapse number (Fig. 4B,C; Table 2).

We also counted the numbers of unpaired synaptic puncta before and after the 100 dB SPL noise exposure (Table 3). In unexposed animals, the mean number of unpaired puncta per IHC was <1. In Vglut3^KO cochleae and in the apical (low-frequency) regions of Vglut3^WT and Vglut3^+/− cochleae, there remained <1 unpaired punctum per IHC at 1 d and 2 weeks after noise exposure. In contrast, in the mid-cochlear and basal regions of Vglut3^WT and Vglut3^+/−, the numbers of unpaired GluA3 puncta and unpaired ribbons changed after noise exposure, showing similar trends. Loss of paired puncta at 1 d after exposure was accompanied by an increase in unpaired puncta (Tables 2, 3). By 2 weeks after exposure, the numbers of unpaired puncta were reduced relative to 1 d after exposure but still greater than in unexposed ears. Unpaired GluA3 puncta were in greater abundance than unpaired ribbons in 24 of 27 cases (3 genotypes × 3 cochlear locations × 3 assessment times). After noise exposure, on average, the sum of paired synapses and unpaired ribbon or AMPA receptor puncta was smaller than the number of paired puncta in unexposed mice, suggesting that some unpaired synaptic puncta had been eliminated from the ROI.

Table 3.

Unpaired ribbons and GluA3 puncta (mean ± SD) per IHC in unexposed, 1 d, or 2 weeks after 100 dB SPL exposure in Vglut3^WT, Vglut3^+/−, and Vglut3^KO in apical, middle, and basal cochlear regions^a

		Ribbons			GluA3
		Unexposed	1 d	2 weeks	Unexposed	1 d	2 weeks
Tonotopic region (kHz)
Apical (9)	Vglut3WT	0.2 ± 0.1	0.2 ± 0.0	0.3 ± 0.0	0.4 ± 0.1	0.3 ± 0.1	0.2 ± 0.1
	Vglut3+/−	0.2 ± 0.1	0.3 ± 0.1	0.5 ± 0.1	0.4 ± 0.1	0.5 ± 0.1	0.5 ± 0.2
	Vglut3KO	0.4 ± 0.1	0.4 ± 0.1	0.4 ± 0.1	0.4 ± 0.0	0.7 ± 0.1	0.4 ± 0.1
Middle (22.5)	Vglut3WT	0.3 ± 0.1	1.8 ± 0.4	0.8 ± 0.1	0.5 ± 0.1	3.2 ± 0.6	1.9 ± 0.6
	Vglut3+/−	0.2 ± 0.1	2.2 ± 0.4	0.8 ± 0.1	0.2 ± 0.1	4.7 ± 0.9	1.6 ± 0.3
	Vglut3KO	0.5 ± 0.0	0.6 ± 0.1	0.6 ± 0.2	0.4 ± 0.1	0.8 ± 0.2	0.9 ± 0.2
Basal (44.5)	Vglut3WT	0.3 ± 0.1	2.5 ± 0.3	0.4 ± 0.0	0.7 ± 0.1	3.1 ± 0.3	1.6 ± 0.2
	Vglut3+/−	0.5 ± 0.3	2.8 ± 0.3	1.4 ± 0.3	0.5 ± 0.3	4.4 ± 0.6	2.8 ± 0.2
	Vglut3KO	0.3 ± 0.1	0.6 ± 0.2	0.7 ± 0.2	0.4 ± 0.1	0.6 ± 0.2	0.6 ± 0.2

^aSee also Figure 4.

Presynaptic ribbon and postsynaptic AMPA receptor volumes before and after 100 dB exposure

Exposure to excitotoxic noise resulted in synapse loss in Vglut3^WT and Vglut3^+/−, but not in Vglut3^KO (Fig. 4; Table 2). Thus, synaptic glutamate release seems to be required for outright synapse loss, but presynaptic Ca²⁺ influx or nonglutamatergic transmission may have measurable effects in the absence of glutamate in Vglut3^KO. In addition, noise exposure may have effects on the properties of surviving or regenerated synapses in Vglut3^WT and Vglut3^+/−. Therefore, we assessed the effects of noise and Vglut3 copy number on synapse morphology by comparing presynaptic ribbon volume and postsynaptic GluA3 volume among mid-cochlear synapses in Vglut3^WT, Vglut3^+/−, and Vglut3^KO IHCs before and 2 weeks after the 100 dB noise exposure. The distributions of synapse volumes for the 12 groups (3 genotypes × 2 conditions × CtBP2 or GluA3) were all non-normal and positively skewed (Fig. 5; Table 4).

