Experiments were carried-out on three rodent species: mice, rats, and guinea pigs. Male serine racemase knockout (Ser^–/–) mice and their heterozygous (Ser^+/–) and wild-type (Ser^+/+) littermates in the isogenic C57BL/6J genetic background were bred at the AnimEX facility (Université Paris-Saclay) and were 2 months old at the time of the experiments. Male adult Wistar rats and C57BL/6J mice were purchased from Janvier Laboratories (Le Genest Saint Isle, France) and used when 2 months old. One- to two-month-old tricolor adult guinea pigs were obtained from Charles River (L’Arbresle, France). Animals were housed on a 12 h light/dark cycle and were provided food and water ad libitum. The animals were bred in pathogen-free animal care facilities accredited by the French Ministry of Agriculture and Forestry (INM: C-34-172-36; December 19, 2014; AnimEX: C92-019-01). Experiments were conducted in accordance with European and French directives on animal experimentation and with local ethical committee approval (agreements C75-05-18 and 01476.02, license #6711).

Adult pigmented guinea pigs, free of middle-ear infection, were anesthetized with urethane (1.4 g/kg i.p; Sanofi, France) and ventilated during the experiment. Supplemental doses (0.35 g/kg, i.p) were administered every 2 h, or more often if the animal withdrew its paw in response to deep pressure. A tracheotomy was performed, the electrocardiogram was monitored, and the core temperature was regulated at 38 ± 1°C by a regulated heating blanket. The pinna and external auditory meatus were resected to ensure good close-field acoustic stimulation.

The method used to monitor the effect of multiple perilymphatic perfusions on sound-evoked potentials has been extensively described previously (Puel, 1995). The cochlea was exposed ventrally, and two holes (0.2 mm in diameter) were gently drilled in the scala vestibuli and scala tympani of the basal turn of the cochlea. Intracochlear perfusions through the hole in the scala tympani flowed at 2.5 μl/min and exited the cochlea through the hole in the scala vestibuli. All perfusions lasted 10 min. The artificial perilymph solution had the following composition (in mM): 137 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, 1 NaHCO₃, 11 glucose, and 10 HEPES (pH 7.4; 300 mOsm/kg H₂O).

Mice were anesthetized by an intraperitoneal injection of a mixture of Rompun 2% (3 mg/kg) and Zoletyl 50 (40 mg/kg). ABRs were recorded from three subcutaneous needle electrodes placed on the vertex (active), on the pinna of the tested ear, and in the neck muscles (ground) of the mice. The acoustical stimuli delivered at a rate of 10/s were generated by a NI PXI-4461 signal generator (National Instruments) and consisted of 10 ms tone-bursts with a 1 ms rise- and fall time. Sound was generated by a JBL 075 loudspeaker (James B. Lansing Sound) positioned at 10 cm from the tested ear in a calibrated free-field condition. Voltage amplification (20,000) was achieved via a Grass P511 differential amplifier, and responses were averaged 1,000 times (Dell Dimensions). Level-amplitude functions of the ABRs were obtained at each frequency tested (2, 4, 6.3, 8, 10, 12.5, 16, 20, 24, and 32 kHz) by varying the level of the tone bursts from 0 to 100 dB SPL, in 5 dB incremental steps. The ABR thresholds were defined as the minimum sound level necessary to elicit a well-defined and reproducible wave II.

CAP was elicited in adult pigmented guinea pigs using tone bursts generated by an arbitrary function generator (type 9100R; LeCroy Corporation) with a 1 ms rise/fall time and a 9 ms total duration. The signals were passed through a programmable attenuator and presented to the ear in the free field via a JBL 075 earphone. Functional assessment was realized at 8 kHz sound frequency, with increasing levels of 5 dB from 0 to 100 dB sound pressure level (SPL) (reference 2.10^–5 Pa). The rate of presentation was 10 bursts per second. Cochlear responses were amplified (gain of 2,000) by a differential amplifier (Grass P511K; Astro-Med), and the signals filtered (bandpass, 100–3 kHz) and averaged (256 repeats) on a Pentium personal computer (Dell Computer Company). The sampling rate of the analog-to-digital converter was 50 kHz, with a dynamic range of 12 bits and 1,024 samples per record. CAPs were measured peak-to-peak between the first negative (N₁) and the following positive value (P₁). The thresholds were defined as the level needed to elicit a measurable response between 2 and 5 μV.

