Editor's evaluation: GluA3 subunits are required for appropriate assembly of AMPAR GluA2 and GluA4 subunits on cochlear afferent synapses and for presynaptic ribbon modiolar–pillar morphology

Abstract

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Cochlear sound encoding depends on α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs), but reliance on specific pore-forming subunits is unknown. With 5-week-old male C57BL/6J Gria3-knockout mice (i.e., subunit GluA3KO) we determined cochlear function, synapse ultrastructure, and AMPAR molecular anatomy at ribbon synapses between inner hair cells (IHCs) and spiral ganglion neurons. GluA3KO and wild-type (GluA3WT) mice reared in ambient sound pressure level (SPL) of 55–75 dB had similar auditory brainstem response (ABR) thresholds, wave-1 amplitudes, and latencies. Postsynaptic densities (PSDs), presynaptic ribbons, and synaptic vesicle sizes were all larger on the modiolar side of the IHCs from GluA3WT, but not GluA3KO, demonstrating GluA3 is required for modiolar–pillar synapse differentiation. Presynaptic ribbons juxtaposed with postsynaptic GluA2/4 subunits were similar in quantity, however, lone ribbons were more frequent in GluA3KO and GluA2-lacking synapses were observed only in GluA3KO. GluA2 and GluA4 immunofluorescence volumes were smaller on the pillar side than the modiolar side in GluA3KO, despite increased pillar-side PSD size. Overall, the fluorescent puncta volumes of GluA2 and GluA4 were smaller in GluA3KO than GluA3WT. However, GluA3KO contained less GluA2 and greater GluA4 immunofluorescence intensity relative to GluA3WT (threefold greater mean GluA4:GluA2 ratio). Thus, GluA3 is essential in development, as germline disruption of Gria3 caused anatomical synapse pathology before cochlear output became symptomatic by ABR. We propose the hearing loss in older male GluA3KO mice results from progressive synaptopathy evident in 5-week-old mice as decreased abundance of GluA2 subunits and an increase in GluA2-lacking, GluA4-monomeric Ca2+-permeable AMPARs. Editor's evaluation Hearing is mediated by hair cells in the cochlea, which synapse onto the primary dendrites of the auditory nerve. This study shows how deletion of a postsynaptic glutamate receptor subtype strongly influences inner hair cell-spiral ganglion cell synapse formation. This work shows that pre- and post-synaptic changes intertwine dynamically, providing insights into how pathological outcomes arise from synaptic perturbations. https://doi.org/10.7554/eLife.80950.sa0 Decision letter Reviews on Sciety eLife's review process Introduction In the cochlea and ascending central auditory system, hearing relies on fast excitatory synaptic transmission via unique α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs) (Raman et al., 1994; Ruel et al., 1999; Gardner et al., 1999; Glowatzki and Fuchs, 2002). AMPARs are tetrameric ionotropic receptor channels comprised of GluA1–4 pore-forming subunits plus auxiliary subunits conferring distinct electrophysiological kinetics, unique molecular structures, and different pharmacological sensitivities (Jackson et al., 2011; Bowie, 2018; Azumaya et al., 2017; Twomey et al., 2018). In the adult brain, most AMPAR tetramers contain an RNA-edited form of the GluA2 subunit that makes the channel relatively impermeable to Ca2+, resulting in Ca2+-impermeable AMPARs (CI-AMPARs; Sommer et al., 1991; Higuchi et al., 1993). AMPARs lacking edited GluA2 are called Ca2+-permeable AMPARs (CP-AMPARs) because they have greater permeability to Ca2+ and larger overall ionic conductance, carried mainly by Na+ (Hollmann et al., 1991; Geiger et al., 1995). The expression of GluA2-lacking CP-AMPARs is