Figure 5.

Ribbon synapses are larger after noise and without Vglut3. CtBP2 volumes (left) or GluA3 volumes (right) for paired synapses of midcochlear IHCs (35 kHz location) in Vglut3^WT, Vglut3^+/−, and Vglut3^KO before and 2 weeks after 100 dB exposure. Tukey's box-and-whiskers: box shows the second quartile (or interquartile range) with median (horizontal line), mean (X), and number of synapses per group (e.g., n = 2407 for unexposed WT); whiskers extend down to minimum and upward for 1.5 × the interquartile range. Outliers are shown as individual points; number of outliers off scale are quantified in parentheses. Bottom, Horizontal bars represent significant differences in the distribution of data between groups (Kolmogorov–Smirnov test, p < 0.05; for exact p values and effect size estimates, see Table 4). Six groups: N = 3–6 cochleae per group; 8–13 images per group; 1305–2407 synapses per group.

Table 4.

Ribbon or GluA3 volume per synapse (μm³, all paired synapses) for unexposed or 2 weeks after 100 dB SPL exposure in midcochlea of Vglut3^WT, Vglut3^+/−, or Vglut3^KOa

Genotype	Unexposed		2 weeks		Kolmogorov–Smirnov test exposure
	Mean ± SD	Median	Mean ± SD	Median
Ribbons
Vglut3WT	0.098 ± 0.079	0.076	0.123 ± 0.08	0.101	p = 9e-33 (0.3)*
Vglut3+/−	0.106 ± 0.095	0.078	0.117 ± 0.092	0.096	p = 2e-7 (0.2)*
Vglut3KO	0.122 ± 0.088	0.099	0.155 ± 0.116	0.127	p = 2e-8 (0.3)*
Kolmogorov–Smirnov test WT ±	p = 0.2		p = 5e-12 (0.1)
Kolmogorov–Smirnov test WT KO	p = 2e-17 (0.3)*		p = 3e-8 (0.3)*
GluA3
Vglut3WT	0.377 ± 0.182	0.355	0.385 ± 0.188	0.354	p = 0.3
Vglut3+/−	0.413 ± 0.209	0.374	0.448 ± 0.225	0.414	p = 2e-4 (0.2)*
Vglut3KO	0.542 ± 0.259	0.52	0.63 ± 0.342	0.575	p = 4e-4 (0.2)*
Kolmogorov–Smirnov test WT ±	p = 1e-4 (0.1)		p = 3e-10 (0.3)*
Kolmogorov–Smirnov test WT KO	p = 6e-73 (0.8)*		p = 5e-50 (0.9)*

^aValues are number of animals-images-synapses unexposed: WT, 4-12-2407; HET, 3-8-1480; KO, 6-13-1305. Number of animals-images-synapses 2 weeks after exposure: WT, 3-8-1365; HET, 4-10-1191; KO, 3-6-515. p value (effect size; Kolmogorov–Smirnov test).

*Comparisons with p values < 0.05 and effect size ≥ 0.2.

For presynaptic ribbons, labeled with anti-CtBP2, the distributions of volumes in Vglut3^KO were larger, and significantly different from those of Vglut3^+/− or Vglut3^WT before and 2 weeks after noise exposure (Fig. 5, left). In contrast, Vglut3^WT and Vglut3^+/− ribbon volume distributions were not significantly different from each other before or after noise. Moreover, the 100 dB noise exposure resulted in significant changes to the ribbon volume distributions relative to unexposed (Table 4). Ribbons were larger 2 weeks after exposure in Vglut3^WT, Vglut3^+/−, and Vglut3^KO. Therefore, noise exposure or loss of both Vglut3 genes was associated with larger ribbon volumes, suggesting a greater number of CtBP2/Ribeye molecules per synapse.