Distortion Product Otoacoustic Emissions (DPOAEs) were recorded in the external auditory canal using an ER-10C S/N 2528 probe (Etymotic research Inc., Elk Grove Village, IL, United States.). The two primary tones of frequency f1 and f2 were generated with a constant f2/f1 ratio of 1.2, and the distortion product 2f1-f2 processed by a Cubdis system HID 40133DP (Mimosa Acoustics Inc., Champaign, IL, United States). The probe was calibrated for the two stimulating tones before each recording. f1 and f2 were presented simultaneously, sweeping f2 from 20 to 2 kHz in quarter-octave steps. For each frequency, the distortion product 2f1-f2 and the neighboring noise amplitude levels were measured and expressed as a function of f2.

Postnatal 1–4-day old rat pups were decapitated and both inner ears removed from the base of the cranium for further dissociation. After the cochleae were extracted from the outer bony labyrinth, the outer ligament/stria vascularis and organ of Corti were dissected away from the central core of the cochlea that contained the spiral ganglion. The samples were then incubated with an enzymatic solution (collagenase type IV, 0.5 mg/ml and trypsin, 2.5 mg/ml) at 37°C for 1 h, shaken every 10 min. The enzymatic dissociation was stopped by a wash in a solution free of Ca²⁺ and Mg²⁺. Spiral ganglia were then dissociated mechanically using a Pasteur pipette whose tip had been flame-polished. The spiral ganglion neurons were centrifuged and then plated onto a poly-ornithine- and laminin-treated glass bottom dish (FluoroDish, World Precision Instrument, Inc., Sarasota, Florida, United States).

Two hours after plating, the spiral ganglion neurons were loaded with 3 μM fura-2 AM in a Locke buffer that had the following composition (in mM): 137NaCl, 3KCl, 2CaCl₂, 1MgCl₂, 10Glucose, 10Hepes, (pH = 7.35; osmolarity 300-303mOsm/kg H₂O) at 37°C for 30 min. Images were acquired with a Cool SNAP CCD camera coupled to an inverted microscope (Axiovert 200; Zeiss). The neurons were illuminated at 340 nm and 380 nm with a monochromator (Lambda DG-4 Diaphot; Sutter Instruments, Burlingame, CA) controlled by the software Metafluor (Universal Imaging Corporation, West Chester, PA). Fura-2 fluorescence images were continuously acquired, giving 4–5 data points per second. The emitted fluorescence images were obtained through a dichroic mirror and an emission filter of 510 nm. After subtraction of the background obtained in a region of interest (ROI) where no neurons were present, the changes in intracellular calcium levels were monitored in neurons defined as ROI. The emission ratio (340/380 nm) was converted into Ca²⁺ concentrations using the equation developed by Grynkiewicz et al. (1985).

SR^–/– mice (n = 6) and their corresponding wild-type (n = 6) counterparts aged at 2 months were placed under anesthesia [Rompun 2% (3 mg/kg) and Zoletil 50 (40 mg/kg)] into a small cage. The cage was suspended directly below the horn of the sound-delivery loudspeaker in a small, reverberant chamber. They were then exposed to noise in the 4–8 kHz band at a level of 100 dB SPL for 2 h. The noise was generated by a PCI 4461 generation card (National instruments) controlled by LabVIEW software. The sound level was calibrated before each exposure session using a 1/4-inch microphone (# 46BE, GRAS Sound and Vibration) controlled by a PCI 4461 card and LabVIEW software, ensuring that there was no more than a 1 dB difference between the center and the edge of the cage.