downregulated in the developing brain (Pickard et al., 2000; Kumar et al., 2002; Henley and Wilkinson, 2016). However, CP-AMPARs persist or even increase with developmental maturation in some neurons of the auditory brainstem where CP-AMPARs enriched in GluA3 and GluA4 subunits are thought to be essential for fast transmission of acoustic signals (Trussell, 1997; Gardner et al., 2001; Lawrence and Trussell, 2000; Sugden et al., 2002; Wang and Manis, 2005; Youssoufian et al., 2005; Lujan et al., 2019). Cochlear afferent projections process fast auditory signals through innervation of the anteroventral cochlear nucleus, at the endbulb of Held synapses onto bushy cells, where the AMPARs are comprised mainly of GluA3 and GluA4 subunits with high Ca2+ permeability and rapid desensitization kinetics (Wang et al., 1998; Rubio et al., 2017). Mice lacking the GluA3 subunit have impaired auditory processing due to effects on synaptic transmission associated with altered ultrastructure of synapses between endbulbs and bushy cells (García-Hernández et al., 2017; Antunes et al., 2020). Mice lacking the GluA4 subunit have altered acoustic startle responses and impaired transmission at the next synaptic relay at the calyx of Held in the brainstem, a high-fidelity central synapse (Yang et al., 2011; García-Hernández and Rubio, 2022). The rapid processing of auditory signals in the brainstem is supported by high-fidelity initial encoding of sound at peripheral synapses between cochlear inner hair cells (IHCs) and spiral ganglion neurons (SGNs) (Rutherford and Moser, 2016; Rutherford et al., 2021), however, relatively little is known about how synapse ultrastructure, molecular composition, and overall abundance of cochlear AMPARs depends on specific pore-forming subunits. In mice, the developmental onset of hearing function begins at the end of the second postnatal week, followed by activity-dependent maturation and neuronal diversification that depends on glutamatergic transmission in the SGNs (Shrestha et al., 2018; Sun et al., 2018; Petitpré et al., 2018; Petitpré et al., 2022). Heterogeneity of the SGNs and the ribbon synapses driving them results in auditory nerve fibers with mutually diverse sound response properties correlated to differences in synapse structure and position of innervation on the IHC modiolar–pillar axis (Merchan-Perez and Liberman, 1996; Ohn et al., 2016). Each primary auditory nerve fiber (i.e., type-I SGN) is unbranched and driven to fire spikes by the release of glutamate from an individual IHC-ribbon synapse driving a single, large postsynaptic density (PSD) of approximately 850 nm in length, on average (cat: Liberman, 1980; mouse Payne et al., 2021). In the mature cochlea, the PSD is populated with AMPARs comprised of subunits GluA2–4 but not GluA1 (Niedzielski and Wenthold, 1995; Matsubara et al., 1996; Parks, 2000; Shrestha et al., 2018). Afferent signaling in the auditory nerve, as well as noise-induced excitotoxicity at cochlear afferent synapses (a form of synaptopathy), depends on activation of AMPARs (Ruel et al., 2000; Hu et al., 2020). However, the dependence of cochlear AMPAR function and pathology on specific pore-forming subunits is unclear. Here, we examined the influence of GluA3 subunits on afferent synapse ultrastructure and on AMPAR subunit molecular anatomy in the PSD of the auditory nerve fiber in the mouse cochlea, with attention to GluA2 and GluA4 flip and flop isoforms and to positions of SGN innervation on the IHC modiolar–pillar axis. At the central auditory nerve projection in the cochlear nucleus, at the endbulb of Held synapse, GluA3 is required for both post- and presynaptic maturation of synapse structure and