For postsynaptic AMPA receptors labeled with anti-GluA3, as in the case for ribbons, the distributions of volumes in Vglut3^KO were larger and significantly different from those of Vglut3^+/− or Vglut3^WT before and 2 weeks after noise exposure (Table 4). Also, the mean and median GluA3 volumes of the Vglut3^+/− synapses were intermediate, between the larger Vglut3^KO and the smaller Vglut3^WT volumes, before and 2 weeks after noise exposure (Fig. 5, right). Thus, fewer copies of the Vglut3 gene were associated with larger GluA3 volumes, suggesting a greater number of AMPA receptors per synapse. Moreover, 2 weeks after 100 dB noise exposure, the distributions of GluA3 volumes were larger and significantly different from unexposed GluA3 volumes in Vglut3^+/− and Vglut3^KO, but not in Vglut3^WT.

In summary, except for GluA3 puncta in Vglut3^WT, the general effect of 100 dB noise exposure on synapse volume after 2 weeks relative to unexposed controls was ribbon and GluA3 enlargement (Fig. 5; Table 4). Greater synapse volume in unexposed Vglut3^KO relative to unexposed Vglut3^WT or Vglut3^+/− suggests glutamate-dependent maintenance of synapse size. In addition, greater volumes of presynaptic ribbons and postsynaptic GluA3 2 weeks after exposure in Vglut3^KO suggests noise-induced glutamate-independent mechanisms of changes in synapse morphology.

Presynaptic Ca_V1.3 before and after noise exposure

Although Vglut3^KO mice are profoundly deaf, Vglut3^KO IHCs support fast Ca²⁺-dependent exocytosis as measured by depolarization-evoked membrane capacitance increases (Ruel et al., 2008). In Vglut3^WT mice, exocytosis is thought to occur predominantly at the ribbon synapses, where voltage-gated Ca²⁺ channels (Ca_V1.3) cluster in the presynaptic membrane (Martinez-Dunst et al., 1997; Rutherford, 2015; Ohn et al., 2016). In Vglut3^KO IHCs, whole-cell voltage-gated Ca²⁺ currents were significantly larger than those in Vglut3^WT IHCs (Seal et al., 2008; Ruel et al., 2008). However, the subcellular localization of Ca_V1.3 in Vglut3^KO IHCs is unknown. Therefore, to ask whether voltage-gated Ca²⁺ channels cluster with ribbons in Vglut3^KO IHCs, we immunolabeled Ca_V1.3 with CtBP2 and GluA3. As in unexposed Vglut3^WT and Vglut3^+/− IHCs, Ca_V1.3 channels were located between presynaptic ribbons and postsynaptic AMPA receptors in unexposed Vglut3^KO IHCs (Fig. 6A,B). Next, we checked for noise-induced synapse loss during the noise exposure.

Figure 6.

Vglut3^KO IHCs cluster voltage-gated Ca²⁺ channels with ribbons; synapses are not lost after noise. A, Unexposed Vglut3^+/−. Presynaptic voltage-gated Ca²⁺ channels are labeled with anti-Ca_V1.3; presynaptic ribbons with anti-CtBP2; and postsynaptic AMPA receptors with anti-GluA3. Scale bar, 5 μm. Bottom, Small boxes are 1 μm × 1 μm. B, Unexposed Vglut3^KO. C, Vglut3^+/− immediately after 2 h of 100 dB noise exposure. D, Vglut3^KO immediately after 2 h of 100 dB noise exposure. Each image is from the 35 kHz region. E, Vglut3^+/− unexposed (left) or 0 h after exposure (right). For each synapse, the intensity of Ca_V1.3 (green markers) and GluA3 (blue markers) is plotted as a function of CtBP2 intensity (a.u., arbitrary units of pixel intensity). Red horizontal arrow indicates ribbon synapses having GluA3 immunofluorescence near 0 a.u. (blue markers along the abscissa) (e.g., see C₁). Green markers along the ordinate represent synapses with near 0 CtBP2 and, typically, near 0 GluA3 such that only Ca_V1.3 remains (red vertical arrow) (e.g., see C₂). F, Vglut3^KO unexposed (left) or 0 h after exposure (right).