1-month old SR^+/+ mice (n = 8) were deeply anesthetized by i.p. injection of sodium pentobarbital (60 mg/kg body weight, Sanofi, Libourne, France), their cochleas were rapidly removed and placed individually in sterile eppendorf tubes in the presence of methanol. Extracted cochlea were then stored at −80°C until analysis. Cochlea were homogenized using a Precellys Evolution bead homogenizer. The supernatants were collected and dried using a Speedvac, and the pellets rinsed with 1,000 μL of purified water. The solution was combined with the original dried supernatant and evaporated to dryness in a Speedvac. All samples were reconstituted in purified water. Protein quantification was performed using the Micro BCA Protein Assay Kit (Thermo Fisher Scientific). The amino acids were pre-column derivatized with Nα-(2,4-Dinitro-5-fluorophenyl)-L-valinamide and analyzed with liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) to quantify D-serine, L-serine, glycine, and L-glutamate levels. This system used an EVOQ Elite Triple Quadrupole Mass Spectrometer coupled with an Elute UHPLC module (Bruker Daltonics) using a reversed-phase Kinetex phenyl-hexyl HPLC column [2.6 μm particle size, 100 Å pore size, 100 mm (length) × 2.1 mm ID (Phenomenex, Torrance, CA, United States)] with mobile phase A: 25 mM ammonium formate, mobile phase B: methanol, and flow rate 300 μL/min. Each analyte retention time and fragmentation parameter were established from the corresponding standard. The resulting chromatograms were analyzed by Data Reviewer 8 (Bruker Daltonics).

Brain and cochleae were homogenized in sample buffer [0.1 M NaCl, 20 Mm Tris, 5 Mm EDTA, 1 Mm PMSF, 1% SDS, with Complete protease inhibitor cocktail (Roche)]. Protein extracts (15 μg/lane) were separated on 10% (vol/vol) SDS-polyacrylamide gel and electro-blotted onto nitrocellulose membranes (0.45 μm, Thermo Fisher Scientific). Blots were incubated overnight at 4°C with goat polyclonal antibody against a peptide mapping the C-terminus of mouse SR (1/2,000, Santa Cruz Biotechnology, United States), or a rabbit polyclonal antibody against porcine kidney DAAO (1/1,000, Nordic Immunological Laboratories, The Netherlands). β-actin (1/10,000, Sigma-Aldrich) served as a loading control. The secondary antibodies used were horseradish peroxidase-conjugated donkey anti-goat IgG antibodies (1/3,000, Jackson ImmunoResearch Laboratories, United States) or donkey anti-rabbit IgG antibodies (1/3,000, Jackson ImmunoResearch Laboratories, United States). Immunodetection was accomplished by enhanced chemiluminescence (ECL) using the WesternBright Sirius femtogram-detection reagent (advansta). Molecular sizes were estimated by separating pre-stained molecular weight markers (10–260 kDa) in parallel (Fisher Scientific). Digital images of the chemiluminescent blots were captured using the chemidoc system (Biorad) with the Labimage acquisition software. To ascertain the expression of SR and DAAO in the cochlea, as a positive control, we used brain tissues known to express high levels of SR and DAAO (Verrall et al., 2007). Demonstration of the specificity of the polyclonal anti-SR antibody was achieved by using protein brain extracts from SR^–/– and WT (SR^+/+) control mice (Supplementary Figure 2). Western-blot analysis required 12 cochleae per species. All experiments were performed in triplicate. All results were normalized by β-actin expression.