function (García-Hernández et al., 2017; Antunes et al., 2020). Therefore, we also examined presynaptic ribbon morphology in relation to position on the IHC modiolar–pillar axis, which is expected to show smaller and more spherical ribbons on the side of the IHC facing the pillar cells and outer hair cells (pillar side) relative to the ribbons on the modiolar side facing the ganglion (Merchan-Perez and Liberman, 1996; Payne et al., 2021). Our findings in young adult male GluA3KO mice include dysregulation of GluA2 and GluA4 subunit relative abundance and alterations in pre- and postsynaptic ultrastructure associated with an increased vulnerability to glutamatergic synaptopathy at ambient, background levels of sound. These structural and molecular alterations at the cochlear ribbon synapses of presymptomatic 5-week-old male GluA3KO mice appear to be pathological, preceding the reduction in ABR wave-1 amplitudes observed at 2 months of age (García-Hernández et al., 2017). Results Cochlear responses to sound and transcriptional splicing of Gria2 and Gria4 mRNA isoforms are similar in 5-week-old GluA3WT and GluA3KO mice The four AMPAR pore-forming subunits GluA1–4 are encoded by four genes, Gria1–4. Here, we studied mice with normal or disrupted Gria3 (i.e., GluA3WT or GluA3KO; García-Hernández et al., 2017; Rubio et al., 2017). We first determined whether the 5-week-old C57BL/6J GluA3WT and GluA3KO differed in cochlear responses to sound. Complete statistical details are included in source data files online. Our ABR analysis showed no differences between genotypes in clicks or pure tone thresholds or wave-1 amplitude or latency (Figure 1A). We note that male GluA3KO and GluA3WT mice at 2 months of age have similar ABR thresholds but GluA3KO mice have reduced ABR wave-1 amplitudes (García-Hernández et al., 2017), suggesting cochlear deafferentation between postnatal weeks 5 and 9. Figure 1 Download asset Open asset ABRs, GluA1 and GluA2 immunolabeling and qRT-PCR in GluA3WT and GluA3KO. (A) Mean ABR thresholds ( ± standard deviation [SD]) were similar between male GluA3WT and GluA3KO mice (F(1, 168) = 2.659, p = 0.11; two-way analysis of variance (ANOVA); GluA3WT n = 13; GluA3KO n = 13). In GluA3WT and GluA3KO, there was a main effect of sound frequency (F(6, 168) = 78.78, p < 0.0001). For ABR wave-1 amplitudes ( ± SD), there was an effect of sound ilevel (F(11, 192) = 49.62, p < 0.0001), but mean amplitudes were similar between genotypes (F(1, 292) = 2.458, p = 0.118; two-way ANOVA). For ABR wave-1 latencies ( ± SD), there was a main effect of sound level in both genotypes (F(1, 288) = 47.11, p < 0.0001) and mean latencies were similar between GluA3WT and GluA3KO mice (F(1, 288) = 0.1273, p = 0.7215; two-way ANOVA). (B) Micrographs show immunolabeling for GluA1, GluA2, and GluA4 on spiral ganglion neuron (SGN) somata, and for GluA1 on the anteroventral cochlear nucleus (AVCN) and cerebellum (Crb) of GluA3WT and GluA3KO mice. Immunolabeling for GluA2 and GluA4 is observed on SGNs of both genotypes. In contrast, immunolabeling for GluA1 was not observed on SGNs nor in the AVCN of GluA3WT or GluA3KO mice, but was observed in the cerebellar Bergmann glia of both genotypes. Scale bars: 20 and 100 µm. (C) Images of Gria2 and Gria4 flip and flop, and GAPDH gels of GluA3WT and GluA3KO inner ears. Histograms show fold change ( ± SD) of qRT-PCR product. Paired t-test, two-tailed; bp: base pairs. Figure 1—source data 1 Data and statistical analysis for the ABR, PCR gels and qRT-PCR for GluA3WT and GluA3KO mice. https://cdn.elifesciences.org/articles/80950/elife-80950-fig1-data1-v2.zip Download elife-80950-fig1-data1-v2.zip Figure 1—source data 2 Raw unedited PCR