As reported above in Figure 4, in Vglut3^KO, noise exposure did not result in loss of IHC ribbon synapse number at 1 d or 2 weeks after exposure (Table 2). Previous experiments in CBA/CaJ^WT mice showed that loss of synaptic elements was nearly complete immediately after a 2 h exposure (Liberman et al., 2015). Thus, to check for immediate synapse loss in the IHC region that may have been repaired or cleared of debris by 1 d after exposure, we looked at 0 h after 2 h of 100 dB SPL exposure. Some Vglut3^WT and Vglut3^+/− synapses were already disintegrated after 2 h of noise exposure (Fig. 6C,E). While Figure 6 shows representative images from Vglut3^+/−, we made similar observations in Vglut3^WT. We observed presynaptic ribbons with Ca_V1.3 that lacked postsynaptic GluA3 (Fig. 6C₁), and we observed lone Ca_V1.3 clusters lacking both the ribbon and GluA3 (Fig. 6C₂). These observations are consistent with previous experiments in CBA/CaJ^WT mice that showed either loss of synaptic elements or downregulation of AMPA receptors immediately after noise trauma (Liberman et al., 2015). In contrast, Vglut3^KO synapses appeared intact (Fig. 6D,F). Postsynaptic GluA3 remained juxtaposed to the presynaptic elements. The presynaptic active zone, defined by a ribbon and Ca_V1.3, appeared to remain intact in Vglut3^KO IHCs, suggesting that presynaptic Ca²⁺ influx alone was insufficient to disassemble the active zone.

Recovery of cochlear function in Vglut3 mice after 94 dB exposure

Physiological and anatomical assessments following the 100 dB SPL exposure demonstrated better recovery of ABR threshold and synapse number in Vglut3^WT mice relative to Vglut3^+/− mice (Figs. 1C, 4B,C; Tables 1, 2). We wondered whether the improved recovery of hearing function in Vglut3^WT relative to Vglut3^+/− would generalize to other levels of exposure. Exposures >100 dB for 2 h are expected to cause irreversible damage to cochlear cells, including the outer hair cells, which might occlude observation of functional recovery due to synapse regeneration (Wang et al., 2002). Therefore, to further investigate this phenomenon, we reduced the exposure level to 94 dB SPL and assessed hearing function by ABR and DPOAE before exposure, and at 1 d, 2 weeks, and 4 weeks after exposure.

For the four assessment times, there were no significant differences in ABR thresholds between Vglut3^WT and Vglut3^+/− mice at any test frequency (Fig. 7A; mixed-model ANOVA, p = 0.98). The persistent elevation of ABR thresholds at 32, 45.2, and 64 kHz was most likely due to aging on the C57BL/6J background, as suggested by our assessment of aging in unexposed control mice in Figure 8 (bottom). Thus, 2 h of 94 dB SPL exposure caused ABR threshold shifts at 1 d that appeared to recover to control levels by 2 and 4 weeks after exposure in Vglut3^WT and Vglut3^+/− mice (Fig. 7A).

Figure 7.

Better recovery of ABR wave I in Vglut3^WT than Vglut3^+/− after 94 dB exposure. A, Mean ABR thresholds in unexposed mice and following 94 dB exposure at 1 d, 2 weeks, and 4 weeks. Eight groups (genotype × time); N = 16–19 animals per group. Legend applies also to B. B, Mean ABR wave I amplitudes versus tone burst level for the same mice in A. C, ABR average waveforms (5 mice per genotype) in response to 85 dB SPL tone pips at 22.6 kHz at 4 weeks after exposure for Vglut3^WT and Vglut3^+/−, and in unexposed Vglut3^KO mice that showed a measurable response. D, ABR wave I amplitude (mean ± SD) over time after 94 dB exposure in Vglut3^WT (N = 10) and Vglut3^+/− (N = 8) for the 8 kHz and 22.6 kHz probe tones.

Figure 8.

Age-related threshold shifts are similar in C57BL/6J Vglut3^WT and Vglut3^+/− mice. ABR threshold shifts from 2 to 9 months demonstrate the well-known progressive hearing loss of the background strain, beginning at the highest frequencies, in Vglut3^WT and Vglut3^+/− mice. Ten groups (age × genotype). Data are mean ± SD; N = 2–10 animals per group.