Adult rat cochleae were prepared according to a standard protocol for fixation and embedding (Dememes et al., 2006). Cochlear slices of 60–80 μm were obtained using a vibratome and collected in ice-cold 0.1 M phosphate buffered saline (PBS). Adult mouse cochlear cryostat sections were prepared (Ladrech et al., 2004), and immunofluorescence procedures carried out as previously described (Dememes et al., 2006). After pre-incubation in the blocking buffer containing 10% normal donkey serum and 0.3% Triton X-100, sections were incubated at 4°C overnight with the primary antibodies including a rabbit polyclonal antibody raised against D-serine cross-linked to BSA with glutaraldehyde (GemacBio, Cenon, France) and diluted in PBS at 1/1,000 for D-serine, goat polyclonal antibody raised against the C-terminal part of the mouse SR (Santa Cruz Biotechnology, United States) diluted at 1/200, or a rabbit polyclonal antibody against porcine kidney DAAO (Nordic Immunological Laboratories, The Netherlands) diluted at 1/500. Anti-Vglut3 (Synapic Systems, #135204, RRID:AB_2619825) diluted at 1/500 and anti-neurofilament (NF 200, Sigma-Aldrich #N0142 RRID:AB-477257) diluted at 1/600 were used to identify the inner hair cells and spiral ganglion neurons, respectively. Rhodamine-conjugated phalloidin (1:1,000, Sigma) was used to label actin. After three PBS rinses, the cochleae were then incubated for 2 h at room temperature in PBS containing the secondary antibodies (Jackson ImmunoResearch Laboratories, PA, United States) diluted at 1/100 in PBS, as follows: Fluorescein isothiocyanate (FITC)-conjugated donkey anti-goat IgGs or FITC-conjugated donkey anti-rabbit IgGs and rhodamine-conjugated phalloidin (1/100 dilution; R-415, Molecular Probes) to double label for actin. DAPI (0.002% wt:vol, Sigma, Missouri, United States) was used to stain DNA. Fluorescence was acquired sequentially using a confocal microscope (LSM 5 Live Duo, Zeiss) with a 40x oil-objective, except for Figures 1E,H that were observed with a 10x-objective. Z stacks of optical sections were then projected onto a single plane using ImageJ and imported into Adobe Photoshop for adjustment of contrast and brightness. These optical sections were 0.25–0.5 μm thick, so that the reconstructed image roughly corresponded to a section thickness of 2.5–3 μm. In control specimens without primary antibodies, no FITC fluorescence signal was observed (Supplementary Figure 1). Immunocytochemistry analysis required 4 animals per species. All experiments were performed in triplicate.

At the end of the in vivo electrophysiological session, cochleae were fixed by intracochlear perfusion of 3.5% glutaraldehyde in phosphate buffer (0.1 M, pH 7.4) through the perilymph compartment for 20 min, then removed and immersed in the same fixative overnight. The next day, fixed cochleae were rinsed in phosphate buffer and post-fixed in a 2% osmic acid solution for 1 h. After two rinses in phosphate buffer, the cochleae were dehydrated in a graded series of ethanol (30–100%), then in propylene oxide, and embedded in Epon resin at 60°C overnight. After trimming the cochleae to separate the different coils and to select samples of various positions along the organ of Corti (from base to apex), thin transverse sections (70–100 nm) were cut using a Leica-Reichert Ultracut E, counterstained with uranyl acetate and lead citrate, and observed using a transmission electron microscope (Hitachi 100) (n = 4 cochleae).

All chemicals and drugs were purchased from Sigma Chemical Company, except for AMPA, which was purchased from RBI and GYKI 53784 (LY303070) which was a generous gift of Margaret H. Niedenthal. Stock solutions were made and stored at -20°c before use, and final concentrations of the drug vehicle for GYKI in the bathing solution was kept at 0.4%.

Data were analyzed using MATLAB software. All values are expressed as means ± the standard error of mean (S.E.M) and were compared by the two-tailed Mann-Whitney Wilcoxon test or Student’s t-test. Significance was assessed at P < 0.05. Symbols used are as follows: *P < 0.05, ^**P < 0.01, and ^***P < 0.001.