acrylamide gels for Gria2 and Gria4 flip/flop in GluA3WT and GluA3KO. https://cdn.elifesciences.org/articles/80950/elife-80950-fig1-data2-v2.zip Download elife-80950-fig1-data2-v2.zip Figure 1—source data 3 Figures of the uncropped PCR acrylamide gels for Gria2 and Gria4 flip/flop in GluA3WT, and GluA3KO with relevant bands and lanes clearly labeled. https://cdn.elifesciences.org/articles/80950/elife-80950-fig1-data3-v2.zip Download elife-80950-fig1-data3-v2.zip We then asked if disruption of Gria3 affected expression of GluA1, GluA2, or GluA4 protein subunits in the cochlear spiral ganglion (the auditory nerve fiber, SGN somata) or the cochlear nucleus (Figure 1B, C). In WT mice, mature SGNs express GluA2, GluA3, and GluA4 subunits of the AMPAR, but not GluA1 (Niedzielski and Wenthold, 1995; Matsubara et al., 1996; Parks, 2000; Shrestha et al., 2018). With immunolabeling, we observed GluA2 and GluA4 in the SGNs of both genotypes, and we confirmed that SGNs lacked GluA1 in GluA3WT mice, as expected. Moreover, we did not observe compensatory GluA1 expression in SGNs of GluA3KO (Figure 1B, left). We also checked immunolabeling of GluA1 on brainstem sections containing the ventral cochlear nucleus and cerebellum. As expected, we found GluA1 immunoreactivity in the cerebellar Bergmann glia of GluA3WT and GluA3KO mice (Matsui et al., 2005; Douyard et al., 2007). In contrast, the ventral cochlear nucleus of 5-week-old mice lacked GluA1 immunoreactivity in GluA3WT as previously shown (Wang et al., 1998). Also similar to the SGNs, we found no compensational expression of GluA1 in neurons of the cochlear nucleus from GluA3KO mice (Figure 1B, right). At ribbon synapses in the cochlea, the PSDs on the postsynaptic terminals of SGNs expressed GluA2, 3, and 4 in GluA3WT as previously shown (Sebe et al., 2017), while PSDs in GluA3KO lacked specific immunolabeling for GluA3 (Figure 2). This confirmed the deletion of GluA3 subunits was effective in SGNs of GluA3KO mice, as previously shown in the cochlear nucleus (García-Hernández et al., 2017; Rubio et al., 2017), and was not associated with compensatory upregulation of GluA1 subunits. Figure 2 Download asset Open asset Immunohistofluorescence of α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR) pore-forming subunits GluA2, 3, and 4 on spiral ganglion neuron postsynaptic terminals in the organ of Corti. Confocal microscope immunofluorescence images of afferent ribbon synapses in organ of Corti whole-mount samples from GluA3WT (left) and GluA3KO mice (right) in the mid-cochlea. Anti-GluA2 (green), -GluA3 (blue), and -GluA4 (red) labels the postsynaptic AMPAR subunits encoded by the Gria2, Gria3, and Gria4 genes, respectively. Each subpanel displays synaptic puncta of approximately 12 inner hair cells (IHCs). Scale bars: 20 µm (A, C); 10 µm (B, D). (A) From top to bottom: GluA3WT in grayscale for anti-GluA2, 3, 4, and the sum of the three. In the GluA2 subpanel, the basolateral membranes of four IHCs are indicated by dashed curves. (B) Merged color image of the region of interest indicated by the dashed rectangle in panel A. Inset on right: enlargement of the dashed rectangular region of interest on left shows five postsynaptic AMPA receptor arrays of ribbon synapses from one IHC. (C) From top to bottom: GluA3KO in grayscale for anti-GluA2, 3, 4, and the sum of the three. (D) Merged color image of the region of interest indicated in panel C. Inset: enlargement of a rectangular region of interest shows several postsynaptic AMPAR arrays of ribbon synapses from one IHC. Two unique isoforms termed flip and flop are generated by alternative splicing of the mRNA encoding each of the pore-forming GluA subunits. In