The amplitude of ABR wave I reflects the number of ANFs simultaneously activated at sound onset, which depends on the number of intact and active IHC-ANF synapses. Figure 7B–D shows that the amplitude of ABR wave I tended to recover faster in Vglut3^WT than Vglut3^+/− following the 94 dB SPL exposure. For the 22.6 kHz tone, in response to all stimulus levels (20–85 dB SPL), ABR wave I amplitudes were reduced at 1 d after exposure relative to amplitudes in unexposed mice for Vglut3^WT and Vglut3^+/− (Fig. 7B). In Vglut3^WT mice, by 2 weeks after exposure, wave I amplitudes were significantly different from unexposed mice for only two of the input levels (40 and 85 dB SPL); by 4 weeks after exposure, only the response to the 20 dB SPL tone level was significantly different from the responses of unexposed ears (Dunnett's post hoc test, p < 0.05). In contrast, Vglut3^+/− mice exhibited significantly reduced wave I response amplitudes relative to unexposed controls for 13 of the 15 probe levels at 2 weeks after exposure, and still for 7 of the 15 levels by 4 weeks after exposure (Dunnett's post hoc test, p < 0.05). Relative to Vglut3^WT at 4 weeks after exposure, Vglut3^+/− mice had significantly reduced ABR wave I amplitudes at 9 of the 15 input levels (Wilcoxon rank sum test: 30 dB, p = 0.0074; 35 dB, p = 0.0019; 40 dB, p = 0.0391; 45 dB, p = 0.0083; 50 dB, p = 0.025; 60 dB, p = 0.0295; 75 dB, p = 0.0479; 80 dB, p = 0.0176; 85 dB, p = 0.0348). For example, in response to 22.6 kHz tones at 85 dB SPL, the ABR wave I amplitude in Vglut3^WT mice was on average ∼27% larger compared with Vglut3^+/− mice at 4 weeks after (Fig. 7C,D). By comparison, for the 8 kHz tone, which probes the less affected apical cochlea, the ABR wave I amplitudes of Vglut3^WT and Vglut3^+/− did not diverge over time after exposure.

Age-related hearing loss in Vglut3^WT and Vglut3^+/− mice

Above, we noise-exposed mice on C57BL/6J background at 9–11 weeks of age. This background strain begins to show progressive high-frequency hearing loss in the ensuing months. To interpret the recovery of hearing function after noise exposure in Vglut3^WT and Vglut3^+/− mice (Figs. 1, 7), we asked whether age-related hearing loss differed between the two genotypes in unexposed mice. Between 2 and 3 months of age, ABR threshold shifts were observed only for the highest frequencies (Fig. 8). By 4–6 months, there were threshold elevations for middle and high frequencies; by 9 months, for all 8 test frequencies. The progressive high-frequency threshold shifts presumably reflect the effects of the Cdh23^753A allele known to be present on the C57BL/6J background (Noben-Trauth et al., 2003). Unexposed Vglut3^+/− mice demonstrated similar age-related hearing loss as seen in Vglut3^WT mice. The assessment in Figure 7A shows a persistent threshold shift at 4 weeks after 94 dB exposure for the highest frequencies, similar to the shifts seen between 2–3 or 3–4 months in Figure 8 due to aging.

Outer hair cell function in Vglut3 mice after 94 or 100 dB exposure

The hallmark of synaptopathic noise exposure is destruction of some synapses while the inner and outer hair cells themselves survive. However, sublethal effects of noise on outer hair cells can affect cochlear function. Low-level cochlear vibrations are amplified by active processes in the outer hair cells that exert forces to counteract viscous dampening in the fluid-filled inner ear (Fisher et al., 2012). This mechanical positive feedback increases hearing sensitivity and enhances frequency selectivity by affecting the excitation of IHCs, which release glutamate onto ANFs at the ribbon synapses. Hearing thresholds thus depend on cochlear amplification by the outer hair cells, which can be damaged by noise exposure.