We first investigated whether D-serine is present in mouse cochlear tissues. Liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) analysis revealed its presence, along with L-serine, glycine, and L-glutamate. Representative chromatograms of D-serine and L-serine in cochlear samples and the standard are shown in Figures 2A,B. Quantification analyses indicated that the level of D-serine (0.68 ± 0.09 pmol/μg protein; n = 8) represented ∼2.3% of L-serine (30.00 ± 5.00 pmol/μg protein; n = 8). Levels of cochlear glycine (31.00 ± 4.00 pmol/μg protein; n = 8) were ∼45 times higher than D-serine, while L-glutamate (89.00 ± 14 pmol/μg protein; n = 7) was by far the most abundant amino acid in the mouse cochlea. These results demonstrated that D-serine is present in the adult mouse cochlea at a relatively low level compared to L-serine, glycine, or glutamate (Figure 2C).

We next examined the cellular distribution of D-serine in the mouse cochleae. We foresaw two possible scenarios. Either D-serine present in some cochlear cells could reflect the site of synthesis, or its presence would correspond to D-serine originating from peripheral sources and taken up by specific cells to be further degraded. In the brain, D-serine is synthetized from L-serine by SR (Wolosker et al., 1999). Intense D-serine immunoreactivity was mainly observed in the cytoplasm of the sensory inner and outer hair cells and their supporting cells (Figure 1A), and of some cells located in the spiral ganglion region (Figure 1B). Serine racemase (SR) immunoreactivity was also pronounced in the cytoplasm of the same types of cochlear cells: i.e., sensory inner and outer hair cells and spiral ganglion neurons, and to a lesser extent, supporting cells of the organ of Corti (Figures 1C,D) of Ser^+/+ mouse cochleae, but not Ser^–/– (Figures 1E,F). These results therefore confirmed the specificity of the polyclonal anti-SR antibody. In rat cochleae, D-serine immunoreactivity was concentrated in IHCs and their supporting cells, and in the supporting cells surrounding the outer hair cells (OHCs) (Figures 1G,H). D-serine was homogeneously distributed in the cytoplasm of most cells, but not in the sensory OHCs (Figures 1H,I). In addition, the immunoreactivities of SR and D-amino acid oxidase (DAAO), the enzyme involved in the degradation of D-serine (Puyal et al., 2006; Pollegioni et al., 2018) were also concentrated in IHCs (Figures 1J,K).

Western-blot analyses revealed the presence of SR in homogenates of rat cochleae and of brain during early development (P0-P10) and an adult stage (P20, Figure 1L). Strikingly, a weak expression of SR was observed during the developmental stages of the cochlea (P0-P5), whereas its expression increased considerably at the onset of hearing (P10; Figure 1L). In addition, we also found SR in adult mouse and guinea-pig cochleae (Figure 1M). We also observed the presence of DAAO in the adult guinea-pig cochlea and brain (Figure 1N). Altogether, these results indicate that D-serine and its synthesis and metabolic enzymes are expressed in the cochlea throughout Rodentia, indicating that it might play important functions.

To uncover the role of D-serine in hearing function, we next took advantage of SR null mutant (SR^–/–) mice, which display 85% reduction of D-serine (Basu et al., 2009). Recording of auditory brainstem responses (ABR), which indicated the information transfer into the auditory pathway showed that ABR thresholds and wave I amplitudes in both SR^–/– (n = 7) and SR^+/– (n = 11) mice were virtually identical to those of SR^+/+ (n = 7) mice (Figures 3A,B). Distortion-product otoacoustic emissions (DPOAEs), reflecting OHC function, showed no significant difference in amplitude in the three strains (n = 5 per strain), suggesting that the function of OHCs is not altered at any of the stimulus frequencies and levels tested (Figure 3C). Taken together, these results indicate that the absence of endogenous D-serine in SR^–/– mice does not affect the normal mechanoelectrical transduction of hair cells nor synaptic transmission mechanisms.