the brain, flip and flop splice variants are expressed in distinct but partly overlapping patterns and impart different desensitization kinetics (Sommer et al., 1990). The chicken and rat cochlear nuclei express predominantly the fast-desensitizing flop isoforms (Schmid et al., 2001; Sugden et al., 2002). With qRT-PCR, we determined whether Gria3 disruption altered posttranscriptional flip and flop splicing of mRNA for GluA2 (Gria2 gene) or GluA4 (Gria4 gene) in the cochlea. We measured the levels of flip or flop for Gria2 and Gria4 mRNA and compared GluA3WT to GluA3KO(Figure 1C). Although GluA3KO exhibited increased variance in transcript abundance among samples for all four isoforms and a trend toward greater mRNA abundance in all four comparisons, we found no significant differences between GluA3KO and GluA3WT. In addition, we calculated the flip/flop ratios for Gria2 and Gria4 in GluA3WT and GluA3KO and found no differences between genotype (Gria2 flip/flop ratio WT: 0.67 ± 0.02, GluA3KO: 0.67 ± 0.01, p = 0.8; Gria4 flip/flop ratio WT: 0.70 ± 0.001, GluA3KO: 0.70 ± 0.01, p = 0.9 paired t-test two-tailed). Thus, Gria3 disruption did not affect hearing sensitivity at 5 weeks of age, in contrast to 8 weeks of age when ABR peak amplitudes were reduced (García-Hernández et al., 2017). Taken together with previous work, this suggests the 5-week-old GluA3KO cochlea may be in a pathological but presymptomatic, vulnerable state. The levels of Gria2 or Gria4 flip or flop mRNA isoforms in cochleae of male mice at 5 weeks of age were similar in GluA3WT and GluA3KO. In both genotypes, expression of Gria2 and Gria4 flop isoforms appeared to exceed expression of the flip isoforms. Pre- and postsynaptic ultrastructural features of IHC-ribbon synapses are disrupted in the organ of Corti of GluA3KO mice Given the similarity of cochlear responses to sound measured by ABR in male GluA3WT and GluA3KO mice at 5 weeks of age, we next asked if the ultrastructure of IHC-ribbon synapses was similar as well. Qualitatively, in GluA3WT and GluA3KO the general structure and cellular components of the sensory epithelia were similar to published data of C57BL/6 mice (not shown; Ohlemiller and Gagnon, 2004). Synapses from the mid-cochlea of both GluA3WT and GluA3KO mice had electron-dense pre- and postsynaptic membrane specializations and membrane-associated presynaptic ribbons (Figures 3 and 4). Figure 3 with 1 supplement see all Download asset Open asset Ultrastructural features of GluA3WT IHC-ribbon mid-cochlear synapses. Transmission electron microscopy (TEM) micrographs of IHC synapses on the modiolar (A) and pillar sides (B). Aff.: afferent; IHC: inner hair cell; Eff.: efferent terminal. Scale bar: 0.5 µm. (A’, B’) Three-dimensional (3D) reconstructions of the IHC-ribbon synapses are shown in A and B. Representative serial electron micrograph images of modiolar- and pillar-side ribbon synapses are shown in Figure 3—figure supplement 1. (C) Plots of the quantitative data of the surface area, and volume of the postsynaptic densities (PSDs) and ribbons obtained from the 3D reconstructions of GluA3WT mice. The error bar corresponds to ± standard deviation (SD). (D) Plots of the quantitative data from single ultrathin sections of the linear length of the PSD, major axis, and circularity of the ribbons, and the average size of synaptic vesicles (SVs)/synapse of GluA3WT mice. The error bar corresponds to ± SD; one-way Anova * p < 0.05, ns: not significant; Mann-Whitney two-tailed U-test, ** p < 0.01, *** p < 0.0001, ns: not significant. Figure 3—source data 1 Data and statistical analysis for the ultrastuctural analysis of GluA3WT mice. https://cdn.elifesciences.org/articles/80950/elife-80950-fig3-data1-v2.zip