To control for outer hair cell effects that might differ by genotype, we measured DPOAE levels as a reflection of cochlear function and outer hair cell integrity over time in unexposed mice and after 94 or 100 dB SPL exposure. Figure 9A, B shows group mean emission levels as a function of frequency (f₂) before, 1 d, and 28 d after exposure to 94 or 100 dB SPL. Vglut3^WT, Vglut3^+/−, and Vglut3^KO mice had similar DPOAE response profiles in unexposed ears, although some cohorts of Vglut3^KO mice tended to have smaller emissions levels at the highest frequencies only (Fig. 9A,B, left). Figure 9C–E shows the group means of the mean emission levels across all frequencies for the 3 groups over time (Fig. 9C: unexposed; Fig. 9D: 94 dB SPL; Fig. 9E: 100 dB SPL). Comparing genotypes at each time, unexposed mice had similar DPOAE levels (Fig. 9C; two-tailed t test with Bonferroni correction, not significant for all four times, p = 0.02–0.68). Within each genotype in the unexposed group, DPOAE levels did not change significantly over the 1 month period after the initial test at 9–11 weeks of age (Fig. 9C; p = 0.06–0.53).

One day after exposure to 8-16 kHz noise at 94 or 100 dB SPL, DPOAE levels in exposed ears were depressed relative to unexposed Vglut3^WT, Vglut3^+/−, and Vglut3^KO mice (Fig. 9A,B, middle, D,E; p = 1e⁻⁶ to 1e⁻¹⁷). For the 100 dB exposure, at 1 d, the Vglut3^WT emissions levels were significantly larger than Vglut3^+/− or Vglut3^KO (Fig. 9E; wt > het, p = 1e⁻⁹; WT > KO, p < 1e⁻⁶). Two and 4 weeks after the 94 dB exposure, DPOAE levels remained significantly depressed relative to unexposed levels in Vglut3^KO (p < 1e⁻⁴), whereas Vglut3^WT and Vglut3^+/− were not significantly different from unexposed levels at 2 and 4 weeks, respectively (Fig. 9D; p = 0.13 and p = 0.0065). Two and 4 weeks after the 100 dB exposure, DPOAE levels remained significantly smaller than unexposed levels for all genotypes (Fig. 9E; p = 0.003–4.2e⁻⁸). By 4 weeks after exposure (Fig. 9A,B, right), there were no significant differences between genotypes for either exposure (Fig. 9D,E; p = 0.03–0.67).

Discussion

Glutamate is required for noise-induced cochlear synaptopathy, but the mechanism is unknown

To directly test the idea that glutamate is required for noise-induced cochlear excitotoxicity, we exposed Vglut3^KO mice to synaptopathic noise. We found that genetic removal of Vglut3 prevented noise-induced loss of synapse number (Fig. 4). The Vgluts are specific for glutamate, unable to transport the similar amino acids and potential neurotransmitters aspartate, glutamine, or GABA (Schäfer et al., 2002). Therefore, as has been presumed based on previous pharmacological studies in vivo and in cell culture (Puel et al., 1994; Wang and Green, 2011), we now can conclude with little doubt that noise-induced cochlear excitotoxicity requires synaptic release of glutamate from IHCs.

The mechanism(s) of glutamate-dependent cochlear excitotoxicity remain unclear but likely involve postsynaptic Ca²⁺. Although they have yet to be directly demonstrated functionally in mammalian ANFs, the existence of Ca²⁺-permeable AMPA receptors at hair cell synapses appears to be evolutionarily conserved in fish, frogs, and mammals (Sebe et al., 2017). Postsynaptic Ca²⁺ influx through AMPA, kainate, or NMDA receptors, or through voltage-gated Ca²⁺ channels may activate calpain-mediated cell death (Chen et al., 2009). On the other hand, postsynaptic Ca²⁺ influx may protect synapses by triggering rapid removal of AMPA receptors via endocytosis during exposure to loud sounds (Chen et al., 2007). Furthermore, glutamatergic activity that is damaging in one context may be beneficial in another. For example, Ca²⁺ influx through NMDA receptors may be required for synapse regeneration after trauma (d'Aldin et al., 1997).