The pharmacological effects of D-serine on gross cochlear potentials were probed in adult guinea pigs (n = 5). The compound action potential of the auditory nerve, which reflects the synchronous activation of the afferent fibers (CAP amplitude: N₁-P₁), the N₁ latency and the cochlear microphonic (CM) that reflects the receptor activity of the OHCs, were recorded after cumulative 10-min intra-cochlear perfusions of increasing doses of D-serine (Figure 4A). We showed that the initial perfusion of artificial perilymph (AP) did not induce significant changes in CAP amplitude or N₁ latency, nor in CM amplitude (Figures 4A–C). The D-serine effects were therefore compared with the cochlear potentials recorded after this initial control perfusion. D-serine (1–100 μM) induced a dose-dependent reduction in CAP amplitude (Figure 4A), but no significant changes in N₁ latency or in CM amplitude (Figures 4B,C). To facilitate the comparison of the effects of D-serine concentrations across animals, CAP amplitudes were expressed as the percentage of the control value obtained after the control perfusion at each level (100–40 dB). These percentages were then averaged across levels to provide an overall average percent change for each concentration of D-serine tested. The dose-response data obtained was then fitted to a curve using a non-linear least-squares logistic equation (Figure 4D) revealing an IC₅₀ for D-serine of 75 μM. However, a saturation effect of D-serine with higher doses i.e., 100 and 200 μM was observed and produced a maximal reduction in the mean amplitude of the CAP only to 54.4 ± 3.4% and 57.5 ± 4.2% at these two doses, respectively (Figure 4D). Finally, CAP amplitudes returned to control values after washing D-serine out of the cochlea with artificial perilymph (data not shown). Altogether, these results suggest that D-serine may act on the cochlear NMDARs.

Because the suppressing effect of D-serine on CAP could reflect cell damage or excitotoxicity, we next used transmission electron microscopy (TEM) to explore possible ultrastructural changes resulting from intracochlear perfusions of D-serine. As shown in Figures 4E,F, both IHCs and OHCs show a normal morphology and cytoplasm with a well-positioned nucleus, rather apically in the cell body of the IHC (Figure 4E) and basally in the OHC (Figure 4F). In both types of hair cells, mitochondria also had a normal aspect. Both types of hair cells showed well-formed synaptic contacts at their basal poles, the IHCs being contacted by dendrites from type I SGNs (Figure 4E). The OHCs essentially showed one or two large synapses with the axons from the medial efferent system (e in Figure 4F). These efferent synapses showed the classical presynaptic vesicle aggregates and an extended postsynaptic cistern. In addition, no evident ultrastructural abnormality was detected in the stria vascularis (Figure 4G). TEM examination of the cochlea perfused with increasing and cumulative doses of D-serine revealed no structural abnormalities such as damage to sensory hair cells, to the stria vascularis or any early signs of excitotoxicity of auditory-nerve dendrites. These data further support the idea that any functional effect of D-serine on hearing function is physiological, caused by the modulation of fast cochlear synaptic transmission.

For exploring the modulation of SGN activity, we next used cochlear neural calcium imaging of SGNs loaded with Fura 2-AM (inserts) (Figure 5A). Pharmacological stimulation with glutamate or AMPA, but not NMDA, induced calcium responses in SGNs. Glutamate (100 μM) application in the vicinity of SGN somata transiently increased the resting [Ca²⁺]_i (59.9 ± 4.9 nM; n = 59; Figure 5B). In contrast, a greater increase in [Ca²⁺]_i of 89.6 ± 4.7 nM (n = 59) was observed in the presence of 100 μM AMPA (Figure 5B). Note that NMDA (100 μM) produced only a very marginal increase in [Ca²⁺]_i of 6.8 ± 0.5 nM (n = 59; Figure 5B). These results are consistent with previous reports, demonstrating that fast cochlear synaptic transmission is preferentially mediated by AMPA but not NMDA receptors (Ruel et al., 1999, 2000, 2008; Glowatzki and Fuchs, 2002).