Download elife-80950-fig3-data1-v2.zip Figure 4 with 1 supplement see all Download asset Open asset Ultrastructural features of GluA3KO IHC-ribbon mid-cochlear synapses. Transmission electron microscopy (TEM) micrographs of IHC synapses on the modiolar (A) and pillar sides (B) of GluA3KO mice. Aff.: afferent; IHC: inner hair cell. Scale bar: 0.5 µm. (A’, B’) Three-dimensional reconstructions of the IHC-ribbon synapses are shown in A and B. Representative serial electron micrograph images of modiolar- and pillar-side ribbon synapses are shown in Figure 4—figure supplement 1. (C) Plots of the quantitative data of the surface area and volume of the postsynaptic densities (PSDs) and ribbons obtained from the 3D reconstructions of GluA3KO mice. The error bar corresponds to ± standard deviation (SD). (D) Plots of the quantitative data from single ultrathin sections of the linear length of the PSD, major axis and circularity of the ribbons and the average size of synaptic vesicles (SVs)/synapse of GluA3KO mice. The error bar corresponds to ± SD; one-way ANOVA, ns: not significant; Mann-Whitney two-tailes U-test, * p < 0.05, ns: not significant. Figure 4—source data 1 Data and statistical analysis for the ultrastuctural analysis of GluA3KO mice. https://cdn.elifesciences.org/articles/80950/elife-80950-fig4-data1-v2.zip Download elife-80950-fig4-data1-v2.zip Ultrastructure in C57BL/6 GluA3WT A total of 29 synapses of GluA3WT mice were analyzed in three dimensions (3D) using serial sections (on average, 7 ultrathin sections per PSD). Of this total, 17 were on the modiolar side and 12 on the pillar side of the IHCs (Figure 3A, B, Figure 3—figure supplement 1). In our sample of the modiolar-side synapses, 11 had one single ribbon whereas 6 had two ribbons, so for the analysis of the PSD we classified the synapses as modiolar-1 and modiolar-2, for single and double ribbons, respectively. All the pillar-side synapses analyzed had a single ribbon. We then compared the PSD surface area and volume among the synapses of modiolar-1, modiolar-2, and pillar sides. One-way analysis of variance (ANOVA) comparison of the PSD surface area was significant (p = 0.007). Pairwise comparisons showed that the PSD surface areas were similar (p = 0.98) for single- and double-ribbon synapses of the modiolar side (modiolar-1 mean = 0.52 ± 0.15 µm2; modiolar-2 mean = 0.56 ± 0.15 µm2). However, in C57BL/6 WT mice, we observed that PSD surface area was larger for modiolar-side synapses compared to pillar-side synapses (p = 0.014 modiolar-1 vs. pillar, and p = 0.03 modiolar-2 vs. pillar; pillar mean: 0.40 ± 0.06 µm2) (Figure 3C, left). We then measured PSD volumes, which were ~2× larger on the modiolar side, on average, but not significantly different (p = 0.051 one-way ANOVA), (modiolar-1 mean = 0.010 ± 0.006 µm3; modiolar-2 mean = 0.008 ± 0.004 µm3; pillar mean = 0.005 ± 0.002 µm3) (Figure 3C, left). One-way ANOVA of the PSD linear length showed no significant differences among synapse type (p = 0.17; modiolar-1, n = 21, mean length: 666 ± 186 nm; modiolar-2, n = 6, mean length: 709 ± 234 nm; pillar, n = 16, mean length: 572 ± 135 nm) (Figure 3D, left). Overall, our analysis shows the PSD surface areas of modiolar-side synapses are significantly larger than those of the pillar side in GluA3WT on C57BL/6 background. Presynaptic ribbon volume of GluA3WT was similar between modiolar- and pillar-side synapses (p = 0.57; modiolar mean: 0.0029 ± 0.001 µm3; pillar mean: 0.0023 ± 0.0009 µm3; Mann–Whitney U-test, two-tailed) (Figure 3C, right). In contrast, the surface area of modiolar-side ribbons was found significantly larger (p = 0.002; modiolar mean: 0.141 ± 0.123 µm2; pillar mean: 0.074 ± 0.02 µm2; Mann–Whitney U-test, two-tailed) (Figure 3C, right). For all pairwise statistical comparisons of ultrastructure in the following, we used the two-tailed Mann–Whitney U-test. Analysis of the major axis and shape of synaptic ribbons in GluA3WT showed that the IHC synaptic ribbons on the modiolar side had longer major ribbon axes (p < 0.0001; mean: 274 ± 75 nm) and less circularity (p < 0.0001; mean: 0.51 ± 0.12) compared to the pillar-side ribbons (mean major axis: 180 ± 54 nm; mean circularity: 0.9 ± 0.06). These data show that ribbons on the modiolar side of GluA3WT IHCs are elongated, while those on the pillar side are more round in shape (Figure 3D, center), as previously shown for C57BL/6 mice at 5 weeks of age (Payne et al., 2021). Analysis of synaptic vesicle (SV) size showed that the SVs of modiolar-side synapses were larger (p = 0.0029) than those of the pillar-side synapses (modiolar mean: 36 ± 3 nm; pillar mean: 33 ± 4 nm; Figure 3D, right). In summary, GluA3WT synapses of the modiolar side had larger PSD surface areas, more elongated and less circular ribbons with greater surface area, and larger SVs compared with synapses of the pillar side. Ultrastructure in C57BL/6 GluA3KO From GluA3KO, a total of 26 synapses were analyzed in 3D with serial sections (on average, 7 ultrathin sections per PSD). Of this total, 16 were on the modiolar side and 10 on the pillar side of the IHCs (Figure 4A, B, Figure 4—figure supplement 1). As with synapses from GluA3WT, for the analysis of the PSDs of GluA3KO we classified the modiolar-side synapses as modiolar-1 (single ribbon; n = 11) or modiolar-2 (double ribbons; n = 5). In contrast to the pillar-side synapses of GluA3WT that had only single ribbons, we found two pillar-side synapses of GluA3KO cochleae with double ribbons (e.g., Figure 4—figure supplement 1). These pillar-side synapses with double ribbons were not included in our analysis. We then compared the PSD surface areas and volumes among synapses on the modiolar (modiolar-1 and modiolar-2) and pillar sides. One-way ANOVA comparison of the PSD surface area was not significant (p = 0.67; modiolar-1 mean: 0.51 ± 0.18 µm2; modiolar-2 mean: 0.55 ± 0.14 µm2; pillar mean: 0.57 ± 0.16 µm2). One-way ANOVA analysis of the PSD volume was not significant (p = 0.65; modiolar-1 mean: 0.0085 ± 0.004 µm3; modiolar-2 mean: 0.0068 ± 0.001 µm3; pillar mean: 0.0083 ± 0.003 µm3) (Figure 4C, left). The PSD linear lengths were not significantly different (p = 0.07, one-way ANOVA; modiolar-1, n = 30, mean length: 629 ± 147 nm; modiolar-2, n = 6, mean length: 543 ± 32 nm; pillar, n = 17, mean length: 684 ± 118 nm) (Figure 4D, left). Analysis of GluA3KO presynaptic ribbon volumes showed that pillar-side synapses (mean: 0.0042 ± 0.001 µm3) had larger volumes than modiolar-side synapses (mean: 0.0032 ± 0.001 µm3; p = 0.047) (Figure 4C, right), in contrast to GluA3WT. Also, unlike GluA3WT, the surface area was similar between ribbons on the modiolar and pillar sides of GluA3KO (modiolar mean: 0.14 ± 0.14 µm2, pillar mean: 0.17 ± 0.25 µm2; p = 0.91) (Figure 4C, right). The major ribbon axes from GluA3KO were similar on the modiolar side (mean: 199 ± 65 nm) and pillar side (mean: 201 ± 89 nm; p = 0.9). Modiolar-side ribbons had less circularity than pillar-side ribbons (modiolar mean: 0.75 ± 0.10; pillar mean: 0.85 ± 0.07; p = 0.01) (Figure 4D, right), but this difference was lesser than the difference observed in GluA3WT. Opposite to the pattern in GluA3WT, SVs of modiolar-side synapses were smaller than those of the pillar side in GluA3KO (modiolar: 35 ± 5 nm; pillar: 38 ± 3 nm; p = 0.04) (Figure 4D, right). In summary, unlike