Glutamate regulates presynaptic and postsynaptic size

IHCs require Vglut3 for quantal transmission and normal ANF activity. As in hair cells of the lateral line organ (Obholzer et al., 2008), Vglut1 and Vglut2 do not compensate in the absence of Vglut3. Despite the lack of glutamatergic transmission, at 9–11 weeks of age, approximately half the normal number of synapses are maintained as AMPA receptors juxtaposed to presynaptic ribbons (Figs. 2, 4). Interestingly, these postsynaptic arrays of AMPA receptors in Vglut3^KO mice were significantly larger than Vglut3^WT or Vglut3^+/− (Fig. 5). AMPA receptor endocytosis triggered by application of AMPA or NMDA in cell culture depends on Ca²⁺ influx and is inhibited by blocking voltage-gated Ca²⁺ channels (Chen et al., 2009). We propose that lack of glutamate-induced Ca²⁺ influx reduced AMPA receptor endocytosis in Vglut3^KO, resulting in enlargement of the array of AMPA receptors labeled by anti-GluA3. Additionally, we observed larger presynaptic ribbons in Vglut3^KO mice (Fig. 5), reminiscent of observations in WT animals showing that larger ribbons belong to less active synapses (Merchan-Perez and Liberman, 1996; Sheets et al., 2012; Nicolson, 2015).

Ribbon synapses regenerate following an initial loss with acoustic trauma in C57BL/6J

After an initial loss of ∼50% of synapses with acoustic trauma, we found that approximately half of those synapses regenerated spontaneously in Vglut3^WT (Figs. 3, 4). This contrasts with the conventional view based on experiments in CBA/CaJ^WT mice showing that ribbon synapse loss is permanent. Regeneration of synapses after noise exposure, or lack thereof, may depend on the animal species, experimental conditions, or the actions of neurotrophic factors (Kujawa and Liberman, 2009; Shi et al., 2013; Wan et al., 2014; Chen et al., 2018). Although it does not imply causality, we found that recovery of IHC ribbon synapse number was correlated with recovery of ABR threshold after the 100 dB exposure (Figs. 1, 4). We found that the enhancement of recovery in Vglut3^WT was not readily explained by differences in DPOAE or age-related hearing loss (Figs. 8, 9).

Two copies of Vglut3 are better than one

Significantly fewer synapses regenerated in Vglut3^+/− than Vglut3^WT after 100 dB exposure, suggesting a role for Vglut3 in regeneration. Vglut3^+/− mice also had poorer recovery of ABR wave I amplitudes after the 94 dB exposure, suggesting a reduction in Vglut3 function detectable when the system is challenged. These phenotypes may result from reduction of glutamate secretion during the recovery period (d'Aldin et al., 1997), if Vglut3^WT IHCs release more glutamate than Vglut3^+/− IHCs after noise trauma. We found no other indirect evidence suggesting increased release of glutamate from Vglut3^WT IHCs: Upon noise exposure, Vglut3^+/− mice lose synapses, such as Vglut3^WT (Fig. 4). In unexposed Vglut3^+/− mice, just one copy of Vglut3 was sufficient to maintain normal auditory thresholds, synapse numbers, and wave I amplitudes (Figs. 1, 4, 7). Perhaps more demanding auditory stimuli requiring more Vglut3 activity would reveal fatigue in Vglut3^+/−. Another possibility is that the Vglut3^+/− haplo-insufficiency relates to reduced binding interactions of Vglut3 with other proteins that support regeneration.

Glutamate-independent effects of noise on the IHC-ANF synapse

Electron micrographs and immunohistofluorescence have been used to illustrate presynaptic loss and sublethal damage to the IHC that accompanies ANF terminal swelling and vacuolization (Robertson, 1983; Kujawa and Liberman, 2009). Like postsynaptic Ca²⁺ overload in the ANF, presynaptic Ca²⁺ overload may damage the IHC. In Vglut3^WT and Vglut3^+/− IHCs, noise exposure resulted in unpairing of some CtBP2 and Ca_V1.3 puncta (Fig. 6A,C,E). In Vglut3^KO IHCs, the proper localization of Ca_V1.3 and the lack of unpairing suggest that active-zone destruction follows glutamate release, and that noise-induced presynaptic Ca²⁺ influx is insufficient for synaptic disintegration (Fig. 6B,D,F).

Unexpectedly, 2 weeks after exposure, we found that Vglut3^KO ribbons and AMPA receptors were significantly different in volume than unexposed Vglut3^KO (Fig. 5), suggesting noise-induced changes that were independent of glutamate. In Vglut3^WT and Vglut3^+/− as well, ribbons appeared to be embiggened by 2 weeks after noise. GluA3 puncta in Vglut3^+/− appeared to become larger by 2 weeks after noise, but less so than Vglut3^KO, and in Vglut3^WT the size of GluA3 puncta was maintained after noise (Fig. 5). We propose a bidirectional control of AMPA receptor number and thus postsynaptic size in which glutamate acts to reduce, and other releasable factors act to increase, AMPA receptor number. In the absence of glutamate release, synaptic activity results in larger synapses.

Our results show that protons and other chemical constituents of synaptic vesicles, such as aspartate, adenosine, and proinflammatory cytokines, if present, are insufficient to cause noise-induced loss of IHC-ANF synapse number. However, this does not exclude the possible involvement of other released substances in glutamate-dependent or -independent mechanisms. Indeed, Vglut3 is identified mainly in neurons coexpressing neurotransmitter transporters for GABA, acetylcholine, or serotonin (El Mestikawy et al., 2011). However, corelease of neurotransmitter from hair cells is unknown. Additionally, synaptic vesicles are acidified, and acidosis may contribute to ANF excitotoxicity or plasticity through activation of the excitatory acid-sensing ion channel-2 in mice (Peng et al., 2004). Indeed, hair cell synapses of the frog amphibian papilla are known to acidify the synaptic cleft by releasing protons, as demonstrated by transient inhibition of the presynaptic Ca²⁺ current (Cho and von Gersdorff, 2014). If vesicular protons are a component of synaptic transmission, then a proton-mediated EPSC is expected; however, application of the AMPA receptor antagonist CNQX completely blocks the EPSC in mammalian ANFs (Glowatzki and Fuchs, 2002), suggesting that protons are not acting as neurotransmitters.

Type I ANF subtype identity in Vglut3^KO mice

Type I ANFs (also called Type I spiral ganglion neurons) differentiate into subtypes in an activity-dependent developmental progression. In the absence of glutamatergic activity, they develop into less-diverse intermediate subtype(s) based on single-cell RNA sequencing (Shrestha et al., 2018; Sun et al., 2018). It is possible that ANFs in Vglut3^KO fail to diversify into the more excitotoxically vulnerable phenotypes. If synapses belonging to the more vulnerable ANFs are already absent in Vglut3^KO, because they failed to differentiate or they already died off, then it remains a possibility they could be damaged by glutamate-independent processes in Vglut3^WT. Resolution of the requirement for glutamate in excitotoxicity of high- versus medium- and low-spontaneous-rate ANFs will require a conditional Vglut3^KO mouse that develops the normal diversified set of ANFs before induction of deafness, followed by noise exposure.

More work is needed to understand the contributions of presynaptic, postsynaptic, efferent, and spike-generator mechanisms to ANF activity under different conditions and in different species (Ohlemiller et al., 2016; Reijntjes and Pyott, 2016). There remain unanswered questions about the influence of IHC-driven activity on ANF identity. Interestingly, spontaneous activity persists in the developing Vglut3^KO cochlea, and the intrinsic excitability of ANFs is enhanced (Babola et al., 2018). There is not a simple phenotypic match between the ANFs remaining in adult Vglut3^KO (remaining in the absence of glutamate) and those in Vglut3^WT that develop normally into a more excitotoxically resistant subtype (i.e., those that remain after the noise-induced glutamate exposure). For example, synaptic ribbon enlargement in Vglut3^KO (Fig. 5) is contrary to the synaptic phenotype of the ANFs in Vglut3^WT that are thought to be more resistant, which have smaller ribbons. We note that the correlation between excitotoxic vulnerability, spontaneous spike rate, and threshold in cat and guinea pig has not been demonstrated in mouse (Merchan-Perez and Liberman, 1996; Furman et al., 2013). It is crucial for future experiments to directly demonstrate the relationships between genetically defined Type I ANF subtypes and physiologically defined ANF phenotypes (Petitpré et al., 2018).