To probe the possible modulatory action of D-serine on cochlear glutamate receptors, SGNs were bathed with classical ionotropic glutamate receptor agonists in the presence or absence of 100 μM D-serine (Figure 5C). Application of D-serine alone in the bathing solution was ineffective in changing [Ca²⁺]_i (Figure 5C), but significantly increased the glutamate-induced [Ca²⁺]_i (60.3 ± 2.5 nM vs. 75.6 ± 5.0 nM, respectively; n = 23; Figures 5C,D). The effect of D-serine on glutamate responses was blocked by bath application of 7-chlorokynurenic acid (7-CKA, 50 μM), a selective antagonist of the glycine site of NMDAR (Zhu et al., 2016) (53.6 ± 3.5 nM; n = 23; Figures 5C,D). Similarly, blockade of D-serine-potentiated glutamate responses was observed in the presence of the competitive glutamate binding site antagonist D-APV (Zhu et al., 2016) (50 μM; n = 25; not shown).

As shown in Figures 5E,F, dose-response curves of glutamate-induced calcium responses obtained in the presence or absence of 100 μM D-serine revealed that the amino acid significantly increased the responses compared to glutamate alone. Thereby, EC₅₀ for glutamate was reduced from 45.8 μM in the absence, to 35.6 μM in the presence of D-serine (n = 212; Figure 5E). By contrast, the AMPA (100 μM)-induced increase of [Ca²⁺]_I (81.3 ± 5.0 nM; n = 29) was not affected by the presence of 100 μM D-serine (82.3 ± 4.6 nM; n = 29; Figures 6A,B). The pharmacological specificity of the AMPAR stimulation in the presence of D-serine was further verified by blocking the AMPA-induced response with the selective AMPAR antagonist GYKI 53784 (Ruel et al., 2000) (40μM; Figures 6A,B). These data suggest that the potentiating effect of D-serine on glutamate responses is most likely mediated by turning on silent NMDAR.

To further examine this scenario, we next investigated the effect of D-serine on NMDA-evoked calcium responses. When applied alone, both NMDA and D-serine did not evoke a significant [Ca²⁺]_i response (Figures 6C,D). In contrast, co-application of D-serine (100 μM) with NMDA (100 μM) evoked a strong increase in [Ca²⁺]_i (57.7 ± 3.6 nM; n = 21) compared to NMDA alone (14.7 ± 1.5 nM; n = 21; Figures 6C,D). The pharmacological specificity of the D-serine to modulate NMDAR was confirmed by blocking its potentiating effects with 7-CKA (50 μM; n = 21; Figures 6C,D). Finally, because NMDAR could be also sensitive to glycine, which can act as a putative co-agonist, we tested the effect of glycine. Application of NMDA (100 μM) in the presence of 20 μM glycine never induced calcium responses in SGNs in the 28 neurons tested. Moreover, increasing concentrations of glycine (0.1–50 μM) did not change basal [Ca²⁺]_i (Figures 6E,F). We thus concluded that cochlear D-serine, but not glycine, unsilences NMDAR responses located on SGN somata.

Because aberrant NMDAR activation has been consistently found to contribute to trauma-induced cochlear damage (Guitton et al., 2003; Ruel et al., 2008; Bing et al., 2015; Sanchez et al., 2015), we next examined the role of D-serine in noise-induced hearing loss (NIHL). Acoustic overstimulation sharply increases ABR thresholds (temporary threshold shift, TTS) equally in both SR^–/– (mean TTS: 64.8 ± 1.2 dB) and control SR^+/+ (mean TTS: 60.4 ± 0.7 dB) mice; regardless of the tested frequency (Figure 7A). In contrast, threshold recovery at the frequencies from 4 kHz to 10 kHz was significantly better (p < 0.05 or 0.01) in SR^–/– (mean PTS: 18.7 ± 1.5 dB) than SR^+/+ (mean PTS: 35.2 ± 0.4 dB) mice (Figure 7B). These results indicate that SR^–/– mice displayed greater capacity for recovery from NIHL than WT, suggesting that deletion of SR might favor the functional recovering of the cochlear nerve endings of SGNs from acoustic damage by preventing the harmful accumulation of D-serine and subsequent over-activation of NMDARs.