GluA3WT, GluA3KO synapses of the modiolar side had similar PSD and ribbon surface areas and ribbon long axes as pillar-side synapses, and smaller SVs, demonstrating disruption of modiolar–pillar synaptic differentiation during development. IHC modiolar–pillar differences are eliminated or reversed in GluA3KO We then compared PSDs and ribbons among GluA3WT and GluA3KO mice on the modiolar and pillar sides (Figure 5A). Overall, the PSD surface area and volume of the modiolar-side synapses (modiolar-1 and modiolar-2) were similar between genotypes (surface area: p = 0.85; volume: p = 0.62; one-way ANOVA). In contrast, the mean surface area and volume of the pillar-side PSDs were larger in GluA3KO than GluA3WT (surface area: p = 0.013; volume: p = 0.007) (Figure 5A, top). The average PSD length was similar between genotypes for the modiolar-side synapses (p = 0.29, one-way ANOVA). In contrast, as with surface area and volume for pillar-side synapses, the mean PSD length of GluA3KO pillar-side synapses was larger than GluA3WT (p = 0.02) (Figure 5B, top). Figure 5 Download asset Open asset Inner hair cell (IHC) modiolar–pillar structural differences in presynaptic ribbon size, ribbon shape, and vesicle size seen in GluA3WT were diminished or reversed in GluA 3KO. (A) Whisker plots show the quantitative data of the surface area and volume of the postsynaptic density (PSD) and ribbon volume of GluA3WT (black) and GluA3KO (gray) mice. The error bar corresponds to ± standard deviation (SD). (B) Whisker plots of the linear length of the PSD, major axis, and circularity of the ribbons of GluA3WT (black) and GluA3KO (gray) mice. Column histogram of the size of synaptic vesicles (SVs) of GluA3WT (black) and GluA3KO (gray). The error bar corresponds to ± SD; one-way ANOVA, * p < 0.05, ** p < 0.01, *** p < 0.005, p < 0.0001, ns: not significant; Mann-Whitney two-tailed U-test, * p < 0.05, ** p < 0.001. Figure 5—source data 1 Data and statistical analysis for the comparison of the ultrastructural analysis of WT vs. GluA3KO mice. https://cdn.elifesciences.org/articles/80950/elife-80950-fig5-data1-v2.zip Download elife-80950-fig5-data1-v2.zip Synaptic ribbon volumes differed among modiolar and pillar groups of GluA3WT and GluA3KO (p = 0.04, one-way ANOVA). Pairwise comparisons showed that ribbon volumes of modiolar-side synapses were similar between GluA3WT and GluA3KO (p = 0.93). In contrast, the pillar-side synapses were larger in GluA3KO than in GluA3WT (p = 0.03) (Figure 5A, bottom right). One-way ANOVA analysis of the ribbon surface area between modiolar- and pillar-side synapses was similar between GluA3WT and GluA3KO (p = 0.39) (Figure 5A, bottom left). Differences between the ribbon major axis were found between GluA3WT and GluA3KO (p < 0.0001, one-way ANOVA). On the modiolar side, analysis of the ribbon major axis length showed that those of the GluA3KO were significantly smaller than GluA3WT (p < 0.0001), whereas pillar-side synapses were similar in major axis length (p = 0.82) (Figure 5B, bottom left). Differences in ribbon circularity were also found between genotypes (p < 0.0001, one-way ANOVA). Paired comparisons showed that modiolar-side ribbons were significantly less circular in GluA3WT (p < 0.0001), whereas pillar-side ribbons were of similar circularity among genotypes (p = 0.62) (Figure 5B, bottom center). SVs size differed between genotypes (p = 0.0008, one-way ANOVA). Data showed that SVs of modiolar-side synapses were similar among genotypes (p = 0.84), while those of pillar-side synapses were significantly larger in GluA3KO (p = 0.0004) (Figure 5D, bottom right). Altogether, our data of 5-week-old male mice show the AMPAR subunit